The Bloomsbury Companion to Syntax 9781472542090, 9781441124609

Citation preview

Abbreviations

1 2 3 A A ABL ABS ACC ACT ADJ ADV AGR, agr AgrP ALL AN ANTIP AP APPL ARG-ST Art ASP Aux B, P BCG BEN C CAUS C-I CIRC CL, CLF CNP CNTXT COM COMP COMPL COND CONN COP CP CVB

first person second person third person adjective agent argument of a transitive clause or agent-like argument of a canonical transitive verb ablative absolutive accusative active adjective adverb, adverbial agreement agreement phrase allative Adjective-Noun antipassive Adjectival Phrase applicative Argument structure (feature) Article aspect Auxiliary, auxiliary constituent classes of Garifuna verb stems Berkeley Construction Grammar benefactive Complementizer, complementizer head causative Conceptual Interface circumstantial classifier common noun phrase Context (feature) comitative complementizer completive conditional connective copula complementizer phrase converb

9781441124609_Pre_Final_txt_print.indd viii

11/28/2012 3:28:27 PM

Abbreviations D DA DAT DE DECL DEF Dem Det DIR DIST DISTR DO DP DTRS DU DUR EA EPIS EPP ERG EST EVID EXCL EXPL F Foc FoFC FPL FPR fr FSG FUT FV G GB GEN GN HD-DTR HPSG I, II I I’ IA IC IE IMP IMPF INCL

determiner (category) Distinguished argument dative Definiteness Effect declarative definite Demonstrative Determiner direct distal distributive direct object Determiner Phrase Daughters (feature) dual durative external argument epistemic Extended Projection Principle ergative Extended Standard Theory evidential exclusive expletive feminine Focus Final over Final Constraint feminine plural Focus Prominence Rule Frame feminine singular future final vowel goal argument of a three-argument clause Government and Binding genitive Genitive-Noun Head daughter (feature) Head-Driven Phrase Structure Grammar classes of Chickasaw agreement markers inflection (Infl) intermediate-level phrase headed by I (Infl) internal argument Independent clause Indo-European imperative imperfective inclusive

ix

9781441124609_Pre_Final_txt_print.indd ix

11/28/2012 3:28:27 PM

Abbreviations IND INDF INF Infl INFL INS INTR INV IO IP IPFV IRR LF LOC M, Ma MP MSU MTR N NN’ NA n-constituent NEG NegP N-F NG NMLZ NOM NP NPST NQA NSR NTC O, OBJ OBL OBV OS OSV OT OV OVS P P

P1

indicative indefinite infinitive inflection (category) inflection node instrumental intransitive inverse indirect object inflectional phrase imperfective irrealis logical form locative masculine Minimalist Program Meaning shift unit Mother (feature) neuter, noun non- (e.g. NSG: nonsingular, NPST: nonpast) intermediate-level phrase headed by N Noun-Adjective negative constituent negation, negative negation phrase non-feminine Noun-Genitive nominalizer/nominalization nominative Noun phrase; Noun-Postposition nonpast Nominal-Quantificational asymmentry Nuclear Stress Rule No Tempering Conditions Object oblique obviative object-subject (constituent order) Object Subject Verb Optimality Theory Object Verb object-verb-subject (constituent order) passive patient argument of a two-argument transitive clause, patient-like argument of canonical transitive verb, Possessor, Preposition/Postposition first position, one-place predicate

x

9781441124609_Pre_Final_txt_print.indd x

11/28/2012 3:28:27 PM

Abbreviations P2 PART PASS PF PFV PG PHON PL PLD PN POS POSS PP PRED PRES PRF PRO Pro PROG PROH PROX PRS PST PTC PTCP PURP PVF Q QUOT R REC.PST RECP REFL REL REM.PST RES S(OV) S(VO) S Sa SAI SBCG SBJ SBJV SEM SG SO

second position; two-place predicate particle passive perfect perfective parasitic gaps Phonology (feature) plural Primary Linguistic Data Preposition-Noun Part-of-speech possessive prepositional phrase predicative present perfect null subject pronoun in a control clause Pronoun progressive prohibitive proximal/proximate present past partice participle purposive, purpose perfective question: particle/marker quotative Relative Clause recent past reciprocal reflexive relative, relitivizer remote past resultative subject object verb Subject verb object sentence, single argument of canonical intransitive verb, Subject single agent role argument of an intransitive clause subject auxiliary inversion Sign-Based Construction Grammar subject subjunctive Semantics (feature) singular subject-object (constituent order)

xi

9781441124609_Pre_Final_txt_print.indd xi

11/28/2012 3:28:27 PM

Abbreviations SOV Sp Spec SR STAT SUB SV SVO SYN T, P, R T t TDPST TNS TOP, Top TopP TP TR , TRANS TVZ UG UTAH VS v V V’ V1 V2 V3 V4 VAL VIS VO VOC VOS vP VP VS VSO WCO X X’ X0 XP XP

subject-object-verb (constituent order) single patient role argument of an intransitive clause Specifier, specifier position (X-bar theory) same reference, switch reference stative subordinate Subject Verb subject-verb-object (constituent order) Syntax (feature) classes of Garifuna agreement markers tense (category), theme argument of a three-argument clause trace today past tense Topic Topic Phrase tense phrase transitive Tlacolula Valley Zapotec Universal Grammar Uniformity of Theta-Assignment Hypothesis verb-subject (constituent order) light verb verb (category) intermediate-level phrase headed by V Intransitive verb Verb second, Transitive verb Verb which occurs with a prepositional phrase Verb which typically occurs with a sentential object (complementizer) Valence (feature) visible Verb Object vocative verb-object-subject (constituent order) light verb phrase Verb Phrase Verb Subject verb-subject-object (constituent order) weak cross over head (X-bar theory), intermediate projection intermediate-level phrase (X-bar theory) head maximal projection phrase headed by X (X-bar theory), X phrase (with X = variable)

xii

9781441124609_Pre_Final_txt_print.indd xii

11/28/2012 3:28:27 PM

Contributors

Mark Baker Rutgers University USA

Leonid Kulikov Ghent University Belgium

William Croft University of New Mexico Albuquerque USA

Terje Lohndal Norwegian University of Science and Technology Norway

Holger Diessel University of Jena Germany

Giuseppe Longobardi University of Trieste Italy

Thórhallur Eythórsson University of Iceland Iceland

Luis López University of Illinois Chicago USA

Maria Freddi University of Pavia Italy Maria Teresa Guasti University of Milano-Bicocca Italy Liliane Haegeman Ghent University Belgium Marja-Liisa Helasvuo University of Turku Finland Paul Hopper Carnegie Mellon University USA Edward Keenan University of California at Los Angeles USA

9781441124609_Pre_Final_txt_print.indd xiii

Silvia Luraghi University of Pavia Italy Caterina Mauri University of Pavia Italy Laura Michaelis University of Colorado Boulder USA Marianne Mithun University of California at Santa Barbara USA Pamela Munro University of California at Los Angeles USA Jairo Nunes University of São Paulo Brazil

11/28/2012 3:28:27 PM

Contributors Francisco Ordóñez State University of New York Stony Brook USA Claudia Parodi University of California at Los Angeles USA Doris Payne University of Oregon USA Cecilia Poletto Goethe-Universität Frankfurt am Main Germany Carlos Quicoli University of California at Los Angeles USA

Luigi Rizzi University of Siena Italy Vieri Samek-Lodovici University College London UK Giuseppina Silvestri University of Pisa Italy Jae Jung Song University of Otago New Zealand Tim Stowell University of California at Los Angeles USA

xiv

9781441124609_Pre_Final_txt_print.indd xiv

11/28/2012 3:28:28 PM

Introduction

Syntax is perhaps the field of linguistics that has generated the biggest number of competing theoretical frameworks, to the effect that mutual communication is o en virtually inexistent. It is therefore a big challenge to try and induce dialogue among researchers and students of different theoretical persuasions. In this book we present a collection of articles that touch on a wide range of aspects of syntactic research, with the aim to bring together different theoretical treatments of partly similar issues. In this way we offer a point of departure for those who are willing to foster intertheoretical dialogue. The book is divided into three parts. The first part, which is focused on methodology, contains papers that discuss the way in which hypotheses are formulated and tested as well as current practices in data collection, in particular in field research and by handling electronic corpora. The second part, which constitutes the bulk of the volume, is an exhaustive collection of papers that discuss a variety of topics, including constituency and constituent structure, word order variation and special word order patterns, such as those in which certain items occur in second position in a sentence, grammatical relations, verbal voice and transitivity, coordination and subordination, negation, syntactic change and language acquisition. All these topics are treated from specific theoretical viewpoints, including formal and functional ones. In the third part we present some current approaches to syntax which we hold to be especially promising, and can provide the reader with some insight in the variety of possible directions for future research. A brief overview of the content of each chapter is given below. In Chapters 1 and 2, “Hypothesis formation” and “Hypothesis testing in the Minimalist Program,” William Cro and Mark Baker take opposite stances on the issue of how hypotheses are formed and tested when doing syntactic, or more in general linguistic research. While Cro suggests a distributional analysis as the basis for hypothesis formation, in which linguistic data play a crucial role in directing the linguist’s theoretical assumptions, Baker favors an approach in which “one first considers the logical consequences of some theory that one finds appealing, and then goes on to test whether those consequences match up with observable data.” Thus Cro favors a bottom up approach, while Baker suggests that a top down view is more promising. The contrast provides the reader with insights in two opposite ways of conceiving of the proper object and methodology of linguistic research. 1

9781441124609_Intro_Final_txt_print.indd 1

11/20/2012 9:25:16 AM

The Bloomsbury Companion to Syntax

The next two chapters, 3 “Field methods in syntactic research” by Marianne Mithun and 4 “Corpus methods in syntax” by Maria Freddi, go deeper into the issue of linguistic data, and show that the way in which they are collected is far from being only a practical matter. In particular, Mithun stresses the importance of avoiding interference from the linguist’s native language when eliciting data from typologically distant languages. She shows that even widely used parallel corpora provided by translations that are commonly considered idiomatic, such as Bible translations, may not reflect actual language usage thus inducing false assumptions regarding language structure. The likelihood of certain topics to occur in spontaneous discourse has consequences on the type of grammatical constructions used by speakers and on their entrenchment in cognition. With data from her own fieldwork, Mithun illustrates a number of syntactic characteristics of Mohawk, including constituency, topic and antitopic construction, word order and subordination. She also shows how independent evidence is gained by the study of intonation, which can help isolate syntactic units. Attention for actual data and their frequency is not only crucial when describing “exotic” languages: much to the contrary, making use of wide collections of data changes one’s understanding of language and syntax in general, by prompting “a novel view of syntax as inherently probabilistic,” “challenging a categorical modeling of language” and replacing “grammaticality judgments with the recognition of variability, gradience and fuzzy phenomena” as Freddi puts it. Freddi further shows how research based on distribution and collocations favors a view of language that refrains from rigid separation of syntax from other linguistic levels, most notably the lexicon. This view is in line with recent approaches to constructions as instantiated in various versions of Construction Grammar (see for example Goldberg 1995, Lancaster 1987). The second part of the book opens with three papers which view constituency form different perspectives. Chapter 5 “Constituents,” by Marja-Liisa Helasvuo, takes a usage-based approach to language, whereby structure is viewed as emerging form usage in actual discourse, a process she calls “emergent constituentization.” In much the same way as Mithun, Helasvuo also stresses the importance of prosody and intonation as means for singling out units in discourse. Based on data from speakers’ repair practices and production of co-constructions in languages such as English, Japanese and Finnish, she further shows how researchers can observe the way in which speakers monitor the emergence of syntactic structures. Jairo Nunes discusses “X-bar Structure and Minimalism” in Chapter 6, thus following what Baker would call a typical top down approach to constituency. Under the assumption that “words combine into larger units with hierarchical structure,” Nunes shows how the Minimalist Program has provided and is currently providing new insights to the issue of constituency as treated within the generative grammar framework. Chapter 7 “The Structure of NPs: Some insights on Case, empty 2

9781441124609_Intro_Final_txt_print.indd 2

11/20/2012 9:25:17 AM

Introduction

categories, and poverty of stimulus” by Giuseppe Longobardi and Giuseppina Silvestri provides a formal treatment of noun phrase structure focusing in particular on the argument structure of nouns. Longobardi and Silvestri set out to describe the mapping of arguments to thematic structure as well as conditions on the formal realization of those phonologically expressed. In this way and based on data from several unrelated languages, they claim to provide “theory-neutral evidence for the existence of phonetically empty categories and for the role of poverty-of-stimulus.” Their method of data collection is in sharp contrast with the method used by Helasvuo: rather than collecting evidence from corpora, Longobardi and Silvestri resort to introspection, either their own or by other linguists. Thus the two papers complement each other by providing an illustration of how different theoretical assumptions lead to different methodologies. The latter, in their turn, have an impact on what linguists actually find in the data and consequently on what they view as relevant to syntactic structure. This also means that what is conceived of as being theory-neutral also crucially depends on theoretical assumptions: an only superficially surprising conclusion. Chapter 8 “Word order” by Pamela Munro provides a typological perspective on various issues connected with word order, including some which are then treated in deeper detail in the following three chapters. Munro starts by tackling the question of what should count as “basic” in the description of word order. She then examines the correlations proposed by Greenberg (1963) and to what extent they can really count as universals. Examples provided by various languages tend to attest to “mixed” or “non-harmonic” patterns, rather than present strict instantiations of Greenberg’s basic types. Under the heading of “variation,” Munro further discusses scrambling, second position phenomena including V2 and the placement of Wackernagel’s clitics, and word order within constituents. In the last part of her paper, Munro presents the reader with an insightful and useful discussion of the way in which word order may be treated under different theoretical assumptions, as well as of some issues related to word order change. Scrambling is then taken over again as the central topic of Chapter 9 “Scrambling” by Francisco Ordóñez, which also constitutes a nice example of how a specific theoretical framework, namely generative grammar, deals with a word order related topic. Ordóñez discusses two approaches to scrambling within generative grammar, that is the base generation approach and the movement approach, and shows how the two approaches relate to one’s assumptions regarding other matters such as configurationality and feature checking. Chapter 10 “On V2 types” by Cecilia Poletto discusses the V2 phenomenon, which is of great relevance for any theory of word order, and has been a traditional topic of research within formally oriented theories. In her paper, Poletto offers a comprehensive survey of 30 years of research on V2 in Germanic and Romance languages, contrasting V2 with scrambling because, 3

9781441124609_Intro_Final_txt_print.indd 3

11/20/2012 9:25:17 AM

The Bloomsbury Companion to Syntax

she maintains, contrary to the latter “there are no approaches that see V2 as basic word order, but only movement approaches, which differ only in the trigger of verb movement.” (This of course only applies if one does not consider possible non-formal accounts of the V2 phenomenon.) Second position is also relevant to clitic placement as shown in chapter 11 “Clitics” by Silvia Luraghi. Luraghi presents a discourse based account of different tendencies in the placement of so-called special clitics (see Zwicky 1977), including P2 or Wackernagel’s clitics and verb-based clitics like those of the Romance languages. She argues that differences in placement rules of pronominal clitics depend on competing motivations. Pronominal clitics refer to shared information and as such tend to occur early in the sentence; they also tend to be attracted by the verb as they normally refer to core arguments. Conflicting motivations are also the cause responsible for change in clitic placement rules. Grammatical relations constitute the main focus of the next two chapters: 12 “Changing the subject: Shi ing notions about subjecthood in generative grammar” by Tim Stowell and 13 “Alignment and grammatical relations” by Doris Payne. Stowell analyzes various modifications regarding the notion of subject in generative syntactic theory over the last four decades, mostly based on data from English. Payne takes a wider look at possible ways in which languages express the basic grammatical relations of subject of transitive and intransitive verbs and of direct object, resulting in different alignment patterns. She starts with a cross-theoretical overview of assumptions regarding grammatical relations and proceeds to present alignment types and their features in typologically different languages, starting with familiar Nominative-Accusative and Ergative-Absolutive system and ending with languages, such as Tibetan, which are held to have no grammatical relations. Valency changing operations are discussed in the next group of chapters, including 14 “Passives and antipassives” by Edward Keenan, 15 “Middles and reflexives” by Leonid Kulikov and 16 “Causatives: Structure, type, and distribution” by Jae Jung Song. Keenan offers a wide ranging typological survey of passive constructions in the world’s languages, with a shorter account of antipassive constructions. He describes types of morphological passives, and goes deeper into a number of issues such as passive constructions with ditransitives, causatives and applicatives, control verbs and object raising, and provides data on impersonal passives as well as on the expression of agents with personal passives. In Chapter 15, Kulikov captures the most relevant features of middles and reflexives within the approach developed by the Leningrad/St. Petersburg Typology Group. Since the categorical domain of reflexives and middles overlaps with such categories as transitivity and voice, he shows how many scholars have analyzed the middle voice as a syncretic merger of several linguistic categories such as reflexive, passive, anticausative and others. In his account of causative constructions Jae Jung Song provides an overview of typological 4

9781441124609_Intro_Final_txt_print.indd 4

11/20/2012 9:25:17 AM

Introduction

research carried out on this topic, in particular regarding the role of grammatical relations and semantic types of causation. He discusses the variety of syntactic causatives and their global distribution based on an innovative typology proposed by Song (1996, 2005a and b). He finds several subtypes of syntactic causative(s) that before his research had been lumped together under a not very revealing syntactic category. The following group of chapters treats complex sentences. In Chapter 17 “Coordination” Caterina Mauri discusses from a functional-typological point of view the main morphosyntactic and semantic features characterizing symmetric and asymmetric coordination across languages, and explores the main types of coordination relation: combination, contrast and alternative. Following this perspective, she assigns great value to the variation attested in the world languages. Subordination comprises three chapters of the handbook: 18 “Wh-movement” by Luis López, 19 “Complementation” by Claudia Parodi and Carlos Quicoli and 20 “Adverbial subordination” by Holger Diessel. López presents an introduction to the syntax and semantics of wh-phrases and wh-movement within the Minimalist Program, a formal model within the Generative framework, as proposed by Chomsky (2000) and developed by Pesetsky and Torrego (2001). In addition, this paper includes an analysis of the complex constructions wh-in situ, multiple wh-movement and partial wh-movement. Parodi and Quicoli also discuss complementation within the Generative framework. They examine different types of complements opposing them to adjuncts and show that while complements enter into Subcategorization and Selection restrictions with major lexical categories such as verbs, adjectives and nouns, adjuncts are exempt from such restrictions. They further examine the interaction of syntactic rules and the principles that are involved in such interactions in complement structures. Diessel takes a functional-typological perspective on adverbial clauses, and shows that they can serve various pragmatic functions which are crucially determined by their position relative to the associated main clause. The particular pragmatic functions of preposed and postposed adverbial clauses are reflected in their morphosyntactic properties and are related to aspects of their meaning and communicative use. Even if adverbial clauses are defined as a consistent grammatical category, the notion of adverbial subordination refers to a very diverse set of constructions that display a vast amount of variation within and across languages. In 21 “Negation” Liliane Haegeman and Terje Lohdal, a er introducing the Generative approach to negation, focus on negative concord, one of the most debated issues on negation. They show that Multiple Agree does not work for negative concord in West Flemish. In this language negative concord, instead, seems to be prone to a binary Agree analysis to account for the data. The last two chapters are devoted to more general topics: 22 by Thórhallur Eythórsson treats “Syntactic change” while in 23 Maria Teresa Guasti discusses 5

9781441124609_Intro_Final_txt_print.indd 5

11/20/2012 9:25:17 AM

The Bloomsbury Companion to Syntax

“The acquisition of syntax.” Eythórsson presents three case studies which illustrate current issues in theoretical discussion about syntactic change: the “personalization,” or Nominative Substitution in Scandinavian, the new passive in Icelandic and the loss of OV order in the VP in Icelandic. In the last part of his paper, Eythórsson combines the Comparative Method with the model of Construction Grammar and shows that reconstruction is feasible in historical and comparative syntax, just as in the other domains of grammar and in the lexicon. The principal goal of linguistic reconstruction is to model the grammar of the prehistoric stages from which the attested languages can be derived. As a consequence, syntactic reconstruction is a necessary part of any serious theory of syntactic change. Regarding language acquisition, Guasti finds that children obey language specific word order rules since their first word combinations. She claims that they exploit phonological properties to figure out whether their language is head-complement or complement-head. Moreover, children move elements in forming questions, but they opt for the non-movement strategy whenever available. The use of relative pronouns takes time, even if movement is applied in relative clauses. This is so because a more economical option exists, which is the generalized use of the complementizer. Relativizing grammatical functions other than objects is costly as well, as the subject in the relative clause may interfere in the relation between the relative head and its copy in the position of interpretation. The interpretation of weak pronouns is hindered by their ambiguity, supportive contexts clarify their meaning. The third part of the book is devoted to new directions in syntactic research. It comprises four cutting edge articles, each of which provides stimulating perspectives on syntax. In 24 “Usage and syntax” Paul J. Hopper presents a model that supports the study of syntax by analyzing spoken conversational, dialogic discourse. This innovative type of study diverges from traditional syntactic studies based on monologic, fabricated sentences by linguists. Spoken conversational, dialogic discourse as proposed by Hopper has as a fundamental building block the turn constructional unit (TCU), which is the short complete utterance characteristic of active dialogue. Hopper clarifies that this model shi s the starting point for the study of syntax and reverses the directionality of long-established syntactic theories. In fact, it places linguistic structure in a temporal interactive framework. In doing so, this chapter provides further theoretical background to Helasvuo’s account of constituency in Chapter 5. In 25 “Construction Grammar and the syntax-semantics interface” Laura Michaelis opposes generative grammar to Sign-Based Construction Grammar (SBCG). She claims that traditional Generative syntax contends that a grammar is organized in the mind of the speaker as a number of hermetically sealed modules, which in the course of a derivation hand off data one to the other. To the contrary, SBCG assumes that grammar contains no such modules, but rather 6

9781441124609_Intro_Final_txt_print.indd 6

11/20/2012 9:25:17 AM

Introduction

that it is an inventory of signs, complexes of linguistic constraints that contain information about form, meaning and use. Such constructions are the means by which simpler signs are combined into more complex signs. The notion of construction, on this view, formalizes in a constraint-based architecture certain concepts of traditional grammar. While transformational-generative syntacticians have sought to import syntactic generalizations into the lexicon, construction grammarians have moved in the opposite direction. For them the lexicon provides a model for the syntax-semantics interface. Lexical-class constructions define classes of lexemes or words, and combinatory constructions define classes of phrases. Rather than seeing the lexicon characterized as a bag of idiosyncratic form-meaning pairings, SBCG proposes a lexicon structured by hierarchically organized lexical classes and extends this model to relations among phrasal classes and constructions. It follows that syntax and semantics are not viewed as independent modules. Chapter 26 “Topic, Focus, and the cartography of the le periphery” by Luigi Rizzi reminds us that until the mid-1990s a widespread assumption was that a single X-bar projection, the complementizer phrase or CP would suffice to structurally express all properties of the sentence, much as a single IP projection had previously been assumed to express the internal structure of the clause. However, the cartographic projects by Rizzi and other drew much attention to the study of the precise structural shape of these zones, showing the necessity of assuming richer representations of syntactic structures. In his chapter Rizzi illustrates some fundamental interface properties of Topic and Focus, and discusses their structural position in a cartography of the le periphery of the clause. His data, based on introspection, mostly come from Italian, a language displaying a particularly rich usage of the initial periphery. He shows that discourse-related Topic (including Comment and Focus) and Presupposition articulations are expressed in the initial periphery of the clause, and that they manifest well-differentiated properties at the syntactic level and at the interpretive interfaces. Moreover, he illustrates the main properties of these articulations, and presents a structural analysis based on results of the cartographic projects. According to this line of research, syntactic representations specify richly articulated structural zones. These syntactic representations explicitly encode information which is used by the interpretive systems to express the special properties of Topic and Focus at the interfaces with sound and meaning, as spectrograms show. Finally, in 27 “Optimality Theory as a possible framework for the Minimalist Program” Vieri Samek-Lodovici examines the relation between Optimality Theory (OT) as proposed by Prince and Smolensky (1993, 2004) and the Minimalist Program (MP) developed by Chomsky (1995, 2000, 2005). He argues that OT offers a useful theoretical tool for pursuing the core goals of the MP. In fact, he shows that OT and the MP are not only logically independent of 7

9781441124609_Intro_Final_txt_print.indd 7

11/20/2012 9:25:17 AM

The Bloomsbury Companion to Syntax

each other, but also compatible with one another. By comparing OT and nonOT analyses of Italian rightmost focus, he demonstrates that the OT analysis best captures the MP’s core goals. From a logical point of view, the MP and OT are independent of each other. Grammar may or may not be entirely reducible to interface constraints as maintained by the MP, and in both cases the constraints of grammar may or may not conflict with one another, as maintained by OT. If constraints at the interfaces do conflict, it is not possible to pursue the MP without a theory of constraint conflict such as OT. The analysis of Italian focused subjects illustrates this point. A conflict-based approach to grammar allows deriving syntactic properties from independently necessary prosodic constraints. The corresponding OT analysis provides a formal account of exactly how prosody determines the syntactic distribution of Italian focus. Other current analyses of the same phenomenon require syntactic accounts of stress-assignment which isolate syntax from the prosodic interface. The book is completed by two further sections: a glossary of key terms as they emerge from the collected papers, and an appendix, “Resources for syntax” by Maria Freddi and Silvia Luraghi. The latter provides readers with information for corpus-based syntactic research by describing a number of current projects aimed to syntactic annotation of electronic corpora. We started planning this book just a er publishing our Key Terms in Syntax and Syntactic Theory (Continuum 2008), in order to provide a wider perspective on the multifaceted field of syntax. We commissioned the first set of articles four years ago and, in spite of some delay, we had a great response. The result is a collection of high standard, cutting hedge articles for which we want to thank and congratulate all contributors. Our gratitude also goes to Chiara Meluzzi, who prepared the final manuscript and the bibliography, and helped us organize the glossary, as well as to Roberto Cirillo, who helped us read the proofs, and to Lizy Moromisato, who corrected the author and language indices.

8

9781441124609_Intro_Final_txt_print.indd 8

11/20/2012 9:25:17 AM

1

Hypothesis Formation William Croft

This chapter describes methods for developing and supporting hypotheses about the analysis of grammatical phenomena. This subject is o en taken for granted as it involves methods of syntactic argumentation rather than explicitly stated theoretical assumptions or explicitly presented empirical linguistic data. Nevertheless, the methods of syntactic argumentation merit explicit discussion as well. The methods differ considerably in different approaches to syntax, and there are a number of usually hidden, unwarranted assumptions in syntactic argumentation that either must be given empirical support or are actually fallacies in argumentation. The basic method for constructing linguistic hypotheses is Distributional Analysis. The name “distributional analysis” was used by the American structuralists to describe the method, but it goes under different names in contemporary theories: applying morphosyntactic (grammatical) tests, satisfying grammatical criteria, possessing grammatical properties, displaying grammatical behavior, and so on. All of these terms essentially describe the same method. In distributional analysis, a hypothesis about the proper analysis of a syntactic unit is formed by examining the occurrence or nonoccurrence of that unit in a range of syntactic structures. A relatively simple example is the analysis of certain argument phrases in English (more precisely, certain phrases functioning as arguments of a verb with which they are combined). Certain argument phrases of verbs occur postverbally and without a preposition in the English Active Voice construction, as in (1): (1) (a) (b) (c) (d)

The boys ate all the pizza. The road extends 10 miles into the mountains. The police tapped my phone. 2010 saw the first hung parliament in Britain for over 30 years.

This is described as the distribution (or a property or behavior) of the phrases in question in a particular position, or role, in the Active construction. Alternatively, occurrence of the phrases in question in the particular role is a called a test or 11

9781441124609_Ch01_Final_txt_print.indd 11

11/20/2012 9:25:56 AM

The Bloomsbury Companion to Syntax

criterion for the phrase to belong to a grammatical category. On the basis of this distribution, we may hypothesize that the phrases in boldface form a single grammatical category. However, we may also compare the distribution of the phrases in (1) in a particular role in another syntactic construction, the English Passive Voice construction, as in (2): (2) (a) (b) (c) (d)

All the pizza was eaten by the boys. *Ten miles is extended by the road into the mountains. My phone was tapped by the police. *The first hung parliament in Britain for over 30 years was seen by 2010.

The distribution (property, behavior, etc.) of the boldface phrases differs in the Passive construction: examples (2b) and (2d) are ungrammatical by introspective judgment. Hence there is a difference in distribution between all the pizza and my phone in (1a and c) respectively and ten miles and the first hung parliament . . . in (1b and d) respectively. The distributional method provides the empirical data that functions as the basis for forming linguistic hypotheses. Before discussion on how hypotheses are formed using distributional data, three observations can be made about distributional analysis. First, the distribution of a syntactic unit is defined over a set of constructions. The “tests,” “criteria,” “properties,” or “behavior” are actually constructions, which are of course syntactic structures in their own right. The distributional method defines a mapping between one set of syntactic structures—the words, morphemes, phrases, etc. in question—and another set of syntactic structures— the constructions used as tests, criteria, etc., or more precisely, the relevant grammatical role in the constructions used. For example, (1) and (2) illustrate a mapping between the set of phrases in boldface and the relevant roles in two constructions: the postverbal NP role in the Active Voice, and the preverbal NP role in the Passive Voice. Hence, the identity of the constructions used and their distinctness from each other is presupposed in the distributional method. Second, the distributional method is generally used to form hypotheses about the syntactic unit that occurs/does not occur in the relevant roles of the constructions used as tests/criteria, etc. Again, the identity of the constructions used as tests/criteria, etc. is taken for granted, and the distributional method is taken to reveal something about the identity and distinctness of the syntactic units, rather than anything about the constructions in which they occur/do not occur. For example, the distribution pa ern in (1) and (2) is taken to tell us something about a difference between all the pizza and my phone in (1a and c) and 10 miles and the first hung parliament . . . in (1b and d), rather than anything about the Active and Passive Voice constructions. 12

9781441124609_Ch01_Final_txt_print.indd 12

11/20/2012 9:25:57 AM

Hypothesis Formation

Third, the mapping revealed by the distributional method is a complex one. Mismatches in distributional pa erns such as the one between the distribution in (1) and the distribution in (2) are ubiquitous. Much of linguistic analysis aims to develop hypotheses that will account for the mismatches in distribution pa erns. Distributional analysis as originally conceived is a pa ern of relations among formal structures. However, distributional analysis may be, and is, extended to include meaning. For example, in examining the distribution of elements that function in the role of complements of the English auxiliary must, one must distinguish between epistemic and deontic meanings. Epistemic meanings allow complements of the form have + Past Participle that denote past events (He must have taken his jacket), whereas deontic meanings only allow complements in the Bare Infinitive form that denote potential future events (You must wear a jacket to the dinner). Finally, one may also compare distributional pa erns of equivalent constructions across languages (i.e. cross-linguistic distributional analysis). There is no “how-to” guide explaining legitimate and illegitimate steps for forming linguistic hypotheses. Nevertheless, there are a number of implicit, unwarranted methodological assumptions that are widely used across linguistic theories in forming hypotheses. The methodological steps are in fact based on certain theoretical assumptions which are taken for granted, at least in some linguistic theories. Yet these theoretical assumptions should not be immune to empirical confirmation. These unwarranted assumptions are briefly described here (see Cro 2010). The account here will also make explicit the theoretical assumptions underlying the methodological assumptions. The first unwarranted assumption is the Free Ride Principle, so named in Zwicky (1970). The Free Ride Principle refers to using a theoretical construct established in one grammatical analysis in some other analysis. Zwicky gives the example of an early transformational analysis of placing the agent nominalizing suffix -er before the verb in underlying form and moving it (i.e. -er + sell → sell-er), since this movement rule was already used in the analysis of English auxiliaries (have + -en + eat → have eat-en). But the use of a theoretical construct must be motivated for each grammatical phenomenon in which it is invoked (Cro 2010: 317). A similar unwarranted assumption is the Global Extension of ConstructionSpecific Analyses. By “global” is meant the generalization of an analysis of a single construction to other constructions in the language, even if the empirical facts differ. A simple example is the analysis of complements of claim and other similar verbs in Minimalism. A sentence like I claimed that she was pregnant is analyzed as containing a Complementizer that that combines with the Complement clause she was pregnant. This analysis of the structure of complements is then extended to I claimed she was pregnant, despite the absence of that. 13

9781441124609_Ch01_Final_txt_print.indd 13

11/20/2012 9:25:57 AM

The Bloomsbury Companion to Syntax

It is therefore concluded that there is a null complement in the la er sentence. But, by a further extension of the original analysis to main clauses, it is suggested that even a main clause such as She was pregnant may contain a null complementizer. Again, the analysis (presence of a complementizer) must be justified explicitly for each construction, or abandoned; one can, for example, simply allow complements to occur without complementizers, as they do more generally in many languages. The theoretical assumption underlying Global Extension of Construction-Specific Analyses is that different constructions should be given the same or similar grammatical analyses wherever possible. One also finds the unwarranted assumption of Cross-linguistic Extension of Language-Specific Analyses in recent work in generative grammar. This is the extension of a grammatical analysis from one language to another. For example, Adger justifies the existence of two syntactic nodes for “cause” and a predicate “see” in English in Emily showed Benjamin himself in the mirror on the basis of the French complex predicate construction Pascale fait manger Georges (Pascal makes Georges eat), even though English lacks the complex predicate construction in this case (Adger 2003: 131–2). This assumption is motivated by the theoretical assumption that all languages have the same syntactic structures for the equivalent sentences. But cross-linguistic diversity in grammatical structure indicates that the methodological assumption is unwarranted and justified only to save the theoretical assumption of universal syntactic structures. A number of unwarranted assumptions in forming hypotheses are based on the more general assumption that the best analysis is the “simplest” or “most elegant.” The theoretical assumption behind this is that the grammar of a language should be maximally simple and elegant, based in turn on certain assumptions about what counts as “simple” or “elegant” (some of these are described below). Yet a grammar should represent the actual knowledge of the speaker (i.e. some mental representation of her knowledge), and there is no psychological imperative for simplicity or elegance. Hence one must be cautious in concluding that the hypothesis formed on the basis of an abstract analysis of the data is actually a hypothesis about the knowledge a speaker has about her language. The first of the simplicity assumptions is Symmetry: an analysis positing a symmetric grammatical pa ern is preferred over one that lacks that symmetry. A simple example is the employment of the features [±N, ±V] as an analysis of the major syntactic categories Noun [+N, –V], Verb [–N, +V], and Adjective [+N, +V]. Since [N] and [V] are binary features, the three major syntactic categories form an asymmetric system: there is no fourth category for the feature combination [–N, –V]. To satisfy symmetry, it is hypothesized that Preposition is the “missing” category (e.g. Adger 2003: 36). However, Preposition is very much unlike the other three categories. It is a closed-class category, unlike Noun and Verb, and is very diverse in its grammatical behavior, unlike Noun and 14

9781441124609_Ch01_Final_txt_print.indd 14

11/20/2012 9:25:57 AM

Hypothesis Formation

Verb. This is because Prepositions and Postpositions are actually the intermediate stage of a grammaticalization process whose sources are Nouns, Verbs, or Adverbs and whose endpoints are case affixes. Thus, the syntactic category system is asymmetric and positing a symmetric set of category features is an unjustifiable assumption in this case. A second simplicity assumption is Nonredundancy in Representation. The motivation for this assumption is a theoretical assumption of storage parsimony (Cro 2010: 327): one should posit the fewest number of basic categories, features, and operations in one’s syntactic analysis. The consequence of this theoretical assumption is that the derivations required for the analysis of a particular sentence structure are quite complex. The converse of this theoretical assumption is computing parsimony: one should minimize the number of operations required for the production of a sentence. The extreme version of computing parsimony is storage of every sentence structure so that production is simply retrieval of the structure from storage. Of course, neither storage parsimony nor computing parsimony are valid assumptions. Empirical evidence must be brought to bear on whether speakers are retrieving entrenched grammatical constructions, or computing a particular sentence structure on the basis of other structures. This is essentially a psychological question, best addressed by psycholinguistic experimentation, although some other types of evidence may be brought to bear on the question. For example, grammatical or semantic idiosyncrasy is a sufficient condition for hypothesizing that a construction is stored. Although shoot the breeze has the Verb–Object order and form that is predictable from the general Transitive Verb construction in English, it is semantically idiosyncratic and we may conclude that it is stored. Also, it is widely hypothesized that constructions that are higher in token frequency in corpora of language use are also stored. Again, although the grammatical structure of I don’t know is predictable from general pa erns of Intransitive Verbs and Declarative Negation in English, it is of high enough token frequency that it is likely to be stored independently. Additional evidence supporting this hypothesis is the degree of phonological reduction in I don’t know, represented orthographically by I dunno and confirmed phonetically by Bybee and Scheibman (1999). A variant of this assumption is the rule/list fallacy (Langacker 1987: 42). The rule/list fallacy is the assumption that a particular syntactic structure is either produced by a rule or stored in a list, but not both: for example, that I don’t know is stored or is produced by a rule, but not both. As Langacker argues, this is a fallacy in hypothesis formation and there is no reason to assume that the two possible analyses are mutually exclusive. Another unwarranted assumption of nonredundancy in representation is Syntagmatic Nonredundancy. An example of syntagmatic nonredundancy is 15

9781441124609_Ch01_Final_txt_print.indd 15

11/20/2012 9:25:57 AM

The Bloomsbury Companion to Syntax

the analysis of so-called agreement, or to use a more neutral term, indexation. Indexation is the expression of features of the Subject (and, in some languages, the Object) referent—for example, person, number, and gender/noun class—by a morpheme that is usually affixed to the verb. In English, it is assumed that a Noun Phrase is the “real” Subject or Object, and inflection on the verb merely “agrees” with the relevant Noun Phrase. In other languages, in which the verb indexes both Subject and Object, and the external Noun Phrase is absent (so-called null anaphora or pro-drop), it has been argued that the affix is the “real” Subject or Object (a so-called bound pronominal). In some languages, however, the external Noun Phrase is absent and there is no “bound pronominal,” yet the verb has a Subject and Object; in many languages, both the index and the Noun Phrase may be present; and in many languages, these different options coexist. The real problem here is the assumption that there is only one referring expression per argument in a clause; this is an example of an assumption of syntagmatic nonredundancy. The diversity of pa erns of expression of arguments suggests that this assumption is unwarranted in this case (see Cro 2001: 226–32). The final simplicity/elegance unwarranted assumption is Generality: a hypothesis that is formulated to cover a larger number of cases must be chosen over a hypothesis that covers a smaller number of cases. This is a particularly widely held assumption in generative hypothesis formation. Examples of generality in generative grammar include the reduction of all syntactic operations to Merge, the reduction of all constituent structures to binary branching structures, and positing the same hierarchy of projections for clause and phrase structure (Adger 2003). Such general hypotheses need to be justified by empirical evidence that the wide range of grammatical phenomena subsumed under the hypothesis are instances of the same general principle. It is frequently the case that the distributional facts in fact vary for each putative instance of the general principle (e.g. there are differences in the structure and distribution of phrases and clauses). Such differences may challenge the hypothesis of generality. It could be that other general principles are interacting to lead to the variation in distributional behavior. But the interactions may end up being more complex than a less general hypothesis would be (i.e. the parsimony in storing just a few general principles may be countered by the complexity in the computation of their interactions). Also, the positing of some highly general principles even if they are allowed to interact in a complex fashion is almost always countered by the positing of some highly particular and idiosyncratic constructs in the same theory. For example, although Minimalism hypothesizes highly general constituent structures and operations, it also posits a wide array of different types of grammatical features that are idiosyncratically associated with particular categories and structures (ibid.). 16

9781441124609_Ch01_Final_txt_print.indd 16

11/20/2012 9:25:57 AM

Hypothesis Formation

There is no a priori reason to posit the most general analysis, nor conversely to posit the most highly specific analyses. Again, if one is constructing hypotheses about mental representations of grammatical structures, then psycholinguistic experimentation is desirable. Otherwise, one must be cautious in extrapolating from one’s analysis to a hypothesis about a real empirical phenomenon, namely a speaker’s knowledge about her language. The next two unwarranted assumptions are based on theoretical assumptions from structuralist linguistics that are widespread in both generative grammar and in nongenerative analyses of grammatical phenomena. These are weak and strong versions of Form-Function Isomorphism, that is, the mapping between overt grammatical form and underlying structure or function. In structuralism, linguistic elements in a grammatical system are defined solely in terms of their contrast with other elements in the system. Hence, the most important properties in the structuralist model are identity (noncontrast) and difference (contrast). Weak Form-Function Isomorphism is based on assigning primacy to difference: differences in form always entail differences in underlying structure or in function. This methodological assumption rules out the possibility that two constructions are either variants of a single construction, or that two morphemes or constructions are synonymous. Yet processes of language change always involve coexistence of the new variant with the old variant, and for at least some period of time the two variants are grammatically and/or semantically equivalent in at least some contexts. Hence one must provide empirical evidence that two forms are actually not variants of the same thing. Strong Form-Function Isomorphism is based on assigned primacy to identity: identity of form always entails identity of underlying structure or of function. For example, English Particles and Prepositions are sometimes proposed to belong to a single category Preposition on the basis of the identity of form (and to some extent, meaning) of many of the Particles/Prepositions: She walked in (the room), She looked up (the downspout), etc. However, the assumption is unwarranted in this case because Particles and Prepositions are not always interchangeable: *She went into; *I went back the apartment. Strong Form-Function Isomorphism is also behind the assumption that the meanings of linguistic forms should always be analyzed as monosemous, so that for example French avoir “have” should have the same meaning when used for ownership (J’ai une voiture) and for physiological states (J’ai froid “I’m cold”). Yet avoir exhibits different behavior in the two cases: J’ai très froid “I’m very cold” but *J’ai très une voiture (for a critique of monosemy, see Cruse 1992). Strong Form-Function Isomorphism has also been applied cross-linguistically, in forming hypotheses about differences in language structures and functions. For example, in some languages such as Samoan, the same morpheme is used to indicate a genitive relation between a head noun and a dependent 17

9781441124609_Ch01_Final_txt_print.indd 17

11/20/2012 9:25:57 AM

The Bloomsbury Companion to Syntax

modifier, and a core argument relation between a predicate and one or more of its arguments. It has been argued that therefore Samoan does not make a noun– verb distinction since the same morpheme is used to express a dependent of a phrase headed by a “noun” and a clause headed by a “verb,” and moreover that Samoan conceptualizes actions to be things (hence the use of the genitive). It is assumed that the morpheme in question is identical in grammatical and semantic function in both contexts of use. However, there are distributional differences that imply that clauses and phrases are not identical in structure. In addition, Cross-linguistic Strong Form-Function Isomorphism gives rise to another unwarranted assumption, described as the Semantic Uncertainty Principle in Cro (2001: 124–6). The hypothesis that Samoan conceptualizes actions as things because their syntax is phrasal is based on the assumption that the Samoan genitive is identical to the English genitive in function. One could instead argue that the Samoan genitive is not the same as the English genitive, in part because it is used in clauses to express predicate–argument relations. Assuming that the genitive is identical in function across languages is unwarranted and would have to be defended. In this case, it is more likely that the genitive is polysemous and has a different range of uses in different languages. The last unwarranted assumptions are perhaps the most deep-rooted assumptions in the formation of linguistic hypotheses. These are described as Methodological Opportunism in Cro (2001, 2010). Methodological Opportunism comes in two varieties: language-internal and cross-linguistic. The example of the distribution of argument phrases in the English Active and Passive constructions presented at this beginning of this chapter may serve as an illustration of Language-Internal Methodological Opportunism. The distribution of argument phrases in the two constructions does not match. As noted above, differences in distribution is the norm in the data of grammatical analysis. The usual methodological strategy is to select one of the constructions used in the distributional analysis as an indicator of the “real” grammatical identity of the syntactic units occurring (or not occurring) in the construction. In (1)–(2), the grammatical identity in question is whether or not the argument phrase is a Direct Object and the Passive construction (more precisely, the Subject role in the Passive construction) is taken as the “real” test or criterion for the Direct Object status of the argument phrase in the Active construction. Hence, ten miles in (1b) and the first hung parliament . . . in (1d) are considered not to be Direct Objects. The theoretical assumption behind Methodological Opportunism is that a syntactic theory should posit a small finite number of categories for syntactic units, and that these categories are the building blocks for larger syntactic structures. Thus, hypothesis formation is the process of identifying what the building blocks are in each sentence and how they are put together. If this were 18

9781441124609_Ch01_Final_txt_print.indd 18

11/20/2012 9:25:57 AM

Hypothesis Formation

in fact true, then we would expect constructions with the same building blocks to be uniform in their structure and behavior. But they are not. Methodological Opportunism a empts to preserve the theoretical assumption by opportunistically selecting the constructions to define the categories in question. But Methodological Opportunism allows for subjectivity in determining which constructions should be considered criteria for establishing categories. As a consequence, conflicts in hypotheses cannot be resolved empirically. For example, some analysts hypothesize that there is no Noun–Verb distinction in Salishan languages, while others hypothesize that there is a Noun–Verb distinction. The analysts used the same data to form their hypotheses, but one group takes certain constructions as criterial and the other group takes certain other constructions as criterial. Methodological Opportunism also occurs in cross-linguistic comparison. Here the assumption is that the same building blocks are used in all or at least most languages. That is, categories such as Noun and Verb, or theory-specific categories, are language universals, valid for the analysis of sentences in all languages, or at least in all languages in which the categories are said to occur (some theories allow some cross-linguistic categories such as Adjective not to exist in particular languages). However, the constructions used to test for the presence of putatively universal categories such as Noun or Verb differ from language to language. Again, the consequence is that presence/absence of a category in a language is subjective, and not subject to empirical (dis)confirmation. For example, a linguist who wants to hypothesize that all languages have a Noun–Verb distinction selects the constructions in Salish that support that distinction, and not the constructions that do not support the distinction. An example of Cross-linguistic Methodological Opportunism is the argumentations for Baker’s (2003) hypothesis of the category Pred. Pred is a theoryspecific category hypothesized by Baker to exist in all languages, at least as a theoretical construct if not as an overt morpheme; for example, Baker argues that the English copula be is not an instance of Pred. Baker invokes several constructions in Edo as support for his hypothesis that Edo yé is an instance of Pred. When Baker turns to Chichewa ndì, however, the equivalent constructions in Chichewa do not behave the same way as they did for Edo yé. Baker then uses other constructions in Chichewa to support his hypothesis that ndì is an instance of Pred. But this selective use of constructions in a language to support an overt Pred morpheme (in Edo and Chichewa) or deny it (in English) is neither rigorous nor justifiable. A genuinely rigorous, justifiable method of forming linguistic hypotheses has the following characteristics (Cro 2009: 158–63): (1) All hypotheses should be cross-linguistically validated. For example, consider the structure of information (WH) questions in English, such as Where 19

9781441124609_Ch01_Final_txt_print.indd 19

11/20/2012 9:25:57 AM

The Bloomsbury Companion to Syntax

are you going? The construction differs from the English Declarative construction in two respects: the interrogative pronoun occurs at the beginning instead of a er the verb, and the position of subject and auxiliary is inverted. The question is, should one form a hypothesis in which these two properties are causally connected? Cross-linguistic evidence indicates that approximately half the world’s languages front the interrogative pronoun, but hardly any invert the subject and auxiliary. Hence a hypothesis based on a causal connection linking these two facts of English is unlikely to be valid. In general, it is impossible to determine which combinations of properties of grammatical constructions are language-specific idiosyncrasies and which are instances of language universals without doing cross-linguistic comparison. (2) Use the same construction(s) across languages for argumentation. This requirement recognizes that the distributional method presupposes the constructions as the basis for describing the distribution of syntactic units. It also ensures that one is looking at the same phenomenon across languages. (On identifying the same constructions; see (5).) This criterion also implies that the use of the distributional method as evidence about the syntactic unit being compared is also invalid. The distributional method actually provides evidence about the constructions used as distributional “tests” or “criteria.” For example, the evidence in (1)–(2) provides evidence about the best analysis of the Active and Passive voice constructions in English, not about the grammatical identity of the argument phrases. Again, explicit acknowledgment of this aspect of the distributional method leads to more sustainable grammatical hypotheses. (3) Examine distributional pa erns in detail. For example, if one examines only (1a and c) and (2a and c) in comparing the distribution of argument phrases in the English Active and Passive constructions, or only in and up in comparing the distribution of English Particles and Prepositions, then one would not discover that the distribution pa erns do not match and one might have formed an invalid hypothesis about the relationship between the distributional patterns of the Active and the Passive or the relationship between Particles and Prepositions. (4) Any additional construction used to account for an anomaly in the data for one’s hypothesis in a particular language must be cross-linguistically validated. This requirement avoids the fallacy of invoking different constructional distributional facts in each language for what is intended to be an analysis of the same grammatical phenomenon. (5) Use cross-linguistically valid criteria to identify potentially universal pa erns. That is, in order to make valid hypotheses about language universals based on cross-linguistic data, one must find cross-linguistically valid criteria to ensure the comparability of the evidence across languages. A major problem is that the constructions employed to identify categories or other syntactic units in each language are specific to that language. Categories defined on the basis 20

9781441124609_Ch01_Final_txt_print.indd 20

11/20/2012 9:25:57 AM

Hypothesis Formation

of language-specific definitions are not cross-linguistically comparable, even if they are given the same names (“Noun,” “Object,” and so on). The most reliable cross-linguistically valid criterion is functional equivalence, that is semantic and discourse–functional (informational–structural) equivalence. For example, it is o en claimed that there is a cross-linguistically valid category Adjective, but that many languages lack this category. However, language-specific constructions are generally used to justify the existence or nonexistence of the category Adjective and linguists differ as to which constructions are relevant, and hence whether Adjectives exist in a language or not. A cross-linguistically valid approach is to compare words that denote property concepts in and across languages, and to compare their distribution in constructions used for predication and modification (a ribution) in and across languages (Cro 2001: chapter 2). The distributional pa erns provide a basis for fomulating hypotheses about the occurrence of property concept words in predicate and a ributive constructions. There are fundamental properties of linguistic form that are not features of language-specific grammatical constructions, and may be used for crosslinguistic comparison. Nevertheless, the employment of these properties for cross-linguistic comparison requires reference to functional equivalents as well. For example, the order of words and morphemes like Demonstrative-Noun or Noun-Demonstrative is a formal property of sentences but to establish word order one must define the Demonstrative and Noun categories across languages, which ultimately requires reference to functional equivalence. The same is true of zero vs. overt coding of, for example, Nominative vs. Accusative or Ergative vs. Absolutive case. Formal grouping (constituency) of syntactic units is also a formal property, but the criteria used to define constituency include languagespecific constructions. Moreover, the constructions used o en do not match even within a language, or across languages where the constructions are functionally equivalent. Morphologically bound vs. free is a formal contrast, but the definitions of wordhood that determine this contrast are based on languagespecific phonological pa erns. As these remarks indicate, the formulation of cross-linguistically valid criteria for universal formal properties of sentences is complex and still not worked out even for some basic theoretical constructions such as wordhood and syntactic constituency.

21

9781441124609_Ch01_Final_txt_print.indd 21

11/20/2012 9:25:58 AM

2

Hypothesis Testing in the Minimalist Program Mark C. Baker

In many respects, hypothesis testing in the Minimalist Program is not very different from what it is for anything that aspires to be a scientific theory of an observable natural phenomenon. The normal practices of science should inevitably apply. Nevertheless, it is true that different sorts of natural phenomena present somewhat different opportunities and challenges, and this may cause good scientific practice to take on somewhat different flavors and emphases in different disciplines. For example, language behaviors are ubiquitous in the lives of most of us, so plenty of potentially relevant data is available. On the other hand, these behaviors are only produced by human beings which have intrinsic value and rights, so there are certain kinds of controlled experiments that might otherwise be desirable that we dare not do (intrusive surgical procedures, raising children in unusual environments, etc.). These factors give linguistics a particular character. More subtly, how hypotheses are tested in the Minimalist Program depends to some degree on what the Minimalist Program takes language to be, and hence what an adequate theory of language will be. In this there can be some differences between the Minimalist Program and other approaches to the study of language that are worth discussing. In this article, then, I say a bit about both facets of the question: both what I take to be good scientific practice applied to natural language, and what may be specific to the Minimalist Program in light of its characteristic take on the subject. At its roots, the Minimalist Program—like earlier versions of generative or “Chomskian” linguistics—takes a language to be a system of knowledge that humans use to construct and interpret sentences (and other units, both smaller and larger). As such, it is inherently about the relationship between the finite and the infinite (Chomsky 1965: 4; 1980: 220–4; 1995: 3).1 It is concerned with the infinite, because most people can easily create and interpret an unbounded number of distinct sentences and sentence types. For example, it is likely that most of the sentences contained in this article are new to both its author and to you its reader; perhaps many of them have never been used before in the history of the world. That does not necessarily make the article particularly hard to read. At the same time, the Minimalist Program is concerned with the finite, 22

9781441124609_Ch02_Final_txt_print.indd 22

11/20/2012 9:26:10 AM

Hypothesis Testing in Minimalist Program

because every person who knows a language learned it from a finite exposure to that language, and each one has finite mental capacity for storing their knowledge of the language. Therefore, we have the challenge of explaining how a finite amount of experience and knowledge can be used to construct and interpret an infinite (unbounded) range of new sentence types. The answer must be some sort of recursive rule system—a generative grammar in the broadest sense. The quest of finding that system underlies the generative program in all its various incarnations, even as its specific terminologies and technologies have changed and developed over time. Chomsky has often summed up this fundamental point, by quoting Wilhelm von Humboldt, that language is “the infinite use of finite means” (Chomsky 1965: 8; 1995: 4, 15). Not all approaches to language see it this way, or at least they do not emphasize this part of the generative project. For example, usage-based approaches to linguistics emphasize that we list much of what we hear, and are rather conservative in how we go beyond the list. Construction grammars emphasize that we learn many specific “constructions” on a continuum from the most narrow and specific to the most general and productive. In certain cases, these approaches might capture well phenomena that generative grammar has tended to neglect or gloss over. But the problem of how to project a finite experience and storage capacity onto an unbounded capacity to produce and interpret new linguistic examples will not go away. It cannot be true that we explicitly store all the sentence types that we use. Computer science teaches us that combinatorial explosion happens very quickly, even with systems of only moderate complexity. If there are more possible chess games than there are particles in the universe, then clearly there are more possible sentences in English or almost any other natural language. So we cannot represent everything we have heard— especially if it is categorized under everything it might count as an exemplar of, at every level of linguistic description. There must be a way of abstracting generalizable knowledge out of experienced data, and of composing listed constructions into new, more complex constructions. In short, there must be some kind of generative grammar. What implications does this have on constructing and testing hypotheses? Consider first the notion that theories should aim to be simple and elegant—a goal that has been more prominent in the Minimalist Program than in other approaches. The claim is that it can be a road to truth and understanding to compare two theories and ask which is the simpler and more elegant. On one level, this is simply sound scientific practice; it is Ockham’s razor that one should not multiply hypotheses needlessly. If five assumptions are enough to account for a body of data, then we should not make nine assumptions. One cannot claim that the data support the extra four assumptions, even if they are consistent with them. Linguists of every sort should presumably acknowledge this, and it calls all of us to think carefully about the logical structure of our 23

9781441124609_Ch02_Final_txt_print.indd 23

11/20/2012 9:26:10 AM

The Bloomsbury Companion to Syntax

theories and hypotheses, and what they entail. But the generativist’s awareness of the relationship between finite knowledge and unbounded capacity perhaps does—and should—put an extra premium on this consideration. It implies that some serious “data compression” is at work in our knowledge of language, that long and complex sentence structures must be the result of interactions and compositions of simpler pieces and patterns. Often it is hypotheses that we consider “simple and elegant” that provide a significant amount of this data compression by successfully teasing out these interactions. We should therefore value such hypotheses and seek them actively (see, for example, Chomsky 1995: 5–29—although when Chomsky discusses the value of simplicity he tends to emphasize how fruitful it has been to pursue it in the past, and linguists who are less impressed by the specific successes of generative grammar might be less moved by this). Consider next the relationship of a theoretical hypothesis to observable data. Again, there is a sense in which this is a very general criterion that any aspiring scientific theory must meet. Its descriptions of the world must agree with our observations of the world wherever they overlap—and they should overlap often. In Chomsky’s term, linguistic hypotheses must be descriptively adequate. Every hypothesis is subject to empirical confirmation and disconfirmation, including the Minimalist Program. This puts a qualification on the values of simplicity and elegance discussed above. The theory should be simple and elegant relative to the phenomena that are described. If those phenomena are very complex in absolute terms, then a hypothesis can be simple and elegant relative to a naïve description of the facts, and still be rather complex in absolute terms. For example, it might have nine assumptions rather than five—but also rather than 20. The zeal of generative linguists for simplicity and elegance may have caused us to oversimplify and ignore complexity on some occasions. This is an occupational hazard. But denying the significance of simplicity and complexity relative to the observed data is not likely to be helpful, especially in areas where the issue of projecting finite knowledge onto complex novel behaviors is salient, including syntax, compositional semantics, and complex productive morphology. We need to recognize that “small and finite” need not mean “less than ten.” But we also need to recognize that the finite is always impressively small compared to the infinite. Nor has generative grammar been guilty of ignoring detail and complexity across the board, I believe. On the contrary, its historical commitment to completeness and explicitness in linguistic theoretical descriptions, not presupposing “common sense” but trying to explicate it, has led it to make many, many discoveries about the details of natural languages that were unsuspected by previous descriptive efforts— including island phenomena, binding conditions, scope interactions, and so on (cf., for example, Chomsky 1995: 4, 8). This continues into the present, with no signs of slowing. 24

9781441124609_Ch02_Final_txt_print.indd 24

11/20/2012 9:26:10 AM

Hypothesis Testing in Minimalist Program

There are some further subtleties concerning testing a hypothesis by whether it matches the data, however, since different approaches might differ somewhat as to what counts as data and how that data should be collected. One issue of current relevance is that people have questioned (again) the value of traditional linguistic elicitation, preferring the study of corpuses of naturally occurring data and/or more formal, controlled experimentation of a psycholinguistic style. Generativists should welcome these “new” techniques as potential new sources of data that could be helpful. But they can and should resist claims that these sources of data are intrinsically superior to the traditional methods of fieldwork, such as constructing new examples and seeing whether native speakers (possibly including the researcher him/herself) accept them or not, and how they interpret them (Chomsky 1965: 18–21; 1986: 36–7). Some might think that if a sentence type does not exist in a large corpus then it may not be real and almost certainly is not worth talking about, but they forget the logic of the finite and the infinite. If we are right that an infinite number of linguistic structures are possible in most languages, then any finite corpus is small, indeed tiny, compared to the total space of legitimate possibilities. So nothing much follows logically from an observation like “X does not occur in a corpus” as long as native speakers’ judgments and interpretations of X are robust and consistent. Of course studying a corpus might be valuable in various ways, but it figures to be a poor and clumsy substitute for direct native speaker judgments across the board. Similarly, it is important to realize that any fieldwork-style interaction is essentially a kind of informal psycholinguistics experiment, seeing how a “subject” responds behaviorally to a stimulus presented by the fieldworker. Knowing this, the sensitive fieldworker should be alert to the possible influence of factors that a good psycholinguist would naturally control for, including the overall length of the stimulus, attentional factors, the order of presentation, and so on. The real question, then, is when will “upgrading” an informal psycholinguistic experiment into a formal one add value. The answer is clearly “sometimes.” For some subtle matters where there are confounding factors to be carefully controlled for, using psycholinguistic techniques can be valuable. But there are many things which can be quickly and accurately determined without going through that laborious process—for example the badness for many of *The child seems sleeping. Doing a laborious psycholinguistic experiment on this would arguably be a waste of valuable tax dollars. And sometimes data from a formal psycholinguistic experiment will be positively misleading—for example, when it averages over a number of subjects who may have different internalized grammars, or when it is impossible to control for all relevant factors and still get enough stimuli to run the desired statistics. In such situations, traditional fieldwork interviews in which a large amount of data is collected from a single speaker who the interviewer gets to know well will give more accurate 25

9781441124609_Ch02_Final_txt_print.indd 25

11/20/2012 9:26:11 AM

The Bloomsbury Companion to Syntax

results. So psycholinguistic data is not intrinsically different from or better than linguistic data, and sometimes one tool will be the best, sometimes another (see Chomsky 1980: 199–202). The bottom line here is that the fact that people are able to construct and interpret novel sentences spontaneously is itself a very important and instructive datum that needs to be accounted for. Traditional grammaticality judgment tasks, elicitation tasks, and interpretation tasks draw on this crucial capacity rather directly. Hence they are valid and ecologically natural things to do. Of course, mistakes can be made in this method. For example, when linguistic elicitation is not double-blind it is possible for researchers to unconsciously skew the results in directions that favor their hypotheses. But we also know what to do about that: we correct the mistakes by the normal scientific practices of replication by others, peer review, looking for converging lines of evidence, etc. There is no inherent need to abandon or discount the technique. (We can even experiment with a style of elicitation which is double-blind, in which say a graduate assistant asks a native speaker for judgments on some paradigm without knowing what the research leader who constructed it predicts.) Generative practice also differs from some others in how it gets hypotheses that are considered worth testing. Some positivistic approaches strongly favor “bottom-up” (inductive) approaches where one starts with the data, organizes it and reflects on it in certain ways, and looks for a theoretical hypothesis to emerge. That is a fine way for a generativist to proceed as well. But it is not the only way. In the Minimalist Program, it is considered equally appropriate to proceed in a “top-down” (deductive) fashion where one first considers the logical consequences of some theory that one finds appealing, and then goes on to test whether those consequences match up with observable data. Most typically, the practicing generative linguist works in cycles first going bottom-up from relatively obvious data that we already have to a promising first hypothesis, then working top-down to find new consequences of that hypothesis and checking them out, often against “second order” (more complex, less common) data, then working bottom-up again to find new hypotheses that help explain the mismatches, and so on. Some individual linguists might make their best contributions working bottom-up and others working top-down. In part, the difference in openness to more top-down, deductive approaches may reflect different readings of scientific practice, some inspired primarily by British empiricism; whereas generativists after Chomsky embrace Cartesian rationalist approaches as well. In particular, generative linguistics values a “Galilean” approach to science, in which prestige is attached to an abstract theory that is explicit and detailed enough to make novel predictions about phenomena which have not yet been observed, which are investigated and found out to be correct (or not, in which case one learns something and tries again) (Chomsky 1980: 2–10, 218–19). But there is also a natural connection here to the generativist’s goal of 26

9781441124609_Ch02_Final_txt_print.indd 26

11/20/2012 9:26:11 AM

Hypothesis Testing in Minimalist Program

explaining how people project a finite experience onto an unbounded range of behavior. From this perspective, people who learn a particular language are often functioning in a kind of top-down fashion, extrapolating their abstracted knowledge of the patterns and rules of the language into “hypotheses” of what someone meant when they uttered a novel sentence and what they themselves can say and be understood by others. So when generative linguists are functioning in top-down mode, they are simply mimicking this natural process that native speakers must be doing on an ongoing basis. We are seeing if our explicit theoretical hypotheses about a language project onto new structures in the same way that speakers’ tacit practical knowledge of the language does. The mental posture of relating what the linguist is doing to what a person learning and speaking the language is doing leads to perhaps the most distinctive way of testing hypotheses in the Minimalist Program. This is the learnability test (see, for example, Chomsky 1965: 25–7, 55–6; 1986: 38); Chomsky often calls it the test of explanatory adequacy, as opposed to the test of “mere” descriptive adequacy. While our theories may be abstract and have complex interactions, it must be possible in principle for every healthy child in a normal linguistic environment to learn the equivalent of those theories from the data they are exposed to. And that is far from trivial. When generative linguists try to choose between two theories which make the same predictions for the obvious surface facts, they typically do so by intentionally checking more obscure facts where the two theories make distinctively different predictions. They do this most efficiently by targeted elicitation. But children do not do this: they do not construct sentences to test them for grammaticality; rather they and their caregivers are obsessed by communication. Indeed, the data that some successful language learners are exposed to is unsystematic, not tagged for its relevance, “noisy” in that it contains speech errors, and relatively simple (compared, say, to the sentence structures attested in the Wall Street Journal corpus). So how do children succeed in learning their language without using an investigative tool that generative linguists find indispensible? One answer that generative grammarians give is that the child does not have to learn which hypothesis is correct; it is given to the child innately in some fashion (e.g. Chomsky 1965: 47–59; 1980: 232–4; 1995: 4–6). But by itself this only works for features that all human languages have in common, given that as far as we know any normal child can learn any human language when raised in the right environment. So if the hypothesis concerns a feature of one language that distinguishes it from another, that hypothesis must in principle be learnable from evidence that any normal child would encounter. For relatively simple and salient properties, like word order, case-marking, and agreement, this criterion is easily met. But languages also differ in much less obvious ways, which linguists only detect in obscure sentences of significant complexity. For example, the conditions under which a question or a relative clause can be formed are rather 27

9781441124609_Ch02_Final_txt_print.indd 27

11/20/2012 9:26:11 AM

The Bloomsbury Companion to Syntax

different in English, Mohawk, and Sakha. But the sentence types are complex, so children would not necessarily expect to hear them and they should not infer from their not hearing them that such sentences are impossible. Nevertheless, children do learn these differences because adults know them. We conclude that they must learn differences like these indirectly: they must be consequences that can be inferred from interactions of universal-innate knowledge and some more obvious feature(s) of the language that a child can reliably learn. This gives us a substantive additional constraint on our hypotheses. A theory of a given domain might successfully predict the grammatical and ungrammatical structures within that domain, but not relate them in any way to simpler, more learnable features of the language. Such hypotheses must be rejected, in favor of hypotheses that posit the right kinds of connections between simpler data and more complex data (Chomsky 1986: 38, 151–2). (For my theory of Mohawk syntax that tries to meet this condition while explaining strange island effects in Mohawk, see Baker 1991, 1996.) This condition exists now mostly in the form of thought experiments and plausibility arguments, because we do not know that much about what the limits of children’s capacities for direct learning really are. But even as a thought experiment it is a useful constraint on hypotheses. And as we learn more details of how language acquisition actually happens across an interesting range of languages, we can hope to test generative hypotheses more and more by studying actual language development as well as hypothetical language acquisition (for some steps in this direction, see Snyder 2007). Returning to more widely shared territory, it follows from this reasoning that generative hypotheses can also be tested typologically. The test of descriptive adequacy discussed above concerns whether a hypothesis about phenomenon X in language Y scales up to other phenomena in language Y; the typological test concerns whether it also succeeds in saying correct things about relevant phenomena in language Z, ideally for all Z. Two subtypes can be distinguished. First, for features of language Y that one wants to say are innate and universal, those should be demonstrably consistent with the observable facts of all other languages, correctly analyzed. These should be valid absolute universals in something like the Greenbergian sense (Greenberg 1963)—although the useful hypotheses might be at a higher level of abstraction, hence somewhat harder to match up with the easily observed facts of a language than Greenberg’s universals were (e.g. Chomsky 1980: 218–19; 1986: 43). Second, for complex features of language Y that are hypothesized to be acquired indirectly via easily learned feature A, we can test whether those features generally occur together in other languages that have feature A—whether our hypotheses give the right “parametric clusters.” These correspond roughly to Greenbergian implicational universals of the form “A implies X”—although again the right level of analysis may be more abstract than the one often used in existing typological literature. 28

9781441124609_Ch02_Final_txt_print.indd 28

11/20/2012 9:26:11 AM

Hypothesis Testing in Minimalist Program

In practice, the typological test is pursued in two somewhat different ways. One is the macro-comparative approach, in which one tests hypotheses of a universal or parametric nature against a set of languages from different families and different parts of the world (Baker 2008, 2010). This is not so different from what functionalist–typologists do in practice, although with a somewhat different idea of what is an attractive hypothesis for testing. The other is the microcomparative approach, in which one tests hypotheses against a set of closely related languages or dialects such as the Romance languages or the Germanic languages (Kayne 2005). Within the Minimalist Program, there have been some debates about which of these approaches is the best to pursue, but those debates are about short-term tactics, about what is most productive to do first. As far as I know everyone agrees that the two approaches complement each other in principle, and that both should be pursued as much as is feasible. For example, the macro-comparative approach will often be the strongest test of a universal hypothesis, whereas the micro-comparative approach may be the strongest test of a parametric hypothesis. Another practical question related to hypothesis testing is when should a hypothesis that has encountered counterevidence be abandoned. Generative linguists have sometimes given the impression of holding on to hypotheses too long, in the face of significant counterevidence. No doubt this is true in some cases (although it is not necessarily distinctive; researchers of all kinds tend to have a stubborn streak). But simple and hard-to-avoid assumptions about the nature of language convince us that language is a complex system where many factors interact, and that some linguistic knowledge must be relatively abstract. Therefore, we cannot know immediately whether the empirical problem that has come to light is a problem with the specific hypothesis being tested, or with some other feature of the system. The inescapable fact underlying this is that we cannot really test individual hypotheses one at a time, even though this is what we strive to approximate. Ultimately, given that language is a complex interacting system, we can only test sets of hypotheses, which work together to derive sets of predictions. When the predictions are supported, great. When those predictions are falsified, we must change one or more of the hypotheses, but we cannot immediately know for sure which one to change.2 In the most typical case, some of the predictions will be supported and a few will be falsified. Then we might be encouraged that our system of hypotheses has some important features of the truth, although there are some corrections to be made in it somewhere. Since the interactions need not be linear ones, one cannot necessarily measure how many of the hypotheses in the set are likely to be correct from how many of the predictions are correct, so literal counting of counterexamples is unlikely to be helpful. More generally, a mere fact is not enough to overturn a theory. Rather, it is a fact that has no analysis under that theory but that does have an analysis under a rival theory 29

9781441124609_Ch02_Final_txt_print.indd 29

11/20/2012 9:26:11 AM

The Bloomsbury Companion to Syntax

that truly threatens the first theory—especially if a pattern of such facts begins to accumulate. So generative theories are defeatable, but generally not by mere counterexamples; those can and often enough should be attributed to anomalies or to unknown factors. Rather, generative theories can be defeated by better theories, ones that fit a wider range of observed facts better. A final practical question concerns the gold standard of hypothesis confirmation that we should be aiming for versus acceptable levels of confirmation for intermediate results gleaned along the way. Studying complex interacting systems is hard, so we must be both idealistic and practical in how we go about it. We must have a standard of truth and practice that we are aiming for, and stick to it. But we must also have intermediate standards for research that has not gotten all the way there yet, but could be a serious step on the way. Otherwise we will be unable to make progress given that the object of study is too complex for anyone to figure out in a single step. For example, it seems to me that many of the recommendations proposed by Croft (this volume) would be valid for the Minimalist Program as well, as expressions of the gold standard that the whole community is aiming for over time. I certainly agree that proposals about universal grammar need to be tested cross-linguistically against many unrelated languages. I also agree that theoretical constructs should in principle be justifiable for each language they apply to using evidence gleaned from within that language. But those are high standards that cannot be applied immediately to every intermediate result or we will put an impossible burden on linguistic research. For example, I think it would be a mistake to lay down a standard of the form “no proposal about a language universal will be taken seriously unless it has been tested on at least 50 unrelated languages”—even though that standard may seem very modest to some. For my money, a hypothesis that has a great deal of logical coherence, that makes distinctive correct predictions about complex new examples that were engineered to test it, and that can be justified by facts internal to five unrelated languages could be more promising than a shallow hypothesis that is statistically significant over 50 languages (although even this may be too much to expect for an average journal article, and needs to be broken down further). Promising hypotheses present themselves in different ways, and the gold standard testing of them is a job for the whole community, not a single researcher. Perhaps this is more evident to the Minimalist than to other kinds of linguists if we are focusing on the more complex parts of language, where several things interact, but I imagine that a distinction between the gold standard of truth and a reportable intermediate result is useful to all linguists. In conclusion, this article has reviewed what generative linguists take the essence of language to be, and how this affects the hypotheses they entertain and how they test them. Some of the methods are general to any scientific endeavor, including empirical support or falsification, favoring hypotheses with fewer 30

9781441124609_Ch02_Final_txt_print.indd 30

11/20/2012 9:26:11 AM

Hypothesis Testing in Minimalist Program

assumptions, the ability to make testable correct predictions, and generality (e.g. over many languages). Others are more distinctive, involving particular choices about how to do science, and a particular idea of what a linguistic theory needs to be. These include accepting more abstract hypotheses—which are connected to data but indirectly via other hypotheses—accepting directly elicited data, appreciating top-down deductive reasoning in the development and testing of hypotheses, the criterion that a hypothesis must be learnable, and perhaps a different sensibility as to what constitutes a useful intermediate result. It is a good overall package—needing to be practiced well rather than badly, to be sure, but not in need of fundamental change. Linguists of all kinds are invited to use as much of it as they will.

Notes 1. Much of this article is about what practicing generative linguists have internalized from what Chomsky has said over the years about how linguistics is best pursued. His influence on these matters is arguably even greater than on specific analyses. He has made the points in many ways over many years, and the references that I give are merely illustrative; they are not necessarily the earliest or the fullest expositions of these views. References to the rather brief discussion in Chomsky (1995) in particular are included to show that they have been carried forward in the Minimalist Program from earlier work. 2. And in dealing with this, it makes perfect sense to make different choices of which phenomena are “figure” and which are “ground” in different studies. Picking up on one of Croft’s (this volume) examples, in one study one might use different sorts of direct objects to study the phenomenon of passive, and in other one might use the passive as a “test” to study what is a direct object. In studying complex systems, anything can be instrument and anything can be object as we try to bootstrap our way to better overall knowledge. I believe that this is well-understood by practicing generative linguists, although it may not be clearly expressed in introductory textbooks.

31

9781441124609_Ch02_Final_txt_print.indd 31

11/20/2012 9:26:11 AM

3

Field Methods in Syntactic Research Marianne Mithun

Chapter Overview Introduction Simple Sentences Complex Sentences The Dynamicity of Syntactic Structure Conclusion

32 35 42 48 50

1 Introduction Current work in syntactic theory continues to enrich our understanding of language. It also continues to demonstrate that progress can involve more than revising abstract formulae to describe configurations of words between periods. Many syntactic patterns seem arbitrary or chaotic without consideration of the lexical, prosodic, discourse, and interactional factors associated with them. They can also make little sense without a recognition of the dynamism of structure, the fact that constructions are born, extended, remodeled, and replaced over time. At the same time, the value of in-depth studies of a variety of types of languages is becoming ever clearer. All of these considerations have implications for the kinds of data that might spur further progress in the field, and that researchers might, accordingly, consider collecting.

1.1 Data The data underlying current syntactic theory have come from a variety of sources: introspection, grammaticality judgments, questionnaires, sentence elicitation, 32

9781441124609_Ch03_Final_txt_print.indd 32

11/20/2012 9:23:31 AM

Field Methods in Syntactic Research

informal observation, historic and modern corpora, and more. Each has been useful for particular purposes. Here the focus will be on data that emerge from collaboration with speakers of lesser-documented languages. Such languages are often typologically quite different from better-known languages, a fact that only enhances their potential contributions. Their value can be further appreciated when we move beyond checking of established typological features to capturing what is special about each, and to exploring what each might show us that we were not already expecting. For this, elicited translations of isolated sentences from a contact language can take us only so far; much more can be discovered by examining extensive spontaneous speech. Elicitation is certainly useful for getting one’s bearings at the outset, filling in gaps, and exploring basic constructions we predict should exist, such as question formation or negation. But particularly in the area of syntax, elicitation can tend to replicate categories and patterns from the model language. Constructions without counterparts in the model language might never surface. And links between structure and lexicon, prosody, and context can be obscured.

1.2 Methodology Fieldwork methodologies vary widely, as they should, according to the personalities of the individuals involved, their cultures, attitudes toward the language, its vitality, its areas of complexity, and the goals of the project. Especially for syntax, spontaneous connected speech, both monologue and conversation, has much to offer. When a single speaker maintains the floor over a period of time, certain kinds of patterns can emerge robustly, such as those involved in referent tracking and textual organization. Conversation can reveal other kinds of structures, such as those that manage interaction and interpersonal relations. Of course monologue and conversation are not mutually exclusive. Conversation is typically studded with stretches of monologue of varying lengths, all the more valuable because they are directed to an understanding, reactive audience. In my own work, after recordings have been made, I typically work through the material with one or more speakers phrase by phrase, to transcribe, analyze, gloss, and translate it. In the course of such work, speakers provide a check on the acceptability of the forms used and can point out inter-speaker variation. They have insight into the meaning of what is said beyond literal translations. They can untangle reference. They can provide the back story behind discussions that would make little sense otherwise. They may comment on the semantic and social implications of certain structural choices. For me, the most interesting discoveries about syntactic structure tend to emerge from this work. Our discussions are triggered by the constructions that actually occur, 33

9781441124609_Ch03_Final_txt_print.indd 33

11/20/2012 9:23:32 AM

The Bloomsbury Companion to Syntax

constructions that are often much more intriguing than those a linguist could invent. Exploring syntactic structure in this way tends to be a more coherent task for the speakers as well. It tends to produce analyses that better fit the shape of the language. The structures are accurate and appropriate on every level: not only syntactic but also morphological, lexical, prosodic, and pragmatic. They also tend to be those that are central to the language, often those that are more elaborately developed. Of course speakers vary in their interests and talents. Some are fascinated by grammar, while others would rather not hear about it. Some are highly articulate in the language; some are good storytellers; some have extensive vocabularies; some take pride in precise translation. But all have certain kinds of intuitions about their language. It is helpful to be aware of what speakers can be expected to know in their capacity as speakers, in contrast with what they might hypothesize as analysts. Speakers rarely make allophonic or allomorphic “mistakes” in their first language. But they can lack conscious access to the larger patterns they follow so expertly in speech: they cannot necessarily specify the contexts in which different constructions are used, or explain the discourse reasons behind the choice of one construction over another. For highly endangered languages, the issue of intuition is complicated by the fact that speakers are usually bilingual. Their bilingualism can certainly facilitate the process of analysis. But it is easy for bilinguals to tap into intuitions from their second language, unwittingly creating structures that are actually not part of the first and would never be used. Speakers need to know how valued their insights are, including their intuitions about the difference between actual and possible constructions. The documentation of highly endangered languages can present special challenges. Such languages are usually no longer used on a daily basis in all contexts, and speakers may worry about their fluency, correctness, and vocabulary. But their contributions can still be important. If there is even one fluent speaker, conversations involving that speaker and other less fluent ones can yield a record of the language far beyond what might emerge from work with that speaker alone. A fluent speaker talking to people who understand and react typically uses language in richer, more varied, and dynamic ways. Sometimes playing older recordings of fluent speakers can bring modern speakers back into a time when they were thinking in the language. One man, perhaps the last fluent speaker of his language, seemed able only to produce isolated words and phrases. But after he heard recordings of fluent speakers from a half century earlier, he came to life, producing fascinating discussions on a rich variety of topics, day after day. In the following sections, some advantages and disadvantages of different kinds of data for syntactic analysis are considered.

34

9781441124609_Ch03_Final_txt_print.indd 34

11/20/2012 9:23:32 AM

Field Methods in Syntactic Research

2 Simple Sentences We generally expect a basic clause to contain a predicate and one or more arguments, plus perhaps adverbials. The sentence in (1) is the opening of the Biblical story of Jonah, translated from English into Mohawk, an Iroquoian language of Quebec, Ontario, and New York State. The translators are excellent firstlanguage Mohawk speakers. The top line is their Mohawk rendition and the bottom line the modern English version they worked from.  (1) Mohawk (Jon. 1.1) Ne the

Roiá:ner king

tahá:on’ ne he gave him the

aterihwahnotáhtshera’ rumor spreading

“One day the Lord spoke to ne the

Jonah Jonah

né:ne which

Ami ai roièn:’a. Amittai his son

Jonah son of Amittai.” From this example it can be seen that Mohawk shows head-marking structure. Grammatical relations are marked on the predicate rather than the arguments: the verb tahá:on “he gave him” contains a pronominal prefix -haw- .sg/m.sg “he/him,” and the nouns are uninflected for case. The relation between Amittai and his son is marked with the prefix ro- on the head “his son,” while the dependent Amittai is uninflected. We see that the word order is SVO, with relative clauses following their heads. With these typological basics established, we might assume we have captured the essence of the language. To anyone familiar with Mohawk, however, it is immediately obvious that this sentence did not originate in the language. If our understanding of Mohawk syntax were based on sentences like this, we would miss many of the most important contributions the language has to make to our understanding of syntax.

2.1 Lexical Categories The Mohawk translation above contains one verb (“he gave him”) and five nominals (“the Lord,” “the rumor-spreading,” “Jonah,” “Amittai,” and “his son”), yielding a verb/noun ratio of 1/5. The verb/noun ratio in my corpus of unscripted Mohawk speech is a robust 17/1. There are several reasons behind the difference. Mohawk words fall into three clear categories on the basis of their morphological structure: verbs, nouns, and particles. The sets of verb and noun roots

35

9781441124609_Ch03_Final_txt_print.indd 35

11/20/2012 9:23:32 AM

The Bloomsbury Companion to Syntax

are completely distinct, as are the sets of verbal and nominal affixes. Particles are by definition morphologically unanalyzable. Morphological category and syntactic function are not isomorphic, however. Verbs can function as predicates, as expected. They can also serve as complete sentences in themselves since they contain pronominal prefixes identifying their core arguments. In addition, they can serve as arguments with no overt nominalization. Some verbs have become lexicalized with nominal meaning, such as kà:sere “car,” literally “it drags.” Others are used sometimes to predicate, sometimes to refer. The sentence in (2) was from a conversation. The translation was provided later by one of the participants. The word wahonwaia’táta’ is a morphological verb, but it was translated as “the burial.”  (2) Mohawk verbal expression Nè:’e tiathrória’t kí:ken “Let’s talk about this thetèn:re wahonwaia’táta’. thetèn:re wa-honwa-ia’t-a-t’a-‘ yesterday actual-m.pl/m.sg-body-linker-insert-pfv they put his body in burial yesterday.” In many cases Mohawk speakers simply use verbs to express ideas English speakers convey with nouns, like “this morning” below.  (3) Mohawk verbs Wa’katià:tawi’te’ ki’ ni’ nì:’i shiiorhón’ke verb particle particle particle verb I put a sweater on in.fact too myself as it became daylight “I put on a sweater this morning too.” There is pervasive noun incorporation in Mohawk, in which a noun stem is included in a verb stem. Both wahonwaia’táta’ “they buried him” and wa’katià:tawi’te’ “I put a sweater on” in the sentences above contain the incorporated noun –ia’t- “body”: “they bodily inserted him” and “I bodily enclosed myself.” Some ideas conveyed by independent nouns in English are evoked by incorporated nouns in Mohawk.  (4) Mohawk noun incorporation Tanon’ tewawennókeriks, tanon’ te-wa-wenn-okerik-s and incl.a-pl-word-scrunch-habitual “And we shorten our words, 36

9781441124609_Ch03_Final_txt_print.indd 36

11/20/2012 9:23:32 AM

Field Methods in Syntactic Research

tóka’ enionkwate’shennaién:ta’ne’ toka’ en-ionkw-ate-‘shenn-a-ient-a’n-e’ if fut-1p.pl-mi dle-chance-linker-have-inchoative-pfv if we get a chance.” It would be a serious mistake to consider Mohawk incorporated nouns syntactically equivalent to nouns in languages like English. They have different referential, syntactic, and semantic properties. The formation and use of incorporating constructions are conditioned by lexical and discourse factors that analytic constructions are not. The first line of the Jonah translation in (1), “One day the Lord spoke to him,” contained a nominalization “rumor-spreading,” consisting of a verb with nominalizing suffix -htsher-. Such nominalizations are rare in Mohawk speech, though pervasive in translations. Speakers simply use verbs. Overt nominalization is required in just one context: incorporation. As mentioned earlier, the term for “car” is a morphological verb: kà–:sere “it-drags.” It appears in that form as a lexical argument: Kà:sere wa’khní:non’ “I bought a car.” When incorporated, however, it must be formally nominalized with the suffix -ht: wa’ke-‘sereht-ahní:non’ “I car-bought.” Another factor contributes to the difference in verb/noun ratio between Mohawk and English. Mohawk clauses may contain lexical nominals referring to a core argument or specifying time or place, but there are no other obliques. What would be expressed in obliques in other languages is expressed in core arguments of other clauses; the verbs in those clauses indicate their role.  (5) Multiple clauses Ne the

sha’kanénhsta same corn

iakwátstha’ we use it

nó:nen when

eniakwatkátston’. we will soup.make

“We make soup with the same kind of corn.”

2.2 Constituency A major aspect of syntactic structure is constituency. Translations can tend to preserve the constituent structure of the model. But even sentences that originate in the language under study can prompt misinterpretation of constituent structure when only written forms are examined. At first glance, (6) might appear to contain a determiner phrase: “this Indian language.” 37

9781441124609_Ch03_Final_txt_print.indd 37

11/20/2012 9:23:33 AM

The Bloomsbury Companion to Syntax

 (6) Mohawk syntactic constituency? Sénha’ ki’ ietiohnhá:ten’ more in fact it is regrettable “It is all the more regrettable that kí:ken onkwehonwehnéha’ onkwawén:na’ iohtentionhátie’. this genuine people style our language it is leaving this Indian language of ours is being lost.” The intonation with which the remark was uttered reflects a quite different structure. The sentence consisted of two prosodic phrases, each represented in a separate line in (6a). (6a) Mohawk prosodic constituency Sénha’ ki’ ietiohnhá:ten’ kikén:, . . . more in fact it is regrettable this “It is all the more regrettable onkwehonwehnéha’ onkwawén:na’ iohtentionhátie’. genuine people style our language it is leaving that our Indian language is being lost.” The demonstrative kí:ken “this” was grouped prosodically with the initial matrix clause “It is more regrettable” and followed by a (non-terminal) prosodic break and significant pause. The next clause “our Indian language is being lost” began with a pitch reset, visible in the high bump on the stressed syllable of the first word onkwehonwehnéha’. This structure can be seen in the pitch trace in Figure 3.1.

Pitch (Hz)

300

200 150

100

Sénha ki’ ietiohnhá:ten’ kí:ken

onkwehonwehnéha’ onkwawén:na’ iohtentionhátie’

It is more regrettable this

our Indian language is leaving.

0

4.005 Time (s)

Figure 3.1 Mohawk demonstrative construction

38

9781441124609_Ch03_Final_txt_print.indd 38

11/20/2012 9:23:33 AM

Field Methods in Syntactic Research

This is a common syntactic construction in Mohawk, one which speakers exploit to manage the flow of information. In general, speakers tend to introduce just one significant new idea at a time in a prosodic phrase (Chafe 1994). In this construction, the first prosodic phrase lays out the basic idea with a verb (and particles) followed by a demonstrative. The demonstrative serves as a kind of placeholder, signaling that further elaboration is to come. The referent of kí:ken “this” in (6) was further identified in the following prosodic unit “our Indian language is disappearing.” Some constituent structure can be seen in sequences consisting of the article ne plus a nominal. The Jonah translation in (1) shows three examples. The translation obscures the actual distribution of the article in the language, however. The English article the indicates that the speaker believes the hearer can identify a referent, whether from general knowledge (the Chinese), uniqueness (the sun), extralinguistic context, or previous mention in the discourse. Mohawk ne indicates only previous mention: “the aforementioned.” It would not normally occur in the opening line of a narrative as here.

2.3 Constituent Order Another feature generally considered fundamental to basic clause structure is constituent order. The Jonah translation in (1) shows clear SVO order. It is vanishingly rare to find clauses with two lexical arguments in Mohawk, as in many languages. But VO order can be seen in (2), and SV order in the second clause in (6), both from conversation. A further look at spontaneous speech reveals, however, that these are not basic orders at all. Mohawk constituent order is determined not by syntactic relations but by information structure: constituents (apart from discourse particles) are arranged in descending order of newsworthiness. One speaker was describing how clear the river water used to be. When she first mentioned the fish in it, the noun “fish” appeared clause-initially, before the predicate. The verb “see” here served primarily to introduce the fish.  (7) OV Kéntson’ enhshé:ken’ fish you will see them “You could see the fish.” She continued, describing how people used to swim and catch fish. She noted that no one ever bought fish in those days. This time the noun “fish” was clausefinal, after the predicate. At this point the “not buying” was more newsworthy than the established referent, the fish. 39

9781441124609_Ch03_Final_txt_print.indd 39

11/20/2012 9:23:33 AM

The Bloomsbury Companion to Syntax

 (8) VO Iáh newén:ton teiakohní:non not ever has one bought “Nobody ever bought fish.”

ne the

kéntson’. fish

The Mohawk pragmatic ordering principle governs all kinds of lexical constituents. This generalization would be difficult to arrive at on the basis of sentences elicited in isolation. Such sentences tend to reflect either the word order of the model, or some context in the mind of the speaker of which the interviewer may be unaware.

2.4 Nuclear Clause Boundaries An important step in syntactic description is the identification of nuclear clause structure. The sentence in (9) appears to be a nuclear clause.  (9) Mohawk nuclear clause? Akhsótha thónon: shontakahá:wi kahwá:tsire’ thó iè:teronne’. my grandmother there at that time family there she resided “At that time my grandmother lived with the family.” The actual structure is quite different. It contains two topicalized constituents, “my grandmother” and “family.” Each is picked up referentially inside of the nuclear clause, the grandmother by the pronominal prefix ie- “she” in the verb iè:teronne’ “she lived,” the family by the particle thó “there.” Mohawk topicalization constructions show a distinctive prosodic profile. The topicalized constituent appears before the nuclear clause in an intonational unit of its own, ending with a fall in pitch. The following clause begins with a full pitch reset. This pattern can be seen in Figure 3.2.

Pitch (Hz)

300 200 150 Akhsótha

100 My grandmother

thónon: shontakaha wi

kahwá:tsire’

thó iè:teronne’.

there at that time

family

there she lived

5.875

0 Time (s)

Figure 3.2 Mohawk topicalization

40

9781441124609_Ch03_Final_txt_print.indd 40

11/20/2012 9:23:33 AM

Field Methods in Syntactic Research

The prosody of topicalization constructions is quite different from that of nuclear clauses, though they may look alike in print. In (10), which occurred a few moments later in the account, the grandmother is within the nuclear clause. (10) Mohawk nuclear clause N=akhsótha ken: neká: ieiè:teron’. the my grandmother here side she lives there “My grandmother was in her space.” As can be seen from the pitch trace in Figure 3.3, the grandmother is part of the same overall pitch contour as the rest of the clause, with no pause or pitch reset. Still another construction occurred a few sentences after the topicalization in (9). This is an antitopic construction used to confirm the identity of a continuing topic, often when there are other referents under discussion. In (11), the grandmother, at this point an established referent, is mentioned in a nominal phrase following the nuclear clause. (11) Mohawk antitopic construction [“Where we lived, there was an addition on the side.”] Thó iè:teronne’, there she lived “That’s where she lived, ne:’ ne that one the my Tóta.”

Tóta. Gramma

500

Pitch (Hz)

300 200 150 100

N-akhsótha

ken’ neká: ieiè:teron’.

70

My grandmother

lived there. 1.515

0 Time (s)

Figure 3.3 Mohawk nuclear clause

41

9781441124609_Ch03_Final_txt_print.indd 41

11/20/2012 9:23:34 AM

The Bloomsbury Companion to Syntax 500

Pitch (Hz)

300 200 150 100

Thó iè:teronne’,

ne:’ ne Tóta.

There she lived,

Gramma.

70 2.723

0 Time (s)

Figure 3.4 Mohawk antitopic construction

Antitopic constructions also have a characteristic prosodic structure. They are typically pronounced with low pitch and narrowed pitch range, and often creaky voice. They may be separated from the nuclear clause by a brief pause. The pitch contour of (11) can be seen in Figure 3.4. The last syllable ne’ of the nuclear clause “there she lives” was pronounced at the baseline pitch. (The nuclear clause descends steeply below the baseline here because of the lexical tone. The final syllable of the antitopic “Gramma,” which ended on the baseline pitch, is not visible because of the creak.) Both topicalization and antitopic constructions are pervasive in Mohawk speech, though they rarely appear in elicited sentences and would be less noticeable without their prosody. Missing them would leave a serious gap in our understanding of Mohawk syntactic structure.

3 Complex Sentences A good empirical foundation is even more important for understanding complex sentence structures. Direct elicitation of complex sentences can often produce structures much like those in the target language, but they may not display the full inventory of constructions available in the language nor their patterns of use and their ranges of functions.

3.1 What is Subordinated? Navajo, an Athabaskan language spoken in the Southwest, contains a robust subordinate clause construction marked by the enclitic =go. The enclitic, often reduced to =o or =ǫ by some speakers, originated as a derivational adverbializer. 42

9781441124609_Ch03_Final_txt_print.indd 42

11/20/2012 9:23:34 AM

Field Methods in Syntactic Research

It can be seen in (12) below in “at night.” It also sets off adverbial clauses, as in “when it is dark.” (12) Navajo adverbial clause Nihił oolwoł with.us something.is.running tł’ée’-o chahałhiił=o. night-adverbiali er it.is.dark=subordinate “We travelled at night [when it was dark].” But in spontaneous speech, subordination appears in unexpected contexts. One common construction is in (13): “I didn’t know where they were setting out for.” (13) Navajo subordination Háágóóshí ̜í¡ deekai where.in.the.world they.started.to.go doo not

shił to.me

beehózin=o. is.it.known=subordinate

“I didn’t know where they were setting out for.” In the English translation, provided later by the speaker, the subordinate clause is “where they were setting out for.” In the Navajo original, the subordinate clause is “I didn’t know.” This is a common, robust pattern, typical of sentences with verbs of thinking and saying. It also makes sense. The main information is the utterance or the thought itself. The Navajo construction reflects this information structure, subordinating the act of speaking or thinking to the actual message or thought.

3.2 Functions of Subordination Navajo subordination is used even more pervasively in another unexpected way. The passage in (14) was later translated by the speaker as an independent sentence. It had the prosody of a sentence. Yet the verb “he was working” is marked as subordinate. (14) Navajo subordination Neeznáá shínááhai yᾁ ądᾁᾁ’, ten my.years when.in.the.past “When I was ten years old, 43

9781441124609_Ch03_Final_txt_print.indd 43

11/20/2012 9:23:34 AM

The Bloomsbury Companion to Syntax

nléí shizhé’é, that my.father my father Na’nízhoozhídi naalnísh=o. in.Gallup he.was.working=subordinate was working in Gallup, New Mexico.” Morphologically subordinate independent sentences do not normally appear under elicitation, but they can be pervasive in spontaneous speech. Navajo speakers have extended the subordinating enclitic from marking syntactic subordination of a clause within a sentence to marking discourse subordination of whole sentences. Sentences marked as subordinate are used to supply background information, setting the scene in narrative, providing offline commentary, etc. The use of subordinate marking on independent sentences occurs in other languages as well, among them those of the Eskimo–Aleut family. These languages contain rich inventories of dependent clause constructions, marked by inflectional suffixes on verbs. In Central Alaskan Yup’ik, the independent moods are indicative, interrogative, and optative. The dependent moods are subordinative, participial, and various other connective moods with such meanings as “when,” “while,” “whenever,” “before,” “although,” “because,” and “if.” As in Navajo, dependent marking appears just as would be expected in elicited translations of English complex sentences. Such constructions also occur in spontaneous speech. Subordinative clauses are marked with the verbal suffix -lu. (15) Yup’ik subordinate clauses Tangvalallru-a-qa tangvaq-la-llru-a-qa watch-customarily-past-tr.ind-1sg/3pl “I used to watch them cat tuai tegumiaqlutki ca-t then tegumiaq-lu-tki thing-pl then hold.in.hand-subordinative-3pl/3pl holding those things arulallutki. arulate-lu-tki wave-subordinative-3pl/3pl waving them (above their heads).” In spontaneous speech, however, dependent verbs are much more frequent than independent verbs. As in Navajo, many verbs in independent sentences 44

9781441124609_Ch03_Final_txt_print.indd 44

11/20/2012 9:23:34 AM

Field Methods in Syntactic Research

are marked as grammatically dependent. These sentences have the prosody of independent sentences and are translated as independent. In (16) the speaker was describing how wooden masks are made. His account contains long sequences of subordinative clauses, marked with the suffix -lu. Every verb in this passage (and in the material before it) is formally dependent. Some of the clauses have the prosody expected of dependent clauses, ending in a non-terminal fall indicated here with a comma. But others end in a full, terminal fall, indicated here with a period. (Each line represents an intonational unit.) (16) Yup’ik subordinative sentences Tuai-llu ilua kegginam ilua And then its interior face’s its.inside

pirrarluku, first.doing.it

kegginam qengapiara, face’s its.authentic.nose canangluku. starting to carve it “And then, in the face, they first start to carve the nose.” Yugetun humanlike

ayuqqan, if it resembles

qengaliluku nose.making.it tua-i and.so

iililuku, eye.making.it

kegginatun ayuqiluku. face-like similar.making it

“They make it like a human nose, they make the eyes, and so they make it like a human face.”

Pitch (Hz)

Figures 3.5 and 3.6 show the prosodic contours of the two sequences that end in a full, terminal fall.

200 150 100 70 50 30 20

Tuai-llu ilua kegginam ilua pirrarluku, And then in the face they first

kegginam qengapiara, the face’s nose

0

canangluku. they carve. 7.999

Time (s)

Figure 3.5 Yup’ik subordinatives

45

9781441124609_Ch03_Final_txt_print.indd 45

11/20/2012 9:23:34 AM

The Bloomsbury Companion to Syntax

Pitch (Hz)

200 150 100 70 50 30

Yugetun ayuqqan,

qengaliluku iililuku,

If it looks human

they make a nose, they make eyes

tua-i kegginatun ayuqiluku. and so they make it like a human face.

8.638

0 Time (s)

Figure 3.6 Yup’ik subordinatives

The extensive use of the Yup’ik subordinative seems at first similar to that of subordination in Navajo. A closer look shows that the two are used for different purposes. In Navajo, subordination is used for background information, for thoughts incidental to the progression of the discourse. In the Yup’ik in (16), the subordinative relates closely associated events. The Yup’ik subordinative actually has a range of functions. One is interactive. Speakers can use the subordinative to indicate that their comments are pertinent to what has gone before. The sequence in (17) is from a telephone conversation. (17) Yup’ik conversation A. Camek calisit? what are you working at “What is your work?” B. Tua–i–gguq qalamciyarturlua. well apparently I storytelling “Well, I tell stories.” A different Yup’ik dependent construction, the participial, functions more like the Navajo subordinator. Participial clauses can be used much like English relative clauses in complex sentences. They can also serve as independent sentence in themselves, providing additional information that does not move a storyline ahead or contribute key points in a discussion.

3.3 Cross-linguistic Comparisons The similarities and differences between the Navajo and Yup’ik constructions raise an important issue. Corpora of spontaneous speech are ideal for discovering what is special about a language, but they can make cross-linguistic comparison difficult. For purposes of comparability, typological surveys have sometimes been based on Bible translations. The advantage is obvious: Bible 46

9781441124609_Ch03_Final_txt_print.indd 46

11/20/2012 9:23:35 AM

Field Methods in Syntactic Research

translations exist for a very large number of languages. But as we saw at the outset, translations can obscure cross-linguistic differences. An approach that has proven fruitful for a number of projects is to present speakers of different languages with a common stimulus, then ask them to perform similar tasks. In 1975, Wallace Chafe made a 6-minute movie the Pear Film, which shows a sequence of events but no language. Speakers are asked to view it then recount what happened. The film, which can be viewed at http://pearstories.org/, was designed to elicit a variety of potential linguistic structures, with various kinds of events and activities, characters who arrive, depart, and reappear, shifts in scene, etc. It provides an excellent foundation for comparing how speakers of different languages package comparable ideas in their grammars. Navajo and Yup’ik descriptions of the Pear Film allow, among other things, comparisons of how speakers distribute ideas over independent and dependent clauses. The Navajo passage in (18) contains three indicative clauses and three subordinate clauses. The corresponding Yup’ik passage in (19) contains no indicative clauses at all. Of the ten dependent clauses, seven are subordinative, two participial, and one past contemporative (“when”). (Due to space limitations, only the translations can be given here.) Each line represents a clause in the original. Formally dependent clauses are indented. Punctuation reflects prosody. (18) Navajo Pear Story excerpt So then we saw a man there  picking apples climbing around in a tree. And from there he came back down.  There were two baskets sitting down there.  He was putting apples into them.

indicative subordinate indicative indicative subordinate subordinate

(19) Yup’ik Pear Story excerpt In the beginning a man, in a tree, was picking fruit, and putting it into a basket, arranging it carefully. And then, after picking, after filling his bucket, he climbed down from the tree. He placed them carefully, and when one fruit fell, he picked it up again, and replaced it.

participial subordinative subordinative subordinative subordinative subordinative subordinative past contemp participial subordinative 47

9781441124609_Ch03_Final_txt_print.indd 47

11/20/2012 9:23:35 AM

The Bloomsbury Companion to Syntax

4 The Dynamicity of Syntactic Structure Syntactic constructions are not static: they evolve over time. Corpora of unscripted speech can often provide opportunities for investigating processes of development. A construction may arise in specific lexical contexts, then be extended to more contexts until it becomes quite general. An example can be seen in Tuscarora, an Iroquoian language spoken near Niagara Falls, New York. We are fortunate to have documentation of connected speech from the late nineteenth century as well as the present. In the nineteenth century, the interrogative hé̜:we “where” appeared in content questions. It also served as a complementizer with matrix verbs of speech, cognition, and perception, but only where the speaker, thinker, or observer could not identify the location. (20) Tuscarora nineteenth century indefinite complement: Thompson 1889 ms 411, RC 1987: 494 Ęhsahrù:yę’ hé̜:we tikè̜’rę’ ukè̜’ tì:wa’θ thuh you will ask where it sits or it is so much there “Ask it where the greatest numbers are of ste’awè̜:te ha’ θa’neθwé:kih. something the you want whatever kind of game you want.” Hé̜:we was not used with other kinds of matrix verbs, or where the speaker, thinker, or observer could identify the location. A different particle, kę’, was used. (21) Tuscarora nineteenth-century kę’ definite complement: 1888 ms 422, RC 1987: 438–9 Ù:nę wahrá:kę’ kę’ newè̜hryę’. now he saw where they two enter dirt “He was able to see where the children had escaped through the ground.” A century later, the original interrogative hè:we “where” is still used in questions and indefinite complements of speech, cognition, and perception verbs. In addition, it has been extended to contexts in which the speaker, thinker, or observer can identify the location. (22) Tuscarora twentieth-century definite complement: Howard Hill, speaker Thwé:’n ha’ è̜:kwe, kayeyę’né:ri hé̜:we, ękayekúhe’ all the person they know where they will get “Everyone knows where to get sassafras.”

anè̜hsnači’. dark seed

48

9781441124609_Ch03_Final_txt_print.indd 48

11/20/2012 9:23:35 AM

Field Methods in Syntactic Research

The particle kę’ is now used in a more restricted set of contexts, meaning “right there.” Processes of development can sometimes be detected by comparing coexisting structures at differing stages of development. Tuscarora temporal adverbial clauses can be introduced by several different markers. The most common is ù:nę. (23) Tuscarora temporal adverbial clause: Elton Greene, speaker Yahwakuwé:rhu’ ha’ ù:nę yahè̜čyęht. it covered her the when she went down, ”It covered her when she went down.” The probable source of this construction still survives in the language. The particle ù:nę also means “then.” (24) Tuscarora ù:nę “then”: Elton Greene, speaker [“They could not find anyone who would volunteer.”] Ù:nę ha’ rakuwanè̜:hu’y wahrè̜hrę:’, . . . then the head chief he said “Then the head chief said, ‘ . . . ‘ ” The subordinate clause construction apparently developed along the following lines. X. Then Y.  >  X, then Y.  >  X when Y. It is clear that the construction in (23) is now one complex sentence rather than a sequence of independent sentences. The subordinate “when” clause is integrated prosodically into the larger sentence. The matrix clause “it covered her” did not end in a terminal fall, and the subordinate clause “when she went down” did not begin with full pitch reset. There was no pause between the two.

Pitch (Hz)

100

70 Yahwakuwé:rhu’

50

It covered her

ha’ ù:nę yahé:čyęht. when she went down. 2.49

0 Time (s)

Figure 3.7 Tuscarora adverbial clause

49

9781441124609_Ch03_Final_txt_print.indd 49

11/20/2012 9:23:35 AM

The Bloomsbury Companion to Syntax

“When” clauses are often set off by the article ha’ as here, which confirms that they are not predications in their own right. These clauses can now also either precede or follow the matrix.

5 Conclusion Progress in our understanding of syntax can be greatly enhanced by documentation of unscripted speech from a variety of languages. Such material can provide more reliable data for analysis than isolated elicited sentences or translations. It can reveal constructions particular to individual languages that we might not think to elicit because they have not yet become an object of general theoretical interest. It allows the integration of lexical, prosodic, and discourse factors into the analysis. It also provides an excellent basis for further exploration of syntactic structure with speakers, leading to analyses that better reflect the structures of their languages. And it establishes an open-ended resource for investigating issues that we do not yet know enough to investigate.

50

9781441124609_Ch03_Final_txt_print.indd 50

11/20/2012 9:23:35 AM

4

Corpus Methods in Syntax Maria Freddi

Chapter Overview Introduction Syntactic Annotation Parsed Corpora and Research in Syntax Colligation, Meaning Shift Unit, and Pattern Grammar Conclusions

51 52 54 58 61

1 Introduction This chapter concerns the use of corpora in the study of syntax, or rather, how corpus methodology is and can be used for understanding, describing, and perfecting existing theorizations of syntax. Although this perspective takes for granted the role of corpus linguistics as a methodology rather than a linguistic theory, I will try to show how the availability of large principled collections of language in the form of electronic corpora has prompted a novel view of syntax as inherently probabilistic. This view challenges a categorical modelling of language and replaces grammaticality judgments with the recognition of variability, gradience, and fuzzy phenomena (cf. among others, Bybee and Hopper 2001 and Hopper, this volume, sec. 1; Manning 2003; Aarts 2007) and also enables observations on current change (e.g. Mair and Leech 2006; Aarts et al. forthcoming 2012). Likewise, a reconceptualization of syntax, holistically combining other levels of linguistic analysis, lexis, semantics, and pragmatics in turn, has originated from the corpus approach. This will be dealt with under the rubrics of “meaning shi unit” (Sinclair 2012) and “pattern grammar” (Hunston and Francis 2000; Hunston 2003, 2009) respectively.

51

9781441124609_Ch04_Final_txt_print.indd 51

11/20/2012 9:23:52 AM

The Bloomsbury Companion to Syntax

However, a different approach takes computerized text corpora and their technological potential as the ideal locus in which to explore syntactic behavior on a large scale with the purpose of validating or disconfirming theories. Thanks to the ever-increasing information storage capacity and processing power of computers, manual syntactic analyses of small samples of text can be extended to large quantities of data with a high degree of automation, and then encoded in the corpus and stored in a way that they can be reused for further investigations or employed for Natural Language Processing purposes. The encoding of interpretative linguistic information, including syntactic analysis, is known as corpus annotation.

2 Syntactic Annotation Annotation more generally refers to the activity, in accordance with some notational standards,1 of enriching the raw corpus with metadata, which is subsequently retrieved by means of complex queries. The two levels of corpus annotation directly relevant to the study of syntax are part-of-speech (POS) tagging (also grammatical tagging or morpho-syntactic annotation, see McEnery et al. 2006: 34) and syntactic parsing (also treebanking or bracketing). POS tagging means assigning a part-of-speech to each word in a corpus and is by far the most automatically performed type of annotation for many languages, with a good accuracy rate (no lower than 97 percent like, for example, with the CLAWS2 automatic tagger). Obviously, POS tagging relies on word segmentation or tokenization, that is, the segmentation of running text into word tokens. For alphabetical languages like English, word tokens in a written text correspond to uninterrupted strings of characters preceded and followed by a space (with the caveat about multiword expressions, variably spelled compounds and mergers, for example, in fact, course-book, can’t). For languages such as Chinese, tokenization is a more challenging task because delimiting spaces in running text cannot be taken as a criterion. Therefore, lexical matching on a stochastic basis is usually resorted to instead (ibid.: 35). Parsing, on the other hand, consists in analyzing the sentences in a corpus into their constituents (groups or phrases, depending on the grammatical nomenclature, that is, noun phrase, verb phrase, prepositional phrase, etc.), function roles (e.g. Subject, Direct Object), and, occasionally, dependency relations (i.e. argument structure). Since this implies a higher degree of abstraction than with part-of-speech disambiguation and attribution, it is clear how parsing can vary depending on the descriptive and theoretical models subscribed to by annotators. For the same reason, that is, due to the difficulty in establishing higherlevel hierarchical relationships between morphosyntactic categories with one another, sometimes even within the same model, parsing can be automated 52

9781441124609_Ch04_Final_txt_print.indd 52

11/20/2012 9:23:52 AM

Corpus Methods in Syntax

with a precision rate typically much lower than that of tagging and therefore requires extensive manual revision and correction. Annotated corpora, when compared to unannotated ones, have the clear advantage of allowing one to query more abstract and complex patterns like multiword units and periphrastic constructions. However, the higher the abstraction, the more likely one is to find annotation errors (Meurers and Müller 2009: 922). An example of this is represented by a recent study on light verbs in British and American English, which participate in constructions such as have a think, take a look, etc. The structures studied comprised: (i) a semantically “light” verb (namely have and take) immediately followed by (ii) the indefinite article a/an, and then (iii) an optionally modified noun, based on the stem of the verb from which it derives (Beltrami 2010: 62). The syntactic frames thus defined were searched for in two chosen corpora (respectively the British National Corpus (BNC), and the Corpus of Contemporary American English (COCA)) by devising a query that would yield most of the relevant structures without either returning unwanted examples or leaving out pertinent ones (e.g. take a walk, but not take a book). In information retrieval contexts this usually means maximal precision and high recall. In the specific case study, one query alone did not prove sufficient either because it was not maximally precise, that is, it would return many “false positives,” or because the recall was low, that is, some of the structures of interest were missed. The reason for this lack of precision and recall was at least in part due to annotation issues. The main difficulty faced by annotators is whether to ascribe the deverbal nominal element of the construction to the noun or to the verb word class and consistently throughout the corpus data (i.e. is the deverbal noun laugh in the sequence have a laugh to be treated as a verb or a noun, and similar cases must be handled similarly, cf. also Abeillé 2003 on this issue). Such a difficulty is mirrored in the so-called ambiguity tags contained in the BNC, for example, NN1-VVB meaning “ambiguity between common singular noun and finite base form of lexical verbs, with major probability tending to the noun” (cf. CLAWS guidelines from BNC website). In other words, the automatic tagger is not 100 percent certain whether the word in question is a noun or a verb—it leaves the choice open, but assigns a higher probability to the former, and vice versa in the case of VVB-NN1. It is clear how, in empirically grounded linguistic research aiming at generalization and theory validation, limitations like the one described should not be overlooked (cf. also Manning 2003). Conversely, a query relying on a syntactic theory and related annotation scheme, however little precise, can sometimes yield unexpected results concerning rare phenomena, which are neglected in grammatical descriptions or relegated to few paragraphs at best, but are found to occur in natural language. This was the case in the above mentioned study where causative and existential structures of the kind exemplified in (1), (2), and (3) below, and spoken usages, as in (4), were serendipitously retrieved 53

9781441124609_Ch04_Final_txt_print.indd 53

11/20/2012 9:23:52 AM

The Bloomsbury Companion to Syntax

through the query. This occurred in the most disparate registers, cf. Beltrami (2010: 105): (1) Although not everyone can afford to their home, (. . .) (COCA-MAG 2003) (2) But saying this does not answer in concrete terms the question of what important work in ethical criticism does. I recently to me, “So, as an ethical critic you (. . .)” (COCA-ACAD 1998) (3) When I was at Baruch, we the Nobel Prize in economics. (COCA-SPOK 1999) (4) (. . .), so we a decision about who was going to live, who was going to die. (COCA-SPOK 2007) Also related to issues of information representation and retrieval and annotation errors is the choice between “flat vs. full” annotation (cf. Meurers and Müller 2009: 923). As the parser cannot reliably make certain distinctions which entail higher-level abstractions, a flat, less fine-grained annotation is preferred in many current projects and in so-called shallow parsing. They identify the constituents but do not specify their internal structure (e.g. into head and modifiers), nor their grammatical role in the sentence. International forums like the International Workshop on Treebanks and Linguistic Theories, which reached its tenth edition3 in January 2012, are testimony to the growing interest in automatic annotation and syntactically motivated corpus research. In the next section, some of the best-known syntactically annotated corpora are briefly introduced and described together with the research findings concerning syntax derived from them. Special emphasis is placed on English. Special emphasis is placed on English. Details about parsed corpora of other languages are to be found in the Appendix.

3 Parsed Corpora and Research in Syntax Parsing has been tried both on general reference corpora thought to be representative of a state of a language or language variety and on small domain-specific ones, with the aim of positing a grammar that might be restricted to a specific domain or even genre (see also the notion of “local grammars” as developed by Barnbrook and Sinclair 1995). In the former case, annotated corpora have exerted a strong influence on newly compiled reference grammars by offering a reliable means for quantifying the syntactic features of a language and measuring the significance of their distribution according to parameters of variation such as medium (spoken vs. written) and regional dialect (for instance in contrasting the two major varieties of English, British vs. American). One of the most

54

9781441124609_Ch04_Final_txt_print.indd 54

11/20/2012 9:23:53 AM

Corpus Methods in Syntax

interesting outcomes of the analysis of tagged representative corpora (at least for English) are reference grammars like the Longman Grammar of Spoken and Written English by Biber et al. (1999), whose driving criterion is the systematic differentiation between the syntax of speech and writing in light of the distributional differences observed in the corpus. More generally, spoken grammar has gained prominence as a variety on its own, worth describing with no prior modeling of the language, and rather showing the inadequacy of some of the traditional sentence-based categories used for written grammar (see also Conrad 2010: 236)4. Parsed corpora usually take the name of the theoretical framework on which they are based. However, they are also commonly referred to as “treebanks” because they represent syntactic relations in a form similar to tree diagrams. On the whole, there are two competing models of syntax in corpus annotation: constituency-based paradigms (derived from phrase structure grammar) on the one hand, and dependency grammar on the other. While the former breaks down the sentence into constituent groups of words that are in a logical relation with each other, the latter specifies the hierarchical relations between a word (usually the verb) and its dependents (e.g. its arguments). For example, the Penn Treebank makes use of labelled bracketing to assign phrase structures. See the following example from Taylor et al. (2003: 10), (5) ((S (NP-SBJ-1 Jones) (VP followed (NP him) (PP-DIR into (NP the front room) , (S-ADV (NP-SBJ *-1) (VP closing (NP the door) (PP behind (NP him))))) .)) Clearly, the Penn Treebank offers an example of full annotation whereby: (i) the type of sentence is specified, cf. (S as opposed to (S-ADV, standing for simple declarative clause vs. adverbial clause)); (ii) the grammatical role is assigned to phrases, for example (NP-SBJ-1 Jones), that is, singular noun phrase having subject role; (iii) the degree of nesting of the phrases is detailed, see (PP behind (NP him)); (iv) null-constituents are included—the asterisk in (NP-SBJ *-1); (v) some semantic roles are distinguished, cf. ((PP-DIR into (NP the front

55

9781441124609_Ch04_Final_txt_print.indd 55

11/20/2012 9:23:53 AM

The Bloomsbury Companion to Syntax

room))) for direction and trajectory5. Indeed, the marking of semantic relationships between the constituents in a sentence is considered by many a syntactic level annotation (cf. among others, McEnery et al. 2006: 38). Moreover, the identification of null elements and co-indexing with the lexical material they stand for are used to derive a further level of annotation, namely the predicateargument structure. A lighter kind of parsing characterizes the Lancaster Parsed Corpus, which shows the phrase structure of each sentence (Garside et al. 1992). Based on the Lancaster-Leeds Treebank, a 45,000 word subsample of the LOB (LancasterOslo/Bergen) English corpus manually parsed by Geoffrey Sampson, it nowadays prefers skeletal parsing as opposed to full parsing. In other words, it tends to use fewer labels than seen in the Penn Treebank, and thus identifies only major constituent types. Another parsed corpus of British English is the ICE-GB component of the International Corpus of English (ICE) project, which, together with the Diachronic Corpus of Present-Day Spoken English (DCPSE), has been developed at the University of London and SEU (Survey of English Usage) (see Nelson et al. 2002). ICE-GB counts 1 million POS-tagged words, and sentences are parsed following phrase grammar. More in detail, for each sentence (or utterance in spoken texts) a parse tree is provided containing (i) the POS of each word in the tree, (ii) the phrases and clauses to which these words belong, and (iii) the function that they serve, for example, Subject, Object (cf. Wallis 2008). Work based on the corpus thus annotated has brought to light new evidence on gradience and fuzzy categories (Aarts 2007) as well as favored observations on syntactic change over short periods. In particular, with regard to the verb phrase in British English, alternation where there is more than one choice available to speakers can be computed in a corpus, and the significance of the frequency associated with a given choice assessed. Finally, the interaction between choices can be monitored and interpreted in relation to language change over decades rather than centuries. Along these lines, for example, a study by Aarts et al. (forthcoming 2012) has explored the alternation in present-day spoken English between progressive vs. simple tenses in verb phrases that can be progressivized. As a result of the corpus being parsed, this research question can be automated and directly observed on extensive data by measuring the probability of any progressive VP in contexts that can be progressivized against any simple VP, or, in formal terms, P (VP(-ing) | {VP(-ing), VP(-ing)}). Results show the continuing though gradual increase in the frequency of the progressive construction over time as a proportion of alternates. It is obvious that results concerning grammar systems like the ones above were inconceivable before the advent of electronic corpora. Among the parsed corpora annotated for dependency relations is the Czech corpus known as the Prague Dependency Treebank (PDT) (see Böhmová et al. 2003). In this model, each word, typically the verb, is treated in its hierarchical 56

9781441124609_Ch04_Final_txt_print.indd 56

11/20/2012 9:23:53 AM

Corpus Methods in Syntax

interrelations with other words, and dependent elements are identified following the Prague school syntactic tradition. However, in view of the difficulty in separating verbal dependents into freely occurring adjuncts and inherent arguments, no argument/adjunct distinction is made. The annotative work done on the PDT has served as a test for various linguistic issues including underlying sentence structure, information structure, markedness (e.g. discontinuous constituents), as well as textual co-reference (Hajičová 2009; see the Appendix for a visual representation of the PDT). Finally, research into the syntax-semantics interface has also benefited from this kind of methodology6, and long-term projects such as FrameNet carried out at Berkeley, California, are an example of this. Starting from phrase grammar, FrameNet identifies the linking patterns between semantics and syntactic realizations, that is, the range of semantic and syntactic combinatory possibilities (or valences, as in Fillmore 2009a and b, that is, the “pairing of a semantic rule with grammatical concepts”), with the aim of drawing an inventory of frames for each word in each of its senses. Although the major product of this work is the FrameNet lexical database (an index of lexical units with related syntactic patterns), the norm is to separate word meanings if doing so correctly pairs meanings with grammar and with a word’s relationships with other words. Therefore, syntax is a by-product of this research. Moreover, FrameNet serves as the starting point for a new project based entirely on syntactic constructions (cf. Fillmore 2009a). An example of the kind of information FrameNet contains follows. If we take the semantic frame of REPLACEMENT, there are certain lexical realizations of the frame, like the verb “replace” vs. “substitute” and the phrases that are found to occur with them, representing particular frame elements like the Agent, that is, the one who brings about the change, the New and the Old in the following sentence: (6) We must replace a state that’s gigantic, inefficient and corrupt with one that’s modern and austere, he said. in which “we” is the Agent, “a state that’s gigantic, inefficient and corrupt” is the Old and “with one that’s modern and austere” is the New—respectively a pronoun, a noun phrase (NP), and a prepositional phrase (PP). A different realization of the same frame is observable in (7) A new Foreign Secretary, John Major, has replaced Sir Geoffrey Howe. Thus, each word evokes a semantic frame that can be realized by different phrases, or frame elements. There can, of course, be different ways of phrasing information in the same frame. However, the ultimate objective is to disambiguate word senses by recognizing valence differences. With FrameNet the corpus 57

9781441124609_Ch04_Final_txt_print.indd 57

11/20/2012 9:23:53 AM

The Bloomsbury Companion to Syntax

is used to extract examples that match and illustrate the lexical and semantic frames worked out by the linguist-lexicographer. In its successive stages, it is also used to provide frequency information on the alternation of different variants. This kind of approach to sentence structure is not so distant from Sinclair’s “idiom principle” and more specifically from pattern grammar as developed by Hunston and Francis (2000). These will be addressed in the next section.

4 Colligation, Meaning Shift Unit, and Pattern Grammar Some new notions of syntax have developed from corpus studies. One is the concept of “colligation” as theorized by Sinclair (1991, 1998/2004a). Drawn from Firth’s pre-corpus idea of the co-selection of grammatical word-classes by analogy with collocates attracting each other at the lexical level, the Sinclairian view of colligation is one side of the same coin reuniting grammar and the lexicon. The claim is that a target word or word-form (also “node” or “keyword-in-context”) can be studied in a corpus in terms of its being systematically surrounded by high-frequency grammatical classes, that is, its colligational pattern. For example, if we take block as node, the corpus observation of its frequent co-occurrence with closed-class grammatical words within a given span on either side of the node (co-text) is not collocationally neutral and contributes to determining the overall meaning of the expression (Sinclair 1998/2004a: 146; Krishnamurthy 2004: 64). A set of concordances will help here to illustrate the concept: the preposition of and alternatively the noun phrase repeatedly occurring immediately to the right of the node help us separate block as a noun from block as a verb correspondingly. A random set of concordances for node block from the Bank of English (KWIC view, right-sorted) is given in Figure 4.1. Along the same lines, recurrent combinations in the le co-text can be taken into consideration in order to identify the preferred usage and disambiguate the senses of the chosen word. Indeed, a second-order sorting to the le would have helped to better visualize repetitions of the same syntactic pattern. However, even in this display, it is clear how in lines 1, 5, 7, 10, 14, 18 the elements immediately preceding the node are all nouns modifying it, with two hyphenated instances (14 and 18). The pattern is further confirmed if we move one or, occasionally, two slots backwards (depending on the extent of the modification, cf. line 10), where we usually encounter the articles the/a. Turning to the lexical content of the pre-modifier nouns, all but line 5 refer to buildings (see administration, stable, assembly, C-, cell-), whereas road preceded by the indeterminate article a selects a different meaning. Analogous considerations can be made for block as verb (cf. the recurrence of to) to conclude that each lexico-grammatical combination corresponds to semantic specialization. In fact, the verbal use of the word is in the sense of 58

9781441124609_Ch04_Final_txt_print.indd 58

11/20/2012 9:23:53 AM

Corpus Methods in Syntax

“impeding, obstructing, placing an obstacle in the path of someone or something.” The object of the verb, placed to its right, is whatever is being blocked (including a route, both animate and inanimate participants like people and chemicals, and more abstract nouns such as plans). Noun uses of block have several distinct senses, the most common one being “building,” followed by other less frequent ones like “lump” (lines 4 and 9), a quantity of things or sequence of numbers (11 and 22), a way to measure distance along a city street (see line 28 down the block), down to single occurrences such as the proper name in line 2 (Sinclair 2003: 5–7).

 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

 on foot between the administration block and some cells can take up to 25    operations are moves designed to block enemy penetrations. The counter    fee are variable. In 1985, Block filed 10 million tax returns,  the 16th Century, salt was used in block form and scraped off with a knife.  Zulu men for rural areas) and a road block had been set up by young men Ltd. could also find itself on the block if Sir James Goldsmith succeeds in   the livery yard. Although the stable block is in darkness, she knows her own    cross as he led the crowd on a three-block march to police headquarters. He    deep pockets, and setting it upon a block of stone between himself and the      Next to the main assembly block of the shipyard in the Baltic port   you’re a winner! Underneath each block of three numbers is a prize value.    The antagonists fasten onto and block off the receptors so that the    he would chase one leaf half a block or more with his blower, whereupon a yodel, came echoing down from C- block’s Two-tier. Bauman    his state partners would be able to block such a move. A Montedison spokesman   antagonists” he’s developed which block the chemical signals small cell   r have clips or rings put on them to block them. Early techniques   lindfolded man cried out in the cell-block yard: about five guards surrounded  inister, Nikolai Ryzhkov, was on the block yesterday, not that of Mr      effectively took itself off the block yesterday and announced a sweeping     appeared to be the main stumbling block. Yorkshire refused to comment on   off ALL THREE numbers in a single block you’re a winner! Underneath     a landscape, the seascape doesn’t block your sight; it extends beyond it.  a turning point in your life. Do not block your own good; ask for guidance.    Another is to go to extremes to block your neighbours out of your life,    and circumstances that appear to block your path. There is a certain    nationalists today said they will block Yugoslavia’s border crossings with   see it by our eyes. Siegel: Down the block, Ziyad, who runs a souvenir shop

Figure 4.1 A random set of concordances for node block from the Bank of English (KWIC view, right-sorted) Source: Datafile 01_block.doc from Sinclair (2003: 4)

59

9781441124609_Ch04_Final_txt_print.indd 59

11/20/2012 9:23:53 AM

The Bloomsbury Companion to Syntax

In shunting from the grammatical to the lexical to the semantic in a holistic manner, the analysis thus carried out goes beyond the traditional separation between lexical and morphosyntactic levels of language analysis. Conversely, the two levels mutually constrain each other in a patterned fashion. Choice at each point in a text is conditioned by the surrounding text both at the grammatical level (e.g. the choice of block as a transitive verb predicts an object) and at the lexical level (the object will be a person, a vision, etc., or a more abstract noun like plans), hence the notion of the “idiom” principle governing language organization, that is, the phraseological tendency of much of the language we produce and consume daily. This goes together with the descriptive emphasis on the syntagmatic axis of language and consequent backgrounding of the paradigmatic, as in Sinclair and Mauranen’s (2006) “linear unit grammar.” Besides, it replaces the dichotomy between the two axes with complementarity and mutual conditioning and is therefore in contrast with the orientation of most grammars. The idea is that repeated co-occurrence as observable in a set of concordances is the result of the intersection of frequent combinations along the horizontal syntagmatic axis with repeated preference along the vertical paradigmatic axis, the latter allowing for some variation to take place within exact repetition. Emphasis on the role of the co-text in creating meaning goes so far as to include pragmatic meaning, that is, the attitudinal features or “semantic prosodies” that derive from regular communicative use (cf. Sinclair 1996/2004a: 33). Such mutual prediction or, as is o en termed, “strength of attraction” between co-occurring words can nowadays be measured on corpus evidence through statistical methods and is shown to extend beyond idioms in the traditional sense, tending to characterize “meaning shi units” (MSUs)—Sinclair’s last legacy (cf. Sinclair forthcoming 2012). The chunking thus achieved, namely via the data in the corpus and non-theory-driven categorization,7 approximates what happens in the language system and certainly aims to a fully fledged description of the system. However, whether this objective has yet been reached is debatable and it is safer to talk about “pattern grammar,” keeping in mind that extended generalizations are not always available as yet. The ideas informing Hunston and Francis’ (2000) pattern grammar are drawn from Sinclair’s idiom principle, as illustrated in this chapter. They are very similar in that repeated grammatical co-selections observable in a corpus build up shared meanings among the node lexical items, and patterns of use can be identified. This potentially leads to innovative descriptions of grammar. An example is the pattern: (8) V n from n (cf. the coding for the Collins Cobuild English Dictionary 1995)

60

9781441124609_Ch04_Final_txt_print.indd 60

11/20/2012 9:23:53 AM

Corpus Methods in Syntax

which reads “verb followed by a noun followed by the lexical element from, followed by another noun” and applies to protect somebody from something and other cognate lexicalizations such as cushion, insulate, safeguard, shade, shield, etc., all having the same meaning in common. However, sometimes two or more combinations share a lexico-grammatical pattern, but not necessarily the same meaning, showing syntax-semantics alignment or disalignment. To exemplify this, another pattern can be considered: (9) V from n into which both recover from and suffer from fit (Hunston 2003: 356). According to pattern grammar the two combinations are identified as similar, that is, as sharing a meaning group, because they share the same pattern, and the assumption is made that meaning belongs rather to pattern than to word (Hunston 2009). The central discovery is the mapping of grammar elements onto meaning elements in order to identify frequent patterns replacing the single word as the repository of meaning, something which dictionary compilers find inconvenient but which has great implications for language teaching and learning (cf. also Sinclair 2004b). The descriptive enterprise is still in the process of improving the coverage of patterns (see Mason and Hunston 2004). However, what we learn from this perspective on corpus methods in syntax is the extension beyond traditional categorizing of lexicon and grammar in order to reach a new description of the system based on chunks, or MSUs, which are indicative of how we process language and make sense through patterning.

5 Conclusions In conclusion, corpus methods in syntax show enormous potential for improving descriptions of languages. On the one hand, pre-corpus syntactic theories can be verified and perfected thanks to the ever-growing size of searchable repositories of naturally occurring language as are corpora. Moreover, the everincreasing processing capacities of computers allow the researcher to formulate complex queries and fine-tune their descriptive tools accordingly. On the other hand, innovative views of language like frequency-driven patterned choice as exemplified by Sinclair’s theory of phraseology, Hunston’s pattern grammar, and other cited models interfacing syntax with different layers of the meaningmaking potential of language find in corpora unprecedented evidence and a natural pathway toward positing a probabilistic language faculty. The challenge remains to continue filling the gap between extended descriptions of patterns of use and systemic generalization beyond instantiation.

61

9781441124609_Ch04_Final_txt_print.indd 61

11/20/2012 9:23:53 AM

The Bloomsbury Companion to Syntax

Notes 1. The encoding format of syntactic metadata is beyond the scope of the present contribution. However, see Ide and Romary (2003) for a discussion of annotation standards and encoding schemes. Some examples are provided in the Appendix. 2. CLAWS stands for “Constituent Likelihood Automatic Word-tagging System” and has been developed by the Lancaster University Centre for Computer Corpus Research on Language since the early 1980s. The latest version of the tagger has been used to POS-tag the 100 million words of the British National Corpus. 3. The “Tenth International Workshop on Treebanks and Linguistic Theories—TLT10” was held in Heidelberg on January 6–7, 2012, hosted by Heidelberg University. 4. Note, however, that some corpus-based grammars have taken the opposite direction, notably Huddleston and Pullum (2000), who downplay the difference between spoken and written grammar. 5. For a fuller discussion of the annotation and methodology employed in the Penn Treebank, see Taylor et al. (2003). 6. Cf. Kay and Fillmore’s (1999) seminal paper on corpus-based frame semantics. 7. This view presupposes that no parsing is added to the corpus as it is the raw data that drives the researcher to identifying the meaning-bearing patterns.

62

9781441124609_Ch04_Final_txt_print.indd 62

11/20/2012 9:23:53 AM

5

Constituents Marja-Liisa Helasvuo

Chapter Overview What is Constituency? Constituents and Phrase Structure Constituency and Intonation Constituency in the Service of Conversational Activities The Emergence of Constituency Constituency and Frequency of Occurrence

65 66 67 70 72 74

1 What is Constituency? The concept of constituency has been used in linguistics to describe the hierarchic organization of syntactic structure within the clause. There is a widely held assumption in linguistics that clauses are not made out of mere linear sequences of words and morphemes but have a more complex structure which is hierarchically organized. This hierarchic structure is called constituent structure. According to the received view, clauses are divided into major parts, which are then further divided into subparts. The major parts or major syntactic constituents are the subject and the predicate (see, for example, Givón 2001b: 110–11). (1) illustrates a simple clause with two constituents, the subject and the predicate. (1) Hank sleeps. (1) is an intransitive clause consisting of a noun phrase (NP) functioning as the subject (here, a proper name Hank) and a verb (sleeps), which forms the predicate alone. Often the hierarchic structure is more complex, depending on the number of obligatory arguments required by the verb. Furthermore, there may be optional 65

9781441124609_Ch05_Final_txt_print.indd 65

11/20/2012 9:26:14 AM

The Bloomsbury Companion to Syntax

constituents which may, for example, describe the circumstances of the event depicted. The obligatory arguments typically refer to the participants of the event or state of affairs described by the clause. In (1) the predicate is formed by a verb which requires only one obligatory argument, the NP Hank, which refers to the sole participant of the event. (2) exemplifies a case where the clause can be divided into two major constituents, the subject and the predicate which in turn can be divided into several constituents. (2) Hank  is writing a book in his study.   Subject Predicate In (2) the predicate contains a verb (is writing) referring to an event with two participants. These are the obligatory arguments required by the verb. They are represented by two NPs, the agent Hank and the patient a book. In addition, there is an optional constituent, the prepositional phrase in his study. In this chapter, we will first discuss the relationship between constituency and phrase structure. We will then look at constituency in relation to intonation. We will also discuss constituency in relation to pa erns of language use. Finally there will be a discussion of alternative ways of viewing constituency.

2 Constituents and Phrase Structure Constituency is often depicted either in the form of a tree-diagram or as a bracketed structure as in (3); (3) shows the bracketed structure of the clause in (2). These two formats basically represent the same information. The purpose of these representations is, first, to show the hierarchic structure of the clause and, secondly, to provide names for each constituent (see Nunes, this volume). Constituents are named after the grammatical category they represent. (3) [S [NP Hank] [VP [V2 is writing] [NP a book] [PP in his study]]] Verbs can be divided into further subgroups in accordance with their syntactic behavior and the kind of arguments they occur with. The sub-groups may be indicated with sub-indexes or the like. For example, the verb run is intransitive and does not require an object whereas love is transitive and occurs with an object. Intransitives may be marked as V1, and transitives as V2. Likewise, some verbs occur with a prepositional phrase, for example, dash. These verbs can be marked as V3. Furthermore, some verbs (e.g. communication verbs like say or verbs of cognition like think) typically occur with sentential objects (complementizers) and are marked as V4. (See Jacobson 2006: 58.) 66

9781441124609_Ch05_Final_txt_print.indd 66

11/20/2012 9:26:14 AM

Constituents

Depicting constituency in the form of a phrase structure tree helps us see structural similarities between clauses. One of the motivations for constituency is distribution: units of the same rank (e.g. NPs) can be substituted for one another still yielding a structurally similar sentence. Compare (4) to (2) and (3). (4) The boy who won the competition is writing a speech of thanks in his study. In (4) there is a complex NP containing a relative clause the boy who won the competition. In the clause, its function is similar to that of the simple NP Hank in (2) and (3): it serves as a subject. Semantically, both NPs are functioning as agents. Likewise, a speech of thanks in (4) and a book in (2) and (3) serve as objects and are semantically patients. It should be noted, however, that while substitutions may yield a sentence that is similar in its hierarchic structure, some substitutions may result in semantically anomalous sentences. Phrase structure trees and bracketed structures represent constituency in a left-to-right order which is—at least in the simplest cases—isomorphic to the linear order of wri en words in our writing system. The left-to right order suits our conventional conceptualization of the temporal sequence of spoken words as proceeding from left to right (cf., for example, forwarding and rewinding a tape). In some languages phrase structure may be essential in explaining rules governing word order (in the so-called configurational languages). English is an example here. However, in some languages word order is based on pragmatic factors. These languages (the so-called non-configurational languages) typically allow for discontinuous constituents (Hale 1983, see also, for example, Austin and Bresnan 1996 for Australian aboriginal languages, Vilkuna 1989 for Finnish).

3 Constituency and Intonation Many studies on intonation take as their starting point the assumption that intonational grouping reflects syntactic structure directly. In his textbook on intonation, Cru enden (1986: 130) states: “[I]ntonation groups generally correlate with major syntactic constituents, although a good deal of choice is available to speakers concerning which constituents intonational groups should correspond with.” By “major constituent boundaries” Cru enden (1986: 37) refers to boundaries either between clauses or between the subject and the predicate. Speer and Ito (2008) report on psycholinguistic research on prosodic phrasing and its relation to syntactic structure. With regard to production they show that speakers reliably used prosodic cues to signal syntactic structure in the experiment. They also used prosodic cues to convey the intended meaning of syntactically ambiguous sentences. Regarding the processing of prosody, the 67

9781441124609_Ch05_Final_txt_print.indd 67

11/20/2012 9:26:14 AM

The Bloomsbury Companion to Syntax

experiments show how listeners exploited intonational cues when processing contrastive u erances (ibid.). Starting in the late there is a growing body of research on grammar in conversational discourse. There has been a keen interest in the relationship of intonational grouping and syntactic grouping, especially the relationship between clauses and intonational groups or units. Givón (2001a: 258–63) proposed the “one chunk per clause principle” according to which “the majority of sentences/ clauses in connected discourse will have only one chunk [---] under the scope of asserted new information. All other elements in the clause will tend to be topical, background or presupposed old information.” Chafe (e.g. 1979) developed similar ideas but he related them more clearly with intonation. He proposed that there is a limitation on the number of new ideas (later: new information) per intonational unit. Chafe (1994) insisted that the limitation is on the amount of new information, not on clauses per intonational unit. He set out to show that clauses may be spread over several intonational units so that each new piece of information is expressed in its own intonational unit (ibid.: 108–19). The general ideas on the relationship between intonation and syntactic structure presented by Givón and Chafe have formed the basis for a number of more detailed studies of the distribution of intonational groups over syntactic constituents. These studies are typically based on specific sets of data which are studied both qualitatively and quantitatively. To name a few, Croft (1995) looks at English, Tao (1996) at Mandarin, Helasvuo (2003) at English and Finnish, and Wouk (2008) at Sasak. Croft (1995) studied the relationship between intonational units and grammatical structure in a corpus of English oral narratives (the so-called Pear Stories; Chafe 1980). He found that intonational units were almost always also grammatical units (in his corpus of 1989 intonational units, 97 percent were grammatical units). Moreover, over 70 percent of the intonational units were clauses (Croft 1995: 844). In his data, intonational unit boundaries were most often also constituent boundaries, and NPs or PPs were hardly ever split across two intonational units (Croft 1995: 849). Helasvuo’s data from conversational English (consisting of 1,000 intonational units) yielded similar results as 99 percent of intonational unit boundaries were also constituent boundaries (Helasvuo 2003: 24). Helasvuo compared the English data to data from Finnish conversation (consisting of 2,000 intonational units) and found that 98.5 percent of intonational unit boundaries co-occurred with constituent boundaries. Helasvuo (2003) looked at clause cores which she defined as consisting of the predicate and its core arguments (subjects and objects) and found that in her data from conversational English 95.2 percent of the clause cores were produced in a single intonational unit (see also Croft 1995). In Helasvuo’s Finnish data, the percentage was slightly lower as 90.5 percent of clause cores were

68

9781441124609_Ch05_Final_txt_print.indd 68

11/20/2012 9:26:14 AM

Constituents

produced as single intonational units. Helasvuo further found that both languages showed an interesting tendency: in cases where the clause core was distributed across several intonational units, it was most often the object that was in a separate intonational unit (Helasvuo 2003: 29–30). This is in contrast with Cru enden’s claim (1986: 130; see above) that intonational unit boundaries should occur between the subject and the predicate. Neither do Helasvuo’s results lend support to the notion of verb phrase as forming an intonational unit if verb phrases are taken to subsume the object phrase. Tao (1996) looked at intonational and grammatical structure in conversational Mandarin. When looking at the distribution of grammatical structures in Mandarin intonational units he found that in his data, less than half (48 percent) of intonational units were clausal and 29 percent of them were nominal (NPs or relative clauses; see Tao 1996: 72). Of these nominal intonational units, 47 percent were a achable to verbal predicates and 53 percent were una ached to any verbal predicate in their contexts (ibid.: 76). Iwasaki and Tao (1993) compared English, Japanese, and Mandarin conversational data and found that in the Japanese data set the percentage of clausal intonational units was 45 percent and that of nominal intonational units was 22 percent. This is in contrast with English where the percentage of nominal intonational units was only 12 percent (Iwasaki & Tao 1993). Tao (1996: 77) concludes that Mandarin does not exhibit a high degree of correspondence between the intonational unit and the clause. He further notes that typical non-clausal intonational units, among them nominal intonational units, are pervasive in the Mandarin data. Wouk (2008) looked at intonation and grammatical structure in Sasak, spoken in Lombok, Indonesia. She compared her results to Tao’s results from Mandarin, and found that the frequency of clausal intonational units in Sasak was very similar to that of Mandarin, but Sasak had far more complete clauses than Mandarin, which prefers elliptical clauses. Wouk (2008: 150) also found that 22.5 percent of intonational units were nominal expressions (NPs or relative clauses, cf. to Mandarin where 29 percent were nominal intonational units). Nominal intonational units were much more likely to be independent than in Mandarin (72 percent of nominal intonational units were independent in the Sasak data compared to 53 percent in Mandarin; ibid.: 154). We may conclude then that intonational grouping does indeed reflect syntactic structure but not in any straightforward manner. Studies on actual data from different languages suggest that there are cross-linguistic differences in the ways in which boundaries of intonational units correspond to syntactic constituency. These differences may be related to differences in the grammatical resources available in a given language (see, for example, discussion in Wouk 2008: 159–60), as in the means of marking argument relations. Cross-linguistic studies discussing constituency and intonation point to a tendency that, among

69

9781441124609_Ch05_Final_txt_print.indd 69

11/20/2012 9:26:14 AM

The Bloomsbury Companion to Syntax

syntactic constituents, the noun phrase emerges as a robust category that functions as a unit also in terms of intonation.

4 Constituency in the Service of Conversational Activities Looking at the temporal ordering of elements and their delivery in prosodically identifiable chunks or units reveals that there indeed is variation in the alignment of the elements into constituents (see Section 3). This alignment can not only be studied with respect to prosodic chunking but also with respect to other ways of packaging elements into units such as constituents. We will now discuss studies on constituency as observed in language use. In these studies, generalizations about constituent structure can be based on observed recurring discourse pa erns that are dependent on the syntactic resources available in a given language but also on differences in conversational practices (see esp. Wouk 2008). As illustrative examples of this, we will look at how constituency is evidenced in practices of conversational repair and in constructions produced jointly by the conversation participants. In all interactions, participants have recurrent problems in producing, hearing, or understanding talk, giving rise to the need for repairing what has just been said in the current turn. There is a wealth of research, especially in the conversation analytic tradition, on repair organization (see especially Schegloff et al. 1977, Schegloff 1979, Goodwin 1981). This research shows that repair is a conversational activity in which the conversation participants show their understanding of the trajectory of the turn (where it is heading) and the intended meaning. This includes also an understanding of the syntactic trajectory of the turn. There are some studies which have specifically looked at the relationship between syntax and repair practices (in English and Japanese: Fox et al. 1996; in German: Uhmann 2001; in Finnish: Helasvuo et al. 2004; Laakso and Sorjonen 2010). In analyzing the syntax of repair practices, constituency comes into play when studying the scope of recycling when self-repair is performed: there may be differences between languages in the extent to which speakers go back in their u erance when they repair something in their prior talk (for more discussion, see Kärkkäinen et al. 2007). Fox, Hayashi, and Jasperson (1996) found that in English conversations, the scope of the backing up was either the constituent that was being produced when the repair was initiated (NP, PP, VP) or the beginning of the clause. The speakers thus did not back up randomly to any word in the prior talk when doing repair but, rather, showed their understanding of the constituent structure of the clause. When studying their Japanese data, Fox et al. (1996) found that Japanese differs from English in that Japanese speakers back up only to the beginning of the constituent that is being produced and not to the beginning of the clause. 70

9781441124609_Ch05_Final_txt_print.indd 70

11/20/2012 9:26:14 AM

Constituents

Fox et al. (1996) conclude that the projection of the organization of a clausal turn-constructional unit is done incrementally and more locally as compared to English (see also Tanaka 1999 on incremental projection in Japanese). Fox et al. (1996) suggest that the differences in the scope of recycling are motivated by differences in syntactic practices in the two languages. In Japanese, all referring nouns can be marked for case, the verb often comes at the end or near the end of a clause, and the order of nouns before the verb is flexible (see Tanaka 1999). The core arguments are often unexpressed and have to be inferred. According to Fox et al. (1996), a clausal turn-constructional unit in Japanese often begins with a connective or a response word, followed by adverbials, followed by a verb, and ending with a final particle, whereas English clausal turn-constructional units, after a possible discourse particle, start typically with a subject followed by a tight syntactic structure schematized as [S + V + O (+ PP/ADVP)]. The conversation participants can use this tight syntactic schema to project the possible syntactic organization of the u erance and its completion. For speakers, it provides a possibility to use the beginning of a clause as well as a constituent as the point to which they back up when doing the repair, while Japanese speakers operate within the constituent they are currently producing. Fox and Jasperson (1995) show that if repair is initiated postverbally, the repairing segment does not recycle back to the verb. They conclude that in English, repair is not organized through the notion of verb phrase (ibid.: 127). When studying the syntax of repair practices in German, Uhmann (2001) found that speakers tend to back up to the beginning of the constituent (to its functional head, in her terminology). Uhmann suggests that this practice is associated with the structural features of German, especially with its relatively free word order and variability in the position of the finite verb. According to her, there is no single typical syntactic schema in German like the one proposed by Fox et al. (1996) for English. Consequently, the recipients have to rely on other (morpho)syntactic cues in the projection of the trajectory of the turn. In German, noun phrases are inflected for grammatical gender, number, and case, and these categories are encoded especially in the form of the pre-modifiers (determiners and other modifiers). For this reason, speakers back up to the beginning of the constituent in order to show exactly what is being repaired and what kind of syntactic construction is under way (Uhmann 2001). Constituency as it emerges in actual language use is also evidenced in co-constructions, that is, clauses that are produced collaboratively by the conversation participants (see Lerner 1991, 1994, 1996; Sacks 1992: 647–55; Ono and Thompson 1996). To put it broadly, they provide a syntactic resource for doing something together that is normally done by one participant only. Ono and Thompson (1996) have looked at co-constructions in English conversation. They divide co-constructions into two syntactic types—expansions and completions. In expansions, speakers build new units together by expanding what one of 71

9781441124609_Ch05_Final_txt_print.indd 71

11/20/2012 9:26:14 AM

The Bloomsbury Companion to Syntax

them has said into a new syntactic unit, for example, by adding a PP to an already complete clause (ibid.: 78). In completions, a speaker completes a syntactic unit that the speaker of the preceding turn has left incomplete (ibid.: 75). Completions concerning single clauses are especially interesting with respect to constituency because the item offered as a completion of a construction under way shows an understanding of the syntactic structure of the clause that is being produced. Helasvuo (2004) has studied co-constructions concerning single clauses in Finnish conversation. Finnish has rich inflection on both verbal and nominal categories. This means that each element in the clause inflects in a form reflecting its syntactic function, and even uninflected forms show their syntactic function through their lack of inflection. In co-constructions that are completions of single clauses, the form of each member of both parts of the co-construction is controlled by the syntactic unit as a whole. To give a simple example, the finite verb must show agreement in person and in number with the subject, and the case marking of the object is controlled not only by the verb but also by the emerging clause as a whole (most notably the presence or absence of an overt subject and the aspect of the clause). Helasvuo (2004) found that in co-constructions it was much more common that the subject and the finite verb or copula were produced by one speaker and the possible object phrase or predicate nominal was produced by another speaker. In other words, the data on Finnish co-constructions showed a subject–verb grouping rather than, say, verb–object grouping as manifested in the notion of verb phrase (cf. Fox and Jasperson 1995 on repair practices in English conversation). To sum up, studies of actual language use give us a viewpoint on constituency as it emerges in discourse. In their repair practices and in the ways in which they produce co-constructions speakers show their understanding of the syntactic structure that is under way. The study of these practices allows researchers to make observations on how speakers monitor the emergence of syntactic structures such as constituents (cf. also Hopper (this volume) on “emergent constituentization” based on the evidence from increments in conversation). Studies on constituency as it emerges in real discourse call for a search for alternative ways of approaching the organization of syntactic structure. We will now discuss two such proposals.

5 The Emergence of Constituency In traditional discussions of constituency, it is viewed as a way to describe the hierarchic organization of the clause. As depicted in tree structure diagrams, it has been assumed to provide a basis for the definitions of syntactic relations such as the subject and the object: subject is the NP immediately dominated by S 72

9781441124609_Ch05_Final_txt_print.indd 72

11/20/2012 9:26:15 AM

Constituents

and object is the NP immediately dominated by VP (see, for example, Chomsky 1965: 68–9). As such, constituency is viewed as being rather rigid and describing only formal syntactic relations. Within the theory of cognitive grammar, Langacker (e.g. 1995, 1997, 2001) has proposed a very different view of constituency that sees it as emerging from more basic phenomena such as conceptual grouping, phonological grouping, and symbolization. He sees constituency primarily as a part–whole hierarchy and relates it to other aspects of human cognition suggesting that constituency is not unique to grammar or to language but that it is a general feature of our cognition (Langacker 1997: 4). Langacker (1997: 1) argues that constituency is “neither fundamental nor essential to grammatical structure.” Instead, our capacity for grouping on the basis of similarity and contiguity is seen as more basic. Linguistic expressions may involve many kinds of conceptual and phonological groupings which serve as candidates for symbolization. Different assemblies may yield different kinds of groupings. Therefore, constituents are emergent in character, and constituency may be variable and fluid (ibid.: 3). Langacker considers the following examples: (5) Alice likes Bill. (6) Bill Alice likes. (5) contains two component structures, likes and Bill form the composite structure likes Bill, and Alice and likes Bill are integrated to form Alice likes Bill. At the conceptual level, Alice functions as the trajector for the relationship profiled and Bill is the landmark of likes. At the phonological level, Bill is in correspondence with the element that directly follows likes in the temporal sequence. At the higher level of phonological correspondences, Alice is the element that directly precedes likes Bill in the temporal sequence. In contrast, (6) is assumed to exhibit O(SV) constituency (the two major components are [Bill] and [Alice likes]). Langacker bases this assumption on linear ordering of the elements and intonational phrasing (ibid.: 6–9). Consider (7), where the bracketed phrase [she hired] forms a constituent (the example is taken from ibid.: 9). (7) The lawyers [she hired] proved very competent at billing. Langacker (1997: 8) presumes that the classic S(VO) constituency is unmarked in English, but alternate groupings involving an SV constituent are possible if there is some structural motivation to them, as in (7). Finally, it is important to note that in Langacker’s view on constituency, constituency hierarchies are not used for a configurational definition of grammatical relations, that is, the position of an element in the constituency hierarchy 73

9781441124609_Ch05_Final_txt_print.indd 73

11/20/2012 9:26:15 AM

The Bloomsbury Companion to Syntax

does not determine its grammatical role. This underlines the conception of these hierarchies as potentially variable and not essential to grammatical structure (ibid.: 7). In a later paper, Langacker (2001: 182) suggests that when variation is possible, constituency serves the packaging and presentation of conceptual structures in discourse.

6 Constituency and Frequency of Occurrence Bybee and Scheibman (1999) propose that constituency is a ma er of degree: elements that frequently collocate show a tighter constituent structure than those that appear less frequently next to each other. They suggest that constituency derives directly from the frequency with which elements are used together and claim: “While semantic and pragmatic factors determine what occurs together in discourse, the actual repetition of stretches of talk triggers the chunking mechanism that binds them into constituents” (ibid.: 575). Bybee and Scheibman (1999) illustrate their claim with a study on the English don’t in a conversational corpus (see also Scheibman 2002). They found that where don’t was used in a frequent phrase, it was more likely to be produced phonologically in its most reduced form. In these contexts, it was often the case that the negative auxiliary did not contribute compositionally to the meaning of the expression (i.e. in the most frequent collocation I don’t know, don’t did not always literally mean do not; see Bybee and Scheibman 1999: 584). They conclude that high-token frequency of particular items in a given construction creates the tightest constituent structure, whereas high-type frequency results in loose bonds to other parts of the construction (ibid.: 594). Bybee and Scheibman (1999: 589–90) also note that within the three-unit phrase they looked at (I don’t know, think, want, etc.) there appeared to be different degrees of constituency. In the most frequent phrases the phonological reduction of the vowel in don’t depended more on the subject than on the following verb. Based on the evidence of phonological fusion and distribution in terms of type and token frequency Bybee and Scheibman (1999: 593) suggest that there is a different constituent structure for I plus don’t than for full NPs plus lexical verbs. In the case of I plus don’t the subject NP (I) merges with the auxiliary into one constituent ([NP+Aux][V]). In Bybee (2002: 112–13) this approach to constituency is formulated as the “Linear Fusion Hypothesis” according to which constituent hierarchies are derivable from frequent sequential co-occurrence: the more often particular elements occur together, the tighter the constituent structure (see also Bybee 2010: Chapter 8 on “gradient constituency”). Bybee (2002: 130) suggests that constituency is based on repetition of sequences of units. Most often the frequently repeated chunks behave like traditional constituents: they function as units that 74

9781441124609_Ch05_Final_txt_print.indd 74

11/20/2012 9:26:15 AM

Constituents

can be used alone (e.g. in answers to questions), they are replaceable by a proform (distribution), and they can be used in various positions in a clause. Bybee points out, however, that there are other kinds of chunks (such as pronoun + auxiliary mentioned above or verb + preposition) which do not fulfill the traditional grammatical criteria for constituency (they cannot be used alone or be replaced by a pro-form). In her view, there are two factors which converge in constituency, sequentiality and hierarchy; constituency itself is not a basic structure but, rather, an emergent phenomenon in language (ibid.). In conclusion, there are competing views on constituency. Bybee’s views are quite radical as she suggests that “constituents of the type proposed for generative grammar which are described by phrase structure trees do not exist” (ibid.). This is in no way representative of “mainstream” thinking in linguistics but nevertheless, many researchers admit that sequentiality does play a role in constituency, at least in the linear ordering of elements (see Section 2) and intonational phrasing (Section 3).

75

9781441124609_Ch05_Final_txt_print.indd 75

11/20/2012 9:26:15 AM

6

X’-Structure and Minimalism Jairo Nunes

Chapter Overview Introduction Some Properties of X’-Structure Bare Phrase Structure Concluding Remarks and Current Issues

76 78 80 86

1 Introduction A fundamental property of natural languages is that words combine into larger units with hierarchical structure. The sentence in (1), for example, does not simply involve a bag of words, for it can be chopped off into smaller grammatically significant units such as the proposal, this hypothesis, on this hypothesis, or relies on this hypothesis. (1) The proposal relies on this hypothesis. One of the major goals of the generative enterprise has been to properly characterize this hierarchical structure. In the Aspects model (Chomsky 1965), for instance, the structure of syntactic constituents like relies on this hypothesis in (1) was captured by phrase structure rules of the type illustrated in (2a), coupled with information regarding the subcategorization features of the lexical items along the lines of (2b). (2a) states that a verb phrase is composed of a verb optionally followed by a noun phrase, which may in turn be optionally followed by a prepositional phrase, whereas (2b) informs that the verb rely selects for a prepositional phrase. The combination of (2a) and (2b) ensures that in grammatical structures, a verb like rely is inserted in a verb phrase that has been rewritten as V PP and not as V NP, V NP PP or simply V.

76

9781441124609_Ch06_Final_txt_print.indd 76

11/20/2012 9:23:43 AM

X’-Structure and Minimalism

(2)

(a) VP → V (NP) (PP) (b) rely: [__ PP]

However successful in describing the broad structure of natural languages, this approach faces a couple of problems.1 On the one hand, it is redundant in the sense that the information that a verb may form a constituent with a PP is encoded twice: in the phrase structure rules and in the lexical entry of verbs like rely. On the other hand, it fails to capture some generalizations that cut across structures. For instance, a verb phrase must obligatorily have a verb in the same way a noun phrase must obligatorily have a noun. Similarly, the verb rely forms a constituent with a PP in the same way the related noun reliance does. However, these two generalizations come out as completely accidental in an approach that adopts mechanisms such as (2), for there is nothing in the format of the rules that has an overarching coverage across categories. The development of X’-Theory within the EST and GB models (see, for example, Chomsky 1981) aimed at overcoming these sorts of problems. Phrase structure rules were eliminated and the subcategorization information of the lexical items was mapped into a general phrasal skeleton along the lines of (3) below.2 In (3) X, Y, Z, W, and K are place holders for any major lexical category (V, N, P, A, etc.), capturing the endocentricity property of natural languages— that is, the fact a verb phrase must have a verb, an adjectival phrase must have an adjective, etc. (3) also captures the fact that the combination of the verb rely with its complement and the combination of reliance with its complement result in structurally comparable structures. (3)

(XP) (KP) (adjunct)

. . . .  (XP) XP

(WP) (adjunct) (ZP) (specifier)

X0 (head)

X’

(YP) (complement)

77

9781441124609_Ch06_Final_txt_print.indd 77

11/20/2012 9:23:43 AM

The Bloomsbury Companion to Syntax

(3) in fact embodies many other claims about syntactic constituents. For instance, it purports to show that syntactic constituents involve binary branching and that there can only be one specifier per head. In the sections that follow, we will consider some of these claims within GB and discuss how and why they have been reanalyzed within the Minimalist Program.

2 Some Properties of X’-Structure 2.1 Bar-levels One distinctive property of the X’-template is that it takes syntactic constituents to be structured around three different levels of complexity: the head (X0), the intermediate projection (X’), and the maximal projection (XP). The bar-level symbols “ 0 ”, “ ‘ ”, and “P” appended to X are taken to be intrinsic features of the syntactic skeleton that determine their behavior in the computation. For instance, heads and maximal projections can be targeted for movement purposes, but intermediate projections cannot. The bar-level features also provide a recipe for interpreting the constituent that is the sister of a bar-level projection. Thus, if a given constituent C is the sister of a zero level category X, C will be interpreted as the complement of X; if the sister of a maximal projection XP, C will be interpreted as an adjunct of X; and if the sister of an intermediate projection, C will be interpreted as the “specifier” of X, a loose notion that encompasses both arguments (as is the case of the specifier of VP, for instance) or other elements that enter into grammatical relations such as agreement (as is the case of the specifier of IP, for example). According to the schema in (3), there are also two important distinctions between arguments and specifiers, on the one hand, and adjuncts, on the other. First, the number of complements and specifiers is restricted to one each per head, but there is no fixed limit on the number of adjuncts. This distinction is empirically based on the observation that clauses prototypically have one subject and transitive verbs prototypically have one complement, for instance, but there is no prototypical number of adjuncts a given projection may take. Second, when complements and specifiers are introduced, the resulting structure has a different bar-level specification (a head combined with a complement yields an intermediate projection and an intermediate projection combined with a specifier yields a maximal projection). By contrast, adjuncts do not change the bar-level specification of their target. If saw the star is a VP, the addition of the adjunct with the telescope leaves the barlevel of the resulting structure unaltered: saw the star with a telescope is also a VP. 78

9781441124609_Ch06_Final_txt_print.indd 78

11/20/2012 9:23:43 AM

X’-Structure and Minimalism

2.2 Binary Branching Another property that (3) encodes is binary branching. Research since the 1980’s triggered by Kayne’s (1984) influential work has led to a wholesale reevaluation of ternary or n-ary branching structures previously assumed for different categories. The structure of the clause, for instance, was reanalyzed from the ternary branching structure [S NP INFL VP] to the binary branching structure [IP NP [I’ I 0 VP]] (see, for example, Chomsky 1986a). Similarly, ditransitive structures previously analyzed as [VP V NP PP] were reconceived as binary branching structures with the format [VP NP [V’ V PP]] with an additional movement of V to the le of NP (see, for example, Larson 1988). Underlying the move in this direction were several empirical observations that pointed to the conclusion that the internal constituents of a given structure must be organized in terms of asymmetric, rather than symmetric c-command and this is compatible with binary branching structures, but incompatible with flat structures. Take the sentences in (4), for instance. If the two arguments of give stood in a mutual c-command relation, both the sentences should be acceptable, for the negative polarity expression headed by any would be locally c-commanded by the negative expression headed by no. The contrast between (4a) and (4b) thus indicates that the first complement must asymmetrically c-command the second one and this can be captured by a binary branching structure along the lines of [VP NP [V’ V NP]] (followed by movement of the verb). (4)

(a)  I gave nobody anything. (b) *I gave anybody nothing.

The two reanalyses mentioned above had far-reaching consequences as they set the basis for additional reanalyses of other projections or the postulation of new projections. It has now become the null hypothesis that for any new head H that one postulates, the projections of H will organize themselves in a binary branching fashion.

2.3 Uniqueness of Mother Nodes The X’-template in (3) also embodies the idea that the structure of the constituents of natural languages is such that it prevents a given element from being immediately dominated by more than one node. That is, (3) excludes structures like (5a) where X takes YP and ZP for complements and projects an intermediate projection in each case, or (5b) where WP is simultaneously the complement of X and the specifier of Y. 79

9781441124609_Ch06_Final_txt_print.indd 79

11/20/2012 9:23:44 AM

The Bloomsbury Companion to Syntax

(5)

(a)

XP X’

X’ X

YP

(b)

ZP

XP YP

X’ X

WP

Y’

Y

2.4 Endocentricity Finally, the X’-template captures the fact that there is an asymmetry between the immediate constituents C1 and C2 of a given syntactic object Σ in the sense that only one of them determines the (categorial) properties of Σ. Take the syntactic object eat apples, for instance. The combination of eat and apples results in a constituent that is verbal and not nominal, as can be seen by the fact illustrated in (6) that distributionally, eat apples may occupy positions that can be occupied by verbs. (6)

I will [run]/[eat apples].

The endocentricity encoded in the X’-template thus ensures that whenever we find syntactic constituents larger than a single lexical item, their properties must be determined by one of its immediate constituents. This in turn led to the complete elimination of phrase structure rules, for even clauses whose phrase structure was exocentric (S → NP INFL VP) were realized as projections of a given functional head ([IP NP [I’ I 0 VP]]), see Parodi and Quicoli, this volume.

3 Bare Phrase Structure With the emergence of the Minimalist Program in the 1990’s, a wholesale conceptual evaluation of the technical apparatus made available in GB took place 80

9781441124609_Ch06_Final_txt_print.indd 80

11/20/2012 9:23:44 AM

X’-Structure and Minimalism

and X’-Theory was one of the components that was targeted for a close examination. The aim of this inspection is, on the one hand, to distinguish the properties of X’-structures that reflect true properties of phrase structure in natural languages from the ones that are tied to the formal notation used and, on the other, to investigate if such properties can be derived from deeper properties of the language faculty. Let us then consider how and why X’-structures came to be reanalyzed in terms of what Chomsky (1995) called bare phrase structure.

3.1 Functional Determination of Bar-levels Consider the X’-structure in (7). (7)

VP V’

NP N’

V

NP

N

loves

N’

Mary

N John

(7) illustrates three properties that may raise minimalist qualms. The first one involves a redundancy between the terminal nodes and the lexical items that they dominate. Arguably, the lexical entry of Mary, for instance, includes the information that it is a noun. Hence, there is nothing that the node N dominating Mary encodes that is not already encoded by the latter. To put it differently, by introducing a distinction between X 0 heads and lexical items, X’-structures bring seemingly unnecessary redundancy to the system. At first sight, the distinction may seem useful in determining projection; a er all, one has to capture the fact that the constituent loves Mary is verbal and not nominal. However, all that is strictly required is that the resulting object has the relevant properties of one of its immediate constituents and not necessarily that its categorial feature must be encoded. The second problem regards vacuous projections. Recall that complements and specifiers are in principle optional (cf. (3)) and their presence is determined by the properties of the lexical item inserted under the X 0 head. In the case at hand, Mary and John do not have the relevant selectional features and, 81

9781441124609_Ch06_Final_txt_print.indd 81

11/20/2012 9:23:44 AM

The Bloomsbury Companion to Syntax

accordingly, they do not have a specifier or a complement. However, under the assumption that bar-levels are intrinsic features of the X’-skeleton, they must still project an N’. This is an unfortunate result, though, for intermediate projections arguably cannot move and both Mary and John can undergo movement. The bar-level themselves are also suspicious from a minimalist perspective. Chomsky (1995) has proposed that the syntactic computations should be subject to the Inclusiveness Condition, which requires that syntactic objects be built from the features present in the lexical atoms that feed the derivation. As mentioned above, the features “ 0 ”, “ ‘ ”, and “P” are tied to the X’-skeleton and, as such, cannot be construed as lexical features. All things being equal, a system that eschews such features is on better conceptual grounds. Chomsky then proposes that the projection status of a given syntactic object should be determined not in terms of extrinsic features but in terms of relational notions along the lines of (8). (8)

(a) Minimal Projections: lexical items that feed the computation. (b) Maximal Projections: syntactic objects that do not project. (c) Intermediate projections: syntactic objects that are neither minimal nor maximal projections.

Given the considerations above, the structure in (7) can be simplified along the lines of (9) below, where (i) there is no redundant distinction between terminal nodes and lexical items; (ii) there are no vacuous projections; and (iii) each syntactic object can have its phrasal status determined in virtue of its relations with the other syntactic objects; hence, Mary and John are minimal maximal projections (they are lexical items that do not project), the lowest instance of loves is a minimal nonmaximal projection, the highest instance of loves is a nonminimal maximal projection, and the instance immediately dominating loves and Mary is an intermediate projection, as it is neither minimal nor maximal. The repetition of loves is a notational device to encode that the lexical item loves is the item that ultimately determines the grammatical properties of the larger syntactic objects loves John and Mary loves John. Derivatively, one may also define complements and specifiers in terms of (8): complements are sisters of minimal nonmaximal projections and specifiers are sisters of intermediate projections. (9)

loves Mary

loves loves

John

82

9781441124609_Ch06_Final_txt_print.indd 82

11/20/2012 9:23:44 AM

X’-Structure and Minimalism

Notice that (9) also captures another property encoded in (3): that specifiers and complements must be maximal projections. In X’-structures this is a stipulative property. By contrast, in a representation such as (9), this follows from the relation among the syntactic objects; since Mary and John do not project, they are necessarily maximal projections. Before closing this section, it is worth observing that the elimination of vacuous projections does not face any obvious empirical challenges. Take the structural difference between the unaccusative and unergative verbs under X’-structures, for instance, which was captured by means of a vacuous intermediate projection of V so that the internal argument of unaccusative structures was mapped inside V’ (e.g. [VP [V ’ arrived John]]) and the external argument of unergative structures was mapped outside V’ (e.g. [VP John [V ’ sneezed]]). With recent developments of the fine structure of VP, the two structures may be distinguished in terms of whether the argument is the sister of V or the sister of a null light verb projection ([vP v [VP arrived John]] vs. [vP John [v ’ v [VP sneezed]]]), with no need to resort to vacuous intermediate projections.

3.2 The Operation Merge In the previous section, we discussed properties of phrasal syntactic objects but le aside the question of how such objects are created once phrase structure rules have been abandoned. Trying to keep the number of assumptions minimal, what we need is an operation that combines lexical items and complex objects built from lexical items and specifies the relevant properties of the resulting structure. Chomsky (1995) calls this operation Merge. Applied to two lexical items α and β, Merge creates the complex syntactic object {γ, {α, β}}, where γ is the label of the resulting structure informing the computation of its relevant grammatical properties. Thus, by merging loves and John, we obtain the syntactic object {loves, {loves, John}}, whose syntactic properties are determined by loves (hence, we say that loves John is a verbal projection). As Merge can apply iteratively, it can target Mary and the complex object {loves, {loves, John}}, yielding as the result {loves, {Mary, {loves, {loves, John}}}}, which is also a verbal projection as its properties are determined by loves. One question that arises is how the label of a given complex syntactic object is determined. We have already seen that from an empirical point of view, it is the case that one of the immediate constituents determines the properties of the resulting structure, but what prevents the system from projecting John when it merges with loves, for example? In the best of possible worlds, the relevant information should be locally available in order for computational complexity to be reduced. Fortunately, there seems to be an independent asymmetry between loves and John which may be used to determine the label of the resulting 83

9781441124609_Ch06_Final_txt_print.indd 83

11/20/2012 9:23:44 AM

The Bloomsbury Companion to Syntax

structure. It is a property of loves that it requires a complement, but it is not a property of John that it requires a verb to be the complement of. If something along these lines is correct, a head will have as many projections as necessary to have its requirements met. This brings up the issue of whether the number of specifiers must be restricted to one. The original empirical motivation for this assumption was that certain designated heads “closed off ” certain projections. For instance, determiners were taken to be specifiers of NP because they were the highest elements of the nominal domain. However, this fact has been reanalyzed in a more streamline fashion in accordance with the X’-template in (3) as a reflex of D taking an NP for its complement and projecting a DP. Thus, in absence of evidence to the contrary, the number of specifiers should not be specified, which is exactly what the bare phrase structure system leads us to expect. And multiple specifiers may in fact be independently required. Chomsky (1995), for instance, has reanalyzed overt object movement as movement to the “outer” specifier of vP. Another question is whether Merge can target any two syntactic objects. Under minimalist inspirations, economy considerations encompassed by Last Resort demand that there can be no superfluous applications of a given operation in a convergent derivation. Applied to Merge, these considerations have two consequences worth noting. First, two syntactic objects can undergo merger only if one satisfies requirements of the other. Given the lexical items Mary, and John, for example, they cannot be merged as neither satisfies requirements of the other. Importantly, this information is made locally available either by the lexical item itself or by the label determined by it in the case of phrasal constituents. The same reasoning prevents the creation of vacuous projections. Under bare phrase structure, vacuous projection can only be derived if a given syntactic object can be merged with nothing, that is, if Merge can apply to a single term. But it is arguably the case that one such application of Merge does not (locally) satisfy any requirements of the term it targets; hence, it is blocked by Last Resort and no vacuous projections are derived. The above discussion leads to the conclusion that a successful application of Merge will target at least two terms. But why must it stick to two? Why can it not apply to three terms, for instance? One possibility is that this is an optimal solution to the question of how much the system needs to create complex syntactic objects and get recursive structures. If two is the minimal number of terms required for Merge to operate with, an optimal solution may simply take the upper limit to be two, as well. If this suggestion is on the right track, binary branching follows. It would be a by-product of the general strategy of trying to do the most with the least. Finally, let us consider the claim that in natural languages a given syntactic object cannot be immediately dominated by more than one node (see 84

9781441124609_Ch06_Final_txt_print.indd 84

11/20/2012 9:23:44 AM

X’-Structure and Minimalism

Section 2.3). Chomsky (1995) has argued that computational complexity may be considerably reduced if Merge can only operate with root syntactic objects (the Extension Condition). If so, Merge cannot target a constituent that is dominated by another and structures such as (5) are ruled out.

3.3 Adjunction In the sections above we have focused on the syntactic representation of structures containing complements and specifiers within the bare phrase structure system. Let us now see how adjuncts are to be represented. It is fair to say that the bare phrase structure system inherited the conceptual problems regarding adjunction found in X’-structures. Let us consider some of these problems by inspecting the X’-structure in (10), formed by adjunction of a moved verb to the Infl (Inflectional) head. (10)

I’ I0 V0

VP I0

. . . tV . . .

Given that further I-to-C movement should piedpipe V, we must conclude that [I 0 V 0 I 0] in (10) is a syntactic constituent. On the other hand, under the standard assumption that a moved element must c-command its trace, it must be the case that the type of constituenthood produced by adjunction is such that it does not prevent the adjunct from c-commanding out of the adjunction structure; otherwise, the moved V in (10) will not be able to c-command its trace. Within GB, this problem was accommodated by distinguishing two types of integrative relations: containment and dominance (see, for example, Chomsky 1986a). Once adjuncts did not change the bar-level of their hosts, the idea was that adjunction did not create new projections but extended the projection of their host in more segments. Under this view, containment is looser than dominance: α contains β if β stands in an integrative relation with some segment of α, whereas α dominates β if β stands in an integrative relation with all the segments of α. If c-command is to be defined in terms of dominance rather than containment, the adjoined V in (10) does c-command its trace as it is only contained but not dominated by I 0. Notice that adjunction is also exceptional in that it apparently need not satisfy the Extension Condition. In (10), for instance, the moved verb does not merge with a root syntactic object. In face of problems such as this, Chomsky 85

9781441124609_Ch06_Final_txt_print.indd 85

11/20/2012 9:23:44 AM

The Bloomsbury Companion to Syntax

(1995) has suggested that the Extension Condition should in fact not hold for adjunction structures. But the most problematic issue is how to capture the idea that adjunction preserves that bar-level status of the relevant host if phrasal status is to be functionally determined, as discussed in Section 3.1. Take the phrase speak Portuguese well, for instance. Merger of the verb and the noun yields the syntactic object K = {speak, {speak, Portuguese}}, which under the functional interpretation of phrasal status is a nonminimal maximal projection. What is now the structure resulting from adjoining the adverb to K? It should at least be a set containing K and the adverb. The question is what kind of label the resulting structure has. We know that it should be determined by speak as it is a verbal projection. However, if the verb projects yielding L = {speak, {{speak, {speak, Portuguese}}, well}}, K within L will be interpreted as an intermediate projection (it is not a lexical item and it projects) and, consequently, the adverb will be interpreted as a specifier and not as an adjunct of K. To keep the distinction between adjuncts and specifiers, Chomsky (1995) in a sense resurrected the distinction between segments and categories and suggested that adjunction involves a different kind of label, which is determined by the head of the construction but leaves the original structure unaltered as far as its phrasal status goes. The notation Chomsky uses for this label is an ordered pair based on the head of the constructions: {, {α, β}}, where γ is again determined by either α or β. In the case at hand, speak Portuguese well should be represented as { {{speak, {speak, Portuguese}}, well}}.3 It is very likely that the two types of complex objects formed by Merge may actually be the output of two related but different operations. Chomsky (2001) has termed the operation that adds a new category to the computation yielding objects of the form {γ, {α, β}} set-merge and the one that only adds a segment to its host yielding objects of the form {γ, {α, β}} pair-merge.

4 Concluding Remarks and Current Issues In this chapter we have seen how X’-structures have been reanalyzed in terms of general architectural properties of the language faculty as conceived of in the Minimalist Program. To recap some of the results, the bar-level features “ 0 ”, “ ‘ ”, and “P” have been abandoned in favor of a functional interpretation of the relations established among the components of a syntactic object. General economy considerations led to the elimination of vacuous projections and the redundancy between terminal nodes and lexical items. Finally, it was suggested that the properties of binary branching and uniqueness of mother nodes can be grounded on general consideration of computational efficiency. 86

9781441124609_Ch06_Final_txt_print.indd 86

11/20/2012 9:23:44 AM

X’-Structure and Minimalism

This change of perspective also le many problems unsatisfactorily handled, as is the case with adjunction structures, and paved the way for the formulation of deeper questions such as why syntactic objects are endocentric. These issues have generated an effervescent area of investigation. To name a few, Kayne (1994) has raised the influential hypothesis that linearization considerations may determine many of the properties of phrase structure in natural languages. Another possibility that has been productively explored is that the movement can be conceived as a sort of internal-merge and as such, it is not restricted to root syntactic objects (see, for example, Citko 2005). More recently, Hornstein (2009) has argued that Merge should in fact be conceived as the output of two different operations: Concatenate and Label. Hornstein and Nunes (2008) have proposed that adjunction only involves Concatenate and not Label (see also Chametzky 2003). There have also been proposals to eliminate labels all together (see Collins 2002). The above is definitely just a sample and is in no way meant to be a comprehensive list of the exciting works on bare phrase structure that have been recently explored (see, for example, Chametzky 2000 and Boeckx 2008 for relevant discussion). All that it does is to illustrate how productive it has been to evaluate the technical apparatus assumed in GB under minimalist lenses.

Notes 1. See, for example, Lyons (1968), Chomsky (1970), and Jackendoff (1977). 2. For purposes of exposition I will put aside for now the possibility of adjunction to heads and intermediate projections (see Sections 3.3 and 4 for discussion). 3. For the relevant definitions of dominance, containment, and c-command under bare phrase structure objects, see Nunes and Thompson (1998).

87

9781441124609_Ch06_Final_txt_print.indd 87

11/20/2012 9:23:45 AM

7

The Structure of NPs: Some Insights on Case, Empty Categories, and Poverty of Stimulus1 Giuseppe Longobardi and Giuseppina Silvestri

Chapter Overview Some Properties of Thematic Structure Varieties of Genitive Syntactic Realization of Arguments Conclusions

89 93 104 116

Owing to the broadness of the domain, in this chapter we will limit ourselves to examine some properties of the argument structure of nouns, disregarding those areas that have come to be considered part of the extended or functional projections of nominal phrases, such as determiners (at least since Szabolcsi 1981; Abney 1987) or adjectival modification (especially since Bernstein 1991; Crisma 1991; Valois 1991). A good deal of information about such areas is anyway available in excellent reference sources such as Plank (2003), Alexiadou et al. (2007), and Keenan and Paperno (2012). Focusing on argument structure, we will consider three major aspects of it: in Section 1 we will deal with the mapping of arguments to thematic structure, in Section 2 with conditions on the formal realization of those phonologically expressed, and in Section 3 with the evidence for the existence of some phonologically null ones. Indeed, this last section will show that nominals provide some of the strongest theory-neutral evidence for the existence of phonetically empty categories and for the role of poverty-of-stimulus considerations in arguing for the latter. 88

9781441124609_Ch07_Final_txt_print.indd 88

11/20/2012 9:25:30 AM

The Structure of NPs

1 Some Properties of Thematic Structure 1.1 Hierarchies of Arguments According to at least three different criteria, the main arguments of a head noun turn out to be hierarchically ordered roughly as in clauses (Thematic Correspondence Hypothesis in Giorgi and Longobardi 1991, possibly a subcase of Baker’s 1988 Uniformity of Theta Assignment Hypothesis, UTAH):2 thematic Subjects (e.g. Agents) are higher than direct Objects (e.g. Themes) and other complements. Nominals peculiarly involve the possibility of a further quasiargument, the R-related phrase of Higginbotham (1983), sometimes improperly called Possessor, which does not exist in clausal structures. Possessors turn out to be higher than Subjects. The conjunction of the two generalizations leads one to assume the following hierarchy: P(ossessor)>S(ubject)>O(bject). All these arguments will be called “direct,” the first two external, the third one internal. The three relevant testing grounds for this hierarchy are the following: (1) (a) possessivization (b) extraction (c) c-command3 The first and older type of evidence for the hierarchy above stems from the observation that many European varieties admit two ways of formally realizing the P, S, and O arguments of the noun (cf. Section 2.2), which somewhat recall the two-Case distinction (e.g. Nominative/Accusative) of clausal structures: (i) through a prepositional form (English of, German von, Romance de or di), steadily postnominal; (ii) by means of either a postpositional affix (e.g. English ‘s) or a special agreeing form (e.g. possessive pronouns). Now, it is very generally the case that type (ii) realization (call it “possessivization,” cf. Section 1.2) is subject to these limitations: (2) (a) if only one among P, S, and O is present, then it will normally be able to assume type (ii) form; (b) if P is overtly present, it will be the only one able to assume type (ii) form; (c) if P is not overtly present and S is, only the latter (i.e. no O) will be able to assume type (ii) form.

89

9781441124609_Ch07_Final_txt_print.indd 89

11/20/2012 9:25:31 AM

The Bloomsbury Companion to Syntax

Very good evidence to exemplify the above generalizations can be provided by French: (3) (a) Mes photographies “my photographs” (b) Mes photographies de Jean “my photographs of Jean” Mes in (3a) has three readings (P, S, or O). The pair of arguments mes, Jean in (3b) also have three, and only three, natural readings: (4) Mes photographies de Jean P S P O S O In Romance at least, the same hierarchy of arguments is replicated by another class of phenomena, namely their accessibility to extraction from the DP through wh-movement or cliticization: (5) Of the phrases in the frame of a head N, only one representing an argument expressible through type (ii) form can be extracted from Nmax. (6) (a) Marie, dont j’ai admiré les photographies . . . “Marie, of whom I admired the photographs . . .” (b) Marie, dont j’ai admiré les photographies de Jean . . . “Marie, of whom I admired the photographs of Jean . . .“ Again, (6a) is three-way ambiguous, and (6b) too, with the only possible readings of the two arguments distributed as in (4): (7) dont . . . . . . . . . de Jean P S P O S O The results of extraction tests, thus, confirm those of possessivization tests (cf. Cinque 1980, Zubizarreta 1980, Milner 1982, Giorgi and Longobardi 1991). Finally, evidence for the same hierarchy comes from such a classical constituency testing ground as c-command relations. Given any pair of arguments among P, S, and O, one containing an anaphorically or quantificationally bound 90

9781441124609_Ch07_Final_txt_print.indd 90

11/20/2012 9:25:31 AM

The Structure of NPs

expression and the other representing the antecedent of such a binding relation, it is invariably the case that P always represents the binder, O always contains the bindee, and S may bind inside O but never inside P. Thus, basically the same hierarchy as in (4) is replicated by the three natural readings of the following example with the Italian possessive anaphor proprio (Giorgi 1983): (8) Ho visto le foto di Maria della propria madre. “I saw the photographs of Maria of self’s mother.” P S P O S O Correspondingly, an R-expression embedded within O or S is necessarily disjointed from P and one embedded within O is disjointed also from S.4 Thus, the hierarchy appears to be structurally represented as follows (order irrelevant): (9) [P [S [O . . . N . . . ]]]

1.2 Supposed “Passivization” Properties The phenomenon of possessivization of O bears some superficial resemblance to passive in clauses: thus it has often been referred to as “passivization” of NPs. This is likely to be a mislabeling; in fact the analogy breaks down in at least four respects: (1) A well established restriction in English (though not in all other languages, cf. Section 3.1) is that some Os cannot appear in the possessivized form even if no overt S or P is present. Most head nouns displaying this restriction are characterized by their assigning a semantically “unaffecting” θ-role to their objects (cf. Anderson 1979): (10) (a) The perception/knowledge of the problem. (b) *The problem’s perception/knowledge. A plausible approach to this class of nouns was suggested in Jaeggli (1986): these heads would be unable to give up the projection of their external argument S, which would then be always obligatorily realized at least syntactically if not phonetically. Such a category, then, can be assumed to compete with an overt possessivized O phrase for one and the same structural position so that the presence of one will exclude the other. The syntactic obligatoriness of 91

9781441124609_Ch07_Final_txt_print.indd 91

11/20/2012 9:25:31 AM

The Bloomsbury Companion to Syntax

the subject with the heads in question seems independently supported by the impossibility of the middle constructions of lexically corresponding verbs: (11)  *The problem perceives/knows easily. Indeed, middle is another environment where the promotion of the object to a subject-like form destroys the possibility of syntactic realization of the underlying subject role, as suggested by the known disappearance of control and binding activity on the part of this latter argument: (12)  (a) The ship was sunk (to collect the insurance).  (b) The ship sank (*to collect the insurance). This does not happen in passives (where some special morphology “retrieves” the subject role: Baker et al. 1989); (2) Correspondingly, in English noun phrases possessivization of the object destroys any trace of syntactic activity of the understood subject role as a controller (Roeper 1987), precisely as happens with middle verbs, though not with passives: (13)  (a) The sinking of the ship (to collect the insurance).  (b) The ship’s sinking (*to collect the insurance). (3) In some languages other than English or French (e.g. Italian, German) the preposition introducing the expression of the agent in nominals with a possessivized object is not the same as the one expressing the agent in verbal passives (cf. Italian da parte di vs. da, German durch vs. von). (4) Languages typically have quite distinct morphological forms for passive verbs, but not for “passive” nouns; again this recalls middle formation for verbs. On the whole, then, possessivization of the object of a noun looks more like clausal middle formation rather than clausal passivization (also cf. Section 3.1).

1.3 Nominative/Ergative Although the two types of formal realization of direct arguments obviously recall the two Cases of the main clausal arguments (cf. Section 2.2), the nominal

92

9781441124609_Ch07_Final_txt_print.indd 92

11/20/2012 9:25:31 AM

The Structure of NPs

system can hardly be assimilated to either a Nominative–Accusative or an Ergative–Absolutive pattern; both markings are in fact optionally possible on the only argument present, in cases like the following: (14)  (a) Mary’s appearance/The appearance of Mary  (b) The president’s arrival/The arrival of the president

2 Varieties of Genitive 2.1 Genitive Several languages use one and the same Case, traditionally termed Genitive, for the arguments of nouns whose verbal thematic correspondents bear instead formally distinct Nominative and Accusative (or Ergative and Absolutive). Such a Case is normally also employed to express P. This does not mean that there cannot be different formal realizations of Genitive in the same language, as we have already seen, resembling the clausal bipartite system also in the sense that such distinctions are mostly positionally determined and not directly connected to thematic interpretations; the retention of a single Case name is probably justified, after all, by considerations like those of Section 1.3. Thus, in most languages one must distinguish different instances of Genitive Case.

2.2 Types of Genitive 2.2.1 “Free” and “Functional” Genitive Formally, a superficially salient descriptive divide separates instances of Genitive Case realized by means of a preposition (as found in most Romance, Germanic, Celtic, and, say, in Bulgarian and some Semitic languages) from the others, many of which (though not all: cf. fn. 6 below) instantiate the forms labeled above as “possessivized.” As for prepositional Genitives, it must be remarked that identifying them unequivocally is less obvious than one may expect, since some arguments bearing the role we called P are often expressed through dative-like marking (such as French à), and some S arguments can be realized through agent-like prepositions (like in English “a book by Chomsky”). As an operational rule we will thus refer to a phrase as prepositional Genitive only if its form may also be used to realize a thematic O argument.

93

9781441124609_Ch07_Final_txt_print.indd 93

11/20/2012 9:25:31 AM

The Bloomsbury Companion to Syntax

As for the non-prepositional realizations, one first notices that among them a maximum of morphological heterogeneity is found, suggesting they can hardly be recognized as a natural class. There are at least five different ways of formal realization: (15) (a) a phrase-final affix (b) a word-final affix (c) an inflectional ending5 (d) φ-feature concord with N

(e) zero-realization

(e.g. English/Scandinavian s) (e.g. German -s, perhaps Arabic -i) (e.g. Latin or Slavic Genitive) (e.g. Romance possessive pronouns, Slavic genitival adjectives, or, in the reverse direction, Hungarian Genitive) (Hebrew construct state Genitive)

Therefore, pure morphology hardly provides especially deep generalizations on Genitives. Syntactically, however, a very consequential distinction arises: prepositional Genitives and some inflectional ones (Latin Genitives, Gianollo 2005; probably also Classical Greek ones, Guardiano 2003) are rather freely iterable, once provided with a natural thematic interpretation (“Leonardo’s famous portrait of Monna Lisa of the Louvre Museum” etc.); in particular, more than one of them may surface in post-adjectival positions. This is not the case with any of the other types mentioned. Slightly oversimplifying, there are two main types, a functionally checked Genitive, bound to precise structural positions, and a free Genitive, the latter licensed simply by its being in (or connected to) a thematically suitable position. However, the distinction cannot be easily learnt from environmental evidence: the characteristic iterability of free Genitive is unlikely to be robustly manifested in spoken corpora, where instead nouns with a single Genitive abound. Therefore, it is not surprising that, correlating with the syntactic distinction (iterable=free/non-iterable=functional), one finds some more detectable peculiarities—circumventing poverty-of-stimulus limitations on the identification of the Genitive type: on one side, only Genitives of the functional type are those which, given appropriate conditions, can and must transfer their definite interpretation to the whole DP they belong to (e.g. Semitic construct state, Celtic Genitives, Germanic Saxon Genitives); this is never the case with free Genitives, despite occasional interactions among definiteness values (e.g. Stavrou and Tsimpli 2011). On the other side, free Genitives, in the syntactic sense, seem to be always realized through robust formal marking in the form of clearly inflectional Case or, analytically, through a dedicated preposition6. Instead, functionally checked Genitive, as we have seen, may be signaled by often reduced (and anyway non-prepositional7) forms.

94

9781441124609_Ch07_Final_txt_print.indd 94

11/20/2012 9:25:31 AM

The Structure of NPs

2.2.2 Free Genitive Summing up, free Genitive seems to be minimally characterized by the following properties: (16) (a) is always formally marked, even in languages where other realizations of Genitive morphophonologically completely unmarked or less robustly encoded; (b) is freely iterable, whenever thematically interpretable; (c) does not suffice to satisfy requirements on definiteness marking of the head nominal. Finally, free Genitive appears to be subject to a general constraint, itself quite useful in enabling property (16a) to serve as a remedy to poverty-of-stimulus limitations: (17) Uniqueness: every language has at most one form to express free Genitive. If such conditions do not hold, we will have to do with “functional” Genitive.

2.2.3 Functional Genitive Cross-linguistically, there are at least (perhaps also at most; but cf. (11)) two positions for functional Genitive, one coming before adjectives and the other after them (though still preceding prepositional Genitives in case of cooccurrence), with the surface position of the head noun cross-linguistically varying without substantially affecting this generalization. The two positions have been descriptively labeled GenS and GenO, respectively, in Longobardi (2001), though without any allusion to thematic function. GenS is best represented by Germanic prenominal -s Genitive and probably by all the agreeing types in (15d). GenO, the lower, basically post-adjectival type, is exemplified by Greek, Slavic, Icelandic, Celtic (prepositionless) Genitives (Duffield 1993, Rouveret 1994), and also by some cases in Old English (Crisma 1999), Old French, and certain sub-standard Romance varieties (Delfitto and Paradisi 2009). It is slightly less straightforward how to class Semitic construct-state Genitives, which surface before adjectives (unlike in Celtic) but could have been raised there from a lower position along with movement of the noun itself, as part of a general process of phrasal raising (“roll up,” Shlonsky 2000). Nothing appears to prevent languages from using both GenS and GenO at the same time, as for example, in German:

95

9781441124609_Ch07_Final_txt_print.indd 95

11/20/2012 9:25:31 AM

The Bloomsbury Companion to Syntax

(18) Marias sorgfältige Beschreibung Ottos “Maria’s accurate description of Otto” In fact, parametrically, languages seem to choose to activate one or the other or both of such positions, and to combine them or not with prepositional (free) Genitive. Therefore, we have languages manifesting GenS alone (Hungarian), GenS and free Gen (English), GenO alone (Greek), GenO and free Gen (Celtic), free Gen alone (most of Romance, if we exclude pronominal possessives from the picture), GenS and GenO (most of Slavic, if we include adjectival Genitives in the picture), and finally all of them (German, probably Latin: Gianollo 2005). This apparent parametric independence may however be subject to some deeper cross-linguistic constraints.

2.3 The Forms of Free Genitive 2.3.1 Consistency First of all, it must be noticed that prepositional Genitive seems incompatible with occurrence in a high structural position, in particular with prenominal surfacing (cf. below). This restriction might follow from general independent constraints, conceivably from Giorgi and Longobardi’s (1991) Consistency Principle, primarily meant to ban internal right-recursion from prenominal adjective phrases and itself perhaps to be ultimately derived now from some version of Biberauer et al.’s (2010) FoFC. Technically, a linear Consistency Principle, though descriptively correct (also cf. Section 2.6.1) could be insufficient, for, it does not forbid prepositional realization for Genitive in postnominal functional positions, say in GenO. However, prepositional realization seems intrinsically impossible for a functional Genitive anyway, for, as a matter of fact, so far no definiteness inheritance (defining property (16c) for non-free Genitive) has been discovered with clearly prepositional Genitive.

2.3.2 Uniformity Whatever the ultimately correct axiomatization of the Consistency restriction, we must stress that it need not be the case in all languages that free Genitive is formally distinct from the functional one: for example, synthetic morphology may in principle satisfy conditions on either type of Genitive. A plausible corollary, explored in Gianollo (2005), is that a language like Latin, which has clearly free (by (16a and b)) Genitive realized in an inflectional, non-prepositional form, seems to use the same Genitive morphology prenominally in both the positions (pre-and post-adjectival) in which non-free Genitives are used in other

96

9781441124609_Ch07_Final_txt_print.indd 96

11/20/2012 9:25:32 AM

The Structure of NPs

languages providing a surface appearance of wide (though not unconstrained) freedom. This prompts the further hypothesis that if a language has a form of free Genitive which is not incompatible with occurring also in functional positions, it always does so, determining what we may call uniform Genitive: (19) Uniformity: if the form of free Genitive is compatible also with functional checking, then it is used across the board (i.e. also in all functional positions). This hypothesis seems supported by Classical Greek as well (Guardiano 2011), and, pending wider philological investigation, might tentatively be taken to describe a peculiar feature of ancient Indo-European noun phrases (apparent freedom of multiple Genitive occurrence both to the left and right of nouns and attributive adjectives).

2.4 Conditions on Functional Genitive 2.4.1 The Licensing of Functional Genitive Descriptively, there seem to exist three common ways to license a functional Genitive: (20) (a) in postnominal position (b) in prenominal position with Gen-N agreement (c) in prenominal position with determiner-like function We examine them in turn. Postnominal position of functional Genitives is always the result of noun- (or noun projection-) raising over GenO or GenS. Of course, the most frequent and uncontroversial cases are those in GenO, that is, in postadjectival position8. The second cross-linguistically common way of licensing a functional Genitive is through φ-feature concord. Longobardi (1996) noted a tendential complementarity between the forms of prenominal and postnominal Genitives: the former must often agree in φ-features with the head noun, the latter can dispense with this requirement, a phenomenon with some clausal parallelisms in the relation between pre-/post-verbal subjects and verbs, already noted by Greenberg (1963b, Universal 33). This is particularly clear in the case of visible languageinternal alternations, as shown, for example, by Somali in (21) (Lecarme 1994) or Catalan in (22) (with the special raising noun casa, Longobardi 1996):

97

9781441124609_Ch07_Final_txt_print.indd 97

11/20/2012 9:25:32 AM

The Bloomsbury Companion to Syntax

(21) (a) Hormúud-ka kulliyád-da Dean-the faculty-the (b) Kulliyád-da  hormúud-k-eeda Faculty-the Dean-the- oss.3f.sg “the Dean of the faculty” (22) (a) La meva/*meu casa The my(-f)/my(-ø) home (b) Casa (/Ca’) meva/meu Home my(-f)/my(-ø) “My house“ Thus, some prenominal Genitives are definitely licensed just under feature copying with the head (essentially “free riding” of Case on φ-features, if we adopt Chomsky’s 1995 terminology9); this comes in the two logically possible varieties: copying from the head onto the argument and in the reverse direction (argument features copied on the noun). The first variety of agreement is manifested in at least three subcases: (23) (a) restricted to possessive personal pronouns, as in many IE languages; (b) extended to adjectives derived by many nouns and displaying full argument properties (e.g. also binding anaphors etc.), as in most Slavic languages; (c) used with all free Genitives and morphologically exposed on the adpositional marker of the latter (e.g. in Hindi, some Arabic varieties). The last subcase does not concern the licensing of functional Genitives, of course, but rather a free one, while the first two do and seem to occur precisely in GenS: GenO is more easily raised over by N or N-projection, of course, therefore agreement is often not necessary as a licensing device, and actually is not normally observed. It is an interesting question whether this is so on general and principled grounds or agreement may sometimes also concern GenO. The second variety of agreement is well exemplified by the Somali example above or by Uralic languages; Hungarian displays such an agreement, variable for person, in (24): (24) te  vendég-e-d -e-d (a) A  te the  you  guest-poss-2sg “your guest” 98

9781441124609_Ch07_Final_txt_print.indd 98

11/20/2012 9:25:32 AM

The Structure of NPs

ter könyv-e v-e (b) A Péter the Péter book-poss-ø(=3sg) “Péter’s book” Finally, the third salient case of functional Genitive is that of the Germanic languages, which is licensed without any visible agreement with the noun—a fact tentatively attributed in Longobardi (1996) to covert noun (or, perhaps, noun projection) movement, which would restore the required N-Gen order perhaps producing for English (25a) an LF representation (25b) that would be a phonetically invisible counterpart (and literal translation) of Hebrew construct state (25c): (25) (a) The teacher’s new book about syntax. (b) Book the teacher’s new t about syntax. (c) Sefer ha-more he-xadaš t ‘al tabxir. The idea is motivated by the fact that there is an important semantic correlate of this construction that indeed recalls construct state, that is, the determiner-like function of the Genitive itself: Germanic non-agreeing prenominal Genitives appear to transmit count and definiteness properties to the whole DP (Crisma 1999, 2011). Longobardi (1996) has tried to set up a theory of covert N movement in such a way as to account for this correlating property as well. Whatever the correct theoretical account, the determiner-like function of such Genitives seems a necessary ingredient of their licensing.10 Before concluding we must notice that all the three licensing conditions above may probably be dispensed with in languages with uniform inflectional Genitives (cf. Section 2.3.2); this is suggested by the fact that Ancient Greek displays some apparent instances of prenominal/preadjectival non-agreeing Genitives without any interaction with the definiteness interpretation of the whole DP (Guardiano 2011).

2.4.2 The Form of Functional Genitive A second issue is whether a Genitive construction must always be signaled formally or can be realized by no morpheme at all, as is the case, for example, for classic structural Cases such as Accusative in languages like English and many others (Chomsky 1986). Prepositional, postpositional, or affixal/inflectional Genitives all formally manifest Genitive in the phonological form, of course, and seem to display some degree of mutual historical substitutability (e.g. loss of Latin Case-inflection gave way to Romance prepositional Genitive). Genitives occurring in the category we defined “free” appear to all fall into these classes, as noticed. Among functionally checked Genitives, there are instances of formally 99

9781441124609_Ch07_Final_txt_print.indd 99

11/20/2012 9:25:32 AM

The Bloomsbury Companion to Syntax

unmarked or very limitedly marked phrases, however: for example, Hebrew construct Genitive, Romance (ancient or non-standard) juxtaposition Genitives, as witnessed, for instance, by Old French (Delfitto and Paradisi 2009: 296): (26) la niece le duc the niece the duke “the duke’s niece” The Hebrew case could be understood as an instance of marking on the head noun (cf. Section 2.6.2) rather than on the argument, speculating on the idea that the morphophonological modifications of the former (its construct state) is the reflex of some suprasegmental morpheme signaling “possessee” condition (Ritter 1988). The Romance cases, definitely a GenO occurrence, when occurring in varieties that have no overt residue of morphological Case contrasts might reduce to a lexically restricted (wrt to the type of possible head nouns) construction, licensed and, in a sense, marked precisely by the nature of the noun governing it (Silvestri forthcoming).

2.4.3 GenS and GenO In this line, a typologically plausible guess is that GenS is always (i.e. in all languages and constructions, without any specific parametrization) used to check Genitive Case, whenever one of the conditions allowing it holds: agreement (in either direction), overt N-Gen order, or covert N-Gen order. Instead, for GenO some parametrization about the activation of its Case-checking capabilities must probably be stipulated: for, there seems to be no obvious way to predict, for example, that GenO is nowadays used in German (cf. (18)), though not in English or Romance.

2.5 A Theory of Genitive A preliminary plausible construal of the whole pattern so far observed is that [+genitive] is a formal feature of DPs: normally it must correspond to at most one (cf. Uniqueness in Section 2.2.2) salient phonological translation into the Articulatory–Perceptual system (e.g. some overt inflection or a specified preposition that lexically means “Genitive”). This form of Genitive will be largely free in syntactic distribution, depending however on a suitable thematic interpretation. If there is no such a phonological translation in the language, or the translation is incompatible with some positions (cf. Section 2.3.1 for remarks on the distribution of prepositional forms), then the feature [+genitive] will be technically inexpressible and will have to be deleted in the course of the derivation. In these cases, [+genitive] may only be checked and eliminated in designated syntactic positions of the extended functional structure of the noun (GenS and 100

9781441124609_Ch07_Final_txt_print.indd 100

11/20/2012 9:25:32 AM

The Structure of NPs

GenO). Ideally, elimination of the feature will result in zero realization, but also commonly in some reduced form defined by the syntactic context or morphological constraints (cf. some discussion in Section 2.4.2); anyway, prepositional Genitive is structurally “too complex” to ever count as checked, while some inflections and postpositional affixes seem more acceptable as the realization of checked and deleted [+genitive] in functional positions, at least in languages that are otherwise basically prepositional. As noticed in Section 2.4.1, the checking of the feature [+genitive] must take place not only in designated positions but also under specific syntactic conditions, informally described as follows (cf. Longobardi’s 1996 Case Checking Principle for a similar proposal): (27) Checking: A Case feature on a category α is checked by a designated head γ iff (a) α is a member of the internal domain of a CH headed by γ or (b) α shares φ-features with γ.

2.6 Some Further Questions 2.6.1 Postpositional Genitives We have already seen that in certain languages (English and Scandinavian) postpositional phrasal markers are used on prenominal functionally checked Genitives, while free Genitives occur in prepositional form (the man’s pictures/ pictures of the man). There are definitely languages in which postpositional phrases instantiate apparently free Genitives, for example, Basque in Europe (Etxepare 2003; Trask 2003; Artiagoitia 2010) and typically Indo-Aryan among IE languages: (28) Basque en kotxea hautsi du. (a) Mikelek Jonen Mikel. rg Jon.gen car break aux.tr “Mikel broke Jon’s car.” ren zezenaren ren hiru irudiak (b) Picassoren Picasso.gen bull.art.gen three picture.art “Picasso’s three pictures of the bull”

(Etxepare 2003: 419)

(Artiagoitia 2010: 3)

(29) Hindi: (a) John kaa durghaṭana kaa ek vyavran John-of accident-of one description “a description of the accident by John”

101

9781441124609_Ch07_Final_txt_print.indd 101

11/20/2012 9:25:32 AM

The Bloomsbury Companion to Syntax

Postpositional Genitives seem always to occur in a very high structural position (e.g. in Indo-Aryan they precede demonstratives and indefinite determiners) and in particular always before the head noun. Therefore, with respect to such Genitives a wider bidirectional generalization appears to hold: (30) Generalized Consistency: Prepositional Genitives always surface after the head noun, postpositional ones always before it. It is an interesting, though non-trivial, research program to investigate if even this generalized version of directional Consistency may derive from a wider phrase-structure principle of the type of FoFC. A possible hint in this direction is provided by the fact that Basque also appears to display some corresponding evidence for a generalization of the Consistency constraint with respect to attributive APs, which essentially surface only in postnominal position and cannot contain pre-head complements (Trask 2003). These observations could even suggest, in a broadly Kaynian spirit (Kayne 2002), that adpositional free Genitive might be universally fronted to and licensed in a very high pre-DP position (connected to some appropriate thematic position) and obligatorily crossed over by the whole DP if and only if the language chooses prepositional realization. This approach however raises some problems with another intriguing tentative generalization presented in Section 3.3 below. Another important question about free postpositional Genitives, in need of further investigation, is whether they may alternate with others in functional positions, as is common with prepositional ones (cf. above). They do not seem to do so in the languages mentioned (Basque and Indo-Aryan) nor to observe the Uniformity condition above in the varieties in which in principle it could hold: for example, in Hindi the order “Gen Dem/Numeral AP N” is normal, “Dem/Numeral AP Gen N” is ungrammatical. As noticed most forcefully by Hawkins (1984), postpositional Genitives seem to correlate in one direction with prenominal relative clauses being implied by the existence of the latter (cf. Dryer 1998: 302–3). It is instead as yet unclear and worth investigating to which extent they correlate with prenominal occurrence (and postpositional realization) of other, indirect (i.e. non-Genitive) complements of the noun, that is, if it is meaningful to talk of any uniform parametrization governing the “head-complement/complement-head” status of nominal phrases.

2.6.2 Head Marking As hinted at above, the adnominal relation is often morphologically marked on the Genitive argument, but sometimes on the head noun itself (cf. Nichols

102

9781441124609_Ch07_Final_txt_print.indd 102

11/20/2012 9:25:32 AM

The Structure of NPs

1992 and references cited) or, more plausibly in theoretical terms, on a functional morpheme incorporated by the noun. To some extent the two strategies may cooccur, for instance a Genitive affix in Arabic cooccurs with the morphological modifications on construct state nouns. A subtler typological question is whether the head-marking argument can be in free Genitive, in GenO, or just in GenS position. The distinction between head marking and argument marking partly crosscuts with issues about agreement. As noted in Section 2.4.1, agreement in φ-features may go in two directions. Agreement from Gen to N is attested in head-marking languages like Hungarian, and actually the agreement marker on the noun is itself a form of head marking anyway; agreement from N to Gen in non-head-marking languages is especially widespread with personal pronouns (so-called possessives) and occurs in other cases as well, as we have seen; it is less clear if it may cooccur with φ-neutral head marking.

2.6.3 Genitive and Definite Suffixes Some languages with definite markers suffixed to the head noun, such as Rumanian (Grosu 1988, Cornilescu 1995), Bulgarian, Norwegian display some non-prepositional (i.e. apparently functionally checked) Genitive in the position right-adjacent to the N+suffix complex, in a form typically reserved to personal pronouns (so-called possessives) in Norwegian and Bulgarian, but open to basically all Genitive DPs in Rumanian, where the latter are more distinctly inflected. In Rumanian the marker al/a (and corresponding plural forms), doubling φ-features of the nominal head under certain circumstances, appears to perform the same function in a lower, apparently post-adjectival, position: (31) (a) cartea lui novă/cartea studentului novă book-the his new/book-the the student’s new “his/the student’s new book” (b) o carte novă a  lui/o a book(FemSg)  new a(FemSg)  his/a carte  novă a  studentului book(FemSg) new a(FemSg) the student’s(MaSg) A natural interesting question arises here: namely whether such two contexts of Genitive checking in Rumanian with the two more general checking ones (GenS and GenO) identified above. Another general issue, of course, concerns the possible licensing role of definiteness in many of these processes.

103

9781441124609_Ch07_Final_txt_print.indd 103

11/20/2012 9:25:32 AM

The Bloomsbury Companion to Syntax

2.7 Conclusions As a result and a summary of some of these observations, the previous scheme (9) could be embedded in the more complex structure (32): (32) [1 GenS 2 AP* 3 GenO 4 [a P [S [O . . . N . . . ]]a]] In (32) the numbered positions 1 through 4 set out some cross-linguistically possible surface positions for the noun, GenS and GenO the high and low functional positions for possessivized Genitive, respectively, and AP* a potentially iterated base position for attributive APs.11

3 Syntactic Realization of Arguments 3.1 Order and Number of Arguments We have already hinted at some generalizations concerning the linear order of arguments of nouns relative to the head noun itself and to attributive adjectives. As for the ordering of arguments with respect to each other, there seem to hardly exist very sharp generalizations: in languages where free Genitives can be rather productively stacked, like Romance, the structural P>S>O hierarchy argued for above is only vaguely and reversely encoded in the linear order of postnominal free Genitives in precisely the controversial preference often given to P under a flat intonation/information structure, and as occurring as the last prepositional genitive in a cluster of two or three. Even less sharp seems the preference for NOS over NSO, the latter order being instead probably more neutral when the complement is an indirect object PP, rather than a Genitive PP. A most important parametric generalization on the argument structure of nouns concerns, instead, the number of external positions for arguments. An observational difference between English and Romance or German is that only one of P and S is overtly expressible in English, while in the other varieties both may occur simultaneously. In other words, it seems that only one external position is syntactically available for a genitive phrase in English nominals, (at least) two in German and Romance (this type represented through Italian examples in what follows): (33) (a) *Mary’s book of my favorite novelist (b) il mio libro del mio romanziere preferito “my book of my favorite novelist” 104

9781441124609_Ch07_Final_txt_print.indd 104

11/20/2012 9:25:32 AM

The Structure of NPs

If possessivization of O is actually movement through a syntactic external argument position, then a whole typological cluster of other differential properties can be parametrically tied to the previous observation about the number of external positions (cf. Giorgi and Longobardi 1991). The properties in question fall at least into three domains: (34) (a) Constraints on unaffecting nouns (b) Subject control phenomena (c) Subject binding phenomena It was noticed above (Section 1.2) how in English, according to now classical literature, the understood subject of a noun can be syntactically active, but only if no possessivization of the O takes place.The same results obtained for control can be replicated in English with binding of an anaphor by an understood subject of a noun. Instead, in all the three processes of (34), a phonologically null but syntactically active subject position remains available in Romance and in German, irrespectively of whether the O is possessivized or not: (35) (a) l’affondamento della nave per riscuotere l’assicurazione “the sinking of the ship to collect the insurance” (b) il suo affondamento per riscuotere l’assicurazione “its sinking to collect the insurance” (36) (a) la sperimentazione di tale farmaco su se stessi “the testing of such a drug on oneself” (b) la sua sperimentazione su se stessi 12 “its testing on oneself” (37) (a) la percezione/conoscenza del problema “the perception/knowledge of the problem” (b) (a proposito del problema) la sua percezione/conoscenza “(speaking of the problem) its perception/knowledge” The facts seem to be interpretable as follows. The existence of the contrasts of O and analogous cases in English provides straightforward evidence for two assumptions: 105

9781441124609_Ch07_Final_txt_print.indd 105

11/20/2012 9:25:33 AM

The Bloomsbury Companion to Syntax

(38) binding and control are syntactic phenomena, that is, their antecedency requirements cannot be satisfied just by a syntactically understood but undischarged theta-role in the semantic structure of a lexical head, but demand a structurally realized, though possibly empty, phrase; (39) in English an empty S competes with a possessivized O for the same syntactic slot. On the whole, these facts confirm the English close parallelism between nominals with possessivized objects and clauses with middle and unaccusative verbs, outlined before (Section 1.2). As noted, in German and Romance the same facts do not hold: at first sight this might speak for the non-universality of the intriguing conclusion in (38); but we have also seen that the difference can be due to the independently attested availability of more than one external argument position in nominals, that is, to the fact that Romance and German parametrically falsify (39) rather than (38). In Section 3.2 we will see that Romance provides precisely further and more intricate evidence for the interplay of movement and empty categories, and for the universality of (38).

3.2 A-Positions and the Evidence for Empty Categories It is widely assumed and well supported in clausal syntax that binding and control take place from so-called A-positions, that is, basically thematic positions and derived subject positions. Thus, left-peripheral positions hosting wh-, topic, and focus phrases do not serve as valid antecedent sites for binding and control. For example, raising the supposed antecedent to an A-position in (40b) rescues the binding violation due to lack of c-command in (40a): (40) (a) *It seems to each other’s friends that those girls are smart. (b) Those girls seem to each other’s friends _ to be smart. But a similar violation is not rescued by moving the antecedent to a non-Aposition below: (41) (a) *I think each other’s friends like those girls. (b) *I think those girls each other’s friends like _. (c) *Which of those girls do you think each other’s friends like _? Within NPs, at least thematic subjects and objects of nouns seem to be basically associated with A-positions, since they are able to regularly bind and control; 106

9781441124609_Ch07_Final_txt_print.indd 106

11/20/2012 9:25:33 AM

The Structure of NPs

for sentence subjects, see Stowell, this volume. Now, it has been argued above that Romance nominals provide more than one external position, allowing cooccurrence of a raised (possessivized) O with a PRO subject. One may wonder whether all such positions qualify as A- or just those of thematic subjects do, with the others to be assimilated more to left-peripheral clausal positions. Three types of considerations prove the latter answer to be correct. Suppose, first that O is in a configuration where it cannot act as a controller from its base position, say for lack of structural prominence (c-command), but could from a higher (subject) A-position. This is exactly the case in active/passive clausal structures with control into an adverbial infinitival sentence: (42) (a) Giannii fu condannato dopo PROi aver subito un regolare processo. “Gianni was convicted after facing a regular trial.” (b) *Hanno condannato Giannii dopo PROi aver subito un regolare processo. “They convicted Gianni after facing a regular trial.” A roughly analogous contrast (albeit sometimes slightly less sharp) can be found within nominals: (43) (a) La suai condanna dopo PROi aver subito un processo irregolare rimarrà un’infamia. “His conviction after facing an irregular trial will remain a shame.” (b) ?* La loro condanna di Giannii dopo PROi aver subito un processo irregolare rimarrà un’infamia. “Their conviction of Gianni after facing an irregular trial will remain a shame.” In (43b), in order to acquire proper controller status, O must be in (or rather have passed through) a higher A-position, presumably that of S, iff no other high position qualifies as A-. Therefore, in these situations a subject empty category should be forbidden in Romance as well and, consequently, no binding ability on the part of an understood S should remain available. Patterns like the following (in particular the ungrammaticality of (45c)) confirm this point (cf. Giorgi and Longobardi 1991 for discussion): (44) (a) ?* La loro sperimentazione del nuovo farmaco su Maria senza PRO essere stato prima approvato è stato uno scandalo. “Their testing of the new drug on Maria without being previously approved was a scandal.” 107

9781441124609_Ch07_Final_txt_print.indd 107

11/20/2012 9:25:33 AM

The Bloomsbury Companion to Syntax

(b) A proposito del nuovo farmaco, la sua sperimentazione su Maria senza PRO essere stato prima approvato è stato uno scandalo. “Speaking of the new drug, its testing on Maria without being previously approved was a scandal.” (45) (a) La sperimentazione del nuovo farmaco su se stessa è stato il più coraggioso atto di Maria. “The testing of the new drug on herself was Maria’s most courageous deed.” (b) A proposito del nuovo farmaco, la sua sperimentazione su se stessa è stato il più coraggioso atto di Maria. “Speaking of the new drug, its testing on herself was Maria’s most courageous deed.” (c) *A proposito del nuovo farmaco, la sua sperimentazione su se stessa senza PRO essere stato prima approvato è stato il più coraggioso atto di Maria. “Speaking of the new drug, its testing on herself without being previously approved was Maria’s most courageous deed.” In fact (44) shows that O must raise to a higher (A-)position in order to control PRO, (45) that this process interferes with the otherwise possible binding of an anaphor by the understood subject. Second, if O, in certain configurations like (46a), is able to control from its base position, when possessivized it will not need to raise through an A-position and a null syntactic subject will be available; indeed the latter may show its presence by itself performing the function of controller: (46) (a) La condanna di Gianni a PRO scontare tre anni di carcere senza PRO avergli dato la possibilità di difendersi mi ha scandalizzato. “The conviction of Gianni to serve three years in prison without giving him a chance to defend himself scandalized me.” (b) La sua condanna a PRO scontare tre anni di carcere senza PRO avergli dato la possibilità di difendersi mi ha scandalizzato. “His conviction to serve three years in prison without giving him a chance to defend himself scandalized me.” Here O controls the PRO subject of scontare “serve” and the understood arbitrary S of condanna “conviction” may control the other PRO subject of the adverbial without-clause. Hence both S and O are syntactically active. Finally, recall that “unaffecting” head nouns were argued, as before, never to dispense with a syntactically realized S role (cf. Section 1.2); therefore, if the 108

9781441124609_Ch07_Final_txt_print.indd 108

11/20/2012 9:25:33 AM

The Structure of NPs

only high A-position is reserved precisely for thematic subjects whenever realized, that is, even in the form of an empty category, with unaffecting nouns the object should never improve its control capabilities through possessivization. This prediction is also borne out: (47) (a) *La conoscenza dell’algebra senza essere studiata bene è una cosa rara. “Knowledge of algebra without being studied well is a rare thing.” (b) *La sua conoscenza senza essere studiata bene è una cosa rara. “Its knowledge without being studied well is a rare thing.” Thus, the data suggest that even in languages admitting more than one external position, like, for example, Romance languages and German as parametrically opposed to English and Scandinavian (cf. Section 3.1), only one of them counts as an A-position. The others should then count as non-A, like the left-peripheral ones involved, for example, in (41b and c). More than that, the pattern shows that even the Romance languages provide evidence for syntactically active though phonologically null subjects of N and for the potential universality of (38).

3.3 Raising to n There is another tempting generalization about the parameter governing the licensing of one or multiple external argument positions: consider that, in parallel to the v-V structure postulated to provide a base external argument position for agentive subjects, it is plausible that nominals display a corresponding n-N structure (Radford 2004, 367ff ), with the Specifier position of nP functioning as an A position for external arguments. Being affixal, the light noun n° should always be licensed through raising of N. Under this approach, having multiple external argument positions would mean for a language to be able to split n° into many sub-heads and recursively embed nP within nP (only upper-bounded by available R-related interpretations), with all instances of n° ultimately licensed by successive-cyclic N-raising: languages with just one (non-recursive) n° could perhaps be understood as languages in which n° can only probe the simple stem N, not the N+n complex. A consequence of this analysis would be that, if arguments of nouns may sometimes surface in their base position, languages with split (recursive) n° (like Romance and German) should be able to freely exhibit arguments interpreted as external (Ss and Ps, in the previous terminology, cf. Section 1.1) to the right of the spelled-out noun, while this should normally be impossible in languages licensing just one n° (like English); at least in some simple cases, 109

9781441124609_Ch07_Final_txt_print.indd 109

11/20/2012 9:25:33 AM

The Bloomsbury Companion to Syntax

the prediction is apparently borne out, with English limiting itself to an internal argument interpretation of an of-phrase, and Romance displaying full ambiguity: (48) (a) I bought the picture of Mary. (b) Ho comprato la fotografia di Maria. “I bought the picture of Maria.” “I bought Maria’s picture of something/someone else.” This is so, of course, unless other processes intervene (rightward shift, sensitive to heaviness/inanimacy, or agentivization through a by-phrase), but at least constitutes some initial evidence for this suggestive hypothesis.

3.4 Poverty of Stimulus The cross-linguistic data discussed in Section 3.2 are important because they provide some of the strongest evidence ever for empty pronominal categories and one of the clearest poverty-of-stimulus arguments in favor of a nativist view of certain aspects of syntactic knowledge. The main conclusions suggested by the patterns of Sections 3.1 and 3.2 can thus be so summarized: (49) (a) Control (and binding) in English do not purely depend on the lexical (thematic) meaning of head nouns but crucially on strictly syntactic factors (movement of Object to a prenominal A-position); (b) the same is true in Romance, though in a different and subtler way; (c) the two generalizations can be reconciled by supposing the existence of phonologically empty categories and, of course, of the distinction between A- and non-A-positions, combined with an independently observable difference in the possible number of external arguments. Now, a theory of an innate UG defining a notion of empty category and A-position and a parameter about the number of external arguments of nouns, combined with exposure to simple primary data (such as those in (48)), suffices to predict basically all the facts of Sections 3.1 and 3.2. Consider, instead, how an organism supposedly NOT equipped with such a theory should converge on the correct grammars: on the analogy of a wellformed paradigm in English like 110

9781441124609_Ch07_Final_txt_print.indd 110

11/20/2012 9:25:33 AM

The Structure of NPs

(50) (a) The sinking of the ship to collect the insurance (b) The ship’s sinking (c) The ship’s sinking by the owner to collect the insurance it would most naturally but incorrectly predict the acceptability of (51) *The ship’s sinking to collect the insurance unless exposed to explicit contrary instruction. The correspondent of (51) is indeed grammatical in Romance (cf. (35)); but hardly as a result of the simple inductive reasoning above, because the latter would certainly fail to predict the complex paradigms of (44)–(47). In conclusion, if no such evidence as hinting at the ungrammaticality of (51) and of the unacceptable examples of (44)–(47) is shown to be potentially available in the primary corpora of all English or Italian speakers sharing the relevant patterns, any theory of acquisition should assume some variant of the innate UG sketched before. Now, it is widely agreed that direct negative evidence (explicit correction) plays no relevant role in acquisition: then, the only way to learn the patterns in question from the stimulus is indirect negative evidence, that is, for the learner to find a significant number of the grammatical instances of the paradigm vs., essentially, none of the ungrammatical ones. But here is the crucial significance of the nominal patterns in question: for, even the corresponding positive evidence, that is, the grammatical examples of the Italian paradigms (44)–(47) seem very unlikely to occur outside the contexts of grammarians’ discussions. Thus, all these facts strongly reinforce the argument for syntactically realized empty positions and their role in coreference phenomena, and ultimately for their unlearnt and universal nature. We consider such empty categories in more detail directly.

3.5 A Synopsis of Empty Pronominals 3.5.1 Null Subjects Thus, the argument structure of nouns may include empty pronominal categories. This conclusion is independently supported by other evidence, pointing to the existence of a PRO-like category as subject of nouns. Such evidence is presented again on the basis of Italian and English data. It falls into two categories: (52) (a) evidence that some sort of PRO may occur (b) evidence that some sort of PRO must occur, with certain nouns 111

9781441124609_Ch07_Final_txt_print.indd 111

11/20/2012 9:25:33 AM

The Bloomsbury Companion to Syntax

3.5.2 Optional PRO Type (51a) evidence, in turn, comes in three subtypes, illustrated below: (53) (a) evidence from binding (b) evidence from arbitrary interpretation (c) evidence from construct state Type (53a): first, in many different languages, exemplified here by English and Italian, there are cases of binding of an anaphor embedded within an NP by a DP-external apparent antecedent that does not satisfy one or more of the conditions normally imposed on antecedents of anaphors: structural prominence (c-command), uniqueness (non-split nature), locality, or subjecthood (where the latter applies). In all such cases it turns out that the phonetically unrealized subject argument role of the noun is understood as coreferential with the anaphor/ antecedent: (54) La descrizione di se stessa inviata a quella ditta è stata di grande giovamento alla carriera di Anna. “The description of herself submitted to that firm was very helpful for Anna’s career.” Furthermore, the environments in which this type of situations arises are exactly those in which infinitives with controlled PRO subjects could grammatically replace the head noun in question: (55) Descrivere se stessa in quel modo è stato di grande giovamento alla carriera di Anna. “Describing herself that way was very helpful for Anna’s career.” Type (53b): some DP-internal anaphors have arbitrary reference without depending on any overt arbitrary binder: again, this only arises when they occur in complement position and are read as bound by the understood subject role of the noun. Thus, some equivalent of PRO could be the primitive source for arbitrariness and occur as subject of N: (56) Una buona conoscenza di se stessi è cosa rara. “A good knowledge of oneself is something rare.” Type (53c): the third subtype of evidence is of a slightly different nature: in the so-called restricted Romance construct state construction with a head noun

112

9781441124609_Ch07_Final_txt_print.indd 112

11/20/2012 9:25:33 AM

The Structure of NPs

like casa “home” (cf. Longobardi 1996), a genitive argument is obligatorily realized non-prepositionally and adjacent to the head noun, giving rise to surface N+DP+(AP) order: (57) (a) Casa Rossi nuova Home Rossi new “Rossi’s new home.” (b) *Casa nuova Rossi Home new Rossi however in some cases a N+(AP)+P(=di)+DP sequence appears: (58) Casa nuova di Rossi Home new of Rossi Both apparent irregularities are regularized if the latter sequence is analyzed as actually constituted of N+PRO+(AP)+P+DP, with PRO linked in a chain to the lower genitive PP (=P+DP) and satisfying the condition on adjacency and nonprepositional realization: (59) Casa PRO nuova di Rossi This analysis is strongly suggested by the fact that the Semitic (e.g. Hebrew, examples (60) and (61)) construct state construction, which displays significant analogies with the Romance one in question, also exhibits the double pattern of (56)–(57), but precisely with an overt pronoun in the position of the Romance supposed PRO (Ritter 1991; Siloni 1994): (60) (a) Beyt Dan ha-gadol House Dan the-big (b) *Beyt  ha-gadol  Dan House the-big Dan “Dan’s big house.” (61) (a) *Beyt ha-gadol šel Dan House the-big of Dan ha-gadol  šel Dan (b) Beyt-oo House-his the-big of  Dan “Dan’s big house.” 113

9781441124609_Ch07_Final_txt_print.indd 113

11/20/2012 9:25:33 AM

The Bloomsbury Companion to Syntax

The visible/null pronominal contrast here between Semitic and Romance would then be nothing other than the analogous one according to which the Genitive of a noun like Romance casa can be often null (and understood as arbitrary, syntactically or pragmatically controlled), while Semitic construct state always requires the phonological spell-out of the Genitive of a construct noun: the contrast was tentatively attributed in Longobardi (1996) to independent differences in the licensing function of adnominal construct Genitive (also following Borer 1994, in Semitic it seems required to transmit a definiteness value to the whole DP, and PRO could be formally deprived of such a feature). In subcase (53c), then, there is some reason to structurally locate the Romance empty pronominal argument in a more precise place, that is, a syntactically high position close to D, to which the noun has been apparently raised.

3.5.3 Obligatoriness of PRO Type (ii) evidence is of the same sort as already seen in (54)–(56) above, though it is obtained replacing the anaphor by a pronoun or name: if with a certain nominal an empty subject must occur, it will be disjoint from the pronoun/name by virtue of binding principles B or C; this is exactly the case, for example, in the interpretation of (62) La conoscenza di lui/Gianni esibita in quell’occasione (ha molto giovato alla sua carriera). “The knowledge of him/Gianni exhibited on that occasion (was very helpful for his career).” Again, the facts parallel those holding with control infinitives/gerunds: (63) Conoscere lui/Gianni (ha molto giovato alla sua carriera). “Knowing him/Gianni (was very helpful for his career).” In neither (62) nor (63) can the understood subject be coreferential with the object pronoun/name, as made clearer by the impossible coreference of lui/ Gianni with an external controller of the subject position, such as sua, if the latter is added. This suggests that some PRO must syntactically represent it. Not all nouns behave this way, though: (64) (a) Il ritratto di lui/Gianni esposto al museo (ha molto giovato alla sua carriera). “The portrait of him/Gianni exhibited at the museum (was very helpful for his career).” (b) Ritrarre lui/Gianni (ha molto giovato alla sua carriera). “Portraying him/Gianni (was very helpful for his career).”

114

9781441124609_Ch07_Final_txt_print.indd 114

11/20/2012 9:25:33 AM

The Structure of NPs

Here no parallelism holds with the corresponding infinitive. Thus, with ritratto “portrait” coreference between lui/Gianni and the understood agent (author of the (self-)portrait) is not excluded. This suggests that the understood role is not obligatorily realized as an empty category, which would induce a binding violation, as is actually the case in (62b). Between the two classes is a third one, which in both English and Romance shares with the knowledge-class the obligatoriness of a syntactic subject and with the portrait-class the option of not assigning the external θ-role. This class is well exemplified by action nominalizations with “affected” objects, such as, for example, destruction: (65) (a) His/The president’s moral destruction (b) The moral destruction of him/the president was certainly not helpful for his career. In (65b), the understood agent of destruction is necessarily disjoint from the president. Thus, (65a) suggests that the subject position can be obliterated by the raised object, (65b), and that unless the object raises the understood subject must be syntactically represented. The contrast between (65b) and (64) leads to the statement of the following tentative generalization: (66) Event nominals require a syntactic external position (occupied by either S or raised O), object nominals do not. If correct, (66) draws the most salient syntactic boundary between the two much-debated types of nominals in question. It remains true that O may always be possessivized in nominals (except for unaffecting nouns), while it cannot always enter a middle construction in the corresponding verbal structure: (67) (a) The president’s moral destruction (b) *The president morally destroyed Such a difference could perhaps be imputed to Case theory, that is, to a wider optionality of free Genitive marking for nominal O as opposed to a lexically conditioned optionality of Accusative marking by verbs (middle formation).

115

9781441124609_Ch07_Final_txt_print.indd 115

11/20/2012 9:25:34 AM

The Bloomsbury Companion to Syntax

4 Conclusions To sum up, the argument structure of nominal phrases is governed by a number of probably universal principles, which are largely shared with clausal structures; among these, principles concerning: (68) (a) the structural hierarchy and obligatoriness/optionality of thematic arguments; (b) the existence of two distinct Case positions for non-adpositional arguments; (c) the access of the different arguments to such positions; (d) the licensing of empty categories. Some major domains of parametric variation in this area concern instead: (69) (a) the number of external argument positions allowed (one or more n°s); (b) the type of Genitive licensed (free or not and, if the former, uniform or not, pre- or post-positional); (c) the number of active functional positions for Genitive checking (active GenO or not); (d) the actual forms of Case realization (argument vs. head marking, agreement, its direction, and its targets). The relevant parameters appear at first sight rather unrelated to the form and settings of those found in the clausal domain.

Notes 1. We are indebted to C. Guardiano, S. Luraghi, and C. Parodi for helpul comments on a first draft of this chapter. 2. Of course, if a language happens to be able to license only one instance of Genitive Case, this will filter out the possibility predicted in principle by UTAH of a transitive head noun realizing both direct arguments; a more radical extension of this formal filter might be instantiated by Hungarian, in which not only is just one Genitive licensed but also indirect complements, which occur adpositionally marked with verbs and are not admitted by nouns. In such a case, apparently, nouns, unlike verbs, seem to inhibit all mechanisms of formal licensing of semantically selectable arguments, except for one Genitive. 3. C-command is originally defined by Reinhart (1976, 1981, 1983) as follows: “A node A c(onstituent)commands a node B iff the first branching node a that dominates A either dominates B, or is immediately dominated by a node a´ which dominates B, where a and a´ are of the same category type (e.g. S and S’)” (Reinhart 1983: 50).

116

9781441124609_Ch07_Final_txt_print.indd 116

11/20/2012 9:25:34 AM

The Structure of NPs 4. Cf. Giorgi and Longobardi (1991), with results confirmed by Siloni (1990) and Taraldsen (1990) among others. 5. By inflectional Case ending, we refer to one that spreads to all suitable (Caseagreeing) elements within the dependent DP (as, for example, in Latin): for example, noun, adjectives, determiners, etc. 6. The reverse must not be inferred from the above, that is, that inflectional Genitive is always free and cannot be merely functionally licensed: the inflected Genitive of Modern Greek and several Slavic languages appears to be normally non-iterable and in a presumably fixed postnominal position. It is inflectional Genitive occurring as free that can be argued, instead, to always extend also to functional positions (cf. Section 2.3.2). 7. If the latter restrictive formulation is indeed correct, one must not consider strictly “prepositional” the Genitives introduced by general markers such as e of Farsi or u in Wolof, which in fact are extensively used to introduce also other types of nominal modifiers, including adjective phrases. 8. Some cases of N-Gen-AP orders could also be, under certain conditions, the result of GenO (rather than GenS) being pied-piped under N-projection raising (cf. Section 2.3.2). 9. Heavily inflectional languages, such as, for example, Latin, Slavic, etc. show precisely that the realization of the genitival relationship under ˳-feature agreement includes actual Case concord with the head noun. 10. One may wonder if it is by chance that Icelandic, the modern Germanic language where an overt determiner is not required by singular count arguments and definiteness can be inherited also from postnominal Genitive, is also the only one without prenominal Genitive. 11. Obviously, this reconstruction is likely to be far from exhaustive: nothing is said, among other things, of a formally Genitive, pre-nominal but post-adjectival, position, reserved in present-day English to non-argument expressions and especially studied by Munn (1994), such as instantiated, for example, in “A fancy blue woman’s hat.” In earlier stages of English a superficially identical position could also host arguments, although of a highly restricted type, mostly designating God or divine persons. It is another interesting question if this non-argument Case position of Modern English has something in common with that identified by Rutkowski (2007), Rutkowski and Progovac (2005) for the special category of classifying adjectives of some Slavic and Baltic languages. 12. It seems thus possible in Romance for an O to raise over S provided that the latter is a null pronominal, apparently violating the possessivization hierarchy of Section 1.1. This may suggest that Chomsky’s (1995) equidistance principle must be relativized, perhaps in the sense of limiting it to overtly Case marked categories, thus excluding PRO from its scope.

117

9781441124609_Ch07_Final_txt_print.indd 117

11/20/2012 9:25:34 AM

8

Word Order1 Pamela Munro

Chapter Overview Introduction Basic Word Order Word Order Typology Word Order Variation Word Order Typology and Theoretical Syntactic Analysis Word Order Change

118 118 124 132 137 140

1 Introduction “Word order” refers to the order of the words in some phrase or construction, most o en a sentence. In this chapter, I introduce the concept of basic word order (Section 2), discuss the notion of typologizing languages according to their word order (Section 3), and survey aspects of word order variation (Section 4). Next, I consider some ways that word order and word order variation have been relevant for linguistic theory (Section 5); finally, I look briefly at how historical linguists deal with word order change (Section 6).

2 Basic Word Order Linguists, especially typologists, o en discuss “basic word order,” referring specifically to the position of the subject, object, and verb words or phrases in a language’s transitive sentences—in the least marked order used for affirmative active main-clause statements with nominal rather than pronominal arguments. 118

9781441124609_Ch08_Final_txt_print.indd 118

11/20/2012 9:26:34 AM

Word Order

2.1 What Counts as Basic? If more than one word order can be used in a simple statement, for example, the “basic” order is the one speakers would regard as neutral or unmarked, or that occurs most frequently in discourse. English normally uses the order Subject-Verb-Object—generally referred to by linguists as SVO—in sentences such as (1a and b): (1) (a) Mary ate pickles. (b) That girl has eaten some pickles. Both these sentences illustrate SVO order, but as (1b) shows, subject, object, and verb can be indicated by multi-word phrases as well as by single words. Although the basic word order of English is certainly SVO, simple sentences with OSV (Object-Subject-Verb) also occur: (2) These honored dead Newport mourns. (Inscription above a list of names on Newport, Rhode Island, city hall warmemorial) (3) —Who ate the pickles? Was it John? —John ate the olives. The pickles Mary ate. Although the topic fronting in the inscription in (2) and the last sentence in (3) are perfectly natural in context, no native speaker of English would consider these sentences neutral or unmarked. These examples raise an important point: even in languages with a consistent basic word order, speakers may use variant orders for pragmatic effect. Negative sentences in some languages may deviate from the basic word order: for example, in negative transitive sentences in the Bafut language of Cameroon, the object precedes rather than follows the verbs:2 (4) Bafut (Niger-Congo: Miestamo 2005: 155, citing Chumbow and Tamanji 1994) (a) bó lɨŋ kó mbà. they tdpst catch animal “They caught an animal (today).” (b) kāā bó lɨŋ wā’ā mbà kó. neg they tdpst neg animal catch “They did not catch an animal (today).”

119

9781441124609_Ch08_Final_txt_print.indd 119

11/20/2012 9:26:34 AM

The Bloomsbury Companion to Syntax

Passive sentences in languages such as English show that grammatical subject (in (5), pickles) and “doer” or agent (here, Mary) may not be identical. A sentence like (5) still has subject-initial word order, and can be used to talk about the same event reported in (1a); but the two sentences differ structurally, and their subjects are not the same (English passive sentences are not “basic”): (5) The pickles were eaten by Mary. Aspectual features, particularly active/stative, may also be relevant. For example, as Kramer (2009) shows, Middle Egyptian active sentences use a basic VSO order, but stative sentences, which are tenseless, use SVO order: (6) Middle Egyptian (Afroasiatic: Kramer 2009: 2–3) (a) iw rḫ.n ddi sḫr. part learn.pst Djedi plan “Djedi learned the plan.” (b) ddi rḫ.w sḫr. Djedi learn.stat.3sg.m plan “Djedi knows the plan.”/“Djedi knew the plan.” (I return to this case in Section 5.3) Word order may vary between main and subordinate clauses, for example in German (see also Sections 4.3 and 5.2): (7) German (a) Hans kennt die Frau. Hans knows the woman “Hans knows the woman.” (b) Ich weiß, dass Hans die Frau kennt. I know that Hans the woman knows “I know that Hans knows the woman.” Questions exhibit changes in word order in many languages: English yes–no questions, for example, begin with an (auxiliary) verb (8a), while wh questions (in English as in many other languages) begin with the question word (8b): (8) (a) Did Mary eat the pickles? (b) What did Mary eat?

120

9781441124609_Ch08_Final_txt_print.indd 120

11/20/2012 9:26:35 AM

Word Order

Pronouns o en appear in different positions from corresponding nouns. In Spanish, for example, an object noun typically follows the verb, as in (9a), while a clitic object pronoun precedes it, as in (9b):3 (9) Spanish (a) Juan comió la manzana. Juan ate the apple “Juan ate the apple.” (b) Juan la Juan bj.3sg.f “Juan ate it.”

comió. ate

Thus, basic word order can be naively defined as the least marked order used for affirmative active transitive main-clause statements with nominal rather than pronominal arguments.

2.2 Word Order, not Morpheme Order Discussions of word order virtually always refer to the order of independent words, not bound elements. Bound pronominal markers do not always appear in the same position as corresponding independent words. The Chickasaw language of Oklahoma, for example, has a basic SOV word order (discussed further in Section 4.1), as in (10): (10) Chickasaw (Muskogean) John-at ihoo-a ithána. John-nom woman-acc know “John knows the woman.” Most one-word Chickasaw sentences with non-third-person pronominal agreement affixes show the same subject-verb-object order of morphemes: (11) Chickasaw Is-sa-thána. 2sg.i-1sg.ii-know “You know me.” However, not all inflected Chickasaw verbs show this affix order, since the usual first-person singular transitive subject affix is a suffix. Thus, one-word sentences with an “I” subject, such as (12), have an object-verb-subject morpheme order:

121

9781441124609_Ch08_Final_txt_print.indd 121

11/20/2012 9:26:35 AM

The Bloomsbury Companion to Syntax

(12) Chickasaw Chi-thána-li. 2sg.ii-know-1sg.i “I know you.” There are many other such cases. Like Chickasaw, the Tolkapaya Yavapai language of Arizona has a basic word order SOV: (13) Tolkapaya Yavapai (Yuman) John-che vqí spó. John-nom woman know “John knows the woman.” But in one-word sentences, segmentable subject and object affixes appear in the order object-subject-verb: (14) Tolkapaya Yavapai Ny-m-spó. 2.obj-1.sbj-know “You know me.” This variability is not restricted to languages with basic SOV order. The Ineseño Chumash language of California has the basic word order VOS, as in (15)—but verbal affixes (on the first word here) appear in the order subject-verb-object: (15) Ineseño Chumash (Chumashan: Applegate 1972: 460) s-uleqpey-uswun ha-weselu ha-mɨy. 3.s bj-chase -3pl.obj art-calf art-wolf “The wolf chases the calves.”

2.3 Categorizing the Types of Basic Word Order Above we have seen examples of sentences with basic word orders (as so far defined) SOV (Chickasaw, Tolkapaya Yavapai), SVO (English, Spanish, German), VSO (Middle Egyptian), and VOS (Ineseño Chumash). The first two word orders, SOV and SVO, are the overwhelmingly most common cross-linguistically basic word orders. VSO is considerably less common, yet still occurs frequently. VOS is the next most common word order. The two object-initial orders, OSV and OVS, are extremely rare as basic word orders— yet these occur frequently as variant orders, o en with specialized discourse functions, as we saw with the English OSV examples in (2–3). 122

9781441124609_Ch08_Final_txt_print.indd 122

11/20/2012 9:26:35 AM

Word Order

The six basic word order categories just listed have been subcategorized in various ways. One possibility is to classify word order types based on the position of only the object or only the subject relative to the verb. Thus, based on the position of the object, we can speak of VO languages (those with word orders SVO, VOS, or VSO) vs. OV languages (SOV, OSV, OVS) (Lehmann 1973; Vennemann 1974). Alternatively, based on the position of the subject, we have SV languages (SVO, SOV, OSV) vs. VS languages (VSO, VOS, OVS) (Dryer 2008a). Sometimes only the position of the verb is considered: this gives verb-initial (VSO, VOS) vs. verb-final (SOV, OSV) vs. verb-medial (SVO, OVS). Since objectinitial orders are so uncommon, in practice verb-final almost always refers to SOV and verb-medial almost always refers to SVO. Languages can have a particular “basic word order” and still allow other orders of subject, object, and verb, most commonly for pragmatic/discourse or grammatical reasons (English topic fronting (2–3) serves a discourse function, while the Middle Egyptian variation between VSO and SVO seen in (6) results from a difference in aspectual features, and so on). Some languages, however, show so much variation that it is actually difficult to classify their word order (see Section 5.1).

2.4 Word Order in Intransitive Sentences The discussion of word order above refers only to transitive sentence order. It is generally true, in fact, that the position of the subject relative to the verb in an intransitive sentence will be the same as in a transitive sentence. Thus, Chickasaw has a basic transitive order SOV, as in (10), and a basic intransitive order SV, as in (16): (16) Chickasaw John-at taloowa-tok. John-nom sing-pst “John sang.” Similarly, the Garifuna language of Belize has a basic transitive order VSO and a basic intransitive order VS, as in (17): (17) Garifuna (Arawakan) (a) Hóu-ti Wán eat.b-t.3sg.m John “John ate bread.”

féin. bread

123

9781441124609_Ch08_Final_txt_print.indd 123

11/20/2012 9:26:35 AM

The Bloomsbury Companion to Syntax

(b) Erémuha-ti sing.b-t.3sg.m “John sang.”

Wán. John

However, there are languages in which the orders of transitive and intransitive subjects and the verb differ. For example, Dryer (2008a) notes that in the Indonesian language Muna intransitive subjects follow the verb, but transitive subjects precede it:4 (18) Muna (Austronesian: Dryer 1008a, citing van den Berg 1989b: 150, 163) (a) no-tende tora dahu. 3sg.realis-run again dog “The dog ran again.” (b) o katogha ne-mbolaku art crow 3sg.realis-steal “The crow stole dried fish.”

kenta fish

topa. dry

3 Word Order Typology The modern study of word order typology began with the work of Greenberg (e.g. 1963),5 who surveyed 30 disparate languages and proposed a provocative collection of 45 linguistic universals (1963: 110–13), 27 of which express correlations between word order (usually, basic word order as defined in Section 2) and other typological features.

3.1 Greenberg’s Findings Greenberg’s universals are stated implicationally. For example, the first one is “In declarative sentences with nominal subject and object, the dominant order is almost always one in which the subject precedes the object” (1963: 110): in other words, if a sentence has a nominal subject and object, the subject will precede the object, regardless of the position of the verb, which is true of the three most common basic orders, SOV, SVO, and VSO, though not of the next most common order, VOS. In addition to the basic word orders SOV, SVO, and VSO, Greenberg considered three additional phrasal features, involving the order of nouns relative to adpositions, genitives, and adjectives: preposition-noun (PN) vs. postpositionnoun (NP), noun-genitive (NG) vs. genitive-noun (GN), and noun-adjective (NA) vs. adjective-noun (AN). Greenberg’s study demonstrated a strong

124

9781441124609_Ch08_Final_txt_print.indd 124

11/20/2012 9:26:35 AM

Word Order

correlation between VSO order and the head-initial phrasal orders PN, NG, and NA (nouns are the heads of noun-adjective and noun-genitive phrases; prepositions are the heads of prepositional phrases) and between SOV order and the head-final phrasal orders NP, GN, and AN. If the verb is considered the head of the sentence, then these patterns are said to be “harmonic” (Hawkins 1979). SVO order is more difficult, though in many cases it seems to share more with the head-initial than the head-final type. Tlacolula Valley Zapotec (a Oaxacan language belonging to the Otomanguean stock), and Japanese are good illustrations of the prototypical VSO and SOV types.

3.2 Tlacolula Valley Zapotec: A Prototypical VSO Language Tlacolula Valley Zapotec (TVZ) has a basic VSO order in sentences like (19): (19) Tlacolula Valley Zapotec (Zapotecan) B-ìi’lly Jwaany li’ebr. pfv-read Juan book “Juan read the book.” TVZ has no case marking on nouns to indicate their sentential role. (We return to case marking in Section 4.5.) TVZ is prepositional: (20) Tlacolula Valley Zapotec B-ìi’lly Jwaany li’ebr làa’any pfv-read Juan book in “Juan read the book in the church.”

ydòòo’. church

Genitives follow the noun in TVZ (21a), as do adjectives in (21b). The genitive possessor in (21a) is introduced by the preposition x:tèe’n “of”; this still is considered an NG order. (21) Tlacolula Valley Zapotec (a) B-ìi’lly Jwaany li’ebr pfv-read Juan book “Juan read Pedro’s book.” (b) B-ìi’lly Jwaany li’ebr pfv-read Juan book “Juan read the new book.”

x:tèe’n of

Beed. Pedro

cweeby. new

125

9781441124609_Ch08_Final_txt_print.indd 125

11/20/2012 9:26:35 AM

The Bloomsbury Companion to Syntax

3.3 Japanese: A Prototypical SOV Language Japanese has a basic SOV order in sentences like (22): (22) Japanese Taroo ga hon o kat-ta. Taro nom book acc buy-pst “Taro bought a book.” Japanese is postpositional, with postpositions indicating grammatical relations as well as adpositional notions: (23) Japanese Taroo ga gakko de hon o Taro nom school at book acc “Taro bought a book at school.”

kat-ta. buy-pst

Genitives and adjectives precede Japanese nouns: (24) Japanese (a) Taroo ga Hanako no hon Taro nom Hanako gen book “Taro bought Hanako’s book.” (b) Taroo ga atarashii hon o Taro nom new book acc “Taro bought a new book.”

o acc

kat-ta. buy-pst

kat-ta. buy-pst

3.4 SVO Order: Italian and Norwegian In an SVO language, the verbal head is in the middle of the phrase, and typological relationships are not so clear-cut. This section presents two IndoEuropean SVO languages that were part of Greenberg’s original sample, Italian and Norwegian. Both illustrate common SVO patterns.6 Italian has basic SVO word order and prepositions, while genitives and adjectives follow the noun: (25) Italian (a) Giovanni legge il giornale. John reads the newspaper “John is reading the newspaper.”

126

9781441124609_Ch08_Final_txt_print.indd 126

11/20/2012 9:26:35 AM

Word Order

(b) Giovanni legge il giornale sulla John reads the newspaper on.the “John is reading the newspaper on the beach.”

spiaggia. beach

(c) Giovanni legge il giornale John reads the newspaper “John is reading Mary’s newspaper.”

Maria. Mary

di of

(d) Giovanni legge un giornale interessante. John reads a newspaper interesting “John is reading an interesting newspaper.” Except for the basic SVO sentence order, then, Italian follows a head-initial pattern, with PN, NA (noun-adjective), and NG order, just as with TVZ. Norwegian also has basic SVO word order and prepositions: (26) Norwegian (a) En gutt jaget katt-en. a boy chased cat-def “A boy chased the cat.” (b) En gutt jaget katt-en rundt hus-et. a boy chased cat-def around house-def “A boy chased the cat around the house.” But genitives and adjectives precede Norwegian nouns: (27) Norwegian (a) En gutt jaget jent-en-s a boy chased girl-def-gen “A boy chased the girl’s cat.”

katt. cat

(b) En gutt jaget den hvite katt-en. a boy chased the white cat-def “A boy chased the white cat.” Thus, Norwegian displays a more mixed SVO pattern: PN order is head-initial, but AN and GN order are head-final. (We return to Norwegian word order in Section 4.3.)

3.5 English: Another Mixed SVO Pattern Like Italian and Norwegian, English has basic SVO order and prepositions,7 as in the translations of the examples above. As in Norwegian, English adjectives

127

9781441124609_Ch08_Final_txt_print.indd 127

11/20/2012 9:26:35 AM

The Bloomsbury Companion to Syntax

regularly precede the noun (though see Section 4.4). English is unusual, however, in having two productive genitive patterns. The GN pattern is usually used for simple possessives, but many genitives (including some possessives) use the NG pattern (with the preposition of) (28a). This last pattern, however, is not possible with simple animate possessors, as (28b) shows. (28) (a) They looked at data from the world’s languages. / They looked at data from the languages of the world. (b) I borrowed John’s book. / *I borrowed the book of John. Greenberg’s typologies generally ignore intransitive patterns. But another difference among the three SVO languages profiled here is that while intransitive sentences are SV in Norwegian and English, Italian intransitive sentences can show VS as well as SV order.

3.6 Word Order and the Category “Adjective” In traditional grammar, the part of speech category “adjective” includes not only qualifying and descriptive adjectives like white, new, and interesting, but also numbers (and other quantifiers such as all and few), articles, demonstratives, and possessive determiners like my and his. Word order patterns provide evidence that these items should not be treated as members of the same lexical category in the grammar. For example, Spanish adjectives follow the noun, while Spanish quantifiers precede it (moreover, Spanish adjectives agree with the noun, but numbers higher than ‘one’ do not): (29) Spanish seis libro-s interesante-s six book-pl interesting-pl “six interesting books” A second example: possessive determiners appear in the opposite position from adjectives (or from genitive nouns) in Italian (30) and Norwegian (31): (30) Italian sua sorella his sister “his sister” (cf. (25d and c))

128

9781441124609_Ch08_Final_txt_print.indd 128

11/20/2012 9:26:35 AM

Word Order

(31) Norwegian katt-en min cat-def my “my cat” (cf. (27b and a))

3.7 Additional Word Order Correlations In addition to the relationships between head position and adposition, genitive, and adjective phrase orders examined above, Greenberg observed correlations with other lexical and morphosyntactic features, such as the following (1963: 79–87): (32) “With well more than chance frequency, when question particles or affixes are specified in position by reference to the sentence as a whole, if initial, such elements are found in prepositional languages, and, if final, in postpositional.” (Universal 9) (33) “When the descriptive adjective precedes the noun, the demonstrative and the numeral, with overwhelmingly more than chance frequency, do likewise.” (Universal 18) (34) “If a language has dominant order VSO in declarative sentences, it always puts interrogative words or phrases first in interrogative word questions; if it has dominant order SOV in declarative sentences, there is never such an invariant rule.” (Universal 12) (35) “In languages with dominant order VSO, an inflected auxiliary always precedes the main verb. In languages with dominant order SOV, an inflected auxiliary always follows the main verb.” (Universal 16) (36) “All languages with dominant VSO order have SVO as an alternative or as the only alternative basic order.” (Universal 6) Greenberg was aware that some of his universals were not, in face, exceptionless, as the caveats included in (32–3) above show. Finer-grained later studies such as those by Hawkins (1983), Dryer (1992), and the WALS project (Haspelmath et al., eds, 2008) have continued to consider Greenberg’s observations, producing an increasing number of further correlations among word order patterns. Such later work draws on a wider data sample than Greenberg used, facilitating the discovery of additional word-order-based relationships. However, the number of true “universals” without exceptions is extremely small. There are counterexamples, for example, to each of the five statements listed above. Lee (2006: 32–3) notes that the TVZ sentence-final question particle èee violates Greenberg’s (32), since the language has VSO word order (19). Cinque (2005: 316) reports a variety of exceptions to the statements concerning the 129

9781441124609_Ch08_Final_txt_print.indd 129

11/20/2012 9:26:35 AM

The Bloomsbury Companion to Syntax

order of nominal modifiers in (33). There are SOV languages with regular Wh-movement, such as Imbabura Quechua (Quechuan; Ecuador: for example, Cole 1982) and Old Japanese (Watanabe 2002): these are counterexamples to (34). The VSO language Garifuna provides counterexamples to both (35) and (36). Despite the claim in (35), inflected Garifuna auxiliaries always follow the main verb, as with the non-future transitive auxiliary umu, which agrees with both the (masculine) subject “John” and the (feminine) object “bread” in (37): (37) Garifuna Hóu l-umu-tu eat.b p3sg.m-transitive-t.3sg.f “John ate the bread.”

Wán John

féin. bread

Also, Garifuna has no alternative basic order (neither SVO nor any other) to the VSO order seen in (17a) (or (37)), despite the universal claim in (36). A Garifuna focused subject (or any other syntactic constituent) may appear before the verb, as in (38), but such sentences are far from basic, involving significant morphological change and the use of a different (wh-related) auxiliary, ba, which here agrees with the object. (In the translation, focus is indicated by underlining.) (38) Garifuna Wán éigi bo-u John eat.p ba-r.3sg.f “John ate the bread.”

féin. bread

These days, typologists less o en propose completely universal claims, preferring to state observed correlations based on word order as tendencies.

3.8 A Non-harmonic Word Order Pattern Greenberg noted, of course, that many languages—and not just those with verb-medial order—have word order patterns that do not show the “harmony” of having all phrasal heads aligned with the verb. Perhaps the most striking lack of correlation is between SOV basic order and the expected head-final AN order, with the adjective following the noun. In Dryer’s survey (2008b) of 488 OV languages (almost all with basic order SOV), well over half (287) had NA rather than AN order.8 For example, the

130

9781441124609_Ch08_Final_txt_print.indd 130

11/20/2012 9:26:35 AM

Word Order

Kunuz language of Egypt has SOV order (39a), but adjectives follow nouns, as in (39b): (39) Kunuz (Nubian: Abdel-Hafiz 1988: 201, 207) (a) sa:b ka:re-g kal-s-u. cat fish-acc eat-pst-3sg “The cat ate the fish.” (b) id adel man good “the good man” Some SOV/NA languages may not have a true category of adjectives at all (Munro 1985, 2006). In a number of languages with SOV order, words expressing adjectival ideas in fact are adjectival stative verbs, taking standard verbal inflection. For example, in Diegueño, a language of California whose basic word order is SOV (40a), the adjectival verb ʔəxan “to be good” can be used with inflectional morphemes such as the verbal second-person subject prefix m- (40b) or the verbal interrogative suffix -a (40c): (40) Diegueño (Yuman: (a) Couro and Langdon 1975: 56; (b–c) Langdon 1970: 141, 161) (a) Hellyaaw-ch hattepaa ewuuw. cottontail-nom coyote see “The cottontail saw the coyote.” (b) /m-ʔəxan/ > maʔxan. 2-be.good “You are good.” (c) ʔəxan-a. be.good-q “Is it good?” When adjectival verbs are used as modifiers within a noun phrase, they are prefixed with another verbal affix, the subject relativizing prefix kw-,9 as in (41): (41) Diegueño (ibid.: 176) ʔ-ku·kap/ /ʔəwa kw-ʔi·ku·-vu house rel-be.big-dem 1-go.around “I went around the big house.”

> ʔəwa kwayʔku·vu ʔəku·kap.

Thus, apparent Diegueño NA phrases are actually subject relative clauses10 (with the same SV order seen in other intransitive sentences)—thus, ʔəwa kwayʔku·

131

9781441124609_Ch08_Final_txt_print.indd 131

11/20/2012 9:26:36 AM

The Bloomsbury Companion to Syntax

“big house” in (41) could also be translated “the house that is big.” Phrases such as (42a and b), with a quantifier (itself also ultimately a Diegueño verb) or an independent demonstrative preceding the kw-plus-verb word, may provide support for this analysis: since quantifiers and demonstratives are otherwise phrase-final (ibid.: 170–1), such phrases follow the order of a clause containing a subject phrase ending in a quantifier or independent demonstrative plus a relativized adjectival verb. (42) Diegueño (ibid.: 171) kw-ʔi·ku·/ > ʔi·kwic ʔəxinkwayʔku· (a) /ʔi·kwic ʔəxin man be.one rel-be.big “one big man” (b) /pəya· kw-məlay/ this rel-die/be.dead “this dead one”

> pəya· kumlay

4 Word Order Variation As we have already seen, word order is not always the same for a given phrase type in a given language: English, for example, allows OSV as a variant of SVO order and has both GN and NG genitive phrases. Additionally, word order changes are o en associated with specific syntactic constructions (such as the two English question types exemplified in (8)).

4.1 Scrambling Some languages allow much more variation. Zigmond, Booth, and Munro (1990: 14) report that an SOV sentence such as (43) in the California language Kawaiisu “might be rephrased with any of five other orderings of its three elements. It is not obvious what effect these different word orders have on the force of a sentence, since consultants claim that in general such changes do not affect meaning, but discourse factors are undoubtedly involved.” (43) Kawaiisu (Uto-Aztecan: Zigmond, Booth, and Munro 1990: 14) taʔnipɨzi momoʔo-a pɨkee-rɨ=ina. man woman-acc see-nml =her “The man saw the woman.” Such “rephrasings” are o en referred to as scrambling. Languages like Kawaiisu that allow multiple word orders are said to have free word order; languages like 132

9781441124609_Ch08_Final_txt_print.indd 132

11/20/2012 9:26:36 AM

Word Order

English that allow only one word order, with perhaps a few clearly dispreferred or discourse-marked variants, have fixed word order. (Ordóñez (this volume) presents a formal analysis of scrambling.) Other languages with relatively free word order may exhibit very specific restrictions. The SOV language Chickasaw (10) allows relatively free scrambling of case-marked subjects and objects, with one exception. While both VS and VO orders are acceptable in transitive sentences with a non-overt argument (44a and b), both VOS and VSO order (44c) are highly questionable:11 (44) Chickasaw (a) Ithána John-at. know John-nom “John knows her,” “John knows him.” (b) Ithána ihoo-a. know woman-acc “He knows the woman,” “She knows the woman.” (c) ?Ithána John-at ihoo-a / know John-nom woman-acc / ?Ithána ihoo-a John-at. know woman-acc John-nom (for “John knows the woman.”)

4.2 Second-position Phenomena Many languages with free word order may have clitic elements (sometimes affixed, sometimes independent) whose clausal position is nonetheless fixed. In Pima, a language of Arizona, for example, a transitive sentence like (45) has six variants, showing all possible orderings of subject, object, and verb: (45) Pima (Uto-Aztecan) (a) Hega’i ‘uvi ‘a-t ha-nolakt ‘iida that woman aux-perf 3pl.obj-buy this “That woman bought these beans.” (b) Hega’i ‘uvi that woman

‘a-t aux-perf

‘iida this

muuñ bean

muuñ. bean ha-nolakt. 3pl.obj-buy

(c) Ha-nolakt 3pl.obj-buy

‘a-t aux-perf

hega’i that

‘uvi woman

‘iida this

muuñ. bean

(d) Ha-nolakt 3pl.obj-buy

‘a-t aux-perf

‘iida this

muuñ bean

hega’i that

‘uvi. woman

133

9781441124609_Ch08_Final_txt_print.indd 133

11/20/2012 9:26:36 AM

The Bloomsbury Companion to Syntax

(e) ‘Iida this

muuñ bean

‘a-t aux-perf

ha-nolakt 3pl.obj-buy

hega’i that

‘uvi. woman

(f) ‘Iida this

muuñ bean

‘a-t aux-perf

hega’i that

‘uvi woman

ha-nolakt. 3pl.obj-buy

However, the second item in the sentence, appearing a er the first word (ha-nolakt) or phrase (hega’i ‘uvi, ‘iida muuñ), is always the same—an “auxiliary” element that indicates the sentence’s aspect (here, perfective), subject (third-person auxiliary agreement is zero), and, sometimes, mood. Pima ‘a-t is a secondposition clitic, whose sentential position is determined structurally rather than by the lexical class or grammatical role of the preceding item. (The multiple word orders in (45) are facilitated by the fact that the syntactic role of the arguments cannot be in doubt, since beans do not buy women, but speakers accept and produce all these orders even in sentences with two human arguments.) In many languages, “second position” can be either a er the first constituent (as in the Pima examples) or a er any initial word. The two possible positions of the (boldfaced) epistemic modal clitic in the TVZ sentences in (46) (where the subject is focused) provide an example: (46) Tlacolula Valley Zapotec zhyi’ ca-yùa’ll. (a) Behts Gye’eihlly=zhyi’ brother Mike=epis prog-sing “Mike’s brother must be singing.” zhyi’ (b) Behts=zhyi’ brother=epis

Gye’eihlly Mike

ca-yùa’ll. prog-sing

The most common items marked by second-position clitic elements are pronominal sentence arguments (particularly subjects and, less commonly, objects) and tense, aspect, and modality elements, but other discourse-related items occur in this position as well. Second-position clitics are sometimes called Wackernagel clitics, and second position may be referred to as Wackernagel’s position, a er Wackernagel’s (1892) analysis of Indo-European data (see further Luraghi, this volume).

4.3 Verb Second One special phenomenon related to second position is verb-second or V2 word order (cf. Poletto, this volume). As shown in (7), German has SVO order in main clauses but SOV order in subordinate clauses. Thus, we might also say that in main clauses like those in (47) the verb appears in second position. In main clauses containing an auxiliary verb such as those in (48), however, the 134

9781441124609_Ch08_Final_txt_print.indd 134

11/20/2012 9:26:36 AM

Word Order

auxiliary appears in second position but the lexical verb (here a past participle) is final: (47) German (a) Hans liest das Buch. Hans reads the book “Hans is reading the book.” (b) Das the

Buch book

liest Hans. reads Hans

(48) German (a) Hans hat das Buch Hans has the book “Hans read the book.” (b) Das the

Buch book

hat has

Hans Hans

gelesen. read gelesen. gelesen

These sentences show that although the order of other constituents in the sentence may vary (here, either the subject or the object may appear initially), the inflected verb of a main clause always follows the first constituent of the sentence. V2 patterns are famously associated with Germanic, but occur sporadically elsewhere in the world, and have been linked to more general secondposition phenomena since the time of Wackernagel. But as Anderson (1993) notes, the placement of the verb in a V2 language is always a er the first constituent (e.g. a er das Buch in the (b) sentences above), never a er the first word of a multi-word phrase. Greenberg’s classification of Norwegian, another Germanic language, as SVO (Section 3.4) is incomplete, since in main clauses Norwegian shows V2 behavior like that of German with other constituents besides the subject allowed to appear before the second-position verb: (49) Norwegian (a) Gutt-en jaget katt-en. boy-def chased cat- def “The boy chased the cat.” (b) I går jaget gutt-en katt-en. yesterday chased boy- def cat-def “Yesterday the boy chased the cat.” However, Norwegian data shows that there are different varieties of V2 languages. Norwegian subordinate clauses again show V2 order, with SOV 135

9781441124609_Ch08_Final_txt_print.indd 135

11/20/2012 9:26:36 AM

The Bloomsbury Companion to Syntax

order unacceptable, in contrast with those of German, which have verb-final order (7b):12 (50) Norwegian (a) Jeg vet at gutt-en jaget I know that boy- def chased “I know that the boy chased the cat.” (b) *Jeg I

vet at know that

gutt-en katt-en boy- def cat- def

katt-en. cat- def jaget. chased

Poletto (this volume) offers a formal approach to the various types of V2 phenomena.

4.4 Variation in the Order of Constituent Elements Usually, the order of words within a syntactic constituent is more restricted than the order of constituents in a sentence: departures from a standard order either are historical relics or represent a specialized construction. For example, although English (with normal AN order) has cases of adjectives following nouns, these are either lexicalized (as with attorney general) or have a specialized semantics (Cinque 2009: 2–4).13 Similarly, in Romance languages (with normal NA order; cf. (25d)), certain adjectives may have a special semantic interpretation when they occur prenominally—for example, in French homme grand means “tall man,” while grand homme is “great man.” Regardless of their order, words within a syntactic constituent normally stay together when word order is scrambled, but not always. Ross (1967: 42) provides an extreme example with the lines of Latin verse in (51): (51) Latin (Horace Ode 1.5) Quis multa gracilis te what.m.sg.nom many.f.sg.abl slender.m.sg.nom you.sg.acc puer in rosa boy.m.sg.nom on rose.f.sg.abl perfusus liquidis urget odoribus drenched.m.sg.nom liquid.m.pl.abl makes.love.to scent.m.pl.abl grato, Pyrrha, sub antro? delightful.n.sg.abl Pyrrha.voc in cave.n.sg.abl “What slender boy, drenched with perfumes, is making love to you, Pyrrha, on a heap of roses, in a delightful cave?” These lines contain four discontinuous constituents, the nominative noun phrase quis puer gracilis “what slender boy,”14 the prepositional phrase in multa rosa “on 136

9781441124609_Ch08_Final_txt_print.indd 136

11/20/2012 9:26:36 AM

Word Order

many a rose,” the ablative noun phrase odoribus liquidis “with liquid scents,” and the prepositional phrase sub antro grato “in a delightful cave.”15 Ross proposes that scrambling is clause-bounded: elements of major constituents may scramble “subject to the restriction that they be in the same clause” (1967: 42).

4.5 Word Order and Case Marking Variable word order is probably most common in languages in which case marking distinguishes all or most subjects and objects. Thus, English and Garifuna, with fixed word order, have no case marking (except on a few English pronouns), while languages like Kawaiisu, Latin, and Chickasaw, which differentiate subject and object (at least) by case marking, allow greater word order freedom. This makes sense, since a hearer has no difficulty parsing a sentence that includes case-marked arguments, regardless of their word order. However, many languages lack case marking but allow quite ambiguous word orders. The VSO language TVZ (19), (52a) has no nominal case marking, but allows both SVO and OVS variant orders. As (52b and c) show, preverbal position indicates focus, not grammatical role: (52) Tlacolula Valley Zapotec (a) Gw-àa’izy Jwaany pfv-hit Juan “Juan hit Pedro.”

Beed. Pedro

(b) Jwaany gw-àa’izy Beed. Juan pfv-hit Pedro “Juan hit Pedro.” (also “Pedro hit Juan.”) (c) Beed gw-àa’izy Jwaany. Pedro pfv-hit Juan “Juan hit Pedro.” (also “Pedro hit Juan.”) It seems to be less common for languages with rich case marking to have fixed word order; however (for example), such Germanic languages as modern Icelandic and Grisons Swiss German have as extensive case marking as German, but have fixed SVO order (Kiparsky 1997).

5 Word Order Typology and Theoretical Syntactic Analysis 5.1 Do All Languages Have a “Basic Word Order”? Languages like Kawaiisu (Section 4.1) and Pima (Section 4.2) seem not to have a basic word order, since speakers produce and accept multiple word orders 137

9781441124609_Ch08_Final_txt_print.indd 137

11/20/2012 9:26:36 AM

The Bloomsbury Companion to Syntax

without restraint. In languages like these, it may be better to think of word order as reflecting discourse factors rather than syntactic roles (as discussed for Pima and the related language Tohono O’odham (Papago) by Smith (2007) and Payne (1987)). On the other hand, if a language has typological features associated with a given word order pattern, a typologist may say that the language behaves like a representative of that pattern even if the pattern is not reflected in the surface word order of main clauses. Thus, Kawaiisu and Pima both belong to (quite different branches of) the Uto-Aztecan language family, many of whose members exhibit similar word order variation to that seen in these two languages. Most languages of the family are postpositional (they use an NP order) and are primarily suffixing, two features o en associated with SOV order, so SOV is o en hypothesized as the actual basic order in these languages.

5.2 Underlying vs. Surface Word Order When a language exhibits different word orders in different syntactic constructions, linguists o en propose analyses that derive one of these from the other. For instance, we saw in (7) that German has basic SVO order in main clauses but SOV order in subordinate clauses. A standard analysis is that SOV is underlying and that the final verb moves to a position a er the first constituent (see Section 4.3) in main clauses. Traditionally, an English sentence is analyzed as having a subject and a predicate: this predicate can be considered a verb phrase (VP) consisting of a verb and various associated elements, especially an object, if present. This approach means that the sentence has two basic constituents, and it is possible to parse three of the four common surface word orders in just this way: [S [VO]], [S [OV]], [[VO] S]. VSO order, however, presents a problem, since there is no way that V and O can be a surface constituent.16

5.3 Non-configurationality Hale (1983) proposed the idea that certain languages are non-configurational, lacking the kind of word (or structure) relationships within and among syntactic constituents seen in prototypical languages—and perhaps, especially, lacking a VP constituent. Such languages typically have free word order and discontinuous constituents (of the sort seen in the Latin example (51)), and allow null anaphora of objects as well as subjects (i.e. with no overt independent pronouns to indicate these constituents, as in examples like Chickasaw (11–2), Tolkapaya Yavapai (14), and Diegueño (40b, c)—and in many other languages 138

9781441124609_Ch08_Final_txt_print.indd 138

11/20/2012 9:26:37 AM

Word Order

not exemplified here, since we are concerned with the order of separate words). Languages analyzed as non-configurational include Hale’s examples of Australian language Warlpiri and the Uto-Aztecan language Tohono O’odham (or Papago, dialectally related to Pima; cf. (45)) as well as the Iroquoian language Mohawk (Baker 1996) and such older Indo-European languages as Old Norse (e.g. Faarlund 1990; Rögnvaldsson 1995) and the ancient languages Latin and Greek (Luraghi 2010). In addition to the three characteristics just mentioned (free word order, discontinuous constituents, and null anaphora), many languages that claimed to be non-configurational have second-position clitics (Warlpiri, Tohono O’odham-Pima, Latin, Ancient Greek).17

5.4 Word Order Typology and Formal Theory Formal linguistic theory has built on the observations contained in word order based typological studies like that of Greenberg (1963) since the 1970s. Early attempts to account for the seeming disharmony of the SVO structure of English included, for example, a proposal by McCawley (1970) that English is underlyingly VSO. In contrast, Chomsky (1970) generalizes Greenberg’s basic word order patterns in what is known as X-bar theory—the idea that all types of phrases have the same structure, where the lexical category for the head of any phrase is represented by “X.” An X-phrase (XP) consists of a “specifier” plus a constituent X’ (read as “X-bar”), which contains the head X plus any complement. Whether the head precedes or follows its complement is determined by a directionality parameter, yielding the orders [specifier [X complement]] for a head-initial language and [[complement X] specifier] for a head-final language. The most familiar example of specifier-head-complement order is the SVO sentence order in a language like English, where the subject is the specifier, the verb is the head, and the object is the complement, with verb and object forming an V’ constituent (earlier called the verb phrase or VP). In the 40 years since Chomsky’s original proposal these ideas have been refined and extended o en (e.g. by Jackendoff (1977)), most strikingly in the work of Kayne (e.g. 1994), who postulated a principle of antisymmetry by which all languages have underlying specifier-complement-head (or, ultimately, SVO) order. Typological facts are adduced to support an antisymmetric approach: for example, second position and V2 are privileged positions in many languages, and these are always calculated from the beginning of the sentence rather than the end. To account for the wide variation in word order among the world’s languages, a Kaynean account must rely on extensive movement of constituents (sometimes as large as a full clause). This tradition has produced many provocative analyses of otherwise recalcitrant data. For example, Kramer (2009) argues that the “basic” surface order 139

9781441124609_Ch08_Final_txt_print.indd 139

11/20/2012 9:26:37 AM

The Bloomsbury Companion to Syntax

seen in Middle Egyptian sentences like (6a) (repeated below) is actually derived by a rule that raises the verb from an underlying SVO order to a higher Tense position, resulting in VSO order. Stative sentences like (6b) retain SVO order because (as the translation shows) they are tenseless. (6) Middle Egyptian (a) iw rḫ.n ddi part learn.pst Djedi “Djedi learned the plan.”

sḫr. plan

(b) ddi rḫ.w sḫr. Djedi learn.stat.3sg.m plan “Djedi knows the plan.”/”Djedi knew the plan.” Although the tools of modern linguistic theory have now been used to analyze many exotic languages, the influential proposals of Chomsky, Jackendoff, Kayne, and their followers have been distrusted by some typologists because of their seeming reliance on English (and other familiar European languages): would the same theories have been developed by scholars who spoke Garifuna or Chickasaw?

6 Word Order Change A major concern of historical and comparative linguists is the question of how word order changes over time, and thus how related languages may have different word orders and different typological properties. Two of the principal factors that facilitate word order change are grammaticization of marked variant word orders and morphological change. Consider Greenberg’s Universal 6, given above as (36): (36) “All languages with dominant VSO order have SVO as an alternative or as the only alternative basic order.” (Universal 6) Such variation in word order normally provides greater prominence for a constituent that is either topicalized or in focus. Thus, for example, the TVZ example (52a), in neutral VSO order (as with almost all languages of this family), has the variant (52b) with SVO order and the subject focused (TVZ object focus gives an OSV variant, as in (52c)). Basic word order change may occur as focus or topic become grammaticalized and the variant word order becomes the norm. Indeed, in the Tlacolula de Matamoros dialect of TVZ SVO is now a more common word order than VSO (Lillehaugen 2008).

140

9781441124609_Ch08_Final_txt_print.indd 140

11/20/2012 9:26:37 AM

Word Order

The most common morphological change that can lead to a change in word order is the loss of nominal case marking. As noted in Section 4.5, languages like English and Garifuna have no case marking but fixed word order, while languages like Kawaiisu, Latin, and Chickasaw, which differentiate subject and object by case marking, have freer word order. We do not have records of earlier stages of Garifuna, Kawaiisu, and Chickasaw, but we know that Germanic ancestors of English had case marking (as modern German still does on full noun phrases) and we know that modern descendants of Latin generally have no case marking and less variable word order.18 Even without earlier written records, comparison of related languages can support postulation of syntactic changes in word order from earlier stages of a language.

Notes 1. My thanks to Felipe H. Lopez (for Tlacolula Valley Zapotec), Maurice Lopez and Anita Lambey Martinez (Garifuna), Catherine Willmond (Chickasaw), Virgil Lewis and the late Etheleen Rosero (Pima), and the late Molly Star Fasthorse (Tolkapaya Yavapai) for teaching me about their languages (examples from which are given in practical orthography). I also thank linguists who kindly provided data and analytical suggestions about particular languages: Richard Applegate (Ineseño Chumash), Brook Danielle Lillehaugen (Norwegian), Silvia Luraghi (Italian), and Marcus Smith (Japanese). This paper is dedicated to my first linguistics teacher, Joseph H. Greenberg. 2. Examples are given in the orthography of the source or in standard orthography. 3. Indeed, one could consider these clitic pronouns to be affixes, even though they are written as separate words; see the discussion in Luraghi (this volume). 4. One could consider this as an Ergative-Verb-Absolutive order, since the class of absolutives includes both transitive objects and intransitive subjects. However, it is rare for ergative languages not to display evidence for a unified class of subjects (Anderson 1976), so most typologists would also treat this as SVO and VS. 5. Greenberg discusses earlier roots of his research (1963: 104–5). 6. Actually, Norwegian main clauses follow a V2 pattern (see Section 4.3), although generally the verb is not final in subordinate clauses as it is in German sentences like 0(b) (Faarlund 2010, Bentzen 2007). Greenberg classifies Norwegian as SVO based on its main clause word order and does not discuss V2. 7. A few English prepositions, such as notwithstanding, can also be used postpositionally. 8. Dryer reports a strong correlation between VO and NA order, on the other hand, with 404 of 505 VO languages showing NA order (however, Dryer’s category VO includes SVO and VOS as well as VSO). 9. Langdon (1970: 142) describes kw-plus-verb formations as nominalizations, but shows that the prefix is a subject relativizer (1970: 176–7). 10. Hale and Danilo (2002: 36–7) present a similar analysis of NA order in SOV languages of the Misumalpan family (Nicaragua, Honduras). 11. One cannot say such sentences are ungrammatical since they are occasionally volunteered. But they are normally rejected in elicitation and occur very infrequently, most likely subject to as yet unexplored discourse constraints.

141

9781441124609_Ch08_Final_txt_print.indd 141

11/20/2012 9:26:37 AM

The Bloomsbury Companion to Syntax 12. Faarlund (2010) discusses comparative aspects of V2 in Scandinavian. 13. Cinque notes that non-lexicalized English postnominal adjectives unambiguously correspond to reduced relative clauses with fewer interpretive possibilities than prenominal ones, as in (1) Olga is a more beautiful dancer than her instructor. (Olga is more beautiful. / Olga dances more beautifully.) (2) Olga is a dancer more beautiful than her instructor. (Olga is more beautiful. / #Olga dances more beautifully.) 14. Perfusus “drenched” also agrees with puer “boy,” but the extraposition of an adjective phrase like perfusus (odoribus liquidis) “drenched (with liquid scents)” might occur in many languages. 15. Pullum (1982: 5) expresses doubt that the word order freedom seen in Horace’s verse “falls within the bounds set by universal linguistic theory,” while Brown (1981: 18) notes Horace’s “impressive . . . use of the latent poetic resources of grammar”: the discontinuous gracilis . . . puer embraces te “you,” and both are surrounded by multa . . . rosa. Spevak (2010: 13) comments, however, that “in the variable constituent order of Latin, there are . . . certain liberties that are not common in modern European languages such as Romance and Germanic, nor in the more flexible Slavic languages. . . . One may note that Latin poetry is much more flexible than prose and shows even more liberties in the placement of constituents.” 16. Schwartz (1972) questions the idea that SOV languages have a VP constituent, arguing that both SOV and VSO languages have “flat” (non-hierarchical) structure, with three rather than two constituents. Anderson and Chung (1977), in contrast, suggest that VSO languages may show evidence for the constituency of verb and object in underlying structure. 17. Following Baker (2000), Luraghi (2010) reports that there are two types of nonconfigurational languages: dependent-marking languages such as Latin (which have nominal case marking) and head-marking languages such as Mohawk (where pronominal arguments are all indicated on the verb, and there is generally no case marking). She argues that the prototypical non-configurational feature of discontinuous constituency (as in Latin (51)) is generally not allowed in head-marking languages. Indeed, Baker notes that discontinuous constituents are not too common in Mohawk (1996: 11–12). 18. Luraghi (2010) notes that nominal case began to be lost in Indo-European as a result of the loss of discontinuous constituents, specifically the requirement that adjectives must be adjacent to the noun they qualify. A further development in Romance was the rise of pronominal clitics associated with the verb, which, Luraghi claims, has led to the development of a new type of non-configurationality in Romance languages like modern spoken French.

142

9781441124609_Ch08_Final_txt_print.indd 142

11/20/2012 9:26:37 AM

9

Scrambling Francisco Ordóñez

Chapter Overview Introduction Base Generation Approaches of Scrambling Movement Approaches of Scrambling Conclusion

143 144 146 152

1 Introduction Scrambling is the term used to refer to the phenomenon of free word order variability in natural languages. Ross (1967) is among the first generative syntacticians to apply this term to languages that permit word order freedom such as German, Latin, and Russian, among many others. The following example from Hungarian, another such language, shows this free word order property. The different orders correlate with different discourse properties of the sentence: (1) (a) Kati Kate

megevett ate

egy szelet tortát. a piece of cake.

(b) Egy szelet tortát A piece of cake

Kati evett meg. Kate ate.

(c) Kati Kate

egy szelet tortát evett meg. a piece of cake ate.

(d) Egy szelet tortát A piece of cake

evett meg ate

Kati. Kate.

143

9781441124609_Ch09_Final_txt_print.indd 143

11/20/2012 9:24:30 AM

The Bloomsbury Companion to Syntax

Thus, the term scrambling has been used to describe any word order variation at the sentential level that appears optional. The topic of scrambling gained considerable theoretical attention with the introduction of a richer theory of movement in the 1980s, and this interest has continued to the present. For instance, scrambling has been the main focus of four edited volumes (Grewendorf and Sternefeld 1989; Corver and Van Riemsdijk 1994; Karimi 2003; Saito and Sabel 2005), three monographs (Richards 2004; Karimi 2005; Hinterhölzl 2006) and innumerable papers (see Munro, this volume). Thus, I will not be able to do justice to all aspects of scrambling in the literature, and I will focus in the aspects and generalizations that have guided the investigation over the last 25 years. These include two major views of scrambling: the base generation approach and the movement approach. We will examine each in the following sections.

2 Base Generation Approaches of Scrambling Base generation approaches to free word order are those that deny any movement or transformation between the different word order possibilities in the language. Early in the 1970s Ken Hale assumed that word order variability in a language like Walpiri was due to the fact that the arguments instantiate a flat structure (a structure in which arguments depend directly from the same sentential node), while languages with strict word order have a hierarchical structure in the sentence. The difference between one type of language and the other has been attributed to the configurationality parameter (Hale 1983). This parameter for free word order could be due to various factors. One is that in a language with a rich inflectional system, all the arguments of the verb are indicated morphologically (i.e. the so called poliysynthetic languages or head-marking languages in Baker 2002). According to Baker (1996) and Jelinek (1984) the arguments of the verb are the agreements, and the overt complement counterparts are adjoined to the sentence linked via a chain. The fact that overt arguments are adjoined in a peripheral position explains why they can be moved easily, since adjunction does not impose a restriction on the order on which elements might appear. Instead of proposing flat structures, other linguists assume that scrambling is the product of a free order of merge of the different arguments of the verb. Thus, in NON scrambling languages merger of arguments is strict because assignment of case and assignment of theta role needs to be done under strict adjacency. However, in scrambling languages case and theta role are freely assigned in alternative D-structures (Neeleman 1994) or they can be assigned in different positions under the same projection line of the inflectional projection (Bayer and Kornfilt 1994). 144

9781441124609_Ch09_Final_txt_print.indd 144

11/20/2012 9:24:31 AM

Scrambling

In the minimalist literature, the base generation approach has taken new forms. Scrambling, in fact, poses two major problems for the minimalist program: (i) It is not clear what kind of feature checking is taking place in scrambling orders and (ii) the change of order appears optional in all these languages, contrary to the guiding idea that all movement is necessarily motivated. Bošković and Takahashi (1997) circumvent the second problem by assuming that scrambled orders are merged in their surface position. Thus in a Japanese sentence both the scrambled order in (2b) and the unscrambled order in (2a) are generated via merge: (2) (a) John-ga [Mary-ga sono hon-o katta to ] omotteiru. John-NOM Mary-NOM that book-ACC bought that thinks “John thinks that Mary bought that book.” John-ga [Mary-ga katta to] omotteiru. (b) Sono hon-oi that book-Acc John-NOM Mary-NOM bought that thinks “John thinks that Mary bought that book.” However, in order to relate both orders Bošković and Takahashi (1997) assume that the scrambled object in (2b) lowers at LF to a case and theta role position adjacent to the embedded verb.1 (3) Sono hon-oi

that book-Acc

John-ga

[Mary-gat

katta to] omotteiru.

John-NOM

Mary-NOM

bought that thinks

From this perspective, theta role assignment is done configurationally at LF. Languages differ as to whether theta roles must be assigned in their theta and case positions, yielding non scrambling languages and languages in which the assignment of theta role and case can be delayed until LF, yielding scrambling languages.2 The parametric difference is due to feature strength of theta-role features. Theta-role features can be strong, in which case they must be assigned configurationally at merge (i.e. non scrambling languages), or they can be weak and just be assigned at LF via a lowering operation (i.e. scrambling languages). The idea of theta roles as features has also been adopted by Fanselow (2001) to explain free word order. He proposes that formal case and categorical features are associated with thematic features. Scrambled and unscrambled orders are all base generated. Theta roles and formal features are assigned in tandem at LF in scrambling languages without the need for any specific configuration for their assignment. This proposal is a theoretical implementation of the idea that 145

9781441124609_Ch09_Final_txt_print.indd 145

11/20/2012 9:24:31 AM

The Bloomsbury Companion to Syntax

free word order is permitted in languages with overt case marking (Argument Marking Languages in Baker 2002’s typology). Finally, Kiss (2008) presents a mixed position. According to her, Hungarian preverbal orders of arguments are fixed, as opposed to postverbal orders. This puzzling state of affairs can be explained if one assumes that preverbal orders must follow a specific locality restriction on checking different functional projections. Thus there is a fixed position for focus, predicates to which different arguments move to preverbally. On the other hand, postverbal orders correspond to flat structures. This asymmetry between configurational structures to the left of the verb and non-configurational ones to the right of the verb is explained if (i) all postverbal orders indicate there is verb movement to the left and (ii) when the verb is moved up to the left, the structure left behind gets flattened. Kiss (2008) has a specific implementation of this intuition in terms of phase theory, which is that when the verb moves up to the next phase the previous phase gets flattened. It is important to signal that the mirror image of Hungarian does not seem to have been attested. Thus, there is no language that shows configurationality diagnostics to the right of the verb and flat structure to the left of the verb.

3 Movement Approaches of Scrambling All the movement approaches to free word order assume the initial order in all languages to be configurational.3 That order is altered via movement of the different phrases to different positions in the sentence or outside the sentence. The questions that arise under a movement approach are the following: (1) What elements are targeted by scrambling. Arguments? Predicates? Adjuncts? Others? (2) What are the locality restrictions of this movement? Can it occur long distance? (3) Does this movement reconstruct? (4) Is this movement motivated for feature checking? (5) What is the landing site of this scrambling movement? There is no consensus in the literature on any of the questions above. In many cases the answers to the previous questions depend on the language or group of languages one is looking at, and at some point to the specific data one is focusing on (Boškovic 2004). However, there seems to be a descriptive taxonomy that linguists seem to agree upon, depending on the locality movement. In that sense, scrambling can be divided into three subtypes: object shift, clause bound scrambling (i.e. internal to the sentence), and long distance scrambling (i.e. scrambling across a finite sentence). 146

9781441124609_Ch09_Final_txt_print.indd 146

11/20/2012 9:24:31 AM

Scrambling

3.1 Object Shift Object shift is the term used to describe the overt movement of objects internal to the sentence. It has been widely described for Scandinavian languages such as Icelandic, Norwegian, and Danish (Holmberg 1991; Vikner 1994; Holmberg and Platzack 1995). The movement is strictly local. It only affects object arguments (as opposed to PP or other arguments of the verb). In some languages, it only affects pronominal object obligatorily as in Danish and Swedish shown in the contrast in (4). However, in Icelandic it might affect DPs as well. Importantly, object shift occurs when there is overt verb movement. This correlation is called Holmberg’s generalization (Holmberg 1986). (4) (a) Han He

Köpte bought it

deni not

(b) *Han Köpte bokeni He bought the book “He did not buy the book.”

inte t

ti.

inte not

ti.

The facts that object shift only affects case marked elements (e.g. pronominals) and that it is very local have led linguists to assume that it might be a case-related movement like an instance of A-movement. Moreover, like A-movement object shift also licenses floating quantifiers (Déprez 1994; Vikner 1994). Contrary to A´-movement object shift cannot license parasitic gaps (Vikner 1994)4.

3.2 Clause Bound Scrambling Probably the most widely spread form of scrambling is clause bound, which affects most types of arguments. In some languages like Dutch, however, it only moves the arguments around adverbs (Neeleman 1994). Possibly, for this reason, the first investigations of scrambling in terms of movement concluded that it was an adjunction operation (Saito 1985; Webelhuth 1992). One of the most important debates on this type of movement has been to establish whether it can be classified as A-movement or A´-movement. Webelhuth (1992) concluded that this type of scrambling in German is a mixture of both A- and A´-movement. Depending on the type of diagnostic, we might reach one conclusion or the other. For instance, if we take Weak Crossover (WCO) as a diagnostic for A´-movement, contrary to overt wh-movement, scrambling of an object

147

9781441124609_Ch09_Final_txt_print.indd 147

11/20/2012 9:24:31 AM

The Bloomsbury Companion to Syntax

quantifier does not create a WCO violation. Instead, scrambling of the quantifier creates a new binding configuration:5 (5) (a) *?Whi [Proi NP] ti V)? wh-movement (b) [QPi] [Proi NP] ti V? scrambling of a QP These facts also indicate that Scrambling does not reconstruct, one of the diagnostics of A´-movement. Another diagnostic is used for A-movement. It is the possibility of a scrambled XP to bind an anaphor from its moved position as in (6) (Mahajan 1990, 1994; Tada 1992; Saito 2003) among others:6 (6) [DPi] .… [ANAPHORi.…] …… ti V (Scrambling of a DP over an anaphor) A third diagnostic that scrambling does not reconstruct is given by the fact that the scrambled XP might circumvent a principle-C effect (Mahajan 1994): (7) (a) *PRONOMINALi [XP DPi]j V (Principle C effect on the merge order) (b) [XP DPi]j PRONOMINALi tj V (scrambling XP =No principle C effect) Finally, the last dignostic for A-movement is that it licenses floating quantifiers.7 (8) DPi YP [FQ ti] V (scrambling) However, not all the diagnostics are equally uniform and do not yield a uniform case against reconstruction effects. In many cases the properties of the binder are crucial to establish whether there is reconstruction or not. Thus, principle-C effects are not circumvented in German, Korean (Lee and Santorini 1994), or Spanish (Ordóñez 1998) when the binder is the subject as shown in the following schema in (9). Thus scrambling makes no difference for grammaticality of examples in which the subject is a pronominal that triggers a principle-C violation before scrambling. (9) (a) *[subject PRONOMINAL i] [DPi] V (b) *[DPi]j [subject PRONOMINAL i] tj V (scrambling over the subject=Reconstruction is requied) Finally, scrambling of a XP containing a pronoun bound by a subject quantifier requires reconstruction (Lee and Santorini 1994): 148

9781441124609_Ch09_Final_txt_print.indd 148

11/20/2012 9:24:31 AM

Scrambling

(10) (a) [subject QPi] [PRONOMINALi DP] V (b) [PRONOMINALi DP]j [subject QPi] tj V (scrambled order) Moreover, Mahajan (1999), Saito (2003), and Tada (1993), show that the scrambling of an anaphor does not create a principle-C effect. Thus, reconstruction is also required:8 (11) (a) DPi ANAPHORi V (b) ANAPHORi DPi tj V (scrambled order) Another diagnostic in favor of A´-movement is the possibility of licensing parasitic gaps. Webelthuth 1992 shows that scrambling of a DP argument can license a parasitic gap in an adjunct clause in the same way that wh-movement licenses parasitic gaps PG. Other authors like Neeleman (1994) question the ability of scrambling to linsence PG: (12) Whi V ti [Adjunct V PGi]? (wh-movement in English) (13) that .… DPi [Adjunct PGi] t i. (Scrambling in German) The diagnostics presented so far show that scrambling is far from being a simple movement amenable to the A-A´-movement distinction. Webelhuth (1992) proposed that scrambling might be characterized as a mixed position, with A- and A´-movement properties. He presented the following puzzling situation: a scrambled XP argument licenses a parasitic gap (diagnostic for A´-movement) and at the same time it creates a new binding possibility, circumventing a WCO violation (diagnostic for A-movement). This has been labeled in scrambling literature Webelhuth’s paradox: (14) that … QPi [Adjunct PGi] t i [Proi DP] V (Webelhuth paradox) This conclusion has been challenged by Hinterhölzl (2006), Mahajan (1994), Müller and Sternefeld (1994), and Saito (2003), among others. Mahajan (1990, 1994) claims that local sentence bound scrambling does not take the XP to a mixed position. Local scrambling might take the argument either to an adjunct position, in which case it shows the diagnostics for A’-movement, or to a specifier of an agreement projection, in which case it shows the diagnostics of an A-movement. From that perspective (14) simply might hide two movements (Mahajan 1990): movement to an A position from where it can license the pronoun in the complement clause, followed by A´-movement, which would license the parasitic gap. This kind of proposal, however, does not easily fit 149

9781441124609_Ch09_Final_txt_print.indd 149

11/20/2012 9:24:31 AM

The Bloomsbury Companion to Syntax

into the minimalist program on feature checking since Mahajan is permitting optionality on possible landing sites for scrambling in Hindi: either a specifier position or an adjunct position. On the debate for A-A’ diagnostics it is important to establish whether scrambled DPs permit scrambling out of them. Bošković and Takahashi (1998) point out that scrambling out of scrambled phrases is permitted in Japanese, which give evidence to the fact that reconstruction occurs. (15) [XP]i [YP t i]j tj V However, Müller has pointed out an interesting restriction on XP from which scrambling has taken place (i.e. remnant movement). A remnant YP out of which scrambling has applied cannot undergo scrambling itself. He attributes this restriction to a principle of Unambiguous Domination (Müller 1996): (16) [XP]i [YP t i] V (Scrambling out of YP) (17) * [YP t i]j [XP]i t j V (No scrambling permitted for the remnant YP) Further restrictions exist on sentential scrambling. Thus, it might only affect arguments in languages like Japanese (see Bošković and Takahashi 1998). In some languages, not all quantifier XPs can scramble. Thus, in German and Spanish wh-words cannot scramble (Muller and Sternefeld 1994; Ordóñez 1998). Finally, scrambling has effects on the interpretation of indefinites and quantifiers. Thus, Diesing and Jelinek (1995) pointed out that an indefinite NP like Kinder in German is interpreted as existential when it is in situ. However when it is moved it is interpreted as generic. Similar effects have been described for Turkish as well (Bayer and Kornfilt 1994): (18) (a) weil Hans meistens Kinder verführt. Since Hans mostly children seduces. (Existencial) (b) weil Hans Since Han

Kinder meistens children mostly

verfuhrt. seduces. (Generic)

In Japanese, quantifier scope possibilities change with scrambling as well (Miyagawa 1997). In a sentence with two quantifier as (19) is unambiguous with a rigid scope interpretation. However, with the scrambling order, the sentence becomes ambiguous with the scrambled DP taking scope over the quantifier in situ (from Miyagawa 2006, among others):

150

9781441124609_Ch09_Final_txt_print.indd 150

11/20/2012 9:24:32 AM

Scrambling

(19) (a) Dareka-ga daremo-o sikatta. someone-nom everyone-acc scolded someone > everyone, *everyone > someone (b) Daremo-oi dareka-ga ti sikatta. Everyone-acc Someone-nom scolded someone > everyone, everyone > someone While sentence-bound scrambling shows this mixture of properties, we will see in the next section that long distance scrambling is more uniformly considered A´-movement.

3.3 Long Distance Scrambling While scrambling inside the clause is widespread, not all languages permit scrambling of DPs out of a finite clause. Thus German and Dutch do not permit scrambling out of a finite CP. Moreover, Dutch does not permit scrambling of XP arguments over the subject (Neeleman 1994). German only permits scrambling out of infinitival clauses to which a process similar to restructuring has occurred (the so called coherent infinitives).9 keiner (20) dass [IP [den Hund]i that the dog-acc nobody-nom “That nobody tried to feed the dog.”

[CP PRO ti zu futtern] versuchte]]] to feed tried

This dichotomy presents an interesting puzzle as to what the source of this parametric difference between languages that permit long distance scrambling versus languages that do not. For one thing, long distance scrambling seems to show the properties akin to A´-movement. The consensus in the literature is that long distance scrambling does not create new binding possibilities (Mahajan 1994; Saito 2003) whereas it does license parasitic gaps as (Mahajan 1990): (21) *DPi [IP1 [DP Anaphori] … … … … [CP … [IP2 ti]]] (no binding of anaphor) *QP i [IP1 [DP pronouni] … … … … [CP … [IP2 ti]]] (no binding by QP) DPi [IP1 [Adjunct PGi] [CP … [IP2 ti]]] (licensing of parasitic gaps) One of the most interesting properties of long distance scrambling involves the so-called radical reconstruction effects discussed by Saito (2003). Thus, a wh-word might scramble outside the embedded sentence selected by a verb that requires wh-words as we saw in (21). This implies that the scrambled

151

9781441124609_Ch09_Final_txt_print.indd 151

11/20/2012 9:24:32 AM

The Bloomsbury Companion to Syntax

wh-word must reconstruct at LF to its embedded clause in order to check the wh-features. Saito (2003) points out that movement of the wh-word outside its embedded CP is not semantically motivated since the wh-feature must be checked in the lower clause.10 John-ga [WH-ISL Taroo-ga ti katta ka]  siritagatteiru. (22) ?Nani-oi what-ACCi John-NOM [WH-ISL Taro-NOM ti bought Q] want.to.know “John wants to know what Taro bought.” Takano (2010), on the other hand, notices that long distance scrambling in obligatory control contexts might create new binding possibilities, contrary to what we saw in (21). These new binding possibilities are created when the binder is the quantifier scrambled in front of a pronominal that is the controller of the embedded clause (23). According to Takano this new binding possibility can be explained if (i) control involves movement and therefore the [CONTROLLER Proi-DAT]j has moved from the embedded CP to the matrix clause and (ii) the scrambled QP-Dat moves before the lower trace of the controller DAT, yielding the proper configuration for pronominal binding: (23) [scrambled DATQP]i XP-NOM [CONTROLER Proi-DAT]j [CP INF ti tj ti V] V Thus, all the pronominal binding relations are established in the embedded CP, given the fact that control involves movement.

4 Conclusion The phenomena labeled scrambling is a pervasive phenomenon in natural languages. It is linked to the important issue of variability of word order (see Munro, this volume), which is an essential part of syntactic theory. For those who work in transformational grammar, scrambling still poses big challenges for the theories of movement. Thus there is still ongoing debate about the following issues: (i) whether movement takes place; (ii) if so, what kind of movement is responsible for this variation (A or A’); (iii) what is the level of representation in which movement applies; and (iv) what motivates this movement.

Notes 1. In that sense, strictly speaking, Bošković and Takahashi (1997) advocate for a movement approach at LF. 2. For criticism of this proposal see Bailyn (2001) and Miyagawa (2006). For a reply to Baylin´s criticism see Bošković (2004).

152

9781441124609_Ch09_Final_txt_print.indd 152

11/20/2012 9:24:32 AM

Scrambling 3. The principle that requires hierarchical configuration is the UTAH (Uniformity Thematic Assignment Hypothesis) by Baker (1988). 4. For a different proposal that object shift is phonetic movement, see Holmberg (1999). 5. The final position of the verb is irrelevant to compute the WCO violation. Müller and Sternefeld (1994) and Grewendorf and Sabel (1999) argue that this diagnostic is not entirely valid for German. 6. Not all anaphors permit binding by scrambled arguments. Thus, purely nominative anaphors reject a new binder in German (Müller and Sternefeld 1994) and Spanish (Ordóñez 1998). This might be reduced to a ban against nominative anaphors in these languages. 7. However, see Ko (2007) for a study of an asymmetry of licensing Floating Quantifiers between the subject and other complements. She resolves this asymmetry in terms of the theory of linearization by Fox and Pesetsky (2004) 8. In order to deal with this contradictory situation between principles A and C, Saito (2003) follows Belletti and Rizzi (1988) and assumes that principle A applies in the derivation. 9. See Grewendorf and Sabel (1994) for an analysis in terms of incorporation at LF. See Hinterhölzl (2006) for an analysis in terms of remnant movement. 10. See Miyagawa (1997), (2005), and (2006) for a minimalist view that solves some of the puzzling conclusions posed by this sentence by Saito (2003).

153

9781441124609_Ch09_Final_txt_print.indd 153

11/20/2012 9:24:32 AM

10

On V2 Types Cecilia Poletto

Chapter Overview Introduction The Standard View on Germanic V2 V2 in the Scandinavian Languages Old Romance V2 Subject Inversion in Cimbrian Concluding Remarks

154 155 158 160 162 163

1 Introduction In this chapter I provide an overview of the V2 phenomenon in a historical perspective. I first present the phenomena related to the V2 restriction, and the standard view derived from the analysis of German and Dutch, the languages where the phenomenon was first discovered. Then I summarize the findings coming from the Scandinavian languages and finally introduce Romance into the picture. At the end of the discussion it will be clear that V2 is by no means a unitary phenomenon as it was originally thought in the early 1980s and that much still has to be done in terms of empirical and theoretical research. The chapter is organized as follows: in Section 2 I discuss the traditional phenomena associated to the V2 property on the basis of German. This association permits to account for some word order facts mentioned in Munro (this volume: examples (10)–(11)), which seem to contradict the idea that there is one basic word order in a language: languages like German look like SVO in main clauses by using SOV in embedded clauses. The relation between word order and the main or embedded status of the clause can be captured through the description of the V2 property. 154

9781441124609_Ch10_Final_txt_print.indd 154

11/27/2012 11:06:19 AM

On V2 Types

In Section 3 I illustrate the problems posed by the Scandinavian languages, which have a different behavior in embedded clauses. Section 4 deals with Old Romance V2, which shows a different set of phenomena related to the V2 property—namely the alternation between enclisis and proclisis and the licensing of null subjects in relation to V2. Moreover, Old Romance allows for V3 orders, though it keeps the typical pattern of subject inversion. Section 5 deals with Cimbrian data, which show that subject inversion can also be banned though the language still has some V2 properties. The conclusion drawn in Section 6 is that V2 is a complex property and that verb movement triggers different consequences depending on the independent factors of the language examined. Thus, contrary to scrambling (see Ordóñez, this volume) there are no approaches that see V2 as basic word order (for a typological approach see Munro, this volume), but only movement approaches that differ in the trigger of verb movement.1

2 The Standard View on Germanic V2 The V2 property was originally defined as a cluster of phenomena that explains the alternation of verb positioning in German and Dutch by den Besten (1983). It can be split into three distinct syntactic phenomena, which are known as the V2 correlates: (i) “subject inversion” of the so-called Germanic type, where the subject occurs between the auxiliary and the past participle, (ii) second position of the inflected verb, which can only be preceded by one single constituent (the so called “linear restriction”), and (iii) “root character” of the phenomenon, which is obligatory in all main clauses but only possible in a very restricted set of embedded clauses selected by a small class of verbs of saying. The three properties are exemplified below on the basis of German: (1) Heute ist Hans angekommen. Today  is Hans arrived “Today Hans arrived.” (2) (a) *Heute das Buch habe ich gelesen. Today the book have I  read “Today I read the book.” (b) Heute habe  ich das Buch  gelesen. Today have I the  book  read (3) (a) Ich habe  heute das Buch gelesen. I  have today the book read 155

9781441124609_Ch10_Final_txt_print.indd 155

11/27/2012 11:06:21 AM

The Bloomsbury Companion to Syntax

(b) . . . dass ich heute das Buch gelesen habe. . . . that I today the book read have . . . “that today I read a book.” (c) *. . . dass ich habe heute das Buch gelesen. . . .  that I have today the book read. In the traditional analysis, which was proposed starting from Chomsky’s (1985) idea that functional features also project syntactic structure in the terms of the X’- theory, these three properties are accounted for by assuming that in V2 languages the C° position constituting the head of the sentence must always be present and lexically filled. In main clauses the inflected verb has to move to the C° position, which would otherwise remain empty. V to C movement results in subject inversion, because the subject position is generally assumed to be SpecT. The ban against V2 in embedded contexts is derived by the fact that C° is already filled by the complementizer. The verb second linear restriction is explained by the fact that there is only one position on top of C°, namely SpecC. Once SpecC is occupied by a single constituent, there is no further syntactic space to add another. Postulating the requirement that C° always has to be filled still does not explain why these languages are V2 and not V1 like, say, the Celtic languages. Therefore, a second constraint has been proposed that ensures movement of one XP into the SpecC position. Several authors have given different interpretations of this constraint, but I will not go into the details here for space reasons (I refer to Tomaselli (1990) and Vikner (1995) for a detailed discussion on this). There are well known exceptions to the V2 constraint, and they are usually split into three types: (1) embedded V2 (2) main clauses with V-final (i.e. embedded word order) (3) verb first main clauses As for embedded V2, this is optionally possible in German only in a restricted set of verbs of saying, known as “bridge verbs,” which are known to constitute a unitary semantic class. When embedded V2 applies, the complementizer is never present; hence we can assume that a special class of main verbs allows for an empty C in its complement, which is then occupied by the inflected verb as in matrix clauses: (4) (a) Er sagt, du wirst kommen. He says, you will come “He says, you will come.” 156

9781441124609_Ch10_Final_txt_print.indd 156

11/27/2012 11:06:21 AM

On V2 Types

(b) Er sagt, dass du kommen wirst. He says that you come will (c) Er bedauert, dass du kommen wirst. He regrets, that you come will “He regrets that you will come.” (d) *Er bedauert, du wirst kommen He regrets you will come As for the possibility of embedding a V2 structure under the complementizer, languages vary with respect to this property; while in German sequences like dass+XP+inflected verb are ungrammatical, the Scandinavian languages require the presence of the complementizer in front of the XP+inflected verb. This has long been noticed (see a.o. Vikner 1995) and generally attributed to the properties of the complementizer in the different languages and more specifically to the position of the complementizer in Scandinavian, which is in a higher position and permits embedding of a whole V2 structure, while this is not the case for German, which has a lower complementizer that blocks verb movement. The second exception concerns cases of main clauses that display verb final word order like embedded clauses: (5) (a) Wenn er nur käme. If he only came “If he only came.” (b) Käme er nur! Came he only Again we notice that there is an alternation between the complementizer (in this case wenn “if”) and the verb in first position. The same analysis of embedded V2 clauses can be applied here too: when the C° position is not occupied by the complementizer, the inflected verb raises to C°. Notice however that the examples above display V1 word order and not V2: this raises the problem of what the distinction between declarative clauses might be, which are generally V2, and other types of clauses like yes/no questions, imperatives, and optatives as in the examples above. The standard explanation for the reason why there are V1 occurrences in V2 languages is that all V1 cases have an empty operator of some sort located in the SpecC position: yes/no questions have an empty operator of the “verum” type, imperatives have an illocutive operator expressing the illocutive force of an imperative clause, and optative clauses also have a modal operator that satifies the V2 constraint. If nothing is either visibly or invisibly present in first

157

9781441124609_Ch10_Final_txt_print.indd 157

11/27/2012 11:06:21 AM

The Bloomsbury Companion to Syntax

position, V2 languages insert expletives in the first position to occupy the SpecC position. The following alternation shows that German es is only a place holder for the SpecC position, because if anything else is moved to SpecC, it disappears from the sentence: (6) (a) Es hat jemand angerufen. It has someone phoned “Someone phoned.” (b) Jemand hat angerufen. Someone has phoned “Someone phoned.” (c) *Jemand hat es angerufen. Someone has it phoned “Someone phoned.” If we only consider the system of German and Dutch, this analysis perfectly covers the whole phenomenology in a straightforward way. However, when the empirical domain is extended to other languages, even inside the Germanic family, various further exceptions are found that require a modification of the standard view.

3 V2 in the Scandinavian Languages The traditional view on V2, which works perfectly well if we restrict ourselves to languages like German and Dutch, has been challenged by work done on the Scandinavian languages. The first problem that emerges in looking at Mainland Scandinavian (from now on MSc) is that in a subset of embedded clauses V2 is indeed possible although the complementizer is present. Vikner (1995: 60) explicitly notices this for Danish: (7) Hun sagde, at vi skulle ikke tage vin med. He said that we should not take wine along “He said that we should not take wine along.” In example (7) it is clear that V2 (shown by the fact that the verb is higher than the negative adverb, while this is not the case in non-V2 clauses) and a complementizer are compatible, and therefore cannot target the same position. In order to explain this cooccurrence two analyses have been proposed: (i) CP is recursive and (ii) V2 targets the IP domain in Scandinavian.

158

9781441124609_Ch10_Final_txt_print.indd 158

11/27/2012 11:06:21 AM

On V2 Types

The second problem emerging from the analysis of Insular Scandinavian like Icelandic (and partially Faroese) is that V2 in Icelandic is “symmetric,” as it is generally allowed also in embedded declaratives even under non bridge verbs like “doubt” (example taken from Rögnvaldsson and Thràinsson (1990: 32)): (8) Jòn efast um að à morgun fari Marìa snemma à fætur. John doubts on that tomorrow will Mary get up early “John doubts that Mary will get up early tomorrow morning.” Embedded interrogative clauses only tolerate V2 if the wh-item is not an argument (like “why”) of the verb; if it is, embedded V2 is also banned in Icelandic. This has been explained by assuming that Yiddish displays exactly the same property (see Santorini (1989)) and again the debate has centered on an analysis in terms of movement to an IP projection or to a (low) CP.2 Authors favoring an analysis in terms of CP projections have shown that the verb crosses high adverb positions (example from Schwartz and Vikner (1996: 65)): (9) Eg held að ì gær hafi Vilhjàlmur örugglega hitt eplið. I believe that yesterday has V. surely hit apple-the “I believe that yesterday V. surely hit the apple.” However, recent work on Northern Norwegian dialects has shown that both verb movement to (a low) C° and to some I° positions are found in Scandinavian (see a.o. Bentzen 2007b). Northern Norwegian displays movement across certain adverbs but not all, showing that the verb stops “halfway” between its original V position and the high C position. Therefore, both analyses are correct depending on the language chosen. However, the problem raised by the fact that at least in Icelandic embedded V2 targets a C° position has been solved by simply admitting that CP recursion is a possible mechanism until work on the left periphery inspired by Rizzi (1997) showed that CP is actually a cover term for a series of projections including ForceP, FinP, and various Topic and Focus projections. On the one hand the idea of a complex left periphery permits a straightforward explanation of cases like the Scandinavian ones mentioned above. On the other hand we lose a very important piece of the original analysis proposed by den Besten (1983), namely that the linear restriction is no longer expected to hold because there is more than one specifier available on top of the C° position where the verb lands. Given that the V2 linear restriction still holds, it cannot be derived from the lack of syntactic space on top of the constituent moved in front of the inflected verb. Actually, there exist V2 languages that do

159

9781441124609_Ch10_Final_txt_print.indd 159

11/27/2012 11:06:21 AM

The Bloomsbury Companion to Syntax

not obey the linear restriction but still have the other two properties. These are the Old Romance languages.

4 Old Romance V2 Old Romance has since long been recognized to share with (modern and old) Germanic languages the V2 character because subject inversion between the auxiliary and the past participle is rather frequent in the old texts, while being completely banned from the modern Romance languages (except for some Rhaetoromance dialects that are still V2, see Poletto (2002)).3 (10) quali denari avea Baldovino lasciati loro. which money had.3sg Baldovino left them (Doc. fior., 437) “How much money Baldovino had left to them.” Following Benincà’s (1984) proposal for Old French and medieval Northern Italian dialects, it is generally assumed that the Old Romance languages were V2, although the root versus embedded asymmetry is not found in Spanish (cf. Fontana (1993)) and Southern Italian varieties. This seems to reproduce the same split between asymmetric and symmetric V2 languages found in Germanic, with German, Dutch, and MSc being like French and Northern Italian dialects, while Icelandic, Faroese, and Yiddish are like Spanish and Southern Italian dialects.4 French and the Northern Italian dialects also display another phenomenon related to the verb position, namely the possibility of omitting the subject pronoun (henceforth pro drop). The modern languages are either pro drop (Italian) or non-pro drop (French) in both main and embedded clauses.5 Old French and Northern Italian dialects are asymmetric in the expression of the subject as they display pro drop only in main clauses, precisely in those contexts in which V to C applies. (11) E così ne provò _ de’ più cari ch’elli avea. and so of-it tested.3sg _ of-the most dear that-he had.3sg (Testi fior., 74) The analysis proposed by Benincà (1984), which has now become standard, is that the verb can only license a null subject from the C position. A third phenomenon related to the V2 character of Old Romance is the so called Tobler-Mussafia law, which states that enclisis of clitic pronouns is enforced when the verb is in first position. This phenomenon is again found only in main clauses and has been traditionally related to the phonological 160

9781441124609_Ch10_Final_txt_print.indd 160

11/27/2012 11:06:21 AM

On V2 Types

weakness of the clitic, which cannot be located in first position. Benincà (1995) shows that the ban against having clitics in first position is not a phonological property, but is related to V2. Benincà’s idea is that proclisis is triggered when a regular V2 structure with a focussed constituent is instantiated and enclisis is instantiated when there is V1 or the element preceding the verb is a Topic. (12) (a) Leggesi di Salamone che … Reads-one of Salomone that

(Novellino, p. 138, r. 1)

(b) Poi  lo fece fuori trarre. then him made.3sg outside take (Novellino, p. 158, rr. 6–7) The fourth phenomenon related to the V2 property is the presence of elements similar to the first position expletives of the type of German es mentioned in Section 2:  la forte fantasia. (Vita nuova 98) (13) … sì mi cessò  so me stopped the strong phantasy “I stopped dreaming.” Although there are good reasons to assume that Old Romance was similar to Old and Modern Germanic in forcing the verb to raise to the left periphery of the clause, it is also well-known that Old Italian did not display the typical “linear restriction” observed in the Germanic domain: in old Italian texts V3 and V4 sequences are common:6 (14) Et dall’ altra  parte Aiaces era uno cavaliere franco. and on the other side A. was a  knight courageous  (BL, Rett., p. 94, r. 7) “And on the other hand A. was a courageous knight.” An illuminating empirical observation is that all exceptions to the V2 restrictions are constituted by one or more Topics occurring in front of a Focus (or wh-item). This is in line with the split CP perspective proposed by Rizzi (1997), and now generally adopted for Romance. The parallel between Old Romance and Germanic is thus based on subject inversion, which becomes the core property defining V2 languages as languages with obligatory V to C movement, since there are both symmetric and asymmetric V2 languages and the linear restriction can also fail to apply. Therefore, including Old Romance in the set of languages that have the V2 property—a hypothesis we are forced to make considering the phenomena seen above— leaves unexplained why the linear restriction is clearly observed by modern 161

9781441124609_Ch10_Final_txt_print.indd 161

11/27/2012 11:06:21 AM

The Bloomsbury Companion to Syntax

Germanic languages, but not by Old Romance (and Old Germanic, see Fuss (2008) on this).

5 Subject Inversion in Cimbrian Cimbrian is a German dialect spoken in a linguistic island in the Italian Alps and has only recently been investigated in its syntactic properties because it maintains very conservative Germanic features and at the same time it seems to have been influenced by the neighboring Romance dialects. Cimbrian still displays V to C, because it displays a clear asymmetry recalling the German one between one type of embedded clauses and main clauses with respect to the negative marker, object clitics, the particle da, verbal prefixes, and has first position expletive z similar to German es (see above). The examples below illustrate the asymmetry on the basis of the position of the negative marker, which occurs after the inflected verb in main clauses, but before the verb in embedded clauses: (15) (a) Dar khimt net. He comes not “He is not coming.” (b) *Dar net khimt. he not comes (16) (a) Dar hat geböllt  azz-e net  vortgea. He has wanted that-I not away-go (b) *Dar hat geböllt azz-e vortgea net. The interesting fact about Cimbrian is that throughout its whole documented history it has never allowed for inversion of a nominal subject, but only for quantifier or subject clitic inversion. Inversion is actually obligatory in all main clauses, but never with nominals (that are generally right or left dislocated): (17) (a) Gestarn  hatt-ar gesekk in has. Yesterday has-he seen a hare (b) *Gestarn hatt dar Giani gesekk in has. Yesterday has the  G. seen  a hare 162

9781441124609_Ch10_Final_txt_print.indd 162

11/27/2012 11:06:21 AM

On V2 Types

This clearly shows that what has been up to now considered as a unitary set of facts has to be split into different phenomena: subject clitics, quantifiers, and nominal subjects do not work in the same way. On the other hand, Bidese (2008) notices that there has indeed been a change in Cimbrian according to the V2 linear restriction: while in the first Cimbrian Catechism from (1602) V2 is always respected (though no cases of DP subject inversion are found), modern Cimbrian does not display the V2 linear restriction and V3 cases are indeed possible. This shows that there is a dissociation between the two properties of subject inversion and linear restriction, as one can change while the other remains stable.

6 Concluding Remarks The standard view based on German and Dutch data that V2 is one single property with three distinct correlates, which always appear together, namely: subject inversion, linear restriction, and main vs. embedded clauses asymmetry is not empirically correct. There are Germanic languages like Icelandic where the asymmetry is not found, and V2 is consistently found in embedded clauses as well. Old Romance languages only have two of the three correlates, namely subject inversion and asymmetry, but no linear restriction, which must be dissociated. Languages like Cimbrian also show a split among the correlates, as the subject inversion property remains constant while the linear restriction is lost. Even more, subtler distinctions have to be made as the very same correlate of subject inversion is a complex phenomenon since not all subject types behave alike. The old idea that V2 can be captured by V to C movement followed by movement of one single constituent to SpecC is also not tenable from a theoretical point of view, as much recent work done on the left periphery of the clause has shown that what used to be conceived as a single syntactic projection, namely CP, is actually an ordered set of distinct positions that host different types of elements. The V2 property can be restated by assuming that in all V2 languages the inflected verb reaches the left peripheral domain, but not all languages target the same position, and according to the position selected, each language will have embedded V2 or not, V3 cases or not, and subject inversion with all subject types or not. Much empirical work remains to be done on other language groups to see whether similar phenomena are found outside the domain of Indo-European languages (but see Bhatt (1999) on Kashmiri) and on language acquisition to see how children acquire the V2 property in general and the V2 type of their target language in particular (see Bentzen and Westengaard (2007) and Guasti in this volume). 163

9781441124609_Ch10_Final_txt_print.indd 163

11/27/2012 11:06:22 AM

The Bloomsbury Companion to Syntax

Notes 1. V2 and one instance of scrambling, namely object shift, interact in an interesting way in the Scandinavian languages, see Ordóñez (this volume: Section 3.1). 2. See Vikner (1995) for a set of arguments showing that also in Icelandic V2 is movement to the C domain. 3. See Benincà (1984) and (2006) for a detailed analysis of the phenomena related to V2 in Romance. 4. This is probably to be related to other phenomena of the area known as Charlesmagne area. See van der Auwera (1998) on this, who introduces the concept of Standard Average European for a set of European languages that share several core syntactic phenomena. 5. The modern Northern Italian dialects have developed a mixed system, where some persons are pro drop like modern Italian, others are not, like modern French (see Poletto (2000) a.o. on this). 6. The possibility of iterating Topics is a typical property of the Old Italian varieties, and is found both in the North and in the South. An example is: in questo Pittagora  sì cominciò . . . in this Pittagora  so began “Precisely then, P. started . . .” (Fiori e Vita de Filosofi, 104) Limited cases of V3, but still more robust than in the modern languages, are also observed in the Germanic domain (see Fuss (2008) on V3 cases in Old High German).

164

9781441124609_Ch10_Final_txt_print.indd 164

11/27/2012 11:06:22 AM

11

Clitics1 Silvia Luraghi

Chapter Overview Introduction Words, Affixes, Simple Clitics, Special Clitics: A Discourse Perspective Between Clitics and Affixes: Romance Clitic Pronouns Clitic and Postpositives in P2 Differences Between Enclisis and Proclisis? Conclusions

165 170 174 179 189 191

1 Introduction Clitics have given rise to an enormous number of studies. In particular, they have been (and still are) an extremely popular topic of research in various versions of generative grammar, including Minimalism, and other formal theories such as Optimality. Even just listing relevant contributions on clitics would take much more than the space allotted for this chapter, so the interested reader is referred to Nevis et al. (1994), and, for more recent publications, to other works cited in the references.2 Here I limit myself to analyzing some aspects of two types of clitics, viz. Romance pronominal clitics and P2 clitics as instantiated in some ancient Indo-European languages, with a focus on Hittite and on Ancient Greek. As remarked above, available treatments of clitics are mostly based on formal syntactic criteria; in addition prosodic treatments have also been worked out. In this chapter, I give a usage-based account of clitics, following an approach that views prosody as reflecting discourse organization, which I outline in Section 2. Section 3 is devoted to some aspects of Romance pronominal clitics, including 165

9781441124609_Ch11_Final_txt_print.indd 165

11/27/2012 11:18:26 AM

The Bloomsbury Companion to Syntax

clitic doubling (Section 3.1), clitic climbing (Section 3.2), and the rise of pronominal clitics out of former unaccented pronouns (Section 3.3). In Section 4 I discuss P2 clitics in two languages that display different tendencies regarding their placement: Hittite, in which P2 positioning of clitics is almost exceptionless (Section 4.1), and Classical Greek, which presents more varied patterns, highly conditioned by discourse structure (Section 4.2). The remainder of this section is devoted to various issues related to clitics, including the difference between “simple” and “special” clitics (Section 1.1), different placement possibilities (Section 1.2), possible types of host (Section 1.3) and order of clitics in clusters (Section 1.4).

1.1 Simple vs. Special Clitics There is little consensus on what clitics are, how they can be classified into types, and even on their existence. According to Joseph (2002), for example, all items identified as clitics in current research on Modern Greek (which are largely similar to Romance clitics discussed in Section 3) are best understood if considered as either words or affixes. In this chapter, I follow the more widespread opinion that clitics are some sort of in-between items, which share some features of bound and some of free morphemes.3 Indeed, clitics are items that cannot stand independently of other (“real”) words in a sentence as they most often do not bear an accent. Prosodically, they are “deficient,” and constitute a prosodic unit, or clitic group, with their host.4 In the meantime, clitics can display a morphological behavior typical of words as, for example, to be inflected in relevant grammatical categories: this is the case of pronouns or auxiliaries, two lexical classes that often belong in the inventory of a language’s clitics. Thus, one way of characterizing clitics is by saying that they are not phonological words, but that morphologically they must be considered words in their own right. Since Zwicky (1977) it is common to distinguish between “simple” and “special” clitics. According to Zwicky’s definition, simple clitics are unaccented variants of “real” words that lose their lexical accents in specific conditions as, for example, the unaccented form [əm] of English third person non-subject pronoun him. On the contrary, special clitics are not simply unaccented variants of lexically accented words: they are lexically unaccented items, which have a distribution of their own, and, crucially, peculiar placement rules which appear to be cross-linguistically limited to a small number of options. Romance pronominal clitics and second position, or P2, clitics best illustrate special clitics in this sense, partly on account of the fact that Zwicky’s definition is mostly worked out in order to accommodate specificities of Romance clitics; as we will see in Section 4.2, some P2 clitics have a somewhat more controversial status as 166

9781441124609_Ch11_Final_txt_print.indd 166

11/27/2012 11:18:27 AM

Clitics

they are lexically accented though confined to second position or at least ruled out from first position.

1.2 Two Options for Clitic Placement Cross-linguistically, clitic placement is determined by two possible tendencies: (a)

clitics attach phonologically to specific constituents (constituent-based, or constituent-hosted clitics); (b) clitics attach to a specific position in a sentence, typically P2.5 Klavans (1985) considers cliticization as a primarily syntactic phenomenon, with phonology only playing a role as a last factor. She considers the domain of cliticization to be determined syntactically, and draws a distinction between the structural host of a clitic, that is, the constituent to which the clitic belongs syntactically, and its phonological host, that is, the constituent to which the clitic attaches phonologically. In Klavan’s terms, type (a) clitics attach to the same constituent both phonologically and syntactically, while P2 clitics may attach to two different hosts (i.e. a phonological host and a distinct structural host at the same time). Among clitics mentioned in Section 1, Romance pronouns belong to type (a), as they always attach to the verb (parameters of possible variation include enclisis vs. proclisis and the position relative to auxiliaries and modal verbs). Other constituent-based clitics include focalizers, which typically attach to the word or constituent that they have in their scope, as Ancient Greek ge in example (1): (1) sù tò prosrēthēsómenon orthôs ámeinon hēdonês ge 2sg.nom art.acc say:fut.ptcp.mid.acc rightly better:acc pleasure:gen.f ptc agathòn eînai   noûn. good:acc be:prs.inf mind:acc “You assert that the good which is rightly to be called better even than pleasure is mind.” (Pl. Phlb. 19d). NP coordinators can be clitic and attach to one of the coordinated items, as the conjunction -que in Latin (see Wanner 1987 on this and other constituent-based clitics in Latin): (2) quae in amicitiam populi Romani dicionem= que rel.nom.pl in alliance:acc people:dat Roman:dat subjecthood:acc and essent. be:subj.impf.3pl “which are allies and subjects of the Roman people.” (Cic. div. in Caec. 66). 167

9781441124609_Ch11_Final_txt_print.indd 167

11/27/2012 11:18:27 AM

The Bloomsbury Companion to Syntax

Possessives also tend to attach to the NP that constitutes their head, as shown in Hittite: (3) nu= us appa ishi= ssi pennai. conn 3pl.acc back master.d/l 3sg.poss.d/l drive:prs.3sg “(S)he takes them (= the oxen) back to their (sg.) owner.” HG § 79 (=Friedrich, 1959). (Note that such Hittite possessives are adjectives, and agree with their head noun in gender, number and case. Possessive clitics realized as genitive or dative forms of pronouns in languages with P2 clitics may display a preference toward NP cliticization as in Serbo-Croatian, see Tomić (1996), but they can also be placed in P2 as in Hittite, see further Section 4.1.) Type (b) clitics are typically enclitic and are placed in P2: apparently, no other dedicated position for clitics is available even though, as I argue in Section 5, a case can perhaps be made for first position, or P1. For such clitics, the sentence is the cliticization domain in terms of Anderson (1993). Anderson further defines two other parameters of clitic placement: dominance and precedence. Dominance specifies whether the clitic attaches to the initial or final hedge of the domain, while precedence specifies whether it precedes or follows the host specified by dominance, that is, whether it is proclitic or enclitic (this parameter corresponds to phonological liaison in Klavans 1985). It follows that P2 clitics all attach to the initial hedge of their domain: they follow the item placed in P1, thus taking the leftmost position for enclitics that, by their nature, cannot stand in first position. Focusing on clitics that can occur as type (a) or type (b) cross-linguistically as, for example, pronouns, it can further be remarked that the two placement strategies can be explained as owing to competing motivations. In the first case, clitics lean phonologically on the constituent to which they belong in terms of construction and argument structure, which helps understand their function (grammatical relation or semantic role) in a sentence. In the second case, clitics tend to occur early in a sentence because of their function in discourse. The two types of motivation are in conflict with one another and drive clitics to attach to different hosts; such competing motivations can also be the cause of change in clitic placement, as discussed in Section 4.2.

1.3 Possible Hosts for P2 Clitics What counts as first position, and consequently how second position must be understood, varies. Wackernagel’s Law is usually stated in terms of words: it is normally said that P2 clitics attach to the first accented word in a sentence. However, this statement needs further refinements. In the first place, languages 168

9781441124609_Ch11_Final_txt_print.indd 168

11/27/2012 11:18:27 AM

Clitics

vary as to whether they admit discontinuous constituents, and accordingly as to whether clitics can be placed after the first word of a more complex constituent. Diesing (2010) divides languages into three types, depending on whether P2 is the position after the first accented word as in Hittite and Old Bulgarian, after the first constituent as in Czech, Slovenian, and Malagasy, or whether it can be both, as in Serbo-Croatian, Ngiyambaa, and Warlpiri (see further Marušič 2008 on Slovenian, and Boškovič 2001 on Serbo-Croatian). Even languages in which discontinuous constituents occur may display a tendency for certain constituents to build one single unit for the purpose of clitic placement. One such language is Classical Greek, in which determiners tend not to be separated from their noun by clitics, as I show in Section 4.2. In addition, evidence from various languages and different language families indicate that P2 clitics can be hosted by words (normally sentence connectives) which are elsewhere considered proclitic. One such example is provided by Hittite, which I discuss in Section 4.1; Halpern (1995: 215) discusses similar evidence from Bulgarian. It appears that in such cases the initial proclitic can receive an accent, or that the complex cluster proclitic+enclitic(s) can count as a clitic group. A possible explanation is that while P2 enclitics are mostly special clitics, initial proclitics in these clusters are simple clitics, which can have an accented variant depending on the prosodic surrounding. In any case, this shows that placement of clitics in P2 cannot simply be accounted for by saying that clitics follow the first accentually strong position in a sentence: this can be the case when clitics follow an autonomous accented word, but when they contribute prosodic material for an otherwise unaccented item one has the feeling that the whole initial unit has a prosodic status of its own. I elaborate on this issue further in Sections 4.1.2 and 5.

1.4 Order of Clitics in Clusters Clitics follow rather strict placement rules in clusters. Though such rules are language specific they often display general tendencies, partly motivated by the scope of co-occurring clitics, partly by other factors. Pronominal clitics typically appear in fixed slots, connected with their morphology or with person. Notoriously, there are restrictions on the co-occurrence of certain pronominal forms in clusters, typically involving second person singular accusative and third person singular dative. Compare the following Italian examples (see Russi 2008 for more examples and discussions): (4) Lo presento a te. / Te lo presento. “I introduce him to you.” (5) Ti presento a lui. / *Gli ti (*ti gli) presento. “I introduce you to him.” 169

9781441124609_Ch11_Final_txt_print.indd 169

11/27/2012 11:18:27 AM

The Bloomsbury Companion to Syntax

Similar restrictions are reported from Modern Greek and Bulgarian, as well as from various non-Indo-European languages. They are reviewed in Haspelmath (2004b), where it is suggested that the motivation is ultimately to be found in discourse-based frequency effects. When clusters include both constituent-based and P2 clitics, the former are expected to precede the latter, and they often do as, for example, in Hittite (Luraghi 1990a and b), Sanskrit (Hale 1987), and other Indo-European languages. However, this is not always the case as Pashto is reported to have P2 clitics inserted between a possible initial verb and verb-hosted deictic clitics (Tegey 1977). Since Pastho also has P2 clitics inserted between verb bases and ancient preverbs no longer analyzable as such, it seems that Wackernagel’s Law can split up prosodic and even phonological words in this language (see Tegey 1977 for a description and Hock 1996 for discussion; a recent account in the framework of LFG is given in Bögel 2010). In chains of P2 clitics it is often the case that discourse particles precede pronominal clitics: arguably, this happens because discourse particles have the whole sentence in their scope, and attach to the left sentence boundary not on account of being P2 clitics but rather because that is their syntactic domain of cliticization (structural host in terms of Klavans 1985). Again, Hittite, Ancient Greek and Sanskrit provide evidence for this tendency (see Section 4), but other orders are also attested as in Pashto (Tegey 1977) and Serbo-Croatian (Radanović-Kocić 1996). Given the limits of this chapter, it is impossible to even attempt here to exhaustively survey all possible orders found in clitic chains. Knowing the origin of certain clitics does not seem sufficient to shed light on specificities of their position in clusters, as argued in Steele (1977); however, diachronic developments can indicate different degrees of grammaticalization reflected in stricter or less strict cliticization. As discussed in Section 4.1, some Hittite particles have enclitic and apparently free-standing variants, which possibly reflect a later development toward cliticization with respect to other items that only have enclitic forms.

2 Words, Affixes, Simple Clitics, Special Clitics: A Discourse Perspective Before moving on to the analysis of specific clitics, the issue concerning the categorial status of these items needs to be addressed. As remarked above, clitics typically bear no lexical accent in spite of displaying various morphosyntactic features typical of free rather than bound morphemes. This characteristic is puzzling only inasmuch as one’s theoretical orientation forces one to work with discrete, rather than fuzzy categories, and if one fails to consider the type of information conveyed by clitics of different sorts, together with 170

9781441124609_Ch11_Final_txt_print.indd 170

11/27/2012 11:18:28 AM

Clitics

their role in discourse. Clitics are mostly items that convey grammatical, rather than lexical information, such as auxiliaries or modal particles; as repeatedly remarked, pronouns also typically have clitic forms or variants. Thus, clitics have a status close to that of bound morphemes: in a language like English, for example, auxiliaries have the same grammatical function as inflectional affixes in morphologically richer languages. In much the same way, clitic pronouns can be compared, and in fact often are, to cross-reference-bound morphology in incorporating languages (Anderson 2003). Still, with respect to such bound morphemes, clitics display some bigger degree of freedom, and their peculiar placement rules are not such that they can easily be equated to bound morphology. As is often the case, looking at things in a diachronic perspective can provide enlightening insights. In such perspective, clitics can be shown to be located on a grammaticalization cline that leads from free forms to bound morphology. This is the case if we consider the development of pronominal clitics in various Romance varieties (see Section 3.3). Clitics that express grammatical categories are likewise undergoing a grammaticalization process that might lead them to eventually become bound morphemes: as well-known, verbal morphology can be shown to have originated in this way in some well-documented instances, such as in the case of the Romance future tense (see Hopper, Traugott 2003). As noted by Bybee: The variation and gradience in the category of “grammatical morpheme” is a direct result of the processes of change that affect morphemes and shape their properties of form and meaning. Lexical morphemes can become grammatical morphemes in the process of grammaticalization (as when the lexical morpheme go becomes part of the future construction be going to), and in this process gradually become more dependent upon and eventually fused with surrounding material. (2010: 4) Thus, wordhood is better understood as a gradience phenomenon, while drawing a sharp distinction between categories looks pointless, as shown by the following quote from Zwicky: This accentual test [i.e. beign unaccented, SL] is probably the most popular rule-of-thumb for distinguishing clitics from independent words, but it is most unreliable; it should never, I think, be used as the sole (or even major) criterion for a classification, though it can support a classification established on other criteria. The test has two problems, one minor and one major. . . . The major problem is that many clearly independent words—e.g. English prepositions, determiners, and auxiliary verbs of English—normally occur without phrasal accent. (1985: 287) 171

9781441124609_Ch11_Final_txt_print.indd 171

11/27/2012 11:18:28 AM

The Bloomsbury Companion to Syntax

In spite of Zwicky’s flat statement, it is not clear at all why, for example, English prepositions are independent words more than other simple (or even special) clitics: on what criterion is wordhood established? This, as well as other similar arguments in Zwicky’s paper, gives a strong impression of circularity. Clitics have in common a low degree of communicative saliency and a high degree of expectedness: for example, clitic pronouns do not introduce new information into the discourse as they rather serve the purpose of tracking referents which have already been introduced; they are highly expected because they are by the most part verb arguments. In a similar way, auxiliaries do not contribute semantic content to the predicate: rather, a typical initial step in the change from main verb to auxiliary is desemanticization. They specify some sort of grammatical category of the main verb, and are expected because they typically occur with a verb form that needs to be specified as a participle or an infinitive in English. In the flow of discourse, items receive different types of accentuation depending on their communicative saliency. This simple fact, which in various forms has been re-discovered over and over again by linguists over decades and even centuries, leads to a possible re-visitation in discourse terms of the distinction between simple and special clitics. As remarked above, following Zwicky, simple clitics are unaccented variants of lexically accented words. Several other authors have pointed out that some of Zwicky’s simple clitics have a different distribution with respect to their accented counterparts, as in the case of English auxiliaries discussed in Anderson (2003), and thus present some characteristics of special clitics. This seems to suggest that the distinction between simple and special clitics is not sharp: as pointed out above, there is a gradience between lexically accented words that get unaccented or are weakly accented in discourse, and lexically unaccented items such as special clitics. This should come as no surprise: in general, linguistic phenomena have a gradient nature, something which is especially clear when one takes a discourse perspective. Following this approach, one can find a unified explanation for other unaccented or weakly accented items as well. Indeed, lexical items that are not typical clitics also tend to lose their accent in discourse: for example, the finite verb is usually unaccented in English (or weakly accented in comparison to its arguments), unless it is focused. Compare: (6) (7)

Jóhn eats cákes. Jóhn éats cákes (rather than baking them, for example).

Indeed, as is known at least since Wackernagel’s classic paper on P2 clitics (Wackernagel 1892), the Sanskrit finite verb in main clauses displayed some features of clitics without being a clitic stricto sensu. In sum, words are given a different prosodic relevance depending on their discourse saliency; some items that very frequently happen to have low saliency, such as pronouns, may 172

9781441124609_Ch11_Final_txt_print.indd 172

11/27/2012 11:18:28 AM

Clitics

develop special variants that occur when they do not need to be given accentual relevance. Eventually, they may also become bound morphemes and develop into agreement markers. As noted above, the distinction between simple and special clitics is gradient: while Romance pronouns are certainly special clitics, English auxiliaries mostly conform to the definition of simple clitics in spite of having some features of special clitics. As we will see, among P2 clitics, too, there are items that conform more or less closely to Zwicky’s definition of special clitics. In the perspective adopted here, accentuation and deaccentuation are emergent phenomena, which happen in actual events of communication. Items that have lexically accented and lexically unaccented variants are those for which both variants are needed in discourse (such as pronouns); some items only have unaccented forms because they never need to be given prominence (some modal particles, for example), other only have accented forms, because they are virtually always salient enough to be accented ( for example personal names). Following such an approach, even the issue regarding the status of clitics as free or bound morphemes loses relevance: clitics are items that frequently co-occur with other items or in special positions; as such they are “chunked” together with their hosts, and may eventually end up being reinterpreted as bound affixes as a consequence of chunking. Chunking is a fundamental principle of language production and processing. As pointed out by Bybee (2010: 33–4), “[w]hen two or more words are often used together, they also develop a sequential relation . . . The principal experience that triggers chunking is repetition.” Thus, frequent collocations lead to entrenchment of constructions and low saliency items tend to acquire a bounded status in construction: a first hint of this is their close prosodic connection with more salient items. Prosody is an important reflex of discourse organization not only in terms of alternation between what is accented and what is not but also because prosodic units reflect information structure in sentences. In this respect, intonational phrases are especially relevant because they represent units of information; extraposed topics or focused constituents can best be indicated in discourse in terms of intonational phrases. As Nespor and Vogel (2007: 187–8) remark, intonational phrases are highly variable. In fact, a sentence is normally a single intonational phrase, but, depending on the style, and especially on the discourse relevance of specific constituents, it can be split into more intonational phrases, as in (example from ibid.: 194): (8) (a) [IMy friend’s baby hamster looks for food in the corners of its cage]I (b) [IMy friend’s baby hamster]I [Ilooks for food in the corners of its cage]I (c) [IMy friend’s baby hamster]I [Ilooks for food]I [Iin the corners of its cage]I 173

9781441124609_Ch11_Final_txt_print.indd 173

11/27/2012 11:18:28 AM

The Bloomsbury Companion to Syntax

As we will see in Section 4.2, this and other prosodic notions can partly account for the placement of P2 clitics in Classical Greek.

3 Between Clitics and Affixes: Romance Clitic Pronouns Romance pronominal clitics are perhaps the best known and thoroughly described group of special clitics, and research regarding their status and behavior in different Romance varieties, including different historical stages, abound. They are typically hosted by the verb, and, in the modern Romance languages, are most often proclitic; they may be enclitic with some non-finite verb forms and imperatives. Clitic pronouns are not only direct and indirect objects but can also indicate various local relations and beneficiary (dativus ethicus); however, their interpretation is always dependent on the verbal predicate: in this sense, the verb constitutes their domain.6 Some examples from Italian are below:  (9) Non me ne è venuto niente di buono. not me(cl) from.it(cl) is come nothing of good “Nothing good has come out of this for me.” (10) Paolo va a scuola tutti i giorni, Maria ci va quando le pare. Paul goes to school all the days Mary there(cl) goes when her(cl) seems “Paul goes to school every day, mary goes when she likes to.” (11) Mia sorella ci ha  cucinato una torta, sua cugina ne ha cucinate due. my sister us(cl) has cooked a cake her cousin part(cl) has cooked two “My sister baked a cake for us, her cousin cooked two.”

3.1 Clitic Doubling Some Romance varieties, most notably French, also feature subject clitics. In standard French, such clitics are in complementary distribution with full NPs or accented pronominals, as shown in the following example: (12) Il m’ a donné un livre. / Jean m’ a donnè un livre. he(cl) me(cl) has given a book / John me(cl) has given a book “He gave me a book. / John gave me a book.” In some other varieties, including the Norther Italian Gallo-Italic vernaculars, subject clitics have gone a step further in grammaticalization, as they are no longer omittable and co-occur with possible subject NPs as, for example, in Piedmontese: 174

9781441124609_Ch11_Final_txt_print.indd 174

11/27/2012 11:18:28 AM

Clitics

(13) It vade a  ca. / Ti’ t vade a ca. / *vade a ca. you(cl) you-go to home / you you(cl) you-go to home / you-go to home “You are going home.” Clitic doubling provides evidence for increasing grammaticalization of Romance pronominal clitics, and constitutes an argument in favor of an analysis that considers them close to bound morphemes. Note however that Romance clitics have a bigger autonomy than bound morphemes even in varities that display clitic doubling: for example, the Northern Italian vernaculars all allow subject clitic drop at least to some extent.7 This ongoing tendency is possibly a consequence of the increasing role of standard Italian as a prestige variety even among speakers of vernaculars. It seems to indicate that such clitics, in spite of the high degree of boundedness reached, never really achieved the status of bound morphemes, and are still analyzed by speakers as featuring a lower degree of boundedness than, for example, verbal endings.8 Clitic doubling is not limited to subjects: in Spanish, for example, recipients are most often doubled by clitics, even though this is not obligatory, as shown in:9 (14) Le doy el libro a  Juan. / Doy el libro  a  Juan. him(cl) I-give the book to John / I-give the book to John “I give John the book.” A construction that triggers clitic doubling in many Modern Romance languages is left dislocation of topicalized constituents, as in:10 (15) Il giornale l’ ho letto   dopo  cena. the newspaper it(cl) I-have read after dinner “I read the newspaper after dinner.” In Modern Italian, clitic doubling is obligatory in such sentences, which are uttered without intonational breaks between the left dislocated constituents and the rest of the sentence (i.e. they normally constitute a unique intonational phrase). Such occurrences show that clitic doubling serves the purpose of indicating an initial direct object which is not focused; it can be considered a way to deal with free word order by means of a head marking strategy, rather than by means of the dependent marking strategy (case) typical of Latin.

3.2 Clitic Climbing Another remarkable phenomenon connected with Romance clitics is climbing. In cases in which an infinitive is governed by a restructuring verb (modal or 175

9781441124609_Ch11_Final_txt_print.indd 175

11/27/2012 11:18:28 AM

The Bloomsbury Companion to Syntax

quasi-modal), clitics syntactically hosted by the infinitive should attach to it as enclitics; but often they are realized as proclitic and precede the governing verb, as in: (16) Incominciava a  dirmelo. / Me lo incominciava a dire. s/he-started to tell me(cl) it(cl) / me(cl) it(cl) s/he-started to tell “S/he started telling me that.” Following the usage-based approach adopted here clitic climbing can have two explanations: in the first place, it indicates at least an incipient degree of auxiliarization on the side of the governing verb. While the sentence above is not such that one would easily find it in the written language, with modal verbs, which are much closer to auxiliaries, climbing is the norm. In the following example the verb sapere, “know” has the modal meaning “be able to,” and favors climbing in all registers: (17) Non te lo saprei dire. / Non saprei dirtelo. Not you(cl) it(cl) I-would.know say / not I-would.know tell you(cl) it(cl) “I wouldn’t be able to tell you.” The second reason why clitics should tend to climb is connected with their information status: as already noted, clitic pronouns convey low saliency information, already shared by the speaker and the hearer. For this reason, they tend to appear early in the sentence, or at least at the left hedge of their phonological host. Not surprisingly, this is also the motivation for P2 pronominal clitics to occur where they do, as I argue in Section 4. In much the same way as subject clitics drop in Northern Italian vernaculars, clitic climbing also indicates a looser degree of coalescence of clitics with their hosts than the degree of coalescense displayed by bound affixes with respect to their lexical base. Again, clitics retain their in-between status and defy attempts at rigid categorization.11

3.3 From Latin Weak Pronouns to Romance Clitics In Latin, pronouns that did not convey new or salient information were most likely simple clitics, just as English unaccented variants of personal pronouns. As Latin is basically a free word order language, it is not as easy as for English to compare their distribution with that of accented pronouns. Diachronic studies on the development of Romance clitics from unaccented or weakly accented Latin pronouns mostly focus on first and second person forms (see for example, Wanner 1987 on Romance languages in general, and Salvi 2004 on Italian); for this reason, in what follows I describe the evolution of third-person pronouns.

176

9781441124609_Ch11_Final_txt_print.indd 176

11/27/2012 11:18:28 AM

Clitics

Even though sentence initial position was normally reserved to topicalized or contrasted constituents in Latin (see Spevak 2010), it is not completely clear that unaccented or weakly accented versions of pronouns, especially in the case of third person, could not stand in initial position. Consider the following examples from Sallust: eum natura mortalium avida  imperi. (18) Terrebat frighten:impf.3sg 3sg.acc nature:nom man:gen.pl eager:nom power:gen “(Though he was at first gratified with these circumstances, Micipsa, considering that the merit of Jugurtha would be an honor to his kingdom, yet, when he reflected that the youth was daily increasing in popularity, while he himself was advanced in age, and his children but young, he was extremely disturbed at the state of things, and revolved it frequently in his mind.) The very nature of man, ambitious of power, gave him reason for apprehension.” Sall. Iug. 6.3; omnis praeter Turpilium inter epulas obtruncant. (19) Eos 3pl.acc all.acc except Turpilius:acc among banquet:acc.pl slaughter:prs.3pl “(When the time came, they invited the centurions and military tribunes, with Titus Turpilius Silanus, the governor of the town, to their several houses.) They butchered them all, except Turpilius, while they were dining.” Sall. Iug. 66.3. Indeed there appears to be no difference in the information status of eum in (18) and eos in (19), and if one were to translate these passages into a Romance language one would use clitics in both cases: spaventava la natura umana stessa, avida di potere. (20) Lo him(cl) frightened the nature human itself eager of power massacrarono tutti, tranne Turpilio, mentre cenavano. (21) Li them(cl) butchered all except Turpilius while  they-dined In the light of the above examples, it is hard to formulate rules regarding the placement of unaccented pronuns in Latin, even though they did indeed most frequently appear close to the left hedge of the sentence. Very often they turned up occurring after the first word (often a conjunction) as in (18), however they could as well appear in initial position as in (19). Clearly, this tendency reflects the information status of weak pronouns: such pronouns convey shared information, which connects a sentence to the preceding discourse. As a consequence, they tend to occur early in order to fulfill their connecting function. In addition, there is a tendency for Latin unstressed pronouns, when they do not follow a

177

9781441124609_Ch11_Final_txt_print.indd 177

11/27/2012 11:18:28 AM

The Bloomsbury Companion to Syntax

connective, to follow a discourse prominent constituent, as pointed out in Adams (1994), who accounts for the placement of such pronouns in terms of sentence focus (and indeed this could be the case for eum following terrebat in 18). Beside overt pronominals, Latin also had null referential anaphoras of nonsubject constituents (Luraghi 1997). They occurred under various circumstances, partly discourse conditioned, partly connected to specific constructions. One such construction is sentence coordination, in which null anaphora of the direct object is obligatory in the second clause if the direct object is the same as in the first, as shown in: filiam Nero  ultra (22) natam sibi ex Poppaea born:acc refl.dat from Poppaea:abl daughter:acc Nero:nom beyond mortale gaudium accepit appellavit= que Augustam. human:acc joy:acc receive:pf.3sg name:pf.3sg and Augusta:acc “Nero welcomed a daughter from Poppaea with more than human joy and named her Augusta.” Tac. Ann. 15.23. With third-person objects, the few occurrences of an overt anaphora (a form of the third-person pronoun is) in the second conjunct of a coordination construction present special pragmatic conditions. Most often, the anaphora is placed in initial position and hosts the clitic conjunction—que, a hint to the fact that it is accented (see Luraghi 1997 for examples and discussion). Thus, Romance clitics did not only substitute Latin unaccented variants of pronouns but also null non-subject pronouns. To sum up, Latin had nothing comparable to Romance clitics, not only on account of the fact that unaccented pronouns in Latin were not special clitics but also because their position was apparently unconstrained: they display a tendency to occur early in the sentence on account of low information load, and most often they did indeed occur after the first word but were not limited to this position; in addition, Latin could feature zero forms which were later replaced by overt clitic pronominals. When we turn to the Old Romance varieties, the scene changes completely: clitics are already there, and their position is highly constrained, as they have to be adjacent to the verb.12 By default, clitics of the Old Romance languages are proclitic, but they follow the verb and become enclitic if the verb stands in sentence initial position. This placement rule is known as Tobler Mussafia’s Law, and is often regarded as a heritage of Wackernagel’s Law (see, for example, Salvi 2004). However, as we have seen above, there is evidence that Latin weak pronouns were not banned from sentence initial position.13 A possible scenario then would be that on their way to become special clitics, such weak forms of pronouns lost their relative placement freedom and became enclitics. As the position after the first word in a sentence was already the most frequent one for weak pronouns, 178

9781441124609_Ch11_Final_txt_print.indd 178

11/27/2012 11:18:29 AM

Clitics

they became P2 clitics. I am not going to elaborate further on this issue, but it still needs to be stressed that this possible development has to be combined with increasing tendency of clitic pronouns to become verb based.14 Salvi (2004: 178) also suggests that the disappearance of null direct objects was one of the triggers of the development of clitics. This may well be the case: null objects were one of the features of non-configurationality displayed by Latin and by the ancient Indo-European languages in general, as shown in Luraghi (2010a). In Old Italian, null objects could still occur in cases of VP coordination and even sentence coordination (Luraghi 1998; Egerland 2002), while in Modern Italian object clitics can be omitted only in case of verb coordination and are felt as marginally acceptable by some speakers. As already remarked in Section 3.2, clitics provide a head marking strategy to signal grammatical relations, which partly substitutes the dependent marking strategy of Latin. This has been argued by several authors especially with regard to frequent clitic doubling in spoken French, see Bossong (1998) and Luraghi (2010a) for discussion.

4 Clitics and Postpositives in P2 In the course of the nineteenth century, Indo-European scholarship repeatedly pointed out that the ancient Indo-European languages featured an array of unaccented or weakly accented items that consistently occurred in second position in sentences. It was Jacob Wackernagel who best described what is now known as Wackernagel’s law in his 1892 article15 in which he also offered a possible explanation for the Germanic V2 phenomenon.16 Wackernagel mostly relied on data from Homeric Greek and Vedic Sanskrit, two languages in which the tendency for clitics to be placed in P2 is quite well attested, but that also display a number of other patterns, as I will discuss below. A couple of decades later, when Hittite was deciphered in 1916, new and more compelling evidence appeared. Indeed, Hittite is cross-linguistically one of the best examples of strict obedience to Wackernagel’s Law as it features long clitic chains in P2 and shows almost no exceptions. For this reason I start by illustrating the Hittite data.

4.1 Clitic Chains in Hittite Let us start by examining the following passage from a Hittite text: (23) (a) nu= mu memir conn 1sg.obl speak: pret.3pl

179

9781441124609_Ch11_Final_txt_print.indd 179

11/27/2012 11:18:29 AM

The Bloomsbury Companion to Syntax

(b) paiueni= war=an= kan kuennumeni. go: prs.1pl ptc 3sg.acc ptc kill: prs.1pl (c) nu= wa=tta  SAG.DU-an utumeni. conn ptc 2sg.obl head:acc bring: prs.1pl “They told me: ‘we shall kill him and shall bring his head to you.’,” StBoT 24 ii 24–6 (=Otten, 1981). In all three sentences we find clitic pronuns in P2: -mu 1sg.obl, -an 3sg.acc, and -ta 2sg.obl. In addition, (23b) and (23c) also feature the direct speech particle -wa(r), and (23b) contains one of a set of so-called local particles, -kan. Other frequently found items in P2 include the modal particle, the reflexive, and some sentence connectives.

4.1.1 Slots in the Clitic Chain Below I give a chart of Hittite P2 clitics and their position relative to each other (see Luraghi 1990a and 2000 for further details): a. sentence connectives and conjunctions: -(y)a-, coordinator; -ma-, -a-, adversative particles, man- modal particle (which also has a variant that can occur sentence-initially and host other encltics; it can co-occur with the connective -ma-.); b. -wa(r)-, direct speech particle; c. nominative or accusative of third person pronoun singular or plural; d. oblique forms of first and second person singular and plural or dative of third person singular or plural (in the plural dative enclitic, pronouns normally precede possible nominative or accusative enclitics); e. -z(a)-, reflexive particle; f. -kan, -(a)sta, -san, -an, -(a)pa, so-called local particles. Regarding connectives in slot (a), it must be noted that they are in complementary distribution with the sentence introducer nu (and with two other such connectives, ta and su, mainly found in Old Hittite). When present, the latter particle hosts clitics, as in (23a) and (23c). I return to the prosodic status of this particle in Section 4.1.2. The modal particle man figures in slot (a) mainly on account of its non-co-occurrence with nu, but, as noted, it can co-occur with the enclitic connective -ma, and it can stand clause intially and host other clitics. Thus, P2 cliticization in the case of man seems to be a recent phenomenon, as the particle retains a double status: in practice, we can regard it as a simple clitic, rather than as special one as other P2 clitics. When standing in initial position, the particle may well have been proclitic, similar to the connective nu, as discussed in Section 4.1.2.

180

9781441124609_Ch11_Final_txt_print.indd 180

11/27/2012 11:18:29 AM

Clitics

The direct speech particle -wa(r)- in slot (b) derives from a grammaticalized form of a verb of speech, cognate to the root of English word and Latin verbum (Luraghi 1996). It does not have clitic counterparts in the other ancient Indo-European languages; however, the tendency for verbs of speech to occur after the first word inside a stretch of reported direct speech is well attested. Compare Latin: (24) Hic . . . “Nihil,” inquit, “de eorum sententia dicturus dem.nom.m nothing:acc tell:prs.3sg of 3pl.gen opinion:abl say:ptcp.fut.nom sum . . . be:prs1sg “This one said: ‘I am not going to talk about the opinion of those . . . ‘ “ Caes. G. 7.77. . . . Comparative evidence points toward late grammaticalization and cliticization of this particle, which must have changed into a P2 clitic after P2 had already been established as the locus of enclisis at least for pronouns, which occur in P2 in the other ancient IE languages as well. The grammaticalization of this verb form into a particle casts interesting light on clitic climbing as attested with serialized motion verbs. An example is paiueni “we go” in (23b).17 In example (23), I have not given an independent translation of this verb form. Indeed, this verb functions as a sort of auxiliary with respect to the second verb form in the same sentence, kuennumeni “we kill.” Remarkably, it hosts the third person accusative clitic -an, which syntactically belongs with the second verb and is its direct object, giving rise to a constuction that has been shown to reflect a process of serialization of motion verbs (Luraghi 1993). Serialized motion verbs can occur in first position as here or in “post-P2” or “post-Wackernagel” position, and be preceded by the sentence introducer nu and the clitic chain, as in, (25) (a) nu= mu= kan AMA- ŠU menahhanda para naista conn 1sg.obl ptc mother-his against towards send:pret.3sg (b) n= as= mu uit  GÌR.MEŠ-as katta haliyattat. conn 3sg.nom 1sg.obl come pret.3sg foot-pl:d/l.pl down fall: pret.3sg “He sent his mother to me: she fell to my feet.” AM 70.28–9 (Goetze 1933) In (25), uit “s/he came” follows the clitic chain that hosts the complements of the verb haliyattat, and is placed in the first available position after the clitic chain. Similar to the modal particle man, serialized motion verbs have sentence initial and (post-)P2 variants and thus qualify as simple clitics; a possible further 181

9781441124609_Ch11_Final_txt_print.indd 181

11/27/2012 11:18:29 AM

The Bloomsbury Companion to Syntax

step in their grammaticalization process (if the Hittite language had not disappeared) might have been to develop invariable forms and become special clitics, as the particle -wa(r)- (for further details and more examples see Luraghi 2000: 49–51; on the possible meaning of serialized motion verbs, which is still debated, see Hoffner, Melchert 2008: 324–9; on the origin of the construction, see van den Hout 2010). Slots (c)–(e) of the clitic chain host pronouns and reflexives; some of them have precise cognates among P2 enclitic pronouns of other IE languages. Slot (f) hosts the so-called local particles, a category that has been puzzling Hittitologists for a long time, as they have no straightforward counterpart in the other IE languages. In New Hittite, local particles (notably the particle -kan, as the other are rather typical of the older language) tend to express aspectual meanings; in Old Hittite, however, their original function as spatial deictic was still clear. As discussed especially by Boley (see Boley 1989, 2000), these particles resemble Romance adverbial clitics: noting that—asta tends to co-occur with source expressions, Boley compares it to Italian ablative clitic ne “from there” (cf. (9)); similarly, -san, which often co-occurs with locatives, can be compared with Italian locative clitic ci (cf. (10)). In Luraghi (2001) it is argued that these deictics are etymologically related to the PIE preverbs, which, similar to what happened in Sanskrit, were renewed by phonologically heavier adverbs. Note that the change from spatial deictics to perfectivity markers is consistent with this scenario and indeed represents a common evolution of preverbs in the Indo-European languages and cross-linguistically (see Pinault 1995; Lazard 1995).

4.1.2 Trends in Clitic Clustering Summarizing, we find two types of item in the Hittite initial chain: (i) sentence connectives or other particles that indicate sentence type (modal particle, direct speech particle), and (ii) personal pronouns and deictics. Interestingly, the items in (i) precede those in (ii): far from being a feature of Hittite, this is the same distribution found in the other ancient IE languages, such as Sanskrit and Ancient Greek (see Section 4.2 for Ancient Greek; on Sanskrit, see Hock 1996 and Schäufele 1996).18 The two types of item have different scope; accordingly their placement in P2 has different motivations. Items in (i) have the whole sentence in their scope, so their occurrence at the sentence boundary is not surprising. Indeed, they occur as early in the sentence as possible for enclitic elements, and attach to the left hedge of their host. In terms of Klavans (1985), sentence particles attach to the same host both syntactially and phonologically. Pronouns and deictics have the VP in their scope. They are functionally similar to Romance clitics, but tend to move to the left because, as pointed out in the case of clitic climbing in Italian, they refer to non-focal, shared information. Again, using the structural analysis in Klavans (1985), P2 pronominal and spatial deictic enclitics

182

9781441124609_Ch11_Final_txt_print.indd 182

11/27/2012 11:18:29 AM

Clitics

have the VP as their structural host, but take the sentence as their phonological host. The fact that pronominal clitics in Hittite do not attach to the verb and hence do not tend to become verbal affixes does not indicate a lower degree of grammaticalization. As argued in Luraghi (1998), clitic chains, typical of Hittite and in general of Anatolian at all language stages, tend to acquire more members, become longer, and more frequent after Old Hittite. This happens through a variety of developments such as increasing obligatoriness of the reflexive particle, which takes over many of the functions of a reflexive middle, grammaticalization of the spatial particle -kan as an indicator of telicity, demise of NP based clitic possessives, which are substituted by P2 dative clitics (an example of the latter usage is -mu in (25b), to be compared with the NP hosted possessive -ssi in (3)). When a constituent needed to be left dislocated, clitic chains served the purpose of isolating it by creating a prosodic boundary between the left dislocated constituent and the rest of the sentence. In case no constituent needed to be left dislocated, the sentence introducer nu was needed just for the purpose of hosting clitics. This particle was most likely itself unaccented, as shown by its tendency to become phonologically fused with following enclitics when they started with a vowel, as in nu+as = nas (cf. ex. 25b). Note however that, when occurring without clitics, nu was separated in writing from the next word: in other words, possible proclisis was not noted in writing, while enclisis always was.19 In sentences introduced by nu and a clitic chain, one cannot speak of a properly accented constituent in first position: the whole clitic group is rather likely to be a weak element, perhaps even proclitic to the next accented word. In conclusion, I would like to suggest that Hittite featured two types of clitic groups: one, with an accented word as the host of the clitic cluster, which served some special discourse function by introducing discontinuity into the text,20 and another, with a weakly accented clitic group hosted by a connective (or possibly by the modal particle man; see Section 3.1.1), which appeared in the context of discourse continuity. Discontinuity was introduced into the text by fronting a specific word and separating it from the rest of the clause by a prosodic boundary: this happened with left dislocated topics, but also with preposed local adverbs, initial verbs,21 or other items that did not stand in their habitual position. Very frequently, such clitic clusters included the adversative connective -ma, whose function was precisely to indicate textual discontinuity or unexpected events (see Luraghi 1990a for a thorough discussion of word order types and sentence connectives in Hittite).

183

9781441124609_Ch11_Final_txt_print.indd 183

11/27/2012 11:18:29 AM

The Bloomsbury Companion to Syntax

4.2 Ancient Greek Clitics Ancient Greek is peculiar in having an extremely free word order, mostly sensitive to discourse organization. For this reason, tendencies in clitic placement, as well as in the placement of other pre- and postpositives (see below for explanation) have been regarded as strikingly in contrast with free positioning of accented constituents. What has attracted the attention of scholars for over a century is not only the regularity for certain items to always be excluded from first position and their frequent appearance in second position but also the fact that, in spite of this regularity, P2 clitics often seemed to “skip” some initial element. Prosodic descriptions of this phenomenon reach back at least to Fraenkel’s seminal study on the relation between clitic placement and cola, units defined in prosodic terms (see Fraenkel 1932, 1933 and Goldstein 2010 for discussion). In this section, I give a short and necessarily schematic description of clitic placement in Ancient Greek, and show how discourse factors favored the demise of P2.

4.2.1 P2 Clitics Outside P2 As already noted by Wackernagel, in Greek and Sanskrit certain words could be skipped by P2 clitics, with the consequence that the ensuing word counted as first. Such words were in the first place subordinators and sentence connectives (e.g. Sanskrit athā, see Luraghi 1990b; Hock 1996), as shown in the following examples: (26) kaì líēn se páros  g’ oút’ eíromai oúte metallô. and frequently 2sg.acc before foc neg ask:aor.1sg.mid neg inquire:aor.1sg “In the past I have not been accustomed to inquire nor ask you.” Hom. Il. 1.553; (27) hòs tóte mén moi hupéskheto kaì  katéneusen. rel.nom once ptc 1sg.dat promise:aor.3sg.mid and bow:aor.3sg “Of old he promised me, and bowed his head thereto.” Hom. Il. 9.19. Example (27) features two types of P2 items, that is, the enclitic pronoun moi (1sg.dat) and the postpositive discourse particle mén. Such particles are never found sentence intially, hence the name of postpositive used to refer to them (e.g. Dover 1960); many of them, such as mén, always bear a graphic accent. As accents have been added in Greek text only at a later time, it is not clear what value one should attribute them. Limited evidence can be adduced from Myceanean regarding the unaccented status of dé (Bartonek 2003: 128); note however that the distribution of postpositives in Classical Greek is not the same as the distribution of enclitic pronouns. In particular, discourse particles such as 184

9781441124609_Ch11_Final_txt_print.indd 184

11/27/2012 11:18:29 AM

Clitics

mén, dé, gár frequently occur between the definite article and the noun it determines, while such placement is quite exceptional for enclitic pronouns.22 More in general, the tendency for such discourse particles to follow the first word in a sentence is much stricter than for enclitic pronouns, which can instead occur later, after pragmatically relevant prosodic units. Thus, a typical pattern found in Classical Greek prose is the folowing: (28) ho gár toi paîs me ho Sáturos apédra. art.nom ptc ptc child.nom 1sg.acc art.nom Satyros:nom run.away:aor.3sg “The boy Satyros ran away from me.” Pl. Prot. 310c3. In (28) the particles gár and toi and the pronoun me are in two different types of P2: after the definite article (first word) and after the determiner phrase ho paîs (first constituent). Such usage has the consequence that a sentence may feature more than one second position, depending on how the first position is defined. Very often, it is not only the article that hosts sentence particles: the double P2 can occur with all sorts of lexical items, as in: (29) áneu gàr episēmou oú sphi  nómos estì  ékhein without ptc image:gen neg 3pl.dat custom:nom be:prs.3sg have:prs.inf skêptron. staff:acc “Indeed, it’s not their custom to carry a staff without an image.” (Hdt. 1.195.2) As in the case of Hittite clitics discussed in Section 4.1, Greek clitics and postpositives arguably attach to P2 for different reasons: discourse particles are sentence introducers, and they occur at the leftmost hedge of their host; on the other hand, enclitic pronouns move there because of their low discourse saliency, but, contrary to Hittite, they can also occur elsewhere (Luraghi 1990b). Now, the interesting issue is for what reason they should do this. As argued in Luraghi (1990b) and discussed at length in Goldstein (2010), parts of sentences that occur to the left of the domain of cliticization as in (29) are left dislocated for specific pragmatic reasons: in this case, Goldstein (2010: 154) shows that the phrase áneu gàr episēmou is focused. Goldstein further argues convincingly that what can be separated cannot always be defined in syntactic terms as a constituent, but is best regarded as an intonational phrase.

4.2.2 Prosodic Boundaries Inside Sentences Similar to Hittite, in Ancient Greek, too, P2 enclitics and postpositives created a boundary inside a sentence by isolating a part of it through a cluster of prosodically deficient elements. In Luraghi (1990b) it is further argued that 185

9781441124609_Ch11_Final_txt_print.indd 185

11/27/2012 11:18:29 AM

The Bloomsbury Companion to Syntax

the occurrence of postpositives and enclitics in more than one position within the same sentence had the effect of creating more than one boundary. In this section, I show how such sentence internal boundaries operate by examining a frequently cited passage which has never been explained in a satisfactory manner. Consider the position of the clitic min (3sg.acc) in (30): (30) (a) eselthoûsan dè kaì titheîsan tà heímata go.out:ptcp.aor.acc.f ptc and put:ptcp.prs.acc.f art.acc.pl clothe:acc.pl ethēîto ho Gúgēs. see:impf.3sg.mid art.nom Gyges:nom “When she had come in and was laying aside her garments, Gyges saw her; (when she turned her back upon him to go to bed, he slipped from the room).” (b) kaì hē gunē eporâi min exiónta. and art.nom woman.nom see:prs.3sg 3sg.acc go.out: ptcp.prs.acc “The woman glimpsed him as he went out. (But when she realized what her husband had done, though shamed, she did not cry out or let it be seen that she had perceived anything, for she meant to punish Candaules.)” Hdt 1.10.2. Two different explanations have been given for the occurrence of min in (30) viz. that the verb eporâi is given prominence, or that the NP hē gunē is extraposed for the sake of creating suspense. Goldstein (2010: 15, 145) discusses both. Regarding the first interpretation, he points out that by focusing on the possible pragmatic status of the verb it fails to recognize that the important fact about this occurrence is that the intonational phrase hē gunē is extraposed; the clitic then is placed at the left hedge of the second intonational phrase and it is just by coincidence that the actual word that hosts it is the verb. Proponents of the second explanation claim that the fact that the woman saw Gyges was unexpected, and that by left dislocating the constituent and delaying the P2 clitic Herodotus creates suspense in the narration. Goldstein counters this explanation by quoting a similar passage: (31) kaì hē Ámēstris punthánetaí min ékhousan. and art.nom Amestris:nom learn:prs.3sg 3sg.acc have:ptcp.prs.acc “(But as he [sc. the king] could not move her, he gave her the mantle; and she, rejoicing greatly in the gift, went flaunting her finery.) Amestris heard that she had it. (But when she learned the truth, it was not the girl with whom she was angry. She supposed rather that the girl’s mother 186

9781441124609_Ch11_Final_txt_print.indd 186

11/27/2012 11:18:30 AM

Clitics

was guilty and that this was her doing, and so it was Masistes’ wife whom she plotted to destroy).” Hdt. 9.110.2. In (31), Goldstein argues, the suspense effect is not needed, as the fact that her husband’s lover had the mantel was exactly what she expected. However, this passage must be considered in a wider context, as I will argue after discussing (30). In the first place, the occurrence of min in the above examples also deserves to be considered in a wider framework, and to be explained, as opposed to the possible occurrence of a null object. In very much the same way as we have seen above for Latin (Section 3.3), referential null objects occur in Ancient Greek. Indeed, as discussed in Luraghi (2003), the above examples represent a typical context in which discourse conditioned null objects could occur: this is made clear in (30a) and (30b). In this passage both participants referred to, the woman (Candaules’ wife) and Gyges, are topical; however, in the first part of the example we find a null object referring to hē gunē while in the second part ho Gúgēs is resumed by the clitic min. Note that in both sentences predicative constituents referring to the direct object also occur. As indicated in Luraghi (2003), the difference is indeed connected with the second event being unexpected, and not simply because the woman did not expect to see someone looking at her while she was undressing, but rather because her husband Candaules, who had arranged for Gyges to see her, had not thought of the consequences. In this sense, the event in (31) is also unexpected because the king’s lover who insisted on receiving the mantel had miscalculated the possible consequences. Note that the following context is remarkably similar in both passages. In both cases, the reaction of the woman who unexpectedly discovered someone else’s misdeed had quite dramatic consequences on the person who suffered them. Thus, not only the position, but the very occurrence of the clitic min depends on the unexpectedness of the event. In both examples the subject, which refers to the person who discovers something that should have been concealed, is left dislocated. The occurence of the clitic after the verb identifies the two parts of the sentence as distinct intonational phrases, while in normal conditions there would have been no need to split up the sentence into two intonational phrases. The clitic group constituted by the verb and the clitic stands out as indicating a boundary inside the sentence, which signals that the event marks a turning point in the life of a previously mentioned highly topical participant.

4.2.3 The Demise of Wackernagel’s Law As well-known, Modern Greek features pronominal clitics hosted by the verb. According to some scholars, including Wackernagel (see Goldstein 2010: 11–13 for further reference), the first traces of the tendency for clitics, mostly pronominal, to leave P2 and gravitate around the verb can be detected in Classical 187

9781441124609_Ch11_Final_txt_print.indd 187

11/27/2012 11:18:30 AM

The Bloomsbury Companion to Syntax

Greek prose already. According to Goldstein (2010), this is not the case. In his interpreatation of the data, clitics can be shown to attach to the leftmost word in an intonational phrase; when they are not sentence initial, it is because one or more other intonational phrase(s) have been “skipped” or extraposed to the left for pragmatic reasons, being topical or focal. It may well be the case that Goldstein’s explanation is correct: however, as I have shown above, the consequence was that possible locations for P2 clitics occurred at several points in a sentence. This tendency is specular to what we have seen in Section 4.1 regarding Hittite clitic chains: rather than being strictly confined to a uniquely defined P2, Greek clitics could be scattered in a sentence at a potentially unlimited number of intonational breaks. Note further that the status of pronominal clitics as “special” clitics is far from being clear: I cannot pursue this issue here; suffice it to mention that the third person clitic min, shown in examples (30) and (31) above, does not occur in Attic prose, which constitutes the bulk of classical literature. There, third person is indicated by non-nominative weak forms of the former demonstrative autón, which was most likely a simple clitic similar to Latin eum (see Section 3.3). In addition, as discussed by Goldstein, some pronominal forms must be taken as proclitic to explain certain metrical facts, thus apparently indicating a double status as to phonological liaison. In conclusion, Classical Greek offers a picture in which particles and weak forms enjoyed a remarkable freedom in their position once prosodic conditions were satisfied, thus serving a variety of discourse functions favored by the high degree of word order freedom, including possible occurrence of discontinous constituents of all sorts. Similar to Latin, Greek also displays increasing grammaticalization of word order, ultimately motivated by increasing configurationality (see Devine and Stephens 1999; Luraghi 2010a), with accompanying loss of null objects and the development of a highly grammaticalized system of verb-based clitics. The latter exhibit features of head marking strategies in Modern Greek, as indicated by frequent clitic doubling (see Joseph 2002 with the references therein and Janse 2000 for a comparison between the development from Latin to Romance and from Ancient to Modern Greek). The picture offered by Classical Greek does not yet indicate such late developments: the freedom of clitic placement was a consequence of the high extent to which Greek word order was discourse conditioned. Remarkably, as pragmatic factors gave way to a more syntactically conditioned sentence structure, clitics, which were not all clustering in the same P2 position, had, so to speak, nowhere to go and ended up being attracted to the verb and forming a structurally defined VP with it.

188

9781441124609_Ch11_Final_txt_print.indd 188

11/27/2012 11:18:30 AM

Clitics

5 Differences Between Enclisis and Proclisis? Observing orthographic representation of proclitics and enclitics in the Romance languages, Benincà and Cinque (1983) suggest that the fact that enclitics are often attached to their host graphically whereas proclitics are not may reflect some difference in the relation between the host and the clitic based on the direction of liaison. Remarkably, such graphic differences are not only found in the Romance languages but also in various other completely unrelalted graphic systems, such as that of Hittite and, possibly, Mycenaean Greek:23 it is frequently the case that in graphic representation enclitics are spelled as one word with their host, while proclitics are not. In the light of the discussion of possible proclitics hosting enclitics in Section 4.1, the possiblity that the two types of clitics indeed present prosodic differences deserves to be addressed. Among the evidence for closer binding between a host and an enclitic in Romance, Benincà and Cinque mention the impossibility of omitting a direct object clitic in the second conjunct of coordinated clauses. Recall that this type of omission was obligatory in Latin, and still occurred, even though not systematically, in Old Italian. In Modern Italian, omission can sometimes occur. Acceptabilty judgments vary among speakers, but in any case, omission is restricted to V coordination in cases in which the object in the first clause is a clitic. Compare the following example: (32) L’ esistenza contemporanea non è minacciata . . . dall’ impeto delle the existence contemporaty not is threatened by+the thrust of+the acque profonde, ma dalla bonifica che le  inaridisce e prosciuga. waters deep but by+the drainage which them(cl) wither and dry.up “Contemporary life is not threatened by the outburst of waves from deep waters, but by the drainage which dries them up.” C. Magris, Dietro le parole, Garzanti, Milano, 1978. Such type of omission would be impossible with enclitics: (33) *la bonifica che minaccia di inaridirle e prosciugare / *di the drainage which threatens to wither them(cl) and dry.up / to inaridire e prosciugarle. wither and dry.up them(cl) In this respect, enclitics behave as bound morphemes: each verb requires an overt enclitic, while proclitics can have scope over two coordinated verbs. Benincà and Cinque further mention sporadic occurrences of adverbs intervening between a proclitic and the verb from various Romance varieties including Old Italian (note that the better known occurrences of separation of a clitic 189

9781441124609_Ch11_Final_txt_print.indd 189

11/27/2012 11:18:30 AM

The Bloomsbury Companion to Syntax

from its verb in Old Spanish are also limited to proclitics; see Fontana 1996 and Wanner 1996 for the examples): (34) e se ella il mi pur crede. and if she it(cl) me(cl) even believes “If she belives what I’m saying.” (Boccaccio Decameron viii.6). In addition, proclitics can be coordinated in French and Romanian, while enclitics cannot: (35) Je lui et vous ferais un plaisir. I him(cl) and you(cl) will.do a favor “I will do him and you a favor.” The evidence reviewed thus far points toward some higher degree of autonomy of proclitics. However, only adducing evidence from pronouns has the obvious risk that one might confuse properties of anaphoric reference with properties of clitics. For this reason, the evidence provided by Hittite and other languages in which otherwise proclitic items can host enclitics is very important. In such languages, the clitic cluster can be hosted by a proclitic, but it looks doubtful that it can constitute an autonomous intonational unit in spite of having some autonomy: most likely it leans to the left hedge of the intonational unit on its right. Thus, the initial proclitic rather than receive an accent from the enclitics, might better be described as transmitting its liaison properties to the enclitics, so that the whole cluster becomes proclitic. It could then be the case that initial chains of enclitics, when hosted by an unaccented clause introducer, become proclitic. As a consequence, they can no longer be said to be placed in P2, but should rather be viewed as standing in P1.24 P1 placement is an extreme consequence of the tendency, repeatedly remarked for clitics, to occur early in a sentence. In languages in which clitics tend to cluster in P2, recurrent clitic chains can be expoited for various discourse functions such as that of isolating extraposed words or constituents by means of a prosodic boundary constituted by a string of unaccented items. Association of clitic clusters with emphazised, focused, topicalized constituents and the like emerges in usage as a recurrent pattern; its entrenchment in the speakers’ cognition creates the need for a different sentence pattern in which no constituent is isolated in such a way. The introduction of a proclitic clause introducer has the effect of extracting the internal boundary formed by unaccented items from the clause, and creates a clause type in which, in spite of the occurrence of clitics, no constituent is isolated.

190

9781441124609_Ch11_Final_txt_print.indd 190

11/27/2012 11:18:30 AM

Clitics

6 Conclusions In this chapter I have given a usage-based account of various characteristics of clitics, includuing their very existence and tendencies in their placement rules. I have shown that discourse organization plays an important role in determining peculiar properties of clitics. Regarding clitic placement, it is still worth remarking that the two main strategies described, that is, for clitics either to cluster around a specific constituent or to occur early in a sentence, typically in P2, can be explained as owing to competing motivations. In the first case, clitics tend to lean phonologically to the constituent that they have in their scope or to which they belong syntactically, as in the case of focalizers (sec. 1.2 ) and of pronominal clitics, such as those of the Romance languages. In the second case, clitics tend to occur early in a sentence because of their function in discourse: for example, pronominal clitics carry old information already introduced into the discourse, and serve referent tracking. So clitic placement may have different motivations, which drive clitics to attach to different hosts. Competing motivations can also be the cause of change in clitic placement.

Notes 1. Research for this paper has been generously supported by a grant of the Alexander von Humboldt Foundation in the fall of 2011. 2. An updated comprehensive bibliography of studies on clitics that includes works within all theoretical frameworks is needed. For example, the bibliography in van Riemsdijk (1999) is limited, as it only includes studies based on formal theories. 3. As I argue in Section 2, in a usage-based perspective it makes little sense to try to define words, affixes, even constructions, and, of course, clitics, as sharply separate categories, since linguistic units are better understood as a result of chunking processes based on frequency of collocations; see below, Section 2. 4. Clitics can be shown not to constitute phonological words with their hosts as they normally do not trigger internal phonological processes triggered by bound morphemes (see among others Nespor & Vogel 2007 and Russi 2008: 211–17 on Italian enclitics). The complex constituted by clitics with their host is often called clitic group (e.g. Nespor & Vogel 2007). In his description of Classical Greek clitcs, discussed in Section 4.2, Goldstein (2010) uses the term “prosodic word” in a similar sense. 5. As I argue in Sections 4.1 and 5, a case can possibly be made for P1 as another preferred position for clitics in addition or alternative to P2. 6. Adverbial clitics refer to spatial notions that arguably depend on the verb for interpretation, as shown in (9), in which ne depends on the verb venire “come,” which requires a source expression, and (10), in which ci depends on andare “go,” which requires a goal. In addition, various Romance languages such as Italian and French feature partitive clitics, which occur in argument position or modifying an argument, such as ne in (11). The only type of consituent which is apparently not part of

191

9781441124609_Ch11_Final_txt_print.indd 191

11/27/2012 11:18:30 AM

The Bloomsbury Companion to Syntax

7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

the VP is the beneficiary clitic (including the dativus ethicus type), as ci in (11). This type of beneficiary is always possible, and can occur with any type of verb. In some formally oriented descriptions, it is suggested that its occurrence opens an extra slot in the subcategorization frame of the verb. As indicated in the example, subject drop is not allowed in standard Piemontese for second person singular; note that this does not depend on the need to disambiguate the subject, as second person singular is unambiguously indicated by the verbal ending. On subject clitics in Northern Italian vernaculars see Renzi & Vanelli (1983). In descriptions inspired by generative grammar such subject clitics are considered inflectional affixes, and the Northern Italian varieties are considered pro-drop, contrary to standard French. See, for example, Brandi & Cordin (1989). See Belloro (2007) for a recent discourse-based analysis of clitic doubling in Spanish, which the author views as connected with cognitive accessibility of referents, and for further reference on the topic. Clitic doubling is a common phenomenon in other languages with pronominal clitics, too, and occurs, for example, in Slovenian and Serbo-Croatian, see Marušič (2008), as well as in Modern Greek, on which see Section 4.2. Clitic climbing in Italian is a comparatively neglected phenomenon. A pionering study is Berretta (1985), which however is not corpus based and relies on just a few occasionally collected examples. For a formal account of Italian clitic climbing see Monachesi (1993). Some example of clitics separated from the verb in Old Spanish are discussed in Fontana (1993, 1996); similar occurrences can also be found in other Old Romance varieties, see Benincà & Cinque (1983) on Old Italian. Remarkably, all such occurrence feature clitics that, while they may lean back enclitically on the first word in the sentence, precede the verb rather than follow it. I come back to this issue in Section 5. Some traces of Wackernagel’s Law as inherited from PIE are available from Early Latin, see Wanner (1987); but it does not seem likely in the light of the evidence provided by classical texts that possible P2 enclisis of weak pronouns in Late Latin could be directly connected with them, and that Wackernagel’s Law as inherited from PIE could have fed the Tobler Mussafia’s Law. Theories on how exactly the Tobler Mussafia’s Law originated are more numerous than what I can summarize here: according to Fontana (1996), for example, the reason why clitics moved to the left sentence boundary is ultimately to be seen in the V2 phenomenon, which is thought to have affected the Old Romance languages; see further Poletto (this volume). Goldstein (2010) gives an accurate summary of nineteenth century scholarship regarding the position of clitics, and argues that the clearest statement of Wackernagel’s law is indeed to be found in works by Delbrück (see, for example, Delbrück 1878). Wackernagel observed that the Sanskrit verb is unaccented or weakly accented in main clauses, and suggested that the V2 phenomenon in Germanic could be a reflex of such property, ultimately going back to PIE. I cannot pursue this issue further here; the idea that P2 clitics and V2 may be related has been entertained by several scholars; see among others Anderson (1993, 2003) and Fontana (1996). Koller (forthcoming) treats Hittite serialized motion verbs as restructuring verbs. He, too, views placement of clitics as in example (25) is taken as an instance of clitic climbing; cf. a similar phenomenon in Romance, discussed in sec. 4.1.2. Hock (1996) pursues a prosodic analysis of Wackernagel’s law in Vedic, and proposes a templatic approach to the order of clitics in clusters, whereby accented and unacceted forms alternate. As Schäufele (1996) points out, it may well be the case

192

9781441124609_Ch11_Final_txt_print.indd 192

11/27/2012 11:18:30 AM

Clitics

19. 20.

21. 22.

23.

24.

that Hock’s templatic approach is fundamentally right; however, the fact remains that sentence particles precede pronouns. This is also true of other writing systems, see Section 5. Textual discontinuity could have a stylistic function, or it could indicate an unexpected event or signal clauses that introduced backgrounded information; see Luraghi (1990a). Hittite was a fairly strict SOV language and initial verbs were quite exceptional at all language stages; see Luraghi (1990a). See Taylor (1990: 120). Dative enclitic pronouns can be placed between the definite article and their noun, but apparently only when they function as possessive, and can therefore be considered part of the NP; see Taylor (ibid.) and Goldstein (2010: 93–5) for discussion. In cases in which dé is hosted by a proclitic, such as the definite article or the negation, Golstein argues that it must be regarded as proclitic to the next accented word. In the light of the discussion in Section 4.1.2 regarding possible clitic groups hosted by proclitics in Hittite, one could equally argue for similar proclitic hosts in Greek. According to Adams (1996), even some Late Latin texts provide evidence for enclisis of pronouns by not separating them in writing from the preceding word (mostly the verb). It is worth noting that Agbayani and Golston (2010) reach a similar conclusion within a formal framework, referring to syntactic cliticization of P2 enclitics in Ancient Indo-European languages.

193

9781441124609_Ch11_Final_txt_print.indd 193

11/27/2012 11:18:31 AM

12

Changing the Subject: Shifting Notions about Subjecthood in Generative Grammar Tim Stowell

Chapter Overview Introduction Sentence-initial Position, Nominative Case, and Argument Alignment Syntactic Positions and Syntactic Structures Generalizing the Subject Position Across Syntactic Categories

194 196 205 214

1 Introduction 1.1 What is a Subject? What is a subject? This question has arisen frequently in the history of generative grammar, and the answers to it that have been offered have evolved considerably over the past 50 years. In this article I provide an overview of several generative theories of what counts as a subject. My discussion broadly follows chronology, though in some cases I combine aspects of analyses from different eras or research traditions. I assume a general familiarity with the basics of generative syntactic theory, including argument structure, constituent structure, grammatical case, and syntactic movement. Throughout the early history of generative grammar, the term subject was applied exclusively to a noun phrase (NP) functioning as the subject of a 194

9781441124609_Ch12_Final_txt_print.indd 194

11/27/2012 11:07:53 AM

Changing the Subject

sentence (S). Later on, it was suggested that constituents other than NP may sometimes function as subjects, and that constituents other than sentences may sometimes contain subjects; see Longobardi and Silvestri, this voume, on subjects of nouns). In Section 2, I review three diagnostic descriptive properties that are commonly attributed to the subjects of sentences, involving constituent order, thematic roles, and grammatical case, and in Section 3, I discuss theories that have been proposed in the literature to account for these diagnostic properties. In Section 4, I discuss proposals that all syntactic categories (and not just full sentences) may contain subject positions, and the related proposal that the subjects of full sentences typically originate as subjects of verb phrases (VPs)—the so-called VP-internal subject hypothesis. Before turning to these issues, however, I address the question of what types of phrases may function as subjects.

1.2 Syntactic Constituents as Subjects What kinds of syntactic constituents can be subjects? According to the classical Aspects theory of Chomsky (1965), only an NP can be a subject. An NP may be a proper name (such as John or Jane Smith), a pronoun (such as he, you, them, etc.) or a common noun phrase (CNP). A CNP always contains a (common) noun (N), such as teacher, book, or statement, which functions as the head of the noun phrase; it is called the head noun. It is usually preceded by one or more determiners (articles, demonstratives, quantifiers, and numerals), as in (1a), and (optionally) one or more adjectives, as in (1b). It may be followed by one or more prepositional phrases (PPs), as in (1c), and/or by a subordinate complement clause or relative clause (1d). Various specific combinations of constituents and linear orderings of constituents are: (1) (a) (b) (c) (d) (e) (f)

the teacher, most teachers, three teachers the old teachers, (some) energetic young assistants the man from Glad, a story about Bill the idea that sentences have subjects, many teachers who work hard *the about Bill *teachers old

All types of NPs (proper names, pronouns, and CNPs) may occur in any of the major syntactic positions where NPs may occur; these include the direct object position, immediately following a verb (V) within a verb phrase (VP); the prepositional object position immediately following a preposition (P) within a prepositional phrase (PP); and the pre-verbal subject position within a sentence (S). 195

9781441124609_Ch12_Final_txt_print.indd 195

11/27/2012 11:07:54 AM

The Bloomsbury Companion to Syntax

Since Abney (1987), it has been widely assumed that NPs are usually contained within a larger constituent, called a determiner phrase (DP), consisting of a determiner (D) followed by an NP, as in (2): (2)

DP NP

D N

...

According to this theory, the pre-nominal determiners in (1) are actually external to the NP. Thus, in a simple CNP such as the teacher, the NP consists of just a common noun (teacher). The determiner the combines (merges) with this NP to form a DP. Paul Postal first suggested that pronouns are actually special types of determiners that can form complete DPs by themselves without having to combine with an NP. (On this view, the English third person pronouns are “intransitive” counterparts to the “transitive” definite article, though their overt expression of gender and number features suggests that pronouns may represent a merger of a definite determiner with a covert counterpart to NP.) Proper names, like nouns, originate within NP, but may undergo movement to the D position in the superordinate DP if the D position contains no determiner (Longobardi 1994). In this paper I will assume the DP hypothesis as outlined above, since it is the theory of noun phrase structure that is most widely assumed at the present time;1 I will therefore refer to subject and object noun phrases as DPs, even when discussing older theories about subjects that referred to them as NPs. Thus, according to the DP hypothesis, the subject of a sentence is typically a DP, rather than an NP. In many languages, such as Spanish and Italian, not all sentences contain an overt subject. Sentences lacking overt subjects are normally interpreted as if they contained pronoun subjects; Rizzi (1982) proposed that these apparently subjectless sentences contain phonetically null subject pronouns, synonymous with (and syntactically similar to) their overt counterparts in English. Phonetically null subject pronouns have also been posited in a number of other syntactic constructions, including infinitival control clauses.

2 Sentence-initial Position, Nominative Case, and Argument Alignment How is the subject DP identified? Sentences may contain more than one DP, but only one of these is the subject. The subject DP has three major diagnostic properties. The first diagnostic involves its syntactic position: in English, the subject is the 196

9781441124609_Ch12_Final_txt_print.indd 196

11/27/2012 11:07:54 AM

Changing the Subject

DP preceding the finite verb in a simple declarative sentence. Generative theories generally account for this in terms of the sentence’s syntactic structure and the rules governing it. The second diagnostic involves the type of thematic role that the main verb assigns to the subject: this is generally the Agent thematic role, if there is one; otherwise, the subject typically bears the thematic role of Instrument or Theme. The third diagnostic property involves the grammatical case borne by the subject DP—this is generally nominative, at least in languages that exhibit a contrast between nominative and accusative case. (Things work differently in ergative/absolutive case systems; see Payne, this volume.) In this section I discuss each of these diagnostic properties briefly, focusing on how they distinguish the subject DP from other DPs, such as the verb’s object. In this section, my discussion of these diagnostic properties is mainly descriptive; in Section 3, I will discuss theoretical explanations for them in terms of explicit syntactic structures.

2.1 Constituent Order Let us first consider constituent order. Abstracting away from sentences containing subordinate clauses, every English sentence normally contains at least one DP and a finite verb; some sentences contain two DPs and a verb. If a sentence consists of just a single DP and a verb, the DP must precede the verb, as (3) shows: (3) (a) The package arrived. (b) *Arrived the package. If a sentence consists of a verb and two DPs, one DP must precede the verb, and the other must follow it, at least when the sentence is a simple declarative with neutral intonation: (4) (a) That lion attacked the tiger. (b) *That lion the tiger attacked. (c) *Attacked that lion the tiger. In both (3) and (4), the sentence must contain exactly one DP in pre-verbal position; this is the subject of the sentence. The sentence may also contain an object DP, which occurs after the verb. Thus the subject position can be defined quite simply, in terms of linear order: (5) The subject of the sentence is the (unique) DP that precedes the verb. Since the object DP follows the verb, English is said to exhibit Subject-VerbObject (SVO) constituent order. 197

9781441124609_Ch12_Final_txt_print.indd 197

11/27/2012 11:07:54 AM

The Bloomsbury Companion to Syntax

Superficially similar to the ungrammatical (4b) is the grammatical (6), with Topic or Focus intonation, both orthographically represented by the comma: (6) That lion, the tiger attacked. Here the first DP (that lion) occupies the Topic position, and is interpreted as the object of the verb. The second DP (the tiger) occupies the true subject position. To accommodate such examples, (5) might be restated as (7): (7) The subject of the sentence is the DP that immediately precedes the verb. But topic constructions like (6) are generally assumed to be derived by movement in generative grammar; the Topic originates in another position within the sentence (in this case, the post-verbal object position) and undergoes movement to the Topic position. If this is correct, the existence of sentences like (6) containing topicalized object DPs is not necessarily inconsistent with (5); for example, we can assume that (5) applies to the “underlying” representation of the sentence, prior to the application of the movement rule that derives Topic constructions. Alternatively, the Topic position may be considered to lie outside the boundary of the sentence, as reflected in the marked intonational phrase boundary separating the topic from the subject, so that the subject would still be correctly identified by (5). Many languages resemble English in exhibiting SVO order in simple declarative sentences, but other languages exhibit different canonical orders of subject, object, and verb. The question therefore arises whether (5) is valid only for SVO languages like English, or whether it may be satisfied in a more indirect way in languages with other constituent orders in neutral declarative main clauses. Alternatively, (5) may have to be amended to accommodate such languages, though the scope of variation in constituent order may be too great for any universal diagnostic for subjecthood based on constituent order to accommodate it fully. As shown in Munro (this volume), the overwhelming majority of the world’s languages exhibit SVO, SOV, or VSO order in simple declarative main clauses, and in seeking to expand the empirical scope of (5) beyond SVO languages (see Munro in this volume), I will focus primarily on SOV and VSO constituent orders. (It may be that some of the languages exhibiting OVS and VOS order derive these by leftward displacement of the VP from an underlying SOV or SVO source, though this type of account is almost certainly too simplistic.). Clearly, the existence of languages with VSO or SOV as the canonical neutral constituent order in simple declaratives poses a problem for the universal validity of (5) or (7); in VSO languages, no DP qualifies as the subject by either (5) or (7), and in SOV languages, both subjects and objects satisfy the diagnostic condition in (5), and only objects satisfy the condition in (7). 198

9781441124609_Ch12_Final_txt_print.indd 198

11/27/2012 11:07:54 AM

Changing the Subject

A striking property of the three major constituent order types (SOV, SVO, and VSO) is that they all involve SO order (as opposed to OS) order in canonical neutral declarative sentences. This suggests a simpler alternative to (5) as a diagnostic property of subjects based on basic constituent order: (8) The subject of the sentence is the first (leftmost) DP in the sentence. (8) is simpler than (5), and has the virtue of being able to pick out the subject, not only in SVO languages but also in SOV and VSO languages. Note that, if an English sentence contains only one DP, that DP qualifies as the subject of the sentence, given (8); the sole argument of an intransitive verb is invariably a subject preceding the verb, rather than an object following the verb. But “ergative/absolutive languages” seem to treat the sole argument of an intransitive verb more like an object, at least in some respects. This poses a potential problem for the idea that a diagnostic like (8) is universally valid. Now consider the status of “null subject” languages such as Italian and Spanish, discussed in Section 1.2. When the subject is overt, it normally precedes the finite verb and is correctly identified by (5) and (8). When the subject is non-overt, the question arises whether generalizations like (5) or (8) are even applicable. If such generalizations apply only to phonetically overt DPs, then no DP will be identified as the subject in such cases, expressing the traditional idea that such sentences lack grammatical subjects. On the other hand, if it is assumed that phonetically null pronouns are syntactically present and “visible” to constituent order generalizations, then one can follow Rizzi (1982) in assuming that a phonetically null pronoun DP precedes the verb in such cases and is correctly identified as the subject by (5) or (8). Perhaps the most significant empirical problem for the idea that the subject DP can be universally identified by a diagnostic generalization about constituent order concerns the existence of languages in which word order is often claimed to be entirely free, with no preferred canonical ordering of the subject relative to the verb or object DP. Such languages are often described as “free word order languages” or “non-configurational languages,” reflecting the widely held view that these languages impose no grammatical constraints on constituent order. Classical Latin and Sanskrit, as well as modern languages such as Warlpiri, have been cited as languages of this type. If these languages lack a canonical position for subject DPs, it should be impossible to identify the subject on the basis of a generalization such as (8). Within the framework of early Government-Binding theory, Hale (1981) developed a formal version of this approach, introducing the term “non-configurationality” to describe the phenomenon of free constituent order. Hale’s view was that there are two fundamentally different types of languages with respect to syntactic phrase structure: some languages (like English and Chinese) have it; other languages (like Latin and Warlpiri) do not. 199

9781441124609_Ch12_Final_txt_print.indd 199

11/27/2012 11:07:55 AM

The Bloomsbury Companion to Syntax

It has been suggested, however, that even in putatively non-configurational languages, sentences exhibiting VS and OS order are not really neutral with respect to criteria such as topic and focus, and that many or all of the sentences that have been cited as evidence for completely free constituent order in these languages are actually derived by the application of movement rules similar to those involved in the derivation of Topic constructions in English. If so, the phenomenon of completely free constituent order in these languages may be illusory. At present there is no general consensus on this issue. (See Munro, this volume, on (non)configurationality and word order.)

2.2 Thematic Roles The second diagnostic property of subject DPs is semantic, involving the types of thematic roles associated with a verb’s arguments. Active verbs such as eat, kiss, sleep, arrive, etc. are normally understood to describe particular types of actions, events, or situations. Each type of event has a certain number of key participants; for example, an event of eating minimally involves an entity (typically animate) that ingests something, and another entity (typically food) that is ingested. The verb is said to select phrases referring to these participants as its arguments; thus, if the verb describes an event that has two participants, it selects two arguments. The verb’s arguments can be classified in terms of the type of role that they play in the event the verb describes; these are called thematic roles. For example, if an action is instigated or caused by one of its participants, that participant is classified as the Agent argument; if one of the participants undergoes a change of state or location as a result of the event, that participant is classified as the Theme argument. These classifications of role-types (Agent and Theme, as well as Instrument, Experiencer, Goal, etc.) are called thematic roles. Each referential DP that occurs as a subject or object normally functions as one of the verb’s arguments in the sense that the referent of that DP is understood to be a participant in the event that the verb describes. For example, in (4a), the denotation of the subject DP that lion is interpreted as the Agent of the eating event, and the denotation of the object DP the tiger is understood as the Theme of the eating event. In this case, the verb is said to “assign” the Agent thematic role to the subject DP that lion, and to assign the Theme thematic role to the object DP the tiger. But how do we know that the lion is the Agent, rather than the tiger? The answer lies in the principles governing thematic role assignment. Whenever an active verb selects two or more arguments, each with a specified thematic role, it is generally possible to predict which thematic role is assigned to the subject of the sentence and which thematic roles are assigned to other DPs, such as the object of the verb. For example, if the verb describes a type of 200

9781441124609_Ch12_Final_txt_print.indd 200

11/27/2012 11:07:55 AM

Changing the Subject

action that involves a deliberate or causal Agent, the Agent thematic role is assigned to the subject DP, as in (4a). If the verb also selects a Theme argument, the Theme role is assigned to the object of the verb. If the verb also selects an Instrument argument, the Instrument role is assigned to a DP preceded by the preposition with, as in (9a). Some verbs that select both an Agent and an Instrument allow the Agent to be omitted, in which case either the Instrument role or the Theme role can be assigned to the subject, as in (9b) and (9c), respectively:  (9) (a) Max opened the door with a key. (b) The key opened the door. (c) The door opened with a key. If the verb describes an event-type that has a Theme participant, but no Agent or Instrument participant, the Theme role is assigned to the subject DP, as in (3a), repeated here:  (3) (a) The package arrived. These generalizations, among others, are assumed to follow from a general set of principles governing argument alignment. These principles appear to involve reference to a hierarchy of thematic roles (Fillmore 1968), often called the thematic hierarchy; the Agent ranks highest on this hierarchy, followed by the Instrument and Theme. The thematic role assigned the subject of the sentence always ranks higher on the thematic hierarchy than any of the thematic roles assigned to any other DPs occurring in the sentence. This is the second diagnostic property of subjects: (10) The subject is the DP to which the verb assigns the unique thematic role that ranks higher on the thematic hierarchy than any other thematic role that the verb assigns. Some aspects of the alignment principles, and the putative hierarchy that they are based on, are imperfectly understood at present; for example, psychological predicates selecting Experiencer and Theme arguments appear to be divided on which of these thematic roles is assigned to the subject. The alignment is apparently sensitive to factors involving temporal aspect and causation. Setting aside such complications, the big picture is that the subject of a sentence can generally be identified on the basis of the thematic role that its referent is understood to bear in the verb’s argument structure. 201

9781441124609_Ch12_Final_txt_print.indd 201

11/27/2012 11:07:55 AM

The Bloomsbury Companion to Syntax

2.3 Grammatical Case The third major diagnostic property of subjects concerns grammatical case; in languages exhibiting a contrast between nominative and accusative case, the descriptive generalization in (11) generally holds: (11) The subject of the sentence is the sole DP bearing nominative case. In English, only pronouns (including Wh-pronouns, in some conservative dialects) overtly express grammatical case: (12) (a) She hit him. (b) He hit her. (c) *Her hit he. (d) *Him hit she. (e) The woman hit the man. (f) The man hit the woman. But in many other languages, including German and Russian, nouns, determiners, and adjectives also have morphologically overt case markings, and most types of DPs exhibit overt case distinctions. For English pronouns, and for most DPs in German and Russian, this is the third diagnostic property of subjects. Actually, there are lots of exceptions to (11) in the world’s languages. In languages such as Icelandic, some verbs select subject arguments bearing “quirky” non-nominative case affixes. (See Eythórsson, this volume, on accusative and dative subjects in Icelandic; for further reference on non-nominative subjects, see e.g. Bhaskararao and Subbarao.) In languages such as Japanese, some verbs select two arguments bearing nominative case (typically the subject and the object.) Since these exceptions are limited to sentences containing a limited lexical subclass of verbs, it is generally assumed that a complicating extraneous factor is at work in the derivation of sentences containing such verbs. Another class of exceptions to (11) arises in languages that allow subordinate infinitival clauses to contain overt subject DPs. In many of these languages, including English, the subject of the infinitive bears accusative case, rather than nominative; as in (13a–b): (13) (a) I would prefer for him to leave. (b) I expect them to eat the tiger. I discuss these cases in Section 3.3. 202

9781441124609_Ch12_Final_txt_print.indd 202

11/27/2012 11:07:55 AM

Changing the Subject

In languages such as English, where only pronouns exhibit overt case distinctions, a question arises concerning the status of (11) for non-pronominal DPs. Traditional descriptive grammars generally assume that (11) is simply irrelevant to such DPs, since they appear to lack grammatical case entirely. An alternative approach, explored in depth by Vergnaud (1977) and much subsequent work, is to assume that grammatical case is an abstract syntactic feature that is assigned to all DPs, even though it often fails to be “spelled out” morphologically. The same approach can be adopted to account for languages such as Chinese, where overt morphological case markings are entirely lacking for all DP types.

2.4 Subject-verb Agreement There is a fourth diagnostic property of subject DPs in English and many other Indo-European languages that I have not yet mentioned, namely that the finite verb agrees with the subject DP in morphologically overt features for grammatical person and number. At least in these languages, agreement serves as a fairly reliable diagnostic for identifying the subject DP. The existence of languages such as Japanese and Chinese, lacking any phonetically overt person and number agreement is not necessarily problematic, since one can assume that in such languages agreement is syntactically present though phonetically null, as in the case of null pronouns in null subject languages. More problematic, however, is the fact that many languages exhibit overt agreement with object DPs as well, and in some cases object agreement is overt even when subject agreement is not. For this reason, agreement with the finite verb is not universally viable as a diagnostic property uniquely identifying subject DPs per se. Of the three diagnostic properties of subject DPs discussed in the previous sections, the most central and apparently cross-linguistically invariant generalization is (10), involving the type of thematic role assigned to the subject. This is the diagnostic property that typically guides descriptive field linguists engaged in studying an unfamiliar language. Because of the widely accepted assumption that there is no universally valid cross-linguistic characterization of major constituent order, and because many languages either lack overt morphological case markings entirely or exhibit an ergative/absolutive case system, field linguists generally rely almost exclusively on (10) as a reliable diagnostic for identifying the subject, at least at the preliminary stages of investigating the syntax of a previously undocumented language. In addition, in typological literature, behavioral, as opposed to coding properties are often used as a diagnostic for subjecthood. Some of such properties are illustrated in Eythórsson (this volume). Within the generative research tradition, the idea that constituent ordering restrictions (and perhaps also constituent structure) are entirely lacking 203

9781441124609_Ch12_Final_txt_print.indd 203

11/27/2012 11:07:55 AM

The Bloomsbury Companion to Syntax

in some languages also played a role in inspiring the development of the theory of Relational Grammar by David Perlmutter and Paul Postal during the 1970s; some of the seminal articles developing this approach later appeared in Perlmutter and Postal (1984). Perlmutter and Postal believed that all languages recognize abstract grammatical relations such as subject, object, indirect object, etc., and they argued that many syntactic constructions (passive, dative shift, causative clause union) can be characterized in terms of language-universal rules, or laws, only if these laws are stated in terms of networks of grammatical relations rather than in terms of constituent ordering relations or syntactic structures. They proposed that all sentences have abstract representations in which only these grammatical relations are represented; these “relational networks” are governed by principles that they presumed to be universal. They proposed that languages differ from each other, however, in how they overtly express these grammatical relations. Some languages express them in terms of constituent order; other languages express them in terms of case-marking; other languages express them in terms of verbal agreement. On their view, properties relating to constituent order and case marking are language-specific; only the laws governing argument alignment and grammatical relation-changing rules are universal. Their theory implicitly assumed the validity of (10), though for them grammatical relations such as “subject” and “object” were undefined primitives of the theory. The central motivating insights of Relational Grammar were later adopted by Bresnan (1978) and Kaplan and Bresnan (1982), who expressed them in terms of lexical rules. Nevertheless, most generative theorists in the research traditions of Government-Binding theory (based on Chomsky (1981)) and the Minimalist Program assume that all three diagnostic properties reflect properties of universal grammar, and that the underlying principles responsible for these generalizations are operative universally, even in languages that provide no overt evidence that would enable a language learner to infer their existence purely by induction on the basis of exposure to primary linguistic data. Such theorists believe that these principles do not have to be learned on the basis of induction, since they are provided to the language learner by his or her innate genetic endowment. In the field at large, this view remains controversial, and most non-generative linguists reject it, for a variety of reasons. Many are suspicious of generativists’ universal claims because they believe them to be based on an insufficiently broad survey of the world’s languages; many believe that linguistic rules must all be learned inductively and that the obvious syntactic diversity of the world’s languages therefore results in language-particular grammars that conform to few universal principles; and many are simply suspicious of specific generative analyses because they find them implausibly abstract or complex. My own view is that these concerns are misplaced, but space considerations preclude further discussion of the issue (see Payne, this volume). 204

9781441124609_Ch12_Final_txt_print.indd 204

11/27/2012 11:07:55 AM

Changing the Subject

3 Syntactic Positions and Syntactic Structures 3.1 Hierarchical Structure Perhaps the most fundamental insight of generative grammar is that sentences are not just sequences of words but rather have a complex hierarchical structure within which phrases and words are located. By positing these structures, it is possible to account for most generalizations about surface word order in a given language in terms of the rules and/or principles that collectively define the set of well-formed phrase structures that sentences (and the phrases within them) can have. According to this approach, the linear order of constituents is simply an indirect and incomplete reflection of the invisible phrase structure that determines it. Words belonging to specific syntactic categories (verbs, nouns, adjectives, prepositions, articles, complementizers, etc.) may only occur in certain positions in the structure. Cross-linguistic differences in constituent order arise because of parametrically constrained variation in the rules or principles governing the formation of constituents and/or grammatical operations involving movement of constituents from one position in the sentence’s constituent structure to another. Ultimately, therefore, it should be possible to restate the linear ordering generalizations in (5) and (8) in terms of the position(s) that subject DPs may occupy in the abstract syntactic structure of the sentence. The traditional generative analysis of SVO order in English is that it is based on the following assumptions, simplified somewhat for ease of exposition: (14) (a) Words may combine, or merge, with other words or phrases to form constituents, or phrases, of a particular type. The set of possible merger operations is formalized as a set of context free phrase stucture rules defining the constituent structure of each type of phrase. (b) A verb V may combine, or merge, with a DP to form a VP. The merger operation locates V before the DP. The merger operation is sanctioned by the phrase structure rule for VP. (c) A DP may combine with a VP to form a sentence (S). The merger operation locates the DP before the VP. The merger operation is sanctioned by the phrase structure rule for S. Thus, a simple sentence like (15a) has the structure in (15b): (15) (a) The doctor wrote a book. 205

9781441124609_Ch12_Final_txt_print.indd 205

11/27/2012 11:07:55 AM

The Bloomsbury Companion to Syntax

(b)

S DP

VP V N doctor] [wrote

Det [The

DP a book]

In (15b), the object DP is located inside the VP, after the verb, and the subject DP is located outside the VP, preceding it. This allows for the positions of the subject and object DPs to be distinguished from each other in structural terms, as in (16): (16) (a) The DP directly dominated by the Sentence (S) node is the subject of the sentence. (b) The DP directly dominated by VP is the object of the verb. Actually, the structure in (15b) is an oversimplification. Sentence (15a) contains only a single verb, which is finite, bearing a tense affix, but sentences may also contain one or more auxiliary verbs, as in (17): (17) (a) The doctor has written a book. (b) The doctor should have written a book. (c) The doctor might be writing a book. Chomsky (1957, 1965) located auxiliary verbs in an Auxiliary phrase (Aux) preceding the VP; he proposed that all sentences have a ternary branching structure, consisting of the subject DP, the Aux, and the VP, as defined by the phrase structure rule in (18a), providing a structure like (18b) for a sentence like (17a)2 (Chomsky’s original formulation used the category NP rather than DP). (18) (a) (b)

S → DP – Aux – VP S DP Det [The

VP

Aux N

doctor] has

V

DP

[written a book]

206

9781441124609_Ch12_Final_txt_print.indd 206

11/27/2012 11:07:56 AM

Changing the Subject

Rule (18a) is a context-free phrase structure rule; the grammar of every language was assumed to contain many such rules, defining the hierarchical structure of each type of phrase, and imposing a linear ordering on sister constituents within each phrase or sub-phrase. Alongside the rule for the structure of the sentence in (18a), there were other rules defining the internal structure of each type of phrase, including VP, NP, AP, etc. Many subsequent analyses of English constituent structure retained the idea that the main verb first merges with its complements (including the object DP) before merging with the auxiliary verb(s) and the subject DP, but posited a uniform binary branching structure rather than the ternary branching structure in (18b) defined by (18a). It would take us too far afield to review all of these subsequent structural analyses here; I will therefore discuss just two approaches, according to which the category Aux is replaced by a “functional” category headed by a free-standing category I (Inflection) or T (Tense), which merges with a VP following it to form a phrase IP or TP. The functional category IP, introduced by Stowell (1981), was based on two previous proposals: Chomsky’s (1970) theory of X-bar structure, discussed in more detail in Section 3.2, and Chomsky’s (1981) category Infl or I, posited as a replacement for the old category Aux, but containing only tense and subject agreement features. X-bar theory was intended to capture cross-categorial parallels of syntactic phrase structure for the four major lexical categories (VP, NP, AP, and PP); it was formulated in terms of a pair of general meta-categorial phrase structure rules, employing a categorial variable X to generalize across syntactic categories. These rules are given in (19), incorporating slight modifications of Chomsky’s original formulation. (19) (a) XP → Specifier – X’ (b) X’ → X – Complement(s) These rules define a uniform syntactic phrase structure for all lexical categories, represented schematically in (20), where the Specifier and Complement positions are occupied by phrases headed by independent categorial variables. XP

(20)

X’

ZP X

YP

X-bar theory was originally intended to account only for the internal structure of the four major lexical categories, and this assumption was retained by

207

9781441124609_Ch12_Final_txt_print.indd 207

11/27/2012 11:07:56 AM

The Bloomsbury Companion to Syntax

Hornstein (1977) and Chomsky (1981), who posited (21) as the phrase structure rule for the structure of the category S. (21) S → DP – Infl – VP In Stowell (1981) I argued that the structure of S should be reconceived as a functional category Infl-P, or IP, headed by Chomsky’s (1981) category Infl (I). This was consistent with a more general proposal to entirely eliminate the component of category-specific phrase structure rules, in effect reducing them to the category-neutral X-bar theoretic rules in (19). Under this approach, X-bar theory applies to all syntactic categories, including functional categories such as Infl. The relevant structure looks like (22): (22)

IP I’

DP

VP

I

Within I’, VP functions as the complement of I, just as DP functions as the complement of V within V’. I’ merges with a DP preceding it in the Specifier of IP position to form IP, the new name for the old category “Sentence.” This structure was widely adopted as the structure of the sentence during the mid-1980s, allowing for the subject of the sentence to be defined as in (23): (23) The DP occupying the Specifier position of IP is the subject of the sentence. Pollock (1988) later argued that Chomsky’s category Infl is really a conflation of two distinct functional categories, which he labeled as Agr (Agreement) and T (Tense), corresponding to the grammatical features for tense and agreement expressed morphologically in terms of finite verbal inflection. Agr and T function as the heads of the phrases Agr-P and TP respectively. For the sake of simplicity, I will ignore the category AgrP here, and simply note that (22) and (23) can be trivially modified, substituting the category label T for I, as in (24) and (25): TP

(24)

T’

DP T

VP

208

9781441124609_Ch12_Final_txt_print.indd 208

11/27/2012 11:07:56 AM

Changing the Subject

(25) The DP occurring in the Specifier position of TP is the subject of the sentence. On this view, (15a) has the structure in (26). (26)

TP T’

DP Det

N

[The doctor]

T

PAST

VP V [write

DP a book]

The tense affix adjoins to the verb as a result of a movement process of “affix hopping,” originally proposed by Chomsky (1957). Through the 1990s, a number of other phrase-types headed by functional categories were proposed in the literature, including phrases associated with verbal aspect (as opposed to tense) to account for the placement of auxiliary verbs and participial affixes in sentences such as those in (17).

3.2 Languages with Canonical SOV or VSO Order Now consider the constituent structures of simple finite sentences in languages exhibiting canonical SOV or VSO order. Traditional generative accounts of SOV order simply assumed that the phrase structure rule for VP located the V position at the right edge of VP rather than at its left edge. With the advent of X-bar theory, it was generally assumed that SOV order arises from the fact that the left-right ordering mandated by the X-bar theoretic rule for the structure of X’ in (19b) is subject to parametric variation; languages may vary according to whether X’ is head-initial or head-final. Thus languages can choose between (19b) and (27): (27) X’ → Complement(s) — X Languages choosing (19b) have VO order in VP (resulting in SVO order for the sentence as a whole) while languages choosing (27) have OV order in VP (resulting in SOV order). According to this hypothesis, a VP in an SOV language is essentially a mirror image of its English counterpart. An account of Japanese head-final constituent order based on (27) was developed by Farmer 209

9781441124609_Ch12_Final_txt_print.indd 209

11/27/2012 11:07:56 AM

The Bloomsbury Companion to Syntax

(1982); see also Stowell (1977) for a similar approach to the related language of Okinawan. Throughout the 1970s and early 1980s, it was widely assumed that the parameter distinguishing verb-initial VP order from verb-final VP order was part of a broader parametric variation between head-initial and head-final order in all lexical phrase types. The empirical underpinning of this idea came from a series of descriptive word-order generalizations formulated by Greenberg (1963) and others. Greenberg noted that SOV languages like Turkish and Japanese typically use postpositions rather than prepositions; thus, P typically follows its object in PP. This is expected if the substructure of PP and VP must conform to a uniform directionality setting determined by the choice of (27) over (19b). Early accounts of IP structure in the Government-Binding framework such as Stowell (1981) likewise assumed that head-complement order at the I’ level in IP is subject to the same parameter setting; that is, Infl should precede VP in a head-initial language and follow it in a head-final language. A conceptual problem for this X-bar theoretic approach to variation in headcomplement order was posed by the fact that there is comparatively little evidence for the same kind of variation at the XP level; it seems that (19a) does not alternate with a mirror image structure in which the Specifier position follows X’ instead of preceding it. This was not evident during the 1970s because there was a lot of analytical confusion over the status of the Specifier position—there was no consensus on what kinds of phrases count as specifiers. By the mid1980s, however, a consensus had emerged to the effect that the Specifier of CP is occupied by Wh-phrases in languages with overt Wh-movement, and that the Specifier positions of IP and DP are occupied by subjects, as discussed in Section 4. Especially in the case of CP and IP, there seems to be virtually no cross-linguistic variation: there are no attested cases of rightward Wh-movement to the right edge of sentence, and only a handful of uncommonly studied languages in which the canonical position of the subject DP is to the right of the VP. Thus it seems that (19a) is cross-linguistically invariant: the Specifier always seems to precede X’ in XP. This is surprising if variation in head-complement order arises from the option to choose between (19b) and (27), as noted by Kayne (1994) and others. A more serious empirical problem with the X-bar theoretic account of headcomplement order variation was posed by the fact that the theory led one to expect far more cross-categorial symmetry within a given language than is actually found in the world’s languages. For example, German appears to place the verb at the right edge of VP in subordinate clauses and in main clause VPs containing an auxiliary verb, but verbs in main clauses lacking auxiliary verbs locate the finite verb immediately after the first phrase in the sentence, deriving so-called verb-second constituent order (see Poletto in this volume). The 210

9781441124609_Ch12_Final_txt_print.indd 210

11/27/2012 11:07:56 AM

Changing the Subject

standard account of this, proposed by Thiersch (1978), was that VP is always verb-final, implying that German chooses the head-final X’ parameter (27) over the head-initial option(19b). In main clauses, the finite verb moves out of VP to an empty complementizer (C) position, and some other constituent of the sentence moves to the left of the C position, resulting in verb-second order. This would lead one to expect that German, like Japanese, should have postpositional constituent order in PP, noun-final constituent order in NP, and determiner-final order in DP. None of these predictions is borne out. Thus, it seems that in German, some heads precede their complements, while others follow them. This implies either that head-complement order is not determined by a single parameter setting of X-bar theory for all categories in a given language, or that there are category-specific movement operations that distort an underlying harmonious head-complement order. For example, if German is presumed to be uniformly head-final, one could propose that prepositions, nouns, and determiners undergo leftward movement out of the X’ constituent where they originate. Alternatively, if German is presumed to be uniformly head-initial, one could assume that the object (and certain other complements of V) undergo movement to a position preceding V’ in subordinate clauses. The hypothesis that syntactic movement, rather than an X-bar parameter, is primarily responsible for cross-categorial variation in constituent order has come to be the most widely adopted view. Moreover, this approach has been extended to account for cross-linguistic variation in constituent structure; instead of assuming that languages vary in terms of the initial placement of a head in relation to its complements, it is now widely accepted that languages vary primarily in terms of the application of syntactic movement of verbs, subject DPs, object DPs, and so forth. This has opened up a Pandora’s box of analytical possibilities in terms of how to account for the various major consitituent order types. In this light, consider the status of the subject position in VSO languages such as Classical Arabic and Modern Irish. Since the canonical position of the subject DP in these languages lies between the verb and the object DP, there appears to be no way to account for this solely in terms of an X-bar theoretic parameter. Many early X-bar theoretic analyses of VSO languages assumed that these languages lack an IP constituent entirely, assuming instead that in such languages sentences are formed simply by head-initial VPs, with the subject DP merging directly with the verb and the object DP within VP. This implied that such languages lack a constituent analogous to the English VP, containing a verb and its object DP but not the subject DP. But Anderson (1983) provided evidence for the existence of a counterpart to the English VP in such languages in non-finite clauses, and Stowell (1989) proposed that VSO languages have the same basic clause structure as SVO languages, with the finite verb undergoing movement to C, just as in German main clauses, in which the finite verb moves to P2 (see Poletto, this volume). In 211

9781441124609_Ch12_Final_txt_print.indd 211

11/27/2012 11:07:56 AM

The Bloomsbury Companion to Syntax

French and earlier forms of English, finite thematic verbs also move to the sentence-initial C position in questions in the same way. According to this account, VSO languages have essentially the same structure as SVO languages, and the subject DP can therefore be defined by (23) or (25) as well.

3.3 Case Assignment So far I have focused almost entirely on generative theories of constituent order, which are primarily germane to the descriptive generalizations about subjects discussed in Section 2.1, and I have said comparatively little about how these theories interact with generative accounts of grammatical case assignment and thematic role assignment. I will now address these issues, though my remarks will be brief because of space considerations. The dominant theoretical approach has been to assume that both grammatical case and thematic roles are assigned to DPs on the basis of the positions that they occupy in the grammatical constituent structure. Let us first consider the status of thematic roles. The lexical entries verbs and other predicate-denoting categories such as nouns and adjectives are assumed to contain not only a syntactic category label (N, V, A, etc.) but also a specification of their phonetic form and of their semantics, where the latter minimally includes a list of the thematic roles that they assign. This representation of a verb’s argument stucture is sometimes called a “thematic grid,” following Stowell (1981). Most thematic roles are assigned to complements of the predicative lexical head within X’; thus, if a verb is lexically specified as assigning a Patient thematic role, it will normally select a DP complement to which this thematic role can be assigned. Thus, the complements that a verb or other predicative category selects will normally be a direct reflection of its lexically specified argument structure. As each of the verb’s thematic roles is assigned to a complement that it selects, the verb’s thematic grid is said to become “saturated.” Williams (1980) proposed that there is an important exception to this, however: among the verb’s thematic roles, one is designated as an “external” thematic role, which must be assigned to a constituent lying outside the VP that the verb heads. Thus, when a verb has merged with all of its complements (and specifiers, if any) to form V’ and VP, one of the argument positions in the verb’s thematic grid still remains unsaturated, namely the external argument. Williams assumed that a verb’s argument structure must ultimately be fully saturated, and he proposed that the verb’s external thematic role is assigned by a general rule of “predication,” which assigns the thematic role associated with the “open” argument of the VP to a DP occurring as the structural sister of that VP. Since he assumed Chomsky’s ternary branching sentence structure mandated by (21), this correctly allowed for the external argument to be assigned to 212

9781441124609_Ch12_Final_txt_print.indd 212

11/27/2012 11:07:57 AM

Changing the Subject

the subject of the sentence. A minor adjustment to his predication rule would be required to make it compatible with subsequent theories of the subject DP as the Specifier of IP or TP, for example, by stipulating that the external argument should be assumed to be the closest DP c-commanding the VP. In any case, Williams’s theory of external arguments and predication relied critically on the long-held assumption that subject DPs originate outside VP. Williams did not provide a comprehensive theory of the principles determining which of the verb’s thematic roles is designated as the external thematic role, but he implicitly assumed some version of the thematic hierarchy discussed above in Section 2.2. In this way, his theory provided a formal means of accounting for generalization (10). Generative theories of grammatical case assignment have likewise been based on the idea that case is assigned to DPs on the basis of the positions that they occupy in the hierarchical constituent structure of the sentence. The literature on this topic is too vast to summarize comprehensively here, but I will provide a brief sketch of some leading ideas. Early generative accounts assumed that accusative case is always assigned to a DP (or NP) directly dominated by VP (or V’, in terms of X-bar theory) and that nominative case is assigned to the DP (or NP) directly dominated by S (or IP or TP, in terms of later theories of phrase structure.) With the advent of Vergnaud’s “case theory” in the late 1970s, this view had to be revised in order to accommodate the theory’s account of “NP movement” in constructions involving passive and unaccusative verbs, as well as the distribution of overt and null DP subjects in non-finite (infinitival) clauses. Vergnaud’s central idea was that DPs must occur in structurally casemarked positions in order to be phonetically overt, and that certain DP positions are not case-marked, requiring DPs that originate in them either to be phonetically null (as in the case of the null subjects of infinitival control clauses) or to move to a higher case-marked DP position (as in the case of DPs that originate as the objects of passive or unaccusative verbs.) The crucial contrast relevant to the theory of nominative case assignment concerns the contrast between finite (tensed) and infinitival clauses. Here Vergnaud suggested that not all subject positions are assigned nominative case; only the subject positions of finite (tensed) clauses have this property. More specifically, he suggested that nominative case is assigned by finite inflection (tense and/or agreement) to the DP sister of Infl (assuming the structure in (21)). This enabled his theory to do two things. First, it explained why only finite clauses have nominative case-marked subjects, and why only subjects have nominative case, thus deriving generalization (11). Second, it enabled his theory to assume that infinitival control clauses contain phonetically null subject pronouns (often called PRO), allowing for a simplification in the theory of external thematic role assignment (since it was no longer necessary to assume that infinitives lack a subject position, and that thematic role assignment works entirely differently with infinitives.) 213

9781441124609_Ch12_Final_txt_print.indd 213

11/27/2012 11:07:57 AM

The Bloomsbury Companion to Syntax

4 Generalizing the Subject Position Across Syntactic Categories Although early X-bar theory provided a reasonably compelling account of cross-categorial symmetry in terms of head-complement stucture, its account of subjects was much more messy, as noted by Stowell (1981, 1983). One of the central ideas motivating Chomsky’s original (1970) account of X-bar theory involved parallels between sentences and derived nominals, as illustrated in (28): (28) (a) The enemy destroyed the city. (b) The enemy’s destruction of the city. Chomsky’s theory provided a parallel structure for the V’ constituent destroy(ed) the city and its nominal N’ counterpart destruction of the city, but its account of the “subject” of the derived nominal in (28b) was not in fact parallel to its account of the subject of the sentence in (28a); in (28b) the pre-nominal genitive DP was assumed to occupy the Specifier of NP position, but the subject of the sentence was assumed to occupy a non-Specifier position dominated by the ternary branching Sentence node. One of the central aims of my proposal in Stowell (1981) to replace the ternary branching category S with an IP structure conforming to X-bar theory was to resolve this inconsistent treatment of subjects; under the revised approach, subject NPs (i.e. DPs) always occur in Specifier positions—the Specifier of IP in the case of (28a) and the Specifier of NP in the case of (28b). While this allowed for a partially parallel treatment of the subjects of sentences and derived nominals, it did not in itself establish full cross-categorial parallelism for Specifier positions, and it did not provide a uniform account of external thematic role assignment consistent with Williams’s theory of predication. With respect to the first issue, it was widely assumed that lexical categories such as VP, AP, and PP lack subject positions; and with respect to the second issue, the subject of a derived nominal could hardly be external to the NP if it occupies the Specifier of NP position. Abney’s later (1987) theory of DP structure provided a possible solution to the second problem, if one assumed that prenominal genitive DPs in derived nominals actually reside in the Specifier of DP rather than in the Specifier of NP; both types of subjects (of sentences and derived nominals) could be assumed to originate in the Specifier of the category immediately dominating the phrase that has an external thematic role to assign. On this view, NP would more closely resemble VP in lacking an XP-internal subject position, contrary to Chomsky’s (1970) proposal.

214

9781441124609_Ch12_Final_txt_print.indd 214

11/27/2012 11:07:57 AM

Changing the Subject

To address the issue of the lack of a fully general cross-categorial parallelism with respect to subjects, there were two analytical options available. One approach would have been to accept Williams’s theory of predication and external arguments, and to assume that lexical categories never contain Specifier positions to which “external” thematic roles can be assigned. This would amount to saying that the reason why lexical categories (VP, AP, NP, and PP) cannot contain subject DPs in their Specifier positions is not that they lack Specifier positions per se, but rather because the principles of thematic role assignment demand that one thematic role must always be external to the XP containing the lexical predicative head (V, A, N, or P). On this view, subject DPs should only be able to occur in non-lexical (functional) categories such as IP or DP. However, the theory of functional categories was only in its infancy during the early 1980s (the only functional categories assumed in Stowell (1981) were CP and IP) and a solution along these lines did not suggest itself. The other analytical option was to reject Williams’s theory of external thematic role assignment and to assume that all XPs, including those headed by lexical categories such as V, N, A, and P, can contain subject arguments in their Specifier positions. In Stowell (1981, 1983) I argued that phrases such as VP, AP, and PP actually do contain XP-internal subjects in their Specifier positions, which can be overtly realized in so-called small clause constructions such as those in (29): (29) (a) Max saw [Bill eat an apple]. (b) Frank considers [Susan intelligent]. (c) We want [you off this ship]. Traditionally, the boldfaced DPs in (29) were assumed to be objects of the verb, directly dominated by V’, but I suggested that they actually occupy the Specifier positions of the bracketed constituents, (VP, AP, and PP respectively). My account of these constructions was modeled on Vergnaud’s (1977) account of infinitives containing overt accusative case-marked subjects: although these subjects bear accusative case assigned by the main verbs, they nevertheless occupy the Specifier positions of the phrases headed by the predicative lexical heads that assign thematic roles to them. On this view, Williams’s theory of “external thematic roles” should be revised so as to assume that the external thematic role assigned by a lexical head X is assigned to a DP occupying the Specifier position of the XP headed by X; this DP is thus external to X’ but not external to XP. I also suggested that, in cases where the main verb selecting a “small clause” XP complement is unable to assign accusative case to the Specifier position of

215

9781441124609_Ch12_Final_txt_print.indd 215

11/27/2012 11:07:57 AM

The Bloomsbury Companion to Syntax

its XP complement, the DP originating in that Specifier position must undergo “raising” to a higher subject position, parallel to the application of raising out of infinitival complements of raising and passive verbs, as in (30): (30) (a) Max seems [—to be crazy] (b) Max seems [—crazy] (c) Max is [—crazy] On this view, seem is always a raising verb, and be is as raising verb as well, at least when it selects a small clause complement, as in (30c). Although I assumed that verbs assign their subject thematic role to the VP-internal specifier position in small clause contructions like (29a), I still assumed that they assign their subject thematic role externally to the specifier of IP position if the VP is not the complement of a higher verb, in sentences such as Bill ate an apple. Koopman and Sportiche (1991) resolved this inconsistency by arguing that verbs always assign their subject thematic role to the Specifier of VP, and that the subjects of all sentences containing active verbs originate in this position and then raise to the Specifier of IP position. Their proposal has since been widely adopted, allowing for a fully general X-bar theoretic account of subject DPs: subjects bearing thematic roles always originate in the Specifier position of a lexical category such as VP, AP, or PP, though they may undergo movement to the Specifier position of a functional category such as IP (or TP) in order to be assigned grammatical case.

5 Conclusion The big picture that emerges from this review of the history of the subject position is as follows. The earliest models of generative grammar assumed that only NPs could function as subjects, and that there was only one subject position, defined as the NP directly dominated by S. By the early 1990s, the picture had changed in almost every respect. A variety of categories other than NP were assumed to be capable of functioning as subjects; the categories NP and S had been reconceived, as DP and IP (or TP), respectively; the original subject position had also been reconceived, as the specifier of IP (or as the specifier of TP or AgrP); and the class of subject positions had been vastly expanded, to encompass the specifier positions of all categories, including the lexical categories VP, AP, PP, and NP, as well as functional categories such as AgrP and TP. It was now assumed that subjects bearing thematic roles always originate in a position within VP, AP, PP, or NP, though they may undergo movement to one or more “derived” subject positions such as the specifier position of a VP headed by a 216

9781441124609_Ch12_Final_txt_print.indd 216

11/27/2012 11:07:57 AM

Changing the Subject

raising verb, or the specifier of a functional category such as AgrP or TP, in order to be assigned case and/or to satisfy certain feature-licensing requirements. Subsequent theoretical proposals by a number of authors have further extended this approach in several ways. First, the theory of feature licensing has been considerably elaborated within the minimalist program, revising standard assumptions about the factors that trigger movement to derived subject positions. Second, the class of derived subject positions has expanded above and below Pollock’s AgrP and TP, to include the specifier positions of various categories associated with focus, mood, aspect, and voice (FocP, ModP, PerfP, ProgP, VoiceP, etc.) (Rizzi 1997, 2006, this volume); furthermore, it is now widely assumed that there is no single derived subject position among these (such as the specifier of TP) in which the subjects of all sentences end up, even within a given language (E. Kiss (1996), Rizzi (2004) and others). Finally, non-subject DPs (such as direct objects of verbs) are now widely assumed to have a syntactic derivation almost identical to that of subjects; Larson (1988) suggested that objects originate as specifiers of VP, and various authors have suggested that objects undergo movement to the specifiers of VP-external functional categories. All of these more recent developments have further refined our understanding of what counts as a subject position and complicated the task of distinguishing subjects from other types of arguments, but these are topics for another paper.

Notes 1. I should note in passing that an interesting alternative version of DP theory has recently been proposed by Sportiche (2005), according to which D and NP are not initially merged; rather, D and NP originate in different regions of the phrase structure tree of the clause, and are only combined as a result of syntactic movements. Because of space constraints I will not address the implications of Sportiche’s theory here. 2. A substantial body of evidence uncovered during the early years of generative grammar, based on syntactic processes such as VP coordination, VP movement, and VP deletion established conclusively that these auxiliary verbs lie outside VP (or at least outside the minimal VP containing the main verb).

217

9781441124609_Ch12_Final_txt_print.indd 217

11/27/2012 11:07:57 AM

13

Alignment and Grammatical Relations1 Doris L. Payne

Chapter Overview What are Grammatical Relations? The Formal Side of GRs: Signifiant The Conceptual Side of GRs: Signifié The Typology of GRs: Alignment The Conventionalization of GRs: l’ arbitraire du signe

218 221 226 232 238

1 What are Grammatical Relations? Grammatical relations (GRs) have to do with the nature of the relationship between a predicate or a construction and its core arguments. Since the early twentieth century, views of GRs have changed radically, with remarkable lack of consensus. Pre-generative structuralists took “subject” and “object” as universal, undefined, and un-analyzable primitives (Bloomfield 1933/1984; Gleason 1955/1961; Hockett 1958; Pike 1982). Since Chomskian generativism, most theoretical approaches have worried significantly about GRs. Keenan (1976a) laid out about 30 properties he thought relevant to the notion of “subject.” Building on this since the last quarter of the twentieth century, deconstructionist approaches have dissected GRs into more elementary components. In the process, some have suggested that some languages completely lack GRs. To help anchor the discussion, what most scholars are concerned with under the label of GRs includes subject, nominative, object, accusative, ergative, absolutive, internal vs. external argument, split intransitivity, and sometimes inverse-direct phenomena. 218

9781441124609_Ch13_Final_txt_print.indd 218

11/20/2012 9:25:44 AM

Alignment and Grammatical Relations

1.1 Primitives Perhaps the broadest theoretical distinction concerns whether GRs are taken as primitives, needing no further definition within a theory, or whether they are defined on the basis of other concepts. In Relational Grammar (RG), GRs were explicitly taken as primitives and were given the nondescript labels “1,” “2,” and “3” (Perlmutter and Postal 1984), perhaps to avoid controversy surrounding terms like “subject,” “object,” and “indirect object,” as well as to try to maintain that in their ontology GRs were not semantic. Whether they are primitives in Lexical Functional Grammar (LFG) is a bit unclear. Bresnan (1982a: 5, 385–6) states they are; yet at times they are defined by feature combinations abbreviated as +/−o (“objective”) and +/−r (“restricted” in semantic role) (Bresnan and Kanerva 1989; Bresnan and Moshi 1990). The Chomskian generative tradition asserts that GRs are entirely derivative from constituency and hierarchical dominance (Chomsky 1965, 1986a, inter alia). This is a configurational conception of their ontology (Section 2.2). In a widely cited article mentioned above, Keenan (1976a) surveyed approximately 30 semantic, syntactic, and pragmatic properties of subjects across a variety of languages. He suggested that the more of these properties a given argument had, the better a subject it was. This “prototype” (Rosch 1978) approach to GRs has not been widely followed (but cf. Givón 2001a: 177). However, Keenan’s work was hugely influential in spurring others to tease apart phenomena that enter into understandings of GRs. A range of modern approaches have assumed what I would call a Sausserian, or “sign-like,” view of GRs, taking them as conventionalized relationships between elements of morphosyntax (or even just phonology) and—at least in their diachronic origins—some more functional notion, be it semantic (Actor, Agent, Patient, Instrument, Undergoer, etc.), pragmatic (Topic), cognitive (perspective or attention flow), or a mix of such factors. Such approaches include some versions of construction grammar (Cro 2001) like Role and Reference Grammar (RRG; Van Valin and LaPolla 1997), most practitioners of typological–functional linguistics (Givón 2001a; Palmer 1994, inter alia), and cognitive linguists (Langacker 1987/1991; Tomlin 1995, inter alia). Sections 2 through 5 elaborate the sign-like view, as this is how alignment is generally approached. Like hermit crabs, Optimality theoreticians simply borrow notions from other theories (as their dominant concern is with relative rankings of principles). For example, Pesetsky (1997) adopts a Chomskian configurational view of GRs, while Aissen (1999) works with Silverstein’s (1976) salience features (Section 3.4). Categorial grammars (Wood 1993) and phrase structure grammars (e.g. Head-Driven Phrase Structure Grammar; Pollard and Sag 1994) are exceptions to the modern trend in their relative silence on the ontology of GRs. Rather like the pre-generative structuralists, they largely take GRs for granted. 219

9781441124609_Ch13_Final_txt_print.indd 219

11/20/2012 9:25:45 AM

The Bloomsbury Companion to Syntax

1.2 Universality A second basic issue dividing theoretical approaches concerns universality: Are GRs found in all languages? Is a particular type of GR (e.g. “subject,” “external argument”) universal? Is the type and even existence of GRs language specific? Can the type of GR vary by construction within a single language? Answers to such questions are grounded in one’s definitions and methodology. Within RG, GRs were, in a very Platonic way, said to be “universally available” to languages. Chomskian generativism assumes that “external” versus “internal argument” are universal (see Stowell in this volume), even though this can entail very abstract, essentially semantic claims about D-structure. For typological–functional linguists, the answer to the first two of the questions posed above is increasingly “no,” and the answer to the second two is “yes.”

1.3 Terminology Grammatical relations terminology can certainly be confusing. Sapir (1921) used personal relation roughly for what is now termed grammatical relation (Perlmutter and Rosen 1984; Palmer 1994; Van Valin and LaPolla 1997), grammatical function (Bresnan 1982a: 5–6; Andrews 2007: 165), or syntactic role (Cro 2001). Givón (1984: 135ff ) refers to subject and object as case roles, a term Fillmore (1977), Longacre (1976), and others use for semantic roles (SRs). Later, Givón (2001a: 105) uses grammatical role for what others call grammatical relation (while Palmer 1994: 4–5 uses grammatical role for what others call semantic role). Alignment most generally means that distinct entities (e.g. two car wheels; two semantic concepts) share a formal or behavioral property (e.g. the wheels are spatially parallel; the semantic concepts are expressed with identical forms). Sapir (1917) is acknowledged as the first to clearly recognize alternative linguistic alignment patterns, though he did not use this term. Though they dealt with the relevant concepts, such typological stalwarts as Givón (2001a), T. Payne (1997), Shopen 1985/2007, and Cro (1990) did not list alignment in their indices. The term alignment was apparently initiated by Harris (1990), and is increasingly employed in typological circles as the notion of “GR” is deconstructed into more basic elements, and as more and more patterns relating form and conceptual categories are recognized. By 2003, the second edition of Cro ’s Typology and Universals repeatedly mentions alignment, and three chapters of The World Atlas of Language Structures Online (Haspelmath, et al. 2008) devote articles to the topic.2

220

9781441124609_Ch13_Final_txt_print.indd 220

11/20/2012 9:25:45 AM

Alignment and Grammatical Relations

2 The Formal Side of GRs: Signifiant The basic formal issue concerns what distinctions in grammar correspond to conceptual groupings of arguments (Section 3). Formal phenomena are o en divided into two types, following Keenan (1976a). “Coding” phenomena involve linguistic elements with phonological form, such as case markers, bound-pronominal, and agreement forms on verbs, as well as constituent order. “Behavioral” phenomena involve syntactic processes like control of reflexivization, NP deletion under co-reference, what controls the interpretation of a “floated” quantifier or allows a possessor to be external to the phrase containing the possessed item, etc. In an alternative way of dividing formal properties, Van Valin and LaPolla (1997), following Dixon (1972, 1979), distinguish “pivot” from “control” and other properties. Roughly, control concerns intra-clause phenomena such as which argument(s) can control the interpretation of a reflexive or the reference of a bound-pronominal marker on a verb. Pivot concerns inter-clausal phenomena such as which argument(s) can be elided under co-reference or what can be the head of a relative clause. Before elaborating how grammatical properties can align with concepts (Sections 3 & 4), I briefly illustrate some coding, behavioral, control, and pivot properties.

2.1 Case, Verb Agreement/Bound Pronominals, and Other Morphological Properties Morphological case distinctions are rampant in European, many Native American, non-Niger-Congo African, and Australian languages. Here I illustrate with Chickasaw (Muskogean). In (1), Pam is an argument of the transitive predicate ahoba “think,” while ilbak “arm” is an argument of intransitive kobaffa “break.” Both Pam and ilbak receive the case marker -(a)t.  (1) Chickasaw Pam-at   [ ilbak-at kobaffa ]  im-ahoba. Pam-sbj arm-sbj break 3dat-think “Pam thinks her arm is broken.” In contrast, in (2)–(3), holisso “letter” and itti’ “tree” lack -(a)t even though they are arguments of “write” and “go around.” This asymmetry in distribution of -(a)t is part of what differentiates Chickasaw subject and object GRs.

221

9781441124609_Ch13_Final_txt_print.indd 221

11/20/2012 9:25:45 AM

The Bloomsbury Companion to Syntax

 (2) Chickasaw Holisso holissochi-li-t binni’li-tok. letter write-1sg.a-sr sit- st “I was sitting writing a letter.”  (3) Chickasaw Mali-t itti’ apakfoota-li-tok. run-sr tree go.around-1sg.a-pst “I ran around a tree (once).” Argument-marking on verbs can be restricted to just (some) core participants; it can involve distinct sets of forms for different core arguments. Boundpronominal suffixes in Spanish are controlled exclusively by what is commonly called “subject” in Spanish: in (4), verb suffixes vary for a number of intransitive and transitive subjects, but not for objects.  (4) Spanish (a) La mujer sali-ó. f.sg.the woman exit-pfv.3sg “The woman le .” (b) Las mujeres sali-eron. f.p .the women exit-pfv.3pl “The women le .” (c) La mujer golpe-ó la pelota / a la niña f.sg.the woman hit-pfv.3sg fsg.the ball dat fsg.the girl “The woman hit the ball / the girl.” (d) Las mujeres golpe-aron la pelota / a la niña. dat fsg.the girl f.sg.the women hit-pfv.3pl f.sg.the ball “The women hit the ball / the girl.” (e) La mujer golpe-ó las pelotas / a las niñas. f.sg.the woman hit-pfv.3sg f.pl.the balls dat fpl.the girls “The woman hit the balls / the girl.” Languages may have fairly unique coding (and behavioral) properties. Maa (Nilotic) clauses are generally verb-initial, but arguments can be fronted in marked constructions. If certain arguments are fronted, the verb takes a relative clause prefix: n- if the fronted NP is grammatically feminine (5a–b), ɔ- if it is masculine singular, and ɔɔ- if it is masculine plural. But if certain other NPs are fronted (6–8), these prefixes do not occur. (In)ability to control the occurrence and form of the relative clause prefix in fronting constructions is part of what

222

9781441124609_Ch13_Final_txt_print.indd 222

11/20/2012 9:25:45 AM

Alignment and Grammatical Relations

distinguishes “subject” and “non-subject” categories in Maa. (Glosses “i” and “ii” are discussed in Section 4.1.) (5)

Northern Samburu Maa (a) N-kέráί n-á-ίshɪr-ɪtá. f-child.i f.rel-f.sg.rel.nom-cry-prog “It is the child that is crying.” (b) Párákúóίsho n-á-ίtulȗb wealth.i f.rel-f.sg.rel.nom-cause.to.be.greedy “It is wealth that makes people greedy.”

l-tʊ́ŋáná. m-people.i

(6)

Northern Samburu Maa Sobúá rɔ́nkáί e-tu-dúŋ-ū-á. herding.stick.i thin 3-pf-cut-toward-pf “It is a thin herding stick that he cut down.”

(7)

Southern Kenyan Maa Ɛn-cháshô e-itáyun-yίék-i ɛnk-ájί. 3-make-ins-pass f.sg-house.i f.sg-ra er.i “It is the ra ers that are used to make a house.”

(8)

Southern Kenyan Maa Tɛ=n-kʊtʊ́k áké έ-tá-ár-á á-yɪmákί ɛn-torrónī obl=f.sg-mouth.ii just 3-pf-beat-pf inf.sg-talk.about f.sg-wrong.i ɛnyέ pɔɔkί. his.i all.i “It is only by mouth that he beat him to expose all his wrongdoing.”

2.2 Constituency and Constituent Order Chomsky (1965) defined subject as the NP immediately dominated by the sentence node and object as the NP immediately dominated by the VP node at the level of D[eep]-structure. In Chomsky (1986a), subject was the argument immediately dominated by the I[nflectional] P[hrase], while object was a structural “sister” of V. In Minimalism (Chomsky 1995), all core arguments were said to be within the verb phrase at D-structure, but object was the structural sister of V while subject was higher within the V-bar unit. In all versions of this tradition, GRs are defined simply by constituency relationships. In simple English clauses there is very good evidence that Verb+Object forms a constituent, while the Nominative (Section 4.1) subject is external to that unit. But the theory claimed these were universal patterns, even when the consequence was an overly abstract analysis of some languages. 223

9781441124609_Ch13_Final_txt_print.indd 223

11/20/2012 9:25:45 AM

The Bloomsbury Companion to Syntax

Some Austronesian languages (e.g. Malagasy [Rasolofo 2006], Balinese [Sidhakarya 1995], and Toba Batak [Schachter 1984]) have transitive clauses in which a verb plus an immediately following NP form a constituent: [[Verb IA] EA]. In Malagasy, either the EA or IA can be Actor or Patient (with concomitant changes in verb form). In the following examples from Rasolofo (2006), placement of the adverbial umali shows the boundary of the verb phrase and hence what is the IA.  (9) Malagasy [[VP] A] construction; A is the EA, and P is the IA: (a) n-i-tangu ni pwara umali ilai p-aN-vuli. pst-ea.a-pick def pear yesterday def nml -ea.a-plant “The farmer picked the pear yesterday.” (b) *n-i-tangu umali ni pwara ilai p-aN-vuli. pst-ea.a-pick yesterday def pear def nmlz-ea.a-plant “The farmer picked the pear yesterday.” (10) Malagasy [[VA] P] Construction; P is the EA, and A is the IA: (a) n-u-tangu-ana ilai p-aN-vuli umali ni npst-u-pick-ea.p def nmlz-ea.a-plant yesterday def “The farmer picked some pear yesterday.” (b) *n-u-tangu-ana umali ilai p-aN-vuli ni npst-u-pick-ea.p yesterday def nmlz-ea.a-plant def “The farmer picked some pear yesterday.”

pwara. pear pwara. pear

The EA NP has different grammatical properties than an NP internal to the V+NP unit; for example, only the EA can be questioned, relativized on, or fronted. Further, based on a corpus study, Rasolofo (2006) argues that many Malagasy clauses with transitive verb stems behave as fully transitive both when the EA is Patient and when it is Actor. Hence, it does not make functional sense to call a two-argument clause with an external Patient a “passive” (contra Keenan 1976b and Keenan this volume). We return to the significance of this in Section 4.1. In some Panare (Cariban) constructions, objects may occur before their verbs or a er the Verb+subject constituent. There is good evidence that a verb plus following Nominative subject NP comprise a constituent (Gildea 1993a); rigid V-subject order, ungrammaticality if sentential adverbs are inserted between V and subject, and phonological phenomena reveal the constituency. Such tests are used to argue that two elements comprise constituents in many other languages, so there is no reason to dismiss them for Panare. Essentially all

224

9781441124609_Ch13_Final_txt_print.indd 224

11/20/2012 9:25:45 AM

Alignment and Grammatical Relations

pragmatically unmarked transitive constructions in Yagua (Peba-Yaguan of Peru) have rigid Verb-subject-object order (Payne and Payne 1990), and there are no straightforward morphosyntactic tests supporting a Verb+Object unit. The relevant Panare and nearly all Yagua constructions are evidence against the adequacy of universal configurational definitions of “subject” versus “object.”

2.3 Behavioral and Pivot Properties Behavioral asymmetries include the ability of an argument to: be expressed as the subject of a passive or as direct or primary object, control the interpretation of reflexive pronouns, be deleted from a subordinate clause if co-referential with a matrix clause participant, be “raised” into a matrix clause from an embedded predicate to which it is semantically related, “launch a floated quantifier,” have a possessor in a separate constituent from the relevant possessed argument, be structurally “cle ed” from the rest of the predication, be the head of a relative clause, be the target of anaphoric co-reference in chained clauses, and other language-specific processes (Keenan 1976a and this volume; Schachter 1977b; inter alia). When a behavioral asymmetry involves cross-clause phenomena, Dixon (1979), Van Valin and LaPolla (1997), inter alia, identify as “pivot” that argument which controls or is the target of the phenomenon. In the Chickasaw examples (1–3) above, a same-reference -t operates on the basis of a transitive+intransitive subject pivot (D. Payne 1980). In Malagasy complex constructions, the EA is the pivot. In English, intransitive/transitive subjects can be deleted under co-reference (11), but objects cannot (12). Similarly, English subjects can undergo “raising” (13a–c), but objects cannot (13d). (11) (a) Roberti hopes Øi to sing. (b) Roberti hopes Øi to pick Joyce. (c) Roberti hopes Øi to be picked by Joyce. (d) Joycei got money, Øi went out, and then Øi bought papaya. (12) (a) *Roberti hopes Joyce will pick Øi. (b) *Joycei went out, and then Robert saw Øi. (13) (a) Joycei seems Øi to have gone out. (b) Joycei seems Øi to have picked Robert. (c) Roberti seems Øi to have been picked by Joyce. (d) *Roberti seems Joyce to have picked Øi.

225

9781441124609_Ch13_Final_txt_print.indd 225

11/20/2012 9:25:46 AM

The Bloomsbury Companion to Syntax

Van Valin and LaPolla (1997: 278–85) demonstrate that different grammatical phenomena need not pick out the same argument as the pivot and as the controlee/controller in even a single construction (e.g. Sama, Austronesian). RRG therefore largely abandons the idea of “subject,” taking construction-specific pivot alignments and controller/controlee alignments as the only viable GR-like categories.

3 The Conceptual Side of GRs: Signifié Comparison of passive and active constructions in (11)–(13) shows that argument semantics does not determine what can be deleted in English. Semantics also does not determine pivot and control privileges in Malagasy and many other languages. This brings us to examine whether there is any conceptual content to GRs. In RG, GRs were taken as primitives without meaning; that was, a er all, presumably the point of “grammatical” as opposed to semantic or pragmatic relations. Nevertheless, Perlmutter and Postal (1984: 97) hopefully entertained a non-definitional link between GRs and meaning; the “Universal Alignment Hypothesis” suggested that principles of Universal Grammar should predict the initial GR borne by each nominal based on the meaning of any particular predicate. But at the same time, Rosen (1984) argued from empirical data that the semantics of arguments does not cross-linguistically or languageinternally predict syntactic behavior (at least on any simple synchronic basis; Section 5). If GRs are assumed to derive strictly from constituency (Section 2.2), there is also no meaning per se in their ontology. Nevertheless, Baker (1988) proposed the “Universal Theta Assignment Hypothesis,” which claimed a non-definitional link between semantic (“theta”) role and constituency: at D-structure, Patient (or T eme) is supposedly always linked to an internal argument position, while certain “higher” SRs will always link to an external argument position. On the other extreme, if a language shows a one-to-one mapping between some well-defined and limited meaning (Agent, etc.) or a function (Topic, identifiability, etc.) and some particular form, then by Occam’s razor, positing an extra notion like a “GR” is unnecessary (Section 4.5). Thus, RRG’s position is that syntactic GRs exist in a language only if there is at least one construction that treats some—but not all—semantic or pragmatic roles with identical grammar for syntactic purposes (especially for pivot-dependent processes; Van Valin and LaPolla 1997: 253–5). This approaches a sign-like view of GRs. Functional linguists, and especially construction grammarians, generally take the stance that linguistic forms or grammar should not exist without an essential tie to something conceptual or 226

9781441124609_Ch13_Final_txt_print.indd 226

11/20/2012 9:25:46 AM

Alignment and Grammatical Relations

functional, and that GRs are no different from other kinds of linguistic form (see especially Langacker 1987/1991). Again, the content must be somewhat abstract or semantically broad (or there is no need to invoke a concept like “GR”). But if a grammatical form or process is conventionalized for, say, Agent, E periencer, and even Patient—but not Location—or if the form expresses the highest SR according to some hierarchy (Section 3.2), or a topical participant out of some restricted semantic range, then there is conceptual content to that grammar, albeit limited. The conventionalization is ultimately no different in nature from that involved in English /kæt/ meaning “a four-footed fuzzy creature that meows”: both are signs. In exploring the conventionalized content of GRs, functional–typological linguists have found it helpful to work with etic features, much as a fieldworker begins by recording phonetic elements before determining language-specific emic groupings of sounds. This is where a deconstructionist approach to GRs has really set in, as linguists find in one language a er another that seemingly similar grammatical forms and processes do not always express the same content.

3.1 Valence Perhaps the most basic etic element involved in all alignment models is valence. One can approach valence on semantic or syntactic bases; either—or a mix—are found in particular analyses. Syntactic valence is the number of core arguments of a verb or clause. Semantic valence is the number of necessary participants in a proposition. Situations may require one participant (laugh, die), two (hug), three (give, put), none (rain), etc. Determining semantic valence is not as simple as it sounds. For the most central sense of the English lexeme cut, my elicitation with native speakers reveals they easily intuit three core items involved: a “cutter” (Agent), a “cut thing” (Patient), and “something the cutter uses” (Instrument). Yet because of how English cut behaves syntactically (Mary cut the bread. The knife cut the bread. but *Mary cut the bread the knife. *The knife cut the bread by Mary), many linguists (with some slight of argumentational hand) are wont to say that the prototypical sense of cut has just two core participants, perhaps an Actor (subsuming both Mary and knife in these examples) and a Patient. Goldberg (1995) suggests we can introspect the number of core participants by evaluating how many entities necessarily come to mind when contemplating “X-ing happened” (e.g. “Cutting happened”). But this yields markedly unclear results. Rather by deductive fiat Van Valen and LaPolla (1997: 152–3) hypothesize that German helfen “help” has one “macro”-role, while English help has two, which then is said to “explain” the language-specific syntactic behaviors. 227

9781441124609_Ch13_Final_txt_print.indd 227

11/20/2012 9:25:46 AM

The Bloomsbury Companion to Syntax

3.2 Semantic Roles and Related Concepts The preceding discussion raises what have been called participant roles, semantic roles (SR), and generalized (“macro”-semantic) roles. All such roles are defined by predicates, and do not exist independently of them. Participant roles are finely-grained concepts like “cutter,” “cut thing,” etc., specific to predicates like cut, chop, slice, hack, etc. (Goldberg 1995). SR, as the term is most widely used, refers to more abstract concepts that group compatible participant roles. For example, Agent subsumes “cutter,” “hugger,” “giver,” etc., but not “one who dies” or “thing that rains.” Role compatibility can be approached via lexical decomposition analyses (Jackendoff 1983). For example, cut, chop, slice, hack, as well as hug, give, put, etc. are assumed to contain an abstract lexicalized “DO” predicate; and it is really from “DO” that the notion of Agent derives. Whether or not one accepts a sign-like view of GRs, there is still usually an assumption that SRs must be related in some constrained way to GRs. To account for limits on possible relationships, widely varying approaches assume a hierarchy of SR-like notions. In general, if more than one core role occurs in a predication, the higher role should evidence (or be linked to) privileged syntactic properties called “subject,” “external argument,” “pivot,” etc. Bresnan and Kanerva (1989) propose agent > benefactive > goal > instrument> patient/ theme > locative. Dik (1978: 76) proposes agent > goal > recipient > benefaciary > instrument > location > theme. Givón (2001a: 139) proposes agent > dative/benefactive > patient > locative > instrument/ associative > manner. Givón (2001a: 200) proposes agent > dative/benefactive > patient > locative > other. These purportedly universal hierarchies do not agree in detail or in ranking of particular elements. Some place Instrument below Patient/Theme, and others the reverse. Some propose Experiencer and others Dative. In actuality, research shows that one construction in a given language can operate on a different hierarchy of SRs from another construction in the same language (D. Payne 2008). Some approaches argue that the semantic elements relevant to GRs are fewer and more general than those listed in the preceding hierarchies. In RRG these are Actor and Undergoer. Dowty (1991) proposes Proto-Agent and ProtoPatient. DeLancey (2000) argues for Agent, Theme, and Location.3 Grimshaw (1990)’s “thematic” hierarchy Agent > Experiencier > Goal/Source/Location > Theme interacts with a distinct so-called aspectual hierarchy of Cause > Other. Any Cause argument will be the syntactic external argument; since this assignment is based on a distinct hierarchy from the thematic hierarchy, it allows for situations where Theme can be Subject even if Experiencer is simultaneously present.

228

9781441124609_Ch13_Final_txt_print.indd 228

11/20/2012 9:25:46 AM

Alignment and Grammatical Relations

RRG’s generalized roles subsume narrower SRs. Actor can correspond to Agent (as the prototype Actor), but also Effector, Experiencier, Instrument, Perceiver, Possessor, and Recipient, depending on principles grounded in the semantic decomposition of verbs (Van Valin and LaPolla 1977: 143–6). The Undergoer is essentially whatever participant is “presented as being most affected” (145). Hence, what SR will be Undergoer cannot be fully predicted simply from semantic decomposition for at least some verbs. Undergoer generally corresponds to Patient (as the prototype Undergoer), Theme, Location, Recipient, or Source. The range of SRs which can be treated as Actor or Undergoer can vary narrowly among languages: an Instrument cannot be Actor in either Lakhota or Jakaltek: in Maa, Undergoer must include Possessor (D. Payne 1997). Cro (2001: 134–70) develops a more elaborated generalized role system involving S, A, P, T, G, Ex, and St. Cro ’s S is “a polysemous category made up of a cluster of semantic Participant Roles found in one-participant situations” (136). A and P roles (mnemonic for Agent and Patient, respectively) are involved in two-participant situations. The roles A, T (mnemonic for Theme), and G (mnemonic for Goal) are relevant to three-participant situations. Ex (Experiencer) and St (Stimulus) are necessary because some verbs grammatically treat Experiencer like A, but others treat it like P; and Stimulus may be sometimes treated like A, but other times as P. For example, the Experiencer of English think is treated like a prototypical A (I think [something] and I cut it.) But the Experiencer of Chickasaw yimmi “think” is treated like other Chickasaw Ps, and not—for instance—like the Experiencer of pĩs “see”: (14) Chickasaw (a) Is-sa-pĩs-tok. 2sg.a-1sg.p-see-pst “You saw me.” (b) Pĩs-li-tok. see-1sg.a-pst “I saw him/her/it.” (c) Sa-yimmi 1sg.p-think “I think [something]” Aside from Cro (2001), I know of no one else who considers S to be a “generalized SR.” S is different in nature from A, P, etc. because single arguments of oneparticipant situations can, in every language known, range from prototypical Agent to prototypical Patient depending on particular predicates; and Agent and Patient simply do not behave together as subtypes of a single SR concept.

229

9781441124609_Ch13_Final_txt_print.indd 229

11/20/2012 9:25:46 AM

The Bloomsbury Companion to Syntax

Dixon (2010) seems to embrace morphosyntactic conceptualizations of S, A, and O (rather than “P”), explicitly calling them “syntactic functions.” However, Dixon (2010: 116) affirms that assigning an argument to A or O status has a semantic basis. An A can be practically identified as “that argument whose referent is most likely to be relevant to the success of the activity,” thus including at least Experiencers. O is identified as “that argument whose referent is most likely to be saliently affected by the activity.” For extended intransitive and extended transitive (ditransitive) predicates, Dixon adds E (mnemonic for “ ‘extension’ to core”). In sum, in practice Dixon employs S, A, O, and E as etically useful sometimes-semantic concepts for determining alignment patterns (Section 4). Between Cro ’s (2001) attempt to define all of S, A, and P semantically, and Dixon’s (2010) claim that S, A, and O are all “syntactic functions,” Van Valin and LaPolla (1997: 143) explicitly claim “there is no semantic equivalent of intransitive subject” (S), while Actor (A) and Undergoer (U) are explicitly semantic in nature. Despite alternative conceptions of them, typologists commonly use S, A, and P (O/U) (more rarely T, G, St, Ex) as etic concepts for exploring emic grammatical alignment systems in and across languages. These concepts are etic as many grammatical systems group S+A, or S+P, or some other combination. Nevertheless, an assumption that generalized S, A, and P roles are sufficient for investigating GRs or alignment systems will lead us astray because this etic set provides no way of accounting for systems sensitive to Experiencer separately from Agent, or for Experiencer separately from Patient, as happens in some Daghestanian case-marking languages (Comrie and van den Berg 2006), or for patterns in some Split-S languages (Section 4.2).

3.3 “Pragmatics” and Cognition What GRs are ultimately about may not be expression of semantic concepts, but (the grammaticalization of) cognitive “perspective” (Fillmore 1977; Dik 1978), primary vs. secondary clausal “topic” (Givón 2001a: 138), or psychological figure vs. ground. Langacker (1987, 1991) says what is taken as the figure (“trajectory”) and as the ground (“landmark”) is a matter of perspective. When RRG characterizes “object” as the participant presented as most affected, it recognizes speakers can adopt alternative cognitive perspectives on what is “most affected.” For example, if I am in an accident I might—depending on my communicative purposes—present the car as most affected (That lorry hit my car in the right front fender!), or present myself as most affected (That lorry hit me in the right front fender!). Fillmore (1977) discusses this at some length for English verbs like load whose object cannot be predicted directly off of a SR hierarchy. In 230

9781441124609_Ch13_Final_txt_print.indd 230

11/20/2012 9:25:46 AM

Alignment and Grammatical Relations

They loaded hay onto the lorry, “hay” is presented as perhaps most affected; but in They loaded the lorry with hay, “the lorry” is presented as being fully loaded. In an excellent empirical study in this domain, Tomlin (1995) examines morphosyntactic expression of the cognitive “focus of attention,” which has to do with intensive on-line mental processing.4 He suggests that what linguists have long groped toward via terms like topic and (discourse) theme might correspond to the psychologists’ focus of attention. Using films involving identically sized but differently colored fish approaching and then swallowing one another, a speaker’s attention is drawn to a particular fish via a flashing symbol. The speaker formulates a sentence about the visual state of affairs within a few milliseconds of the flashing symbol. If the time between the flashing symbol and sentence formulation is sufficiently constrained, presumably the speaker’s focus of attention will not wander to anything else before the sentence is formulated. The analyst can then examine the grammar coding the item over which a symbol flashed: English speakers consistently use English-subject grammar for the highlighted item (appropriately adopting active or passive constructions). Using somewhat similar methodology, Forrest (1999) examines situations where one could theoretically say The orange circle is over the yellow triangle, or The yellow triangle is under the orange circle. Here also, English speakers reliably code as subject that item to which attention is visually drawn. For certain other languages, however, Tomlin finds that what is grammatically called the “subject” of those languages does not always correlate with visually highlighted items. He thus concludes that English subject grammar codes focus of attention (within the confines of the experiment), but that we cannot generalize cross-linguistically to say that “subject” universally codes cognitive focus of attention. These studies do not mean that lexicalization of SR features into particular predicates will have no bearing on construction choice (e.g. active versus passive) or on lexical verb choice in particular occasions of speaking (e.g. one might choose Avocados please me immensely over I like avocados immensely, when avocados is in the focus of attention, as *Avocados are liked immensely by me is odd, if not impossible). But they do suggest that the signifié of English “subject” might be less related to SR than has commonly been thought. In any case, cognitive focus of attention should be taken as a prime and as an etic tool for investigating GRs.

3.4 Salience Actually, it has been suggested since at least Silverstein (1976) that the conceptual content of GRs can correspond to “salience” and not SRs. Salience features include identifiability, animacy, humanness, deixis (first and second 231

9781441124609_Ch13_Final_txt_print.indd 231

11/20/2012 9:25:46 AM

The Bloomsbury Companion to Syntax

versus third person; proximate versus distal participants), what is relatively more “topical” in discourse (and ultimately what is in the cognitive foreground vs. the background). In Kapampangan, an Ergative–Absolutive construction is used only for definite arguments (Mithun 1994); definiteness is necessary for Malagasy external argument status (Rasolofo 2006); animacy and definiteness may affect primary vs. secondary object coding (Bantu, Hyman and Duranti 1982; Olutec, Zavala 2000). But salience features generally inhere in referents independently of the relationship between predicates and their arguments. In contrast, S, A, P, and SRs are grounded in valence and in the meaning of lexical predicates/constructions and do not exist independently. Foreground–background are similarly relational notions (there is no foreground without a background). Even focus of attention might conceivably imply that something else is not in the focus of attention; certainly in language use, interlocutors are busy relating what is in the focus of attention to other elements that are being processed at some level. But a participant can be animate or identifiable in and of itself. So if certain grammar responds to identifiability or animacy, then that grammar+meaning combination departs quite significantly from the traditional notion of a grammatical “relation.” Nevertheless, within a single proposition one referent may be animate while another is not; one item might be identifiable while another is not. And such asymmetries can affect what is chosen as the privileged syntactic argument, as primary versus secondary object, as a differentially marked object (Section 4.3), etc. Salience hierarchies play a role in inverse–direct systems (Section 4.4). And surely, saliency features are not divorced from what is most frequently in the cognitive foreground or focus of attention.

4 The Typology of GRs: Alignment We now address the typology of alignment. Most work on alignment has adopted some version of the S-A-P triad as etically relevant. Though he did not use the terms alignment, S, A, or P, the idea of an alignment typology started with Sapir (1917), who rejected interpreting certain Native American languages (Eskimo, Muskogean, Siouan, and others) as being “passive” in ethno-psychological nature, relative to the lens of European accusative systems. Sapir (1917) identified the alignment patterns in Figure 13.1. This Figure names alignment categories with terms used by Comrie (1978), Dixon (1979, 1994), Plank (1979), and others. Ovals indicate that a single coding or behavioral phenomenon applies to the etic entities contained within the ellipse. Thus, in an Ergative–Absolutive system, the A displays one kind of coding or behavior, while the S+P combination displays distinct grammar. In a Neutral system, S+A+P all receive the same grammar. 232

9781441124609_Ch13_Final_txt_print.indd 232

11/20/2012 9:25:46 AM

Alignment and Grammatical Relations

S

S A

P

P

A

Nominative Accusative

S A

P

Tripartite

Ergative Absolutive

Sa

S A

P Neutral

Sp P

A Split-S

Figure 13.1 Common alignment patterns Alignment patterns are o en loosely talked about as applying to languages as wholes (e.g. “Basque is an ergative–absolutive language”). But they should be understood as potentially pertaining to different grammatical subsystems and constructions within languages. For example, Chickasaw case marking is Nominative–Accusative (Section 4.1), but bound-pronominals display an even more-elaborate split-S system than that sketched in Figure 13.1 (Section 4.2). Commonly, the term “split system” is used when different constructions or processes in a language display different alignments.

4.1 Nominative–Accusative, Ergative–Absolutive, Tripartite, and Neutral Nominative–Accusative alignment is common for case marking, bound pronominal, verb agreement, and other language-specific morphological patterns as in Maa fronting constructions (5–8), and for various control, pivot, and definite participant reference phenomena in discourse. The dominant Nominative–Accusative case subtype is to have an overt case-marker for the Accusative (P) category, and none for Nominative (S+A), which is typically also the citation form. A second subtype overtly marks both Accusative and Nominative with different forms. Less frequent is the Marked Nominative type, though it is well-attested in some African families (König 2006). In this subtype, S+A is “marked” in one or both of two senses, while P is unmarked. In the first sense, S+A has an overt case-marker, but P has a zero. This is the basic pattern found in Chickasaw/Choctaw, where Nominative NPs carry -(a)t (1), while 80–90 percent of Accusatives are zero-marked (Broadwell 2006: 74–6) (2–3).5 In the second sense, the P case form has a wider distribution in the overall grammar of the language than does the S+A form. 233

9781441124609_Ch13_Final_txt_print.indd 233

11/20/2012 9:25:46 AM

The Bloomsbury Companion to Syntax

Ergative–Absolutive patterns are less common than Nominative–Accusative, and are typically relevant to coding phenomena like case marking and boundpronominal systems. The dominant case subtype is to have an overt marker for the Ergative category and zero for Absolutive, usually also the citation form. In Dyirbal (Dixon 1972), both pivot and coding phenomena have Ergative– Absolutive alignments. However, most languages with some Ergative– Absolutive coding also have Nominative–Accusatively aligned grammatical subsystems, especially for pivots. In Austronesian languages like Malagasy (Section 2.2), the external argument (EA) is clearly the syntactic pivot. For those who might equate “subject” with a Nominative S+A grouping (DeLancey 2004), the Malagasy EA is clearly not part of “Nominative” in [[V A] P] transitive clauses; nor is the EA part of “Absolutive” in [[V P] A] transitive clauses. The emic pivot category is S+EA (which of course reduces to EA), an alignment type not represented in Figure 13.1. Languages with both Ergative–Absolutive and some other alignment pattern(s) in different parts of their grammars are commonly termed “splitErgative” languages. Split-Ergative systems may be determined by person hierarchies (approximating inverse systems; DeLancey 1981, section 4.4), tense, aspect, main versus subordinate clause status, and other grammatical parameters. Less-frequent typological patterns involving split-alignments do occur. Oirata (Trans New Guinea) is unusual with Accusative coding/control properties and an arguably Ergative pivot (Donohue and Wichmann 2008). Gildea and Alves (2010) argue for Nominative–Absolutive alignment in some Cariban and Je constructions, where A and O have distinct coding properties but S displays a union of A and O properties. In Rosani (Iranian), past-tense clauses very unusually give A and O identical case marking and S a distinct form, while nonpast clauses have a Nominative–Accusative case system (J. Payne 1980). Complete absence of case marking and verb agreement/bound pronouns in Thai constitute Neutral A/S/P patterns. Maa shows Neutral case alignment for fronted NPs, a Nominative–Accusative case alignment for post-verbal NPs, and a consistent Nominative pivot. In (5) –(7) above, Maa nominals and their modifiers have either an “i” or a “ii” form, varying by tone. It makes sense to call the “ii” pattern Nominative and the “i” pattern Accusative, as “ii” is used for S+A following the verb, and “i” for P following the verb. However, both “i” and “ii” forms occur in numerous other syntactic environments, some having nothing to do with GRs. The environments for the “ii” form are fewer than that for the “i” form, and hence Maa is Marked-Nominative in König’s second sense. But all the distributions together show that the language has a case-marking “split” since there are both Neutral and Nominative–Accusative subsystems. Tripartite systems are rare. As T. Payne (1997) points out, in intransitive clauses the single argument does not need to be distinguished from anything else; so it is inefficient to mark S distinctly from both A and P. 234

9781441124609_Ch13_Final_txt_print.indd 234

11/20/2012 9:25:47 AM

Alignment and Grammatical Relations

4.2 Split-S In Split-S systems, some intransitive arguments are treated like transitive A’s, and others like P’s. Split-S systems have also been called “stative-active” (Sapir 1917), “split-intransitive” (Merlan 1985), “fluid-S” (Dixon 1994); Donohue and Wichmann (2008) prefer “semantic alignment,” as split treatment of S can be based on lexical aspect of the verb (various Native American languages, Loma [Mande]), telicity (Georgian), volitionality or agency of S (Muskogean, NakhDagestanian), affectedness of S (Central Pomo), deixis or viewpoint (Pilagá), and pragmatic factors (see further below). Dahlstrom (1983) and Mithun (1991: 542) observe that split-S systems call into question the very notion of “S” as a viable etic category. Chickasaw and Choctaw (Muskogean) display an elaborate Split-S system, distinguishing three intransitive patterns which we could term Sa, Sp, and Sd, the last taking the same bound pronominals as Dative objects (Munro and Gordon 1982; D. Payne 1982). Most Chickasaw/Choctaw intransitive verbs are invariant for whether S takes the same bound pronominal markers as (most) A, P, or Dative participants of (di)transitive clauses. Compare the first person singular affixes: (15) Chickasaw Am-asihĩlha-tok. 1sg.dat-say-pst “He/she said [something] to me.” (ditransitive) (16) Chickasaw (a) Chi-shooli-li-tok. 2sg.p-hug-1sg.p-pst “I hugged you.” (transitive) (b) Is-sa-shooli-tok. 2sg.a-1sg.p-hug-pst “You hugged me.” (17) Chickasaw Aya-li-tok. go-1sg.a-pst “I went” (agentive/active intransitive) (18) Chickasaw Sa-banna. 1sg.p-want “I want it.” (inactive intransitive)

235

9781441124609_Ch13_Final_txt_print.indd 235

11/20/2012 9:25:47 AM

The Bloomsbury Companion to Syntax

However, some Chickasaw roots have “fluid” options depending on participant features such as volitionality; witness the famous Chickasaw triad chokma-li (be. good-1sg.a) “I act good,” sa-chokma (1sg.p-be.good) “I am good,” an-chokma (1sg. dat-be.good) “I feel good.” Fluid-S patterns can also be discourse-driven. In particular, Sp-P alignment can correlate with presentational or “change of scene” function (some South American languages; T. Payne 1992). A French “de-topicalization” construction codes S like a P in that the S occurs postverbally and no longer controls verb agreement (Lambrecht 2000: 643): (19) Il est/*sont venu trois it is/*are come three “There came three women.”

femmes. women

Similar facts are true in many Bantu languages (Creissels 2009). Word order (though not verb agreement) in English inversion constructions, and perhaps presentative constructions cross-linguistically, can be analyzed as falling into this category, as in Here comes my bus.

4.3 Object Alignment Split-S systems show that “S” need not be a unified category. The same can be true of A and P. In differential-object marking systems, elements of the supposed P “category” are treated differently based on saliency factors like specificity, identifiability, animacy, or humanness (many Iranian languages, Bossong 1985; differential coding of Spanish P’s in (4c–e); etc.). In Tibetan, transitive Patient “O’s” (e.g. of bsad “kill”) are unmarked; while Goal “O’s” (e.g. of gzhus “hit”) take Dative forms (DeLancey 1984). Dryer (1986) discussed Primary–Secondary versus Direct–Indirect object systems. In a Primary–Secondary system, the transitive P and ditransitive G (using Cro ’s 2001 generalized role labels) have identical grammar; while ditransitive T has distinct grammar. In a Direct–Indirect object system, transitive P and ditransitive T have identical grammar, while ditransitive G is distinct. Primary– Secondary object alignment is very frequent, surely motivated by saliency factors: participants in G role are generally human or animate, while those in T role are typically inanimate. Finally, transitive P and ditransitive T and G can all have the same grammar. Such “symmetrical object” systems are found in some (but not all) Bantu languages (Bresnan and Moshi 1990). The Yagua symmetrical object system treats Causers, Causees, basic P’s, and applied objects identically when two items are both animate or both inanimate (Payne and Payne 1990; see Song, this volume, for a general discussion of causative constructions). 236

9781441124609_Ch13_Final_txt_print.indd 236

11/20/2012 9:25:47 AM

Alignment and Grammatical Relations

4.4 Inverse–Direct and Hierarchical Alignment Inverse–direct systems involve a hierarchy in which first and second person (singular) outrank third persons (and perhaps plurals, as plurals typically involve third persons) (Zúñiga 2006; T. Payne 1999). Coding of A and P typically involves marking in the verb but sometimes also case and constituent order, and depends on the hierarchical ranking of A and P in a clause. “Direct” grammar results when A is higher on the hierarchy than P. “Inverse” grammar is used when P is higher than A. The hierarchy 1 > 2 > 3 is common. But in Algonquian languages, it is 2 > 1 > 3Proximate > 3Obviative. Both A and P are marked on the verb, but a direct or an inverse marker shows whether the highest argument is A or P. In (20a), note that ni- corresponds to A and -w to P; while in (20b), ni- corresponds to P and -w to A (Dahlstrom 1986: 43). (20) Plains Cree (a) Ni-wa.pam-im-a.-w-a 1sg-see-obv-dir-3-obv “I see his son.” (b) Ni-wa.pam-iko-yi-w-a 1sg-see-inv-obv-3-obv “His son sees me.”

o-kosis-a. 3-son-obv o-kosis-a. 3-son-obv

In Carib of Suriname (Gildea 1994: 192), an inverse–direct option is applicable in a certain verb-initial construction, where the hierarchy is simply 1/2 > 3 as one prefix references both 1A+2P and 2A+1P. If a third person plus speech-actparticipant is involved, distinct A (direct) versus P (inverse) bound pronominals occur, but only one appears in any clause. An A form occurs if A is higher on the hierarchy than P, and a P form occurs if P is higher. S participants take bound pronominals similar to the transitive Inverse/P set, approaching an absolutive alignment pattern (or possibly a historical Split-S pattern). (21) Carib of Suriname (a) kï-kuupi-ya. 1/2-bathe-TNS “I bathe you. / You bathe me.” (b) mi-kuupi-ya. 2A-bathe-TNS “You bathe him.” (direct pattern) (c) a-kuupi-ya-ῆ. 2P-bathe-TNS-EVID “He bathes you.” (inverse pattern) 237

9781441124609_Ch13_Final_txt_print.indd 237

11/20/2012 9:25:47 AM

The Bloomsbury Companion to Syntax

4.5 Languages and Constructions without GRs As noted earlier, some languages may lack GRs like subject, object, etc. Tibetan case markers directly code features like +agentive or +goal (DeLancey 1984, 2004). That is, the case forms neither neutralize across a restricted range of SRs, nor are they sensitive to anything like “focus of attention” or “most topical participant.” In quite another type, LaPolla (1990), following Li and Thompson (1976), argues that Chinese does not have GRs corresponding to a restricted SR range, but rather just a Topic–Comment construction, as grammatical subjects are not consistently marked by position, nor is there agreement or case markers, and they may be entirely omitted. In contrast, topic expressions (italicized in the following free translations) are extremely characteristic of simple sentence structure. (22) Chinese (a) zhè kē shù yèzi hĕn dà. this CL tree leaf very big “This tree, (its) leaves are very big.” (Li and Thompson 1976: 15) (b) nèi bĕn shū chūbǎn le. that CL book publish PFV “That book, (someone) has published it.” (Li and Thompson 1976: 88)

5 The Conventionalization of GRs: l’ arbitraire du signe As more than just Nominative and Accusative alignments have been discovered, linguists have explored why the various alignment patterns should exist. Grammatically splitting the treatment of S has ready explanations in fairly transparent semantic or pragmatic factors. However, split-intransitive patterns can have non-transparent irregularities (Rosen 1984, inter alia). For instance, the Lakhota split-S system treats the S of “be hungry” like an A. But here the historical reason is clear: “be hungry” derives from the transitive root “want” plus an incorporated noun “food,” that is, “want food.” The Nominative S+A group is sometimes said to reflect topicality. But topicality is not true of all S’s in discourse. Hence a Nominative category simply “conventionalizes” all S’s together with A’s. The absolutive S+P group may reflect discourse-new status of participants (Givón 2001a; DuBois 1987, inter alia), though again there must be conventionalization here since not all S’s are new information. When all S’s are treated like A’s or all like P’s, analogical (Harris 1990) or other grammatical re-analysis processes must have set in because S’s are never semantically or pragmatically homogenous. As grammatical systems

238

9781441124609_Ch13_Final_txt_print.indd 238

11/20/2012 9:25:47 AM

Alignment and Grammatical Relations

become more and more expanded, they become less semantically or pragmatically transparent. (Spilt-)ergative patterns have also received various functional explanations, and increasingly these are different in nature. DeLancey (2004) discusses how formal ergative–absolutive patterns can sometimes be sensitive to transitivity (Mizo or the Tibeto-Burman family), but other times more transparently responsive to discourse-pragmatic status (Tibetan) or to deixis factors like speech act vs. non-speech act participant (Sahaptin, of the Penutian family). It was thought for some time that if a split between alignment systems depended on tense-aspect, then Ergative–Absolutive alignment would occur in past/perfective environments and Nominative–Accusative in non-past/ imperfective environments, due to “P orientation” in perfective situations but to “A orientation” in imperfective situations (DeLancey 1981). However, some Cariban languages have Ergative–Absolutive alignment in non-past/ non-perfective environments while Nominative–Accusative patterns surface in past/perfective environments. The modern Cariban distribution is due to how certain possessed nominalizations became reanalyzed as new tense-aspect categories. In that reanalysis, the syntactic possessor in a transitive nominalization was automatically reanalyzed as Ergative (Gildea 1993b, 1998; DeLancey 2004). As Creissels (2008) has perspicaciously said, “In the evolution of languages, some changes may lack functional explanations of their own and instead be mechanically triggered by changes occurring in other areas of the grammar.”

Notes 1. I am grateful to Robert Carlson and Tom Payne for comments on this paper. 2. Relational Grammar explored the “Universal Alignment Hypothesis” (UAH) in the mid-1980s. The UAH was concerned with whether participants could be assigned to “initial” (and primitive) GRs on the basis of semantic role and not with patterns of relationship between conceptual entities and morphosyntax. 3. DeLancey’s categories are best viewed as a type of generalized role; for example, his “Location” is akin to others’ Goal, Recipient, and Experiencer grouped together. 4. This should not be confused with “contrastive” or other “marked focus.” 5. Overt Accusative case markers are conditioned by occurrence of three objects, extraposition of an object, use of a determiner, and non-native phonology.

239

9781441124609_Ch13_Final_txt_print.indd 239

11/20/2012 9:25:47 AM

14

Passives and Antipassives Edward L. Keenan

Chapter Overview Passives Antipassives

240 258

1 Passives In this chapter we discuss passives and, briefly, antipassives in the world’s languages. Passive is prominent in the traditional grammars of Latin (Gildersleeve and Lodge 1913), and Greek (Smyth 1920, 1984) and has played a major role in generative grammar from Chomsky (1957) through all formally oriented theories: Relational Grammar (Perlmutter and Postal 1983b), Arc Pair Grammar (Postal 1986), Lexical Functional Grammar (Bresnan 2001), Unification based theories (Sag et al. 2003), Principles and Parameters (Chomsky 1981; Jaeggli 1986; Baker et al. 1989), and finally, Minimalism (Collins 2004; Lasnik and Uriagereka 2005). In addition there are many useful typological studies of passives, including: Siewierska (1984), Keenan (1985), Haspelmath (1990), Fox and Hopper (1994), Givón (2001a), and Keenan and Dryer (2007). Some source books we have found typologically informative are Dixon and Aikhenvald (2000), Perlmutter and Rosen (1984), and Shibatani (1988). Simple passives are illustrated in (1b)–(3b). (1) (a) Mary hired Susan. (b) Susan was hired (by Mary). (2) Turkish (Perlmutter and Postal 1983b) (a) Hasan-ø bavul-u aç-tı. Hasan-nom suitcase-acc open-pst “Hasan opened the suitcase.”

240

9781441124609_Ch14_Final_txt_print.indd 240

11/20/2012 9:25:11 AM

Passives and Antipassives

(b) Bavul-ø (Hasan tarafindan) aç-ıl-dı. suitcase-nom (Hasan by) open-pass-pst “The suitcase was opened (by Hasan).” (3) Hebrew (a) Ha saba gidel et the grandfather.nom bring.up.act-pst act-def “The grandfather brought up the child.”

ha yeled. the child

(b) Ha yeled gudal al yedei ha saba. the child.nom bring.up.pass-pst on hands the grandfather “The child was brought up by the grandfather.” The passives in (1)–(3) differ in regular ways but also share several properties. Most obvious is that what distinguishes passive from active sentences lies in the verb form. The verb combines with other material to form a one place predicate (P1), which combines with one argument, its subject, to form a sentence. The subject has the same case and pronominal forms as in an active sentence. Generally if you can form passive predicates in a language, the rules you need anyway to form sentences will apply as they do in non-passives. So we treat passive and other valency changing operations as ways of forming (complex) predicates from predicates. Passive clauses (sentences) are defined as ones built from passive predicates. Further, a predicate-level treatment is semantically natural. In general if B is derived from A then the semantic interpretation of B is some function of that of A. This is our basis for understanding novel utterances: we understand what the lexical items that compose them mean and we understand how things built in that way take their meaning as a function of the meanings of what they are built from. The meaning of be hired in (1b) is a function of that of hire in (1a): be hired holds of an individual x if and only if for some individual y, y hired x. It follows that (1a) entails (1b) without its agent phrase. So Mary hired Susan entails Susan was hired since the argument expressions denote individuals. Also this semantics guarantees that Passive does not change theta roles of arguments. Susan is Patient (Theme) in both (1a nad b). But this entailment relation does not extend to active and passive sentences as their argument expressions may not denote individuals. (4a) does not entail (4b), as someone besides Mary may have hired other applicants: (4) (a) Mary hired exactly one third of the applicants. (b) Exactly one third of the applicants were hired. These entailment facts extend to other types of predicates derived by valency changing operations. For example, in Malagasy (Western Austronesian, 241

9781441124609_Ch14_Final_txt_print.indd 241

11/20/2012 9:25:12 AM

The Bloomsbury Companion to Syntax

Madagascar) (5a) is built from an intransitive verb and is entailed by (5b), built from the causative of that verb by prefixing amp yielding a transitive verb: (5) (a) n-i-omehy ny pst-act-laugh the “The child laughed.”

zaza. child

(b) n-amp-i-omehy ny zaza Rabe. pst-caus-act-laugh the child Rabe “Rabe made the child laugh.” But (6a) does not entail (6b) since more than a third of the children may have laughed (some without being caused to do so by Rabe). (6) (a) n-amp-i-homehy ny ampahatelon’ny ankizy fotsiny Rabe. pst-caus-act-laugh the third’of the children only Rabe “Rabe made only a third of the children laugh.” (b) n-i-homehy ny ampahatelon’ny ankizy pst-act-laugh the third’of the children “Only a third of the children laughed.”

fotsiny. only

1.1 Periphrastic Passives (1b) is called periphrastic as it is composed of two (or more) independent words— the auxiliary verb be, and the past participle of hire. A common choice of passive auxiliary cross-linguistically is “be” or “become.” German (7a and b) uses both with somewhat different meanings: (7) (a) Das Haus wird verkauft. The house becomes sold:ptcp “The house is being sold.”

(b) Das Haus ist verkauft. the house is sold:ptcp “The house is sold.”

(7a) has a process interpretation: if it obtains you may still be able to buy the house; but if (7b) holds you are too late. Old English also has both “become” weorþan and “be” wesan/beon passives, but we do not know if the lexical difference correlates with the process vs. state interpretation as in German (Fischer and van der Wurff 2008: 152).

242

9781441124609_Ch14_Final_txt_print.indd 242

11/20/2012 9:25:12 AM

Passives and Antipassives

Participles in passives behave like adjectives (Haspelmath 1990) or as a base from which adjectives are derived (Wasow 1977; Bresnan 1982b). In Latin the participle in the periphrastic passive (limited to tenses derived from the perfect stem) agrees in gender and number but not person with the subject—the adjectival agreement pattern: “I have been loved” is amatus sum or amata sum according to whether “I” is masculine or feminine; and for a plural feminine group we have “they have been loved” amatae sunt, but amati sunt for a mixed or masculine group. Periphrastic passives are dominant in modern Indo-European languages but are also present in non-Indo-European Basque and Quechua (8). (8) Huallaga Quechua (Webber 1989: 244) (a) Pay chay runa-ta salba-ra-n. he that man- acc save- pst-3 “He saved that man.” (b) Chay runa salba-sha that man save- ptcp “That man was saved.”

ka-ra-n. be- pst-3

A second type of periphrastic passive is one in which the analogue of the auxiliary verb in English is a verb of reception, such as get, receive: (9) Welsh (Sadler 1988) (a) Rhybuddiodd y dyn ferch. warned.3s the man girl “The man warned a girl.” (b) Cafodd merch ei rhybuddio (gan y got.3sg girl 3sg warn by the “A girl was warned (by the man).”

dyn). man

(10) Tzeltal (Mayan) La y-ich’ ‘utel (yu’un s-tat) te  Ziak-e. pst he-receive bawling.out (because his-father) art Ziak-art “Ziak got a bawling out (from his father).” (Keenan 1985) A third type of periphrastic structure, common in Southeast Asia (Chinese, Vietnamese, and Thai), uses a verb of experiencing as an auxiliary. (11) Chinese (Hashimoto 1988) Ta bei (taitai) kanjian. he sustain (wife) see “He is seen (by his wife).”

243

9781441124609_Ch14_Final_txt_print.indd 243

11/20/2012 9:25:12 AM

The Bloomsbury Companion to Syntax

Finally, we note the use of “go” as a passive auxiliary in Hindi: (12) Hindi Murgi mari gayee. chicken killed went “The chicken was killed.”

1.2 Strict Morphological Passives Examples (2) and (3) illustrate strict morphological (sm-) passives, in which the active and passive predicates differ just by affixing or internal vowel change.

1.2.1 Asymmetric Passives The Turkish (2) is asymmetric—the passive verb appears to be derived by affixing the active verb. But Semitic passives, as in Arabic and Hebrew, are symmetric. Active gidel and passive gudal in Hebrew (3) are of equal morphological complexity. Both are morphologically derived from a common source, the triconsonantal root GDL (Gimel, Daleth, Lamed); neither derives from the other. Asymmetric passives are widely distributed: Turkish (2), Kinyarwanda (Bantu; Kimenyi 1988), and Kannada (Dravidian; Dryer 1982) as in (13). (13) Kannada (a) candra-nige nan-na dhvani-” Chandra-dat 1.sg-gen voice- nom “Chandra heard my voice.” (b) candra-nige nan-na dhvani Chandra-dat 1.sg-gen voice.nom “My voice was heard by Chandra.”

keeLis-itu. hear-3s.n. keeLis-alpaTT-itu . hear- pass-pst 3s.n.

(13a and b) differ audibly just by a passive suffix. But the structural differences run deeper: the dative in (13a) is selected by the verb whereas it is merely an oblique Agent in (13b). is a common theta grid for Dravidian verbs, generally being used for ones translating need, like, have, learn, know, and others. Huallaga Quechua (Webber 1989: chapter 11) also has an asymmetric sm-passive as (14) and, thus, like Latin has both periphrastic and sm-passives. (14) (a) Hwan wasi-ta rika-n. Juan house-acc see-3 “Juan sees the house.”

(b) Wasi rika-ka-n. house see-pass-3 “The house is seen.”

244

9781441124609_Ch14_Final_txt_print.indd 244

11/20/2012 9:25:12 AM

Passives and Antipassives

Asymmetric passives may also be prefixing, as in (15) from Sre (Mon-Khmer; Manley 1972). Likely the dominance of suffixal passive morphology just reflects the general preference of suffixing over prefixing in bound morphology, Greenberg (1963b). (15) (a) Cal pa? mpon. (b) Mpon gə-pa? mə cal. wind open door door pass open by wind “Wind opened the door.” “The door was opened by the wind.”

1.2.2 Symmetric Passives Symmetric sm-passives are also found in Latin (present indicative): (16) Active: amare “to love” amo amamus amas amatis amat amant

Passive: amari “to be loved” amor amamur amaris amamini amatur amantur

Neither active nor passive forms are naturally derived from the other. Compare 2pl passive amamini “you(pl) are (being) loved” with the active amatis “you(pl) love”; each derives from a common stem. The passive form is nonetheless always heavier then the corresponding active form. Russian also presents both periphrastic passives and sm-passives, conditioned by aspect, with the sm-passive (17) only imperfective, the periphrastic passive (18) only perfective. (I. Kapitonov and N. Korotkova (personal communication)). (17) (a) Specialisty provodili opros. specialist.nom conduct.impf.pst survey.nom “Specialists conducted a survey.” (b) Opros provodilsja specialistami. survey.nom conduct.impf.pst.detr specialist.inst “The survey was (being) conducted by specialists.” (18) (a) Specialisty proveli speecialists.nom conduct.pvf. pst.pl “Specialists conducted a survey.”

opros. survey.nom

(b) Opros byl provedën specialistami. survey.nom be.pst conducted.ptcp.pvf specialists.inst “The survey has been conducted by specialists.” 245

9781441124609_Ch14_Final_txt_print.indd 245

11/20/2012 9:25:12 AM

The Bloomsbury Companion to Syntax

Philippine type languages (Shibatani 1988a) present many sm-passives. (19a and b) (Malagasy) are derived from a common root enjika, not itself a word; (20b) is a prefixal passive, as is (21b). (22a and b) from Tagalog illustrate an infixal passive. (19) (a) n-aN-enjika (nanenjika) ny pst-act-chase the “Soa chased the dog.”

alika dog

I art

(b) n-enjika-ina I Soa (nenjehin’i Soa) pst-chase-pass art Soa.gen “The dog was chased by Soa.”

Soa. Soa ny alika. the dog

(20) (a) n-i-laza izany aho. pst-act-say that 1.sg.nom “I said that.”

(b) voa-laza(-ko) izany. pst-say-1sg.gen that “That was said (by me).”

(21) (a) maN-roso (mandroso) vary pres.act-serve rice “S/he serves rice to the guest(s).”

ny the

vahiny guest

izy. 3

(b) a-roso ny vahiny ny vary. pass-serve the guest(s) the rice “The rice is served to the guest(s).” (c) roso-ana vary ny vahiny. serve-pass rice the guest(s) “The guest(s) are served rice.” (22) (a) s-um-ampal ng lelaki ang act-slap acc man top “The woman slapped a man.”

babae. woman

(b) s-in-ampal ng babae ang lelaki. pass-slap gen woman top man “The man was slapped by a/the woman.” The rich voice system in Philippine-type languages plays a deeper role than Passive does in European languages. In relative clauses (and with one restriction constituent questions and cle sentences) the voice morphology on the

246

9781441124609_Ch14_Final_txt_print.indd 246

11/20/2012 9:25:12 AM

Passives and Antipassives

verb determines the semantic role of the head noun in the action or state expressed by the main verb of the relative clause (Keenan 1972; Keenan and Comrie 1977; Keenan 2008). Thus relativizing the Theme from (21a) or (21c) is ungrammatical. Only the Theme/Patient can be relativized from (21b) as shown in (23). (23) (a) *ny vary (izay) mandroso ny vahiny izy. the rice (that) act.serves the guest(s) 3 “The rice that he is serving the guest(s).” (b) ny vary (izay) a-roso-ny ny vahiny. the rice (that) pass-serve-3gen the guest(s) “The rice that is being served by him to the guest(s).” (c) *ny vary (izay) roso-a-ny ny vahiny. the rice (that) serve- pass-3gen the guest(s) “The rice that is being served by him to the guest(s).” Similarly the Recipient can only be relativized from (21c), not (21a) or (21b). Obliques such as Instrumentals, Locatives and Benefactives can be relativized from the circumstantial form of the verb, formed by suffixing -ana to an active verb, not just the root, as in (24c). (24) (a) n-aN-vono (namono) ny akoho t-amin’ny antsy Rabe. pst-act-kill the chicken pst-with’the knife Rabe “Rabe killed the chicken with the knife.” (b) no-vono-ina-Rabe (novonoin-dRabe) t-amin’ny antsy pst-kill- pass-Rabe.gen pst-with’the knife “The chicken was killed by Rabe with the knife.”

ny akoko. the chicken

(c) n-aN-vono-ana-Rabe (namonoan-dRabe) ny akoho ny antsy. pst-act-circ-Rabe.gen the chicken the knife “The knife was used by Rabe to kill the chicken.” (d) ny antsy (izay) namonoan-dRabe ny akoho. the knife (that) was+killed+with-by Rabe the chicken “The knife (that) Rabe used to kill the chicken.” In the relative clause in (24d) the verb must be in the circumstantial form so that “knife” binds the subject. Replacing the verb with the passive in (24b) or the active in (24a) is ungrammatical (Keenan 2008). So voice marking in Philippine languages serves to feed relativization (and some other extraction processes).

247

9781441124609_Ch14_Final_txt_print.indd 247

11/20/2012 9:25:13 AM

The Bloomsbury Companion to Syntax

1.3 Passives of Three Argument Verbs: Ditransitives, Causatives, and Applicatives If a language can passivize transitive verbs (P2s, two place predicates) to form intransitive ones (P1s, one place predicates) then, as illustrated in (21b and c), it can passivize ditransitive verbs (P3s) to form P2s—the result still requiring two arguments to form a sentence. Languages vary however with regard to which arguments can be the subject of such passives. Malagasy, using distinct morphology in (23 b and c) allows both Theme and Goal to be the subject. The verb -ome “give” takes both Theme and Goal in the accusative in the active, and either can be the subject of the unique passive, omena. This pattern obtains in Kinyarwanda as well (Kimenyi 1980: 127) and in Japanese (Miyagawa 1989). Double passives from causatives are normal in Malagasy. (25) Japanese (a) Daitooryoo ga gakusei ni kunsyoo o ataeta. president NOM student DAT medal ACC give.PST “The president gave a medal to the student.” (b) Kunsyoo ga daitooryoo ni gakusei ni atae-rare-ta. medal NOM president DAT student DAT give-PASS-PST “A medal was given to the student by the president.” (c) Gakusei ga daitooryoo ni kunsyoo o atae-rare-ta. student NOM president DAT medal ACC give-PASS-PST “The student was given a medal by the president.” (26) Malagasy (a) n-amp-an-sasa (nampanasa) ny lamba azy aho. PST-CAUS-ACT-wash the clothes 3ACC 1SG.NOM “I made him wash the clothes.” (b) n-amp-an-sasa-ina-ko (nampanasaiko) ny lamba izy. PST-CAUS-ACT-wash-PASS-1SG.GEN the clothes 3NOM “He was made to wash the clothes by me.” (c) n-amp-an-sasa-ina-ko (nampanasaiko) azy ny lamba. PST-CAUS-ACT-wash-PST-1SG.GEB 3ACC the clothes “The clothes were made washed by him by me.” It is common to be able to passivize a causative verb, as in (26). (27) illustrates this from Huallaga Quechua (Webber 1989: 236). (27) mana wanu-chi-ka:-na-n-paq. not die-caus-pass-sub-3-purp “In order that he not be killed.” 248

9781441124609_Ch14_Final_txt_print.indd 248

11/20/2012 9:25:13 AM

Passives and Antipassives

It is much less common to causativize a passive verb. However, voa- passives in Malagasy may causativize with the weak causative -aha-. (28) (a) voa-sazy izy noho ny ditrany. pass-punish 3nom due.to the mischief.his “He was punished because of his mischief.” (Randriamasimanana 1986: 580) (b) ny ditrany no n-aha-voa-sazy (nahavoasazy) azy. the mischief.his foc pst-caus-pass-punish 3acc “It was his mischief that got him punished.” A more common paradigm with lexical ditransitives is that just the Theme (French) or just the Recipient (Yindjibarndi and Pama-Nyungan, Australia; Wordick 1982, cited in Dryer 1986) can be the P1 argument; see Dryer (1986). (29) French (a) La photo a été offerte the photo.f.sg has been offered.f.sg “The photo was offered to John.” (b) *Jean a été offert Jean.m.sg. has been offered.m.sg “John was offered the photo.”

à Jean. to Jean la the

photo photo

(30) Yindjibarndi (a) Ngaarta yungku-nha ngayu murla-yi. man.nom give-pst 1sg.acc meat-acc “A man gave me the meat.” (b) Ngayi yungku-nguli-nha mjurla-yi ngaarta-lu 1sg.nom give-pass-pst meat-acc man-inst “I was given the meat by a man.” (c) *Murla yungku-nguli-nha ngayu ngaarta-lu meat.nom give-pass-pst 1sg.acc man-inst “The meat was given to me by a man.” Applicatives are a second valency increasing operation, widely used in Bantu. The -er suffix in (31a) derives the passivizable P2 “die-for” from the P1 “die.” (31) Chicewa (a) Yesu a-na-f-er-a anthu. Jesus he-pst-die-ben-fv people “Jesus died-for the people.” (Alsina and Mchombo 1990) 249

9781441124609_Ch14_Final_txt_print.indd 249

11/20/2012 9:25:13 AM

The Bloomsbury Companion to Syntax

(b) Anthu a-na-f-er-edw-a (ndi Yesu). people they-pst-die-ben-pass-fv (by Jesus) “The people were died-for (by Jesus).” Applicatives play a deep role in Bantu syntax. In (32a) from Kinyarwanda the verb passivizes, but only the Theme, not the Instrument, as its subject (32b and c). But in (32d) the Instrumental applicative extends the valency of the verb to take an Instrumental object, and that verb passivizes, (32e), taking a nominative Instrument as subject. (32) Kinyarwanda (a) Umugore a-ra-andik-a ibaruwa n’iikaramu. woman she-pres-write-asp letter with’pen “The woman is writing a letter with a pen.” (b) Ibaruwa i-ra-andik-w-a n’iikaramu n’umugore. letter it-pres-write- pass-asp with’pen by’woman “The letter is being written with a pen by the woman.” (c) *Ikaramu i-ra-andik-w-a ibaruwa (na) n’umugore. pen it-pres-write-pass-asp letter (with) by’woman “The pen is being written a letter with by the woman.” (d) Umugore a-ra-andik-iish-a ibaruwa ikaramu. woman she-pres-write-inst-asp pen letter “The woman is writing-with a pen a letter.” (e) Ikaramu i-ra-andik-iish-w-a ibaruwa n’umugore. pen it-pres-write-inst-pass-asp letter by’woman “The pen is being written-with a letter by the woman.” (32) illustrates that applicatives in Bantu enable obliques—Instrumentals, Locatives, Manners, constructed as PPs in actives—to be treated as Prepositionless arguments of a derived predicate. The active cannot be passivized to make them subjects but the applicative verb can. Moreover, Kinyarwanda does not relativize objects of Prepositions, only unmarked subjects and objects. So applicatives enable Kinyarwanda to relativize (reflexivize, etc.; Kimenyi 1980: 91) semantic obliques. So they play a syntactic role in Bantu like that of voice in Western Austronesian in (24). (33) (a) *Dore ikaramu umugabo y-aándik-a ibaruwa na (yo). look pen  man he-rel.write-asp letter with it “This is the pen that the man uses to write the letter.”

250

9781441124609_Ch14_Final_txt_print.indd 250

11/20/2012 9:25:13 AM

Passives and Antipassives

(b) Dore ikaramu umugabo y-aándik-iish-a ibaruwa. look pen man he-rel.write-inst-asp letter “This is the pen the man uses to write the letter.” (Exceptionally we marked tone, with á, as it is part of the relativization structure.)

1.4 Agent Phrases and Impersonal Passives Agent phrases in passives (constituents with the same semantic role as the subject of the active) are normally optional, sometimes disallowed (Latvian, Lazdina 1966; Kutenai (isolate) Keenan and Dryer 2007). When present they usually occur as obliques or PPs, typically instrumental, locative, or genitive. The use of English by to mark Agent phrases is typical. Independently by has a locative function (He sat down by the window) and a means function (He won by cheating). In European languages with Agent phrases their frequency is rather small: Svartvik (1966), Stein (1979), and Givón (1979) show 80–85 percent of passives in texts lack Agent phrases. But Keenan and Manorohanta (2004) find a lower figure for Malagasy, about 38 percent. The passives illustrated so far are personal, meaning they take a thematic subject. Impersonal passives are passive predicates that either are sentences or take a semantically empty (expletive) subject, as it in It is raining. (34) and (35) illustrate impersonal sm-passives from Sesotho (Bantu; Baker 1992) and Turkish (Özkaragöz 1988). Missing agent phrases in P1 passives are normally understood to be animate. When present they are case marked as in personal passives. (34) Sesotho: Ho-tsama-uo-a ka-maoto expl-walk-pass-ind with-feet “People travel by foot.”

(ke-batho). (by-people)

(35) Turkish: Burada düş-ül-ür. here fall-pass-aorist “Here one falls.” Dutch (36), from Kirsner (1976), and Latin (37), from Postal (1986), illustrate periphrastic impersonal passives, Dutch with an expletive subject, Latin not: (36) (a) De jongens fluiten. (b) Er wordt (door de jongens) gefloten. the boys whistle there becomes (by the boys) whistled “The boys are whistling.” “There is whistling by the boys.” 251

9781441124609_Ch14_Final_txt_print.indd 251

11/20/2012 9:25:13 AM

The Bloomsbury Companion to Syntax

(37) (a) Milites acriter pugnaverunt. soldiers.nom fiercely fought.3pl “The soldiers fought fiercely.” (b) Acriter (a militibus) pugnatum est. Fiercely (by soldiers.abl) fought.ptcp.n is “There was fierce fighting (by the soldiers).” More challenging to generative treatments are cases in which the passivized P1 consists of a P2 + its argument. In Latin many lexical P2s select an argument in a non-accusative case, o en dative. That argument remains dative in the passive, which is thus subjectless and impersonal. (38) generalizes the pattern for lexical P1s in (39). (38) (a) Brutus mihi invidet. (b) Mihi invidetur (a Bruto). Brutus.nom 1sg.dat envy.act.3sg 1sg.dat envy.pass.3sg (by Brutus) “Brutus envies me.” “I am envied (by Brutus).” (39) (a) Marcus currit. Marcus runs.act “Marcus is running.”

(b) Curritur (ab aliquo). run.pass (by someone.abl) “Running is being done.”

Other languages with impersonal passives are German, Welsh, and Hindi. An example outside the Indo-European family is (40) from Tariana (Arawakan; Brazil; from Aikhenvald 2001). (40b) is the personal passive of the P2 in (40a). (40c) illustrates the same passive affix -kana on an intransitive verb. (40) (a) wali-da episi-da-nuka  nu-phya-ka  di-na. new-clf:round iron-clf:round-top.ob 1sg-buy-rec.pst.vis 3sg.n-f-obj “I bought him a new motor.” (b) wali-da episi-da ka-phya-kana-ka di-a. new-clf:round iron-clf:round rel-buy-pass.rec.pst.vis 3sg.n.f-aux “A new motor was bought.” (c) kiaku ma-kama-kana-de-na. strong neg-get.drunk-pass-neg-rem.pst.vis “One did not get really drunk (on a certain type of manioc beer).”

252

9781441124609_Ch14_Final_txt_print.indd 252

11/20/2012 9:25:13 AM

Passives and Antipassives

1.5 Two Passives from the Same Source The pattern in (38) is standard in Latin grammars, but we found one case (in Horace) where “to envy” formed a personal passive: (41) Cur (ego) invideor? Why (1sg.nom) envy.pass.1sg “Why am I envied?” Keenan and Timberlake (1985a and b) offer more productive cases from North Russian: (42b) is the passive of the P2 “slaughter” taking a nominative Patient subject “calf” with which it agrees. (42c) can be understood as the passive of the P1 “slaughter a calf,” with “calf” accusative and postverbal, the verb in a default non-agreement form. (42) North Russian (a) Ja zarezal 1sg.nom slaughter.act.pst “I slaughtered a calf.”

telenka. calf.acc.sg

(b) (U menja) telenok zarezan. by me calf.nom.sg slaughter.pass.m.nom.sg “(By me) a calf was slaughtered.” (c) (U menja) zarezano telenka. by me slaughter.pass.n.nom.sg calf.acc.sg “(By me) slaughtered a calf (There was slaughtered a calf).” In (43a) from Polish we have the Passive of the P2 “cut down” with a nominative Theme subject, and in (43b) the Passive of the P1 “cut down the linden tree” with “linden tree” postverbal and accusative, the verb in a non-agreement form. Ukranian (Sobin 1985) also presents such passives, so it may be an areal phenomenon. (43) (a) lipa ścięta. linden.nom.f.sg cut.down.pass.nom.f.sg “The linden tree was cut down.” (b) scięto lipę. cut.down.pass.n.nom.sg linden.f.acc.sg “There was cut down the linden tree.”

253

9781441124609_Ch14_Final_txt_print.indd 253

11/20/2012 9:25:13 AM

The Bloomsbury Companion to Syntax

1.6 Iterated Passives Our valency reduction analysis of Passive suggests that it could iterate. The passive of the passive of a P2 would be a P0 (sentence). And we do find cases. (44) Turkish (Özkaragöz 1988) Harp-te vur-ul-un-ur. war-loc shoot-pass-pass-aorist “In war one gets shot (by one).” (The Passive suffix is -ɪn following laterals, -n a er vowel-final stems, and -ıl elsewhere. The vowels undergo front-back harmony surfacing as -ül or -ün.) Such double passivization is productive (Özkaragöz 1988). Biktimir (1988) also discusses them, noting that overt agent phrases are not possible with passives of P1s in Turkish. But they are possible with periphrastic iterated passives in Lithuanian (45b). (45) Timberlake (1982) (a) Tas lapelis vèjo nupūstas. that leaf.nom.m.sg wind.gen blown.pass.m.sg “That leaf was blown down by the wind.” (b) To  lapelio būta vèjo nupūsto. that leaf.gen.m.sg been.nom.m.sg wind.gen blown.gen.m.sg “By the leaf there was getting blown down by the wind.”

1.7 Passives with Control Verbs of Aspect, Desire, Intent, and Modality In many cases passive morphology suggests that complex P2s are formed with lexical P2s and verbs of aspect, desire, etc. Turkish (George and Kornfilt 1977), (46b), and Malagasy (Keenan and Polinsky 1998), (47b), are illustrative. (46) Turkish (a) Ahmet kitab-i oku-maya başla-di. Ahmet book-acc read-inf(dat) begin-pst “Ahmet began to read the book.” (b) Kitap (Ahmet tarafindan) oku-n-may başla-n-di. book (Ahmet by) read-pass-inf(dat) begin-pass-pst “The book was begun to be read (by Ahmet).”

254

9781441124609_Ch14_Final_txt_print.indd 254

11/20/2012 9:25:13 AM

Passives and Antipassives

(47) Malagasy (a) n-i-kasa h-aN-vaky (hamaky) io boky io aho. pst-act-intend fut-act-read that book that 1.sg.nom “I intended to read that book.” (b) no-kasa-ina-ko (nokasaiko) ho-vaky-ina (hovakina) io boky io. pst-intend-pass-1.sg.gen fut-read-pass that book that “That book was intended by me to be read (by me).” The active a-Ss above are cases of obligatory control. But in their passives, the two verbs must passivize together. So it seems that Passive has only applied once. This makes sense if “begin/intend to read” are complex P2s. Passive morphology would spread to both components analogous to how case marking may mark all components of nominals like this fat cat. In both Turkish and Malagasy verbs of aspect (begin, finish, continue), intent (intend, plan, try), and desire (want, desire) form such predicates. A similar construction with modality verbs presents another case of double passives in Kinyarwanda (Kimenyi 1988): (48) (a) Abanyeeshuuri ba-shobor-a gu-som-a ibitabo. students they-may-asp to-read-asp books “The students may read the books.” (b) Gu-som-a ibitabo bi-shobor-w-a n’abanyeeshuuri. to-read-asp books it-may-pass-asp by’students “To read books may be done by the students.” (c) Ibitabo bi-shobor-a gu-som-w-a n’abanyeeshuuri. books they-may-asp to-read-pass-asp by’students “The books may be read by the students.”

1.8 Passives of Raising to Object The last passives we consider are those fed by Raising to Object (RtoO). They play a prominent role in generative theories, enlighteningly presented in Davies and Dubinsky (2004). The basic phenomena are illustrated in (49). (49) (a) Sue believes that she is clever. (b) Sue believes her to be clever. (c) She is believed to be clever. In (49a) believe is a P2, taking a complementizer phrase (CP) that she is clever as first argument: she may be interpreted as bound (coreferential) with the matrix 255

9781441124609_Ch14_Final_txt_print.indd 255

11/20/2012 9:25:14 AM

The Bloomsbury Companion to Syntax

subject Sue. The RtoO expression, (49b), is essentially a paraphrase of (49a) but her is accusative and while understood as subject of be clever cannot be understood as bound by she, indicating that it is a clausemate of believe, as in She saw her. In (49c) believe is passivized with she as subject, still understood as the subject of be clever. The passive in (49c) is usually treated syntactically, being fed by RtoO, which is usually treated syntactically. If we passivize believes in (50a) the result can take the CP as subject, (50a and b), but not she, the subject of be clever. (50) (a) That she is clever is believed (by Sue.) (b) That the Earth is flat was once widely believed. (c) *She is believed that is clever. However, a predicate level derivational view is not without merit. RtoO is lexically conditioned. Few verbs in English that take clausal objects can grammatically replace believe in He believed Sue to be clever: *doubted, *shouted, *whispered, *denied, *ignored, *disproved, *?said, *?hoped, etc. The predicate level view treats believe polymorphically as merging with nominalized Pn’s to form Pn+1’s. that (complementizer) nominalizes P0s, and infinitival to nominalizes P1s. So believe that Flo is clever is a P1 and believe to be clever is a P2. (Thomason 1976; Keenan 1980). (51) (a) John [P1 believes [that [P0 Sue is clever]]] (b) John [P1 [P2 believes [to [P1 be clever]]] Sue] So as a P2 believe to be clever passivizes. Several verbs marginal in the RtoO format are fine when passivized. (52) (a) The director was said / thought / claimed to be dishonest. (b) ??They said / thought / claimed the director to be dishonest. This fact receives some explanation on a predicate level view, as passive eliminates a superficially discontinuous P2 believe . . . to be dishonest. Further, some arguments do occur to the right of this constituent, (53a), and it coordinates with lexical P2s, (53b). (53) (a) John believes to be clever every student he has ever advised. (b) He both admires and believes to be smart every student he has advised. There is a regularity here: “Light to the le , heavy to the right” (Hawkins 2001). This pattern shows up in many other expression types as well: Extra/ 256

9781441124609_Ch14_Final_txt_print.indd 256

11/20/2012 9:25:14 AM

Passives and Antipassives

Intra-position, (54a), and Heavy NP Shi , (54b), and verb-particle constructions, (54c). That short arguments would attract to their verbal heads would be another instance. (54) (a) That Flo laughed is strange

↔

It is strange that Flo laughed.

(b) The soldier who was standing guard all night fell asleep The soldier fell asleep who was standing guard all night. (c) ?I summed that argument up

↔

↔

I summed up that argument.

Kinyarwanda (Kimenyi 1988: 379) exhibits RtoO with manipulative verbs such as command, forbid, promise and supports two passives as in (48). In Malagasy Paul and Rabaovolona (1998) cite some 50 verbs that host RtoO. Here (55a) has an active matrix verb and shows VSO order (“Heavy to the right”). (55b) has a passive matrix with the CP as derived subject (quite natural, unlike English). (55c) shows the RtoO structure, where ny jiolahy “the brigand” is replaceable with an accusative pronoun azy “him,” not nominative izy “he.” The matrix verb is active. In (55d) we passivize the matrix verb and the subject of the complex P1 heverin-dRabe ho nosamborin-dRasoa takes ny jiolahy “the brigand” as subject. (55e) extracts that subject. Extraction of “the thief” from any of (55a)–(55c) is ungrammatical, as is extraction from the object position of the embedded clause with an active verb (not illustrated). (55) [P0 nosamborin-dRasoa ny jiolahy]i. (a) mi-hevitra ti Rabe fa pres.act-thinks ti Rabe that captured.pass-Rasoa.gen the brigand “Rabe thinks that the brigand was captured by Rasoa.” ny jiolahy]. (b) heverin-dRabe fa [P0 nosamborin-dRasoa think.pass-Rabe.gen that captured.pass-Rasoa.gen the brigand “That the brigand was captured by Rasoa was thought by Rabe.” (c) mi-hevitra ny jiolahy ho nosamborin-dRasoa Rabe. pres.act-think the brigand comp capture.pass-Rasoa.gen Rabe “Rabe thinks the brigand to have been caught by Rasoa.” (d) heverin-Rabe ho nosamborin-dRasoa ny jiolahy. think.pass-Rabe.gen comp captured.pass-Rasoa the brigand “The brigand was thought to have been captured by Rasoa.” (e) ny

jiolany (izay) heverin-Rabe ho nosamborin-dRasoa. the brigand (that) think.pass-Rabe comp captured.pass-Rasoa.gen “The brigand who was thought by Rabe to have been captured by Rasoa.” 257

9781441124609_Ch14_Final_txt_print.indd 257

11/20/2012 9:25:14 AM

The Bloomsbury Companion to Syntax

2 Antipassives Like Passive, Antipassive (ANTIP) derives intransitive verbs from transitive ones, but now the subject of the derived verb bears the same semantic role as the subject of the corresponding transitive. The object semantic role may be unexpressed or expressed with an oblique. Antipassives are most visible, and were first noticed, in languages with ergative case marking or verb agreements as ANTIP marked the Agent absolutive, usually zero, whereas in actives it was marked ergative, usually non-zero. All our examples below are from ergative languages, but in principle antipassives can be present in nominative–accusative languages as well. (60) Chukchee (Paleosiberian) (Kozinsky et al. 1988) (a) ətləg-e qora-ŋəqərir-nin. father-erg deer-abs.look.for-3sg.3sg.aorist “Father looked for the deer (for some time).” (b) ətləgən ena-rer-gɁe. father-abs antip-look.for-3sg.aorist “Father did some searching.” The verb in (60b) is intransitive, derived by the ANTIP prefix ena. It has only one agreement affix and its subject is absolutive, the case of lexical P1s. In contrast the verb in (61a) is transitive. Its Theme is absolutive and its Agent is ergative, the case of P2 subjects. The verb agrees with both arguments. The Theme in the antipassive is naturally understood as existentially quantified, not one of zero pronominalization. An expressed Theme of an antipassive verb takes oblique case marking. (61) Chukchee (Kozinsky et al. 1988 ) (a) Ɂaacček-a kimitɁ-ən ne-nlɁetet-ən. youth-erg load-abs 3pl-carry-3sg.aorist “The young man carried away the load.” (b) Ɂaacček-ət ine-nlɁetet-gɁet kimitɁ-e. youth-abs.antip-carry-3pl.aorist load-inst “The young man carried away a load.” In addition to Chukchee ANTIP allows omission of the implied oblique Theme in Chamorro (Austronesian; Cooreman 1988), K’iche’ (Mayan; Campbell 2000), and Burushashki (isolate, Pakistan; Morin and Tiffou 1988). And as with the voices in Western Austronesian and applicatives in Bantu, ANTIP may play a syntactic role. We find two such cases. The first is clause chaining (Conjunction

258

9781441124609_Ch14_Final_txt_print.indd 258

11/20/2012 9:25:14 AM

Passives and Antipassives

Reduction) in which ANTIP can be used to force control of a missing argument. Striking examples are found in Dyirbal (Dixon 1972), Yidiny (Dixon 1977), and Warrungu (Tsunoda 1988: 613)—all Australian languages. A common coordinate deletion pattern in Dyirbal is that an absolutive deletes an absolutive, hence (62b) is ungrammatical. But that meaning is expressible using the -ngay ANTIP form. (62) Dyirbal: (a) bayi yara bani-nyu baŋgun dyugumbiru balgan. masc.abs man-abs come.tns, f.erg woman.erg hit “The man came and the woman hit him.” (b) *Bayi yara bani-nyu, balan dyugumbil balgan. m.abs man.abs come-tns f.abs woman.abs hit “The man came and the woman hit him.” (c) bayi yara bani-nyu, bagun dyugumbil-gu m.abs man.abs come-tns, f.dat woman-dat “The man came and the woman hit him.”

balgal-ŋa-nyu. hit-antip-tns

Dyirbal also illustrates the second syntactic role of Antipassive. It is not possible to relativize the ergative argument of a transitive verb, but when the verb is in the ANTIP form that Agent relativizes as an absolutive argument of an intransitive verb. (63) (a) bayi yara bagun dyugumbil-gu balgal-ŋa-nyu m.abs man.abs f.dat woman-dat hit-antip-tns “The man who hit the woman is returning home.”

banagaŋau. return.home

balgan banagaŋu. (b) *bayi yara balan dyugumbil m.abs man.abs f.abs woman.abs hit return-home “The man who hit the woman is returning home.” This role is shared by Antipassive in West Greenlandic (Eskimo; Fortescue 1984) and in Mayan languages such as Jacaltec (Craig 1977). (64) Jacaltec (a) x-ø-s-watx’e asp-3abs-3erg-make “He made this.”

naj hun-ti’. cl/he one-this

(b) x-ø-w-il naj x-ø-watx’e-n. asp-3abs-1erg-see clf/he asp-3abs-make-antip “I saw the man who made this.” 259

9781441124609_Ch14_Final_txt_print.indd 259

11/20/2012 9:25:14 AM

The Bloomsbury Companion to Syntax

(c) *x-ø-w-il naj x-ø-s-watx’e hun-ti’. asp-3abs-1erg-see cl/he asp-3abs-3erg-make one-this “I saw the man who made this.” The absolutive Agent of make in (64b) relativizes since the verb is an ANTIP intransitive. Relativizing the transitive subject directly is ungrammatical, (64c). Comparable facts hold in K’ekchi (Berinstein 1985: 161), (65), and Kiche’ (Campbell 2000). (65) (a) Ha´ li cui:nk qui-ø-sac´-o-c emp  the man.tns-3abs-hit-antip-asp “That’s the man who hit me.”

cu-a. me-dat

(b) *Ha´ li cui:nk qu-in-ix-sac´.  emph the man.tns-1abs-3erg-hit “That’s the man who hit me.” In closing, two general remarks on Antipassive and Passive. First, the presence of an Antipassive or a Passive in a grammar presupposes an active with which it contrasts. The (anti)passive affixes maybe limited to certain verbs, the active is not so limited. And second, Passive and Antipassives may coexist in a given language, as in (66) from Inuktitut (see M. Johnson 1980; examples from Dryer 1986) (66) (a) Piruutisi-up Siisa-ø Brutus- erg Caesar-abs

kapi-vaa. stab-3sg.3sg

(b) Piruutisi-ø Siisa-mik Brutus-abs Caesar-com

kapi-si-vuq. stab-antip-3sg

(c) Piita-up Maali-ø Peter-erg Mary-abs

kunik-paa. kiss-3sg.3sg

(d) Piita-mit Maali-ø kunik-tau-vuq. Peter-abl Mary-abs kiss-pass-3sg

Active Antipassive Active Passive

This concludes our discussion of voice.

260

9781441124609_Ch14_Final_txt_print.indd 260

11/20/2012 9:25:14 AM

15

Middle and Reflexive Leonid Kulikov

Chapter Overview Preliminaries Middle as a Cluster of Deagentivized (Intransitivized) Syntactic Patterns Middle Types/Patterns Middle Voice in Indo-European and Beyond: The Range of Variation Middle and Reflexive in a Diachronic Perspective Concluding Terminological Remarks: Reflexive Sensu Stricto vs. “Reflexive” as a Cover Term

261 265 266 273 276 278

1 Preliminaries 1.1 Historical Notes The term “middle” is inherited from the ancient Greek grammatical tradition (mesótēs, that is, “middle”), where it appears combined with the term diáthesis “disposition.” The formal opposition between three diatheses, enérgeia “performance,” páthos “experience,” and mesótēs literally “central position, mean,” amounts to the morphological opposition between two series of the verbal inflectional suffixes (as, for instance, in the case of 1sg.pres. bouleúō, 2sg.pres. bouleúeis, 3sg.pres. bouleúei, etc. “to take counsel, deliberate” vs. middle: 1sg. pres. bouleú-o-mai, 2sg.pres. bouleú-ei, 3sg.pres. bouleú-e-tai, etc. “to take counsel with oneself”), known in the modern terminology as active and middle (though not quite consistently described in Greek tradition in terms of three functional labels: “active,” “passive,” and “middle”; for details, see Andersen 1994a: 125ff., 1994b; Shibatani 2004: 1149). 261

9781441124609_Ch15_Final_txt_print.indd 261

11/20/2012 9:25:50 AM

The Bloomsbury Companion to Syntax

1.2 Basic Definitions and Terminology The categorical domain of reflexives and middle is one of the most important— and the most intricate—parts of the verbal grammar. It overlaps with (and can to some extent be described in terms of) such categories as transitivity and voice or diathesis, depending on the theoretical framework chosen as basic for linguistic description. For the purposes of the present survey, it will be convenient to capture the most relevant features of the reflexive and middle in question in a broader categorial context—as members of (or, at least, as categories related to) the category of voice. Below I have chosen as basic the approach developed by the Leningrad/St Petersburg Typology Group,1 which offers a useful calculus of relations between two main levels of presentation of the structure of a clause. These are: (i) the level of semantic “macroroles” (cf. the macroroles of Actor and Undergoer in the framework of Role and Reference Grammar; to avoid terminological confusion, I will denote macroroles by means of capital letters X, Y, Z, W); and (ii) the level of syntactic functions (grammatical relations): Subject [S], Direct Object [DO], Indirect Object [IO], and Oblique Object [Obl]. The level of grammatical relations is responsible for the realization of arguments in the clause—in particular, for the three main formal means of encoding grammatical relations, which include case marking, verbal agreement, and word order. Altogether, these three morpho-syntactic devices determine the syntactic structure of the clause. In simple cases, the syntactic functions can be straightforwardly identified in terms of one of these parameters. Thus, the grammatical relations of S, DO, and IO often correspond to the nominative, accusative, and dative, respectively (in nominative–accusative case-marking languages); the clause-initial noun bears the grammatical relation of Subject in many languages; etc. The most important theoretical concept that can be determined in terms of the two levels of representation and enables to capture the rich variety of voices is that of syntactic pattern, determined as a pattern of mapping of semantic arguments onto syntactic functions (grammatical relations).2 The notion of syntactic pattern is closely related to that of verbal valency/valence, which is inherently associated with the set of arguments governed by the verb in question. The pattern where the Actor is mapped onto the Subject and the Undergoer onto the Direct Object is the most common, unmarked way of representing an event and therefore can be regarded as the basic, or the neutral, syntactic pattern3 of a simple transitive verb. This can be illustrated by the following Ancient Greek and Sanskrit sentences (1–2):

262

9781441124609_Ch15_Final_txt_print.indd 262

11/20/2012 9:25:51 AM

Middle and Reflexive

(1) Ancient Greek ho paîs bállei tḕn sfaîran. the: om.sg.m child:nom.sg throw:pres:3sg.act the:acc.sg.f ball:acc.sg “The child throws (active) the ball.” (2) Sanskrit kumāro vr̥kṣaṃ han-ti. prince:nom wolf:acc kill:pres-3sg.act “The prince kills the wolf.” Syntactic patterns can be conveniently presented in a tabular form (which I will use from now on in the present chapter). Thus the syntactic pattern exemplified by (1–2) can be schematized in Table (3): (3) Basic transitive pattern Semantic argument level (role)

X (Actor)

Y (Undergoer)

Syntactic function level (case)

S (nom)

DO (acc)

(ho paîs; kumāro)

(tḕn sfaîran; vr̥kṣaṃ)

Syntactic changes affecting transitivity can be conveniently described in terms of modifications of syntactic patterns. For instance, the modification of the basic (neutral) transitive pattern that results in the passive equivalent of a transitive clause typically suggests the following two (partly independent) syntactic phenomena: (i) the promotion of the initial Direct Object to the Subject (= the Subject of the passive construction); and (ii) the demotion of the initial Subject (usually, an Agent). The demotion of the Subject may amount either to its downgrading to an Oblique Object (passive Agent), or to its removal from the structure. This change in pattern is exemplified by the passive equivalents of (1–2) in (4–5) and presented in tabular form in Table (6): (4) Ancient Greek hē sfaîra bálle-tai hupò toû paidós. the:nom.sg.f ball:nom.sg throw:pres-3sg.mid under the:gen.sg.m child:gen.sg “The ball is thrown by the child.” (5) Sanskrit vr̥kṣo kumāreṇa han-ya-te. wolf:nom king:ins kill-pres.pass-3sg.mid “The wolf is (being) killed by the prince.”

263

9781441124609_Ch15_Final_txt_print.indd 263

11/20/2012 9:25:51 AM

The Bloomsbury Companion to Syntax

(6) Passive pattern X

Y

S

DO

⇒

X

Y

Obl / –

S

In the languages where the correspondence between grammatical relations and case-marking is relatively straightforward, pattern modification can also be formulated in terms of changes in case-marking. Thus, Scheme (7) describes passivization in Sanskrit: (7) Passivization in Sanskrit X

Y

S (nom)

DO (acc)

⇒

X

Y

Obl (ins)

S (nom)

1.3 Syntactic Pattern and Voice The technical concept of syntactic pattern was introduced, foremost, in order to arrive at a more accurate definition and a reasonable formalization of our traditional concept of voice, and, eventually, to sort out its linguistic content. The category of “voice” is determined on the basis of the concept of syntactic pattern as follows: voice is a regular encoding of syntactic pattern through the verbal morphology. Thus, many languages encode the passive pattern outlined above by means of a special verbal morpheme, which, accordingly, is interpreted as the marker of the passive voice. In Sanskrit, these are the present passive suffix -ya- and the middle inflection; in Greek and Latin, the passive voice is expressed by means of a special series of endings (“medio-passive” inflection); cf. 3sg.pass. Gr. -tai ~ 3sg.act. -i ( direct object > indirect object > oblique

For instance, the cross-linguistic variation on relativization can be explained in terms of which grammatical relations are (not) to be relativized on in individual languages. Some languages may relativize on subject only, other languages on subject and direct object only, and so on. Thus, languages, with the “option” of stopping on any point on the hierarchy (below subject, which is relativized on universally), relativize on the hierarchy in a continuous manner (e.g. no languages relativizing on subject and indirect object, while skipping over direct object). By extending this hierarchy to causatives (known as the Case Hierarchy (CH)), in particular morphological causatives, Comrie (1975, 1976) made an attempt to predict what grammatical relation would be borne by the causee NP. His CH theory assumed a biclausal structure for a morphological causative sentence; the higher clause of this abstract structure consists of the causer NP and the predicate of cause, and the lower clause the causee NP and the predicate of effect. Comrie’s prediction is that the causee NP will assume the next highest available position in the hierarchy in (3)—with other NPs, if any, in the lower clause retaining their grammatical relations. If the predicate of effect is intransitive, the causee NP will take up the direct object relation, as it is the next highest available position in (3), with the subject relation taken up by the causer NP (e.g. (4)). If the predicate of effect is transitive (with a subject and an object NP), the causee NP will end up assuming the indirect object (IO) relation, as the subject relation and direct object (DO) relation are taken up by the causer NP and the direct object NP of the lower clause, respectively (e.g. (5)). (4)

Turkish (Comrie 1975: 5) (a) Hasan öl-dü. Hasan die-pst “Hasan died.” (b) Ali Hasan-ɪ öl-dür-dü. Ali Hasan- o die-caus-pst “Ali killed Hasan.”

284

9781441124609_Ch16_Final_txt_print.indd 284

11/20/2012 9:25:37 AM

Causatives: Structure, Type, and Distribution

(5) Turkish (Comrie 1975: 6) (a) müdür mektub-u imzala-dɪ. director letter- o sign-pst “The director signed the letter.” (b) dişçi mektub-u müdür-e imzala-t-tɪ. dentist letter- o director-io sign-caus-pst “The dentist made the director sign the letter.” Comrie’s CH theory, however, ran into empirical problems. Many languages do not behave as predicted by his theory (e.g. Song 1996: 166–9). In Blackfoot and Southern Tiwa, for instance, it is not the next highest available grammatical relation in the hierarchy that the causee NP takes up (Song 1996: 179–80). Rather, in these languages, the causee NP “steals” the direct object relation from the direct object NP of the predicate of effect. Moreover, Comrie’s assumption that morphological causatives have an abstract biclausal structure is not supported by the fact that many languages draw upon general grammatical mechanisms to keep the number of core NPs in morphological causatives at the same level as in simple, non-causative sentences. In Afar, many causativized transitive verbs “lose” the original subject of transitive verbs, that is, the causee NP. In Babungo, it is the object of the transitive verb that must be le out or expressed as an optional adjunct. Thus, the total number of core NPs in morphologically causativized transitive sentences never exceeds that of non-causative transitive sentences (cf. Keenan, this volume, on the interaction between causative and passive).

2.2 Semantic Typology of Causation The definition of the causative construction given in Section 1 embodies two mixed but distinct levels of description: the level of events and the level of participants. The former captures the temporal relationship between the causing and the caused event. The latter concerns the interpersonal relationship between the causer and causee. There are two major semantic types of causation: direct vs. indirect causation; and manipulative vs. directive causation. The first semantic type is based on the level of events, and the second on the level of participants. The distinction between direct and indirect causation rests on the temporal distance between the causing event and the caused event. If the caused event is temporally adjacent to the causing event, without any other event intervening between them, the causative situation is direct. For example, if X makes Y fall into the river by pushing Y, X pushing Y immediately precedes Y’s falling into the river. 285

9781441124609_Ch16_Final_txt_print.indd 285

11/20/2012 9:25:37 AM

The Bloomsbury Companion to Syntax

Indirect causation, in contrast, involves a situation in which the caused event may not immediately follow the causing event. There is at least one event intervening between the two events. For example, X fiddles with Y’s car, and later Y is involved in a car accident due to the failure of the car. The causing event is X’s fiddling with Y’s car, and the caused event is Y’s accident. These events are separated by the intermediary event (the failure of the car).2 Depending on the causer’s relationship with the causee, the causative situation may be either manipulative or directive. If the causer acts physically on the causee, the causative situation is manipulative. The causative situation referred to earlier in order to exemplify direct causation is also manipulative because the causer pushes the causee into the river. The causer may also rely on an intermediary physical process or means in effecting the caused event. For example, if X causes Y to fall by pushing a shopping trolley into Y, the causer effects the caused event through some physical means as in the case of direct manipulative causation. This intermediary physical process also constitutes an independent event intervening between the causing and the caused event. The overall causative situation may thus be indirect and manipulative. The causer may draw upon a non-physical—verbal or social—means in causing the causee to carry out an action or to undergo a change of condition or state. For example, if X causes Y to sit down by giving Y an instruction to do so, the causative situation is directive. This particular situation is also direct in that there is no other event intervening between the causing and the caused event—Y’s sitting down is temporally adjacent to X’s uttering the instruction. Again, directive causation may also be indirect, rather than direct. For example, if X causes Y to write a letter by giving Z an instruction to that effect, then one is dealing with indirect directive causation. The caused event is separated from the causing event by the event of Z telling or asking Y to write a letter.

2.3 Causative Continuum and Causation Types There is a strong correlation between the causative and the causation types. The three causative types—lexical, morphological, and syntactic—form a continuum of formal fusion between the expression of cause and that of effect, as in: (6) lexical

morphological

syntactic

a greater degree of fusion 286

9781441124609_Ch16_Final_txt_print.indd 286

11/20/2012 9:25:37 AM

Causatives: Structure, Type, and Distribution

There is a tendency for manipulative or direct causation to be mapped onto the causative types to the le of the continuum in preference to those to the right. Directive or indirect causation, in contrast, is more likely to be expressed by the causative types to the right of the continuum than by those to the le . This is cited in the literature as a prime example of iconic motivation in language. Iconicity represents the principle that the formal structure of language (e.g. lexical causatives, involving the maximum fusion between the expression of cause and that of effect) should, as closely as possible, mirror the conceptual structure of what is expressed by language (e.g. manipulative causation, involving propinquity between the causer and causee) (Haiman 1985b). This correlation has recently been reinterpreted as that between the degree of difficulty in bringing about the caused event and the degree of transparency in expressing the notion of causation (Shibatani 2001).

2.4 Semantics of Causatives and Case Marking There is a large amount of cross-linguistic evidence in support of the case marking of the causee NP being determined by agency, control, affectedness, or topicality of the main participants of the causative situation (e.g. Cole 1983; Song 2001: 283–6; cf. Payne, this volume, on case marking). In Bolivian Quechua, for instance, the causee NP bears the accusative case if the causee is directly under the causer’s authority and lacks control over the action (i.e. (7)). If, however, the causee has control over his or her action but complies voluntarily with the causer’s wish, the causee NP appears in the instrumental case (i.e. (8)). (7) Bolivian Quechua (Cole 1983: 118) nuqa Fan-ta rumi-ta apa-či-ni. I Juan-acc rock-acc carry-caus-1sg “I made Juan carry the rock.” (8) Bolivian Quechua (Cole 1983: 118) nuqa Fan-wan rumi-ta apa-či-ni. I Juan-ins rock-acc carry-caus-1sg “I had Juan carry the rock.” Thus, the case marking of the causee NP is semantically motivated by the causee’s agency or control in particular. Note that this kind of phenomenon also calls into question the role of the grammatical relations hierarchy in predicting the grammatical relation of the causee NP (e.g. Comrie 1975, 1976; also see Section 2.1). 287

9781441124609_Ch16_Final_txt_print.indd 287

11/20/2012 9:25:37 AM

The Bloomsbury Companion to Syntax

3 Syntactic Causatives Research on causatives tends to concentrate on the morphological causative type, unfortunately, to the underrepresentation of the syntactic type of causative in particular. This lopsidedness has been addressed in Song (1996, 2005a and b). The typology of causatives proposed in Song (1996) was based on a database of 613 languages, and subsequently replicated in Song (2005a and b) by means of the World Atlas of Language Structures 200-language sample (Haspelmath et al. 2008).

3.1 Syntactic Causatives: Sequential and Purposive Subtypes Song (1996, 2005a and b) proposes two types of syntactic causative construction: sequential and purposive.3 In the sequential type, the clause of cause and that of effect are juxtaposed strictly in that order, with or without a linking element between them. The first clause denotes the causing event, and the second clause the caused event. The causative in (9) is an example of the sequential type, without use of a linking element. (9) Kobon (Davies 1981: 165) nipe g-aj-ɨp yad 3sg do-dur-pst.3sg 1sg “He made me cut the tree.”

mab rɨb-pin. tree cut-pfv.1sg

The sequential type, with use of a linking element, is exemplified by (10). The linking element in (10) is a switch reference (SR) affix -ce-. (10) Amele (Roberts 1987: 222) ija od-ude-ce-min na qete-i-a. 1sg do-3sg-sr-1sg tree cut-3sg-today.pst “I made him cut the tree.” The purposive type also involves two clauses, one representing an event carried out for the purpose of realizing another event denoted by the other. The sense of purpose or goal present in this type can be signaled by (i) verbal markings such as future tense, irrealis, subjunctive mood, incompletive aspect, etc.; (ii) dative, allative, or purposive case markers; or (iii) purposive particles. Swahili is an example of the purposive type. In (11), the predicate of effect, which must appear in subjunctive mood, stands in contrast with the predicate of cause in past tense. The subjunctive mood marking is utilized here to signal the sense of purpose. 288

9781441124609_Ch16_Final_txt_print.indd 288

11/20/2012 9:25:37 AM

Causatives: Structure, Type, and Distribution

(11) Swahili (Vitale 1981: 152) Ahmed a-li-m-fanya mbwa a-l-e Ahmed he-pst-him-make dog he-eat-s jv “Ahmed made the dog eat a large fish.”

samaki mkubwa. fish large

In other languages of the purposive type, case markers such as dative, allative, purposive, and the like may be utilized to signal the sense of purpose or goal. Basque is such a language. (12) Basque (Saltarelli 1988: 221) Mikel-ek Jon Edurne-ri liburu-a. Mikel-erg Jon-abs Edurne-dat book-sg.abs eros-te-ra behar-tu z-u-en. buy-nmlz-all make-pfv 3sg.erg-aux-pst “Mikel made Jon buy a book for Edurne.” In (12), the nominalized (subordinate) clause of effect as a whole is “flagged” by the allative case suffix -ra (e.g. parke-ra “to the park”). The purposive type in Modern Greek relies on a subjunctive particle (i.e. na) to signal the sense of purpose or goal, as in: (13) Modern Greek (Joseph and Philippaki-Warburton 1987: 171) ékana ton jáni na fíji. made-1sg.actor the John-acc sbjv leave-3sg “I made John leave.”

3.2 The New Typology of Causatives: Theoretical Implications The structural variety of the syntactic causative, summarized in Section 3.1, offers an empirical basis on which to construct a cognitive model of causation, and also a diachronic model with which to account for the origins of causative affixes (for further discussion, Song 1996: 73–109, 138–48, and also Song 1990). The sequential type expresses two components of causation: a deliberate attempt to realize a desire or wish (or EVENT) and accomplishment of that desire or wish (or RESULT). In this type, the EVENT component corresponds to the clause of cause, and the RESULT component to the clause of effect. The purposive type denotes the EVENT component (i.e. a deliberate attempt to realize a desire or wish), and also one other component, that is, perception of that desire or wish (or GOAL)—the latter component corresponds to the subordinate clause of purpose. These three components can be assembled into the temporally based cognitive model: 289

9781441124609_Ch16_Final_txt_print.indd 289

11/20/2012 9:25:38 AM

The Bloomsbury Companion to Syntax

(14) GOAL  → EVENT  → RESULT  where → represents progression of time Furthermore, recognition as a causative construction par excellence of the purposive type in particular has implications for the diachrony of causative affixes. In a large number of languages, causative affixes can possibly or plausibly be traced back to purposive case markers, verbal markers of purpose, etc.; in other languages, hitherto claimed sources of causative affixes can be reexamined.

4 Areal Typology of Causative Constructions In line with emerging interest in areal typology, the global distribution of the syntactic and nonsyntactic (i.e. morphological and compound) causatives has recently been investigated in Song (2005a and b). There are many gaps in the database concerning the use of syntactic causatives, compared with nonsyntactic ones. While it may seem premature to comment on the areal distributions of the sequential and purposive causatives, a few tentative observations can be advanced.

4.1 Purposive and Sequential Subtypes In the eastern part of Africa, from Ethiopia to South Africa, the purposive type is much more common than the sequential type. In the northwestern part of Africa, in contrast, the sequential type seems to be dominant. In a vast area covering Europe, the Caucasus, and the Middle East, and extending to the doorstep of Central Asia, the purposive type predominates over the sequential type. The dominance of the purposive type in this area may possibly link with that in the eastern part of Africa. In India and Southeast Asia taken together, the purposive type is predominant in the west, with the sequential type being more common in the east. What remains to be seen is whether the western half of this combined area may meet the vast area alluded to earlier, that is, Europe, the Caucasus, the Middle East, and the eastern part of Africa, all of which display a strong preference for the purposive type. In Australia, the purposive type is a clear winner. In New Guinea, the sequential type seems to prevail over the purposive type, although there are gaps in the data. The sequential type in New Guinea also seems to extend eastwards and westwards, although Oceania has its own share of the purposive type. There are too many gaps in the data to make suggestions about North and South America, although there may be a concentration of purposive-type languages in between the bottom of North America and the top of South America. There are a small number of languages 290

9781441124609_Ch16_Final_txt_print.indd 290

11/20/2012 9:25:38 AM

Causatives: Structure, Type, and Distribution

with both the sequential and the purposive type, and potential clusters of such languages exist in Indo-China and the northwestern part of Africa.

4.2 Nonsyntactic Causatives In view of the prevalence of the morphological type in the world’s languages, it is not inappropriate to describe its areal distribution in negative terms, that is, its absence. There are languages lacking the morphological type in all major geographical areas: Africa, Eurasia, Southeast Asia–Oceania, Australia–New Guinea, North America, and South America. Within Eurasia, languages without the morphological type are found only within Europe. In each of the major geographical areas, languages without nonsyntactic causatives are too few or scattered to reveal clear areal patterns except for possible clusters of such languages in Southeast Asia and northwestern Australia.

5 Concluding Remarks This chapter has provided an overview of the research carried out on causative constructions in linguistic typology, in particular the role of grammatical relations and the semantic types of causation. Special emphasis has been placed on the variety of the syntactic causative, neglected in the traditional typology. From this renewed emphasis theoretical implications have been drawn particularly in terms of the cognitive model of causation and the diachrony of causative affixes. The global distribution of the causatives, based on the typology proposed in Song (1996, 2005a and b), has also been briefly discussed. The subtypes of the syntactic causative, as highlighted here, need to be taken into consideration instead of being lumped together under the broad but not very revealing category of the syntactic type, especially when hitherto undocumented languages are investigated.

Notes 1. That is, if (1) is taken to consist of a main and a subordinate clause. 2. Said differently, X’s fiddling causes the failure of the car, which in turn causes Y’s accident. Thus, the intervening event itself is a caused event as well as a causing event. 3. The sequential and purposive type were referred to as the AND and PURP type, respectively, in Song (1996). The lexical and morphological types were subsumed under the COMPACT type in Song (1996), the difference between the two being that of formal fusion between the expression of cause and that of effect.

291

9781441124609_Ch16_Final_txt_print.indd 291

11/20/2012 9:25:38 AM

17

Coordination Caterina Mauri

Chapter Overview The Notion of Coordination Relation Semantic and Morphosyntactic Features of Coordination Coordination across Languages: Some Typological Patterns

292 295 302

1 The Notion of Coordination Relation As Langacker (1987: 472) points out in his discussion on coordination, two coordinated elements “have to be parallel in certain respects not yet fully understood.” At the basis of any definition of coordination lies some idea of symmetry, which is what distinguishes it from dependency relations. The property of being symmetrical has been interpreted in a number of different ways in the literature, but the crucial demarcation line can be drawn between approaches defining coordination at a formal, syntactic level and approaches adopting a functional definition. Formal approaches focus on the identification of an abstract formal structure that defines coordination universally. Scholars like Johannessen (1998), Camacho (2003), and Rebuschi (2005) consider coordinate structures as headed constructions. In particular, coordinate structures are treated like conjunction phrases, in which the coordinator is the head, the first conjunct the specifier, and the second conjunct the complement. As Borsley (2005) points out, this conception is widely accepted within Principles and Parameters theories, but it is rejected within other frameworks. Borsley (2005) himself rejects the idea that coordinate structures are conjunction phrases and argues that they are rather to be analyzed as adjunct phrases. This different conception of coordination is held by Bresnan (2001), Yuasa and Sadock (2002), and Cormack and Smith (2005). Generative approaches aim at the identification of the formal structure 292

9781441124609_Ch17_Final_txt_print.indd 292

11/20/2012 9:25:01 AM

Coordination

characterizing coordination, regardless of the conceptual relations expressed and of the cross-linguistic variation in the coding of such relations, which are instead central topics in functional approaches. In this chapter a functional perspective will be discussed, which defines coordination in functional terms and assigns great value to the variation attested in the world’s languages. For this reason the debate within the generative framework and the relevant literature will not be addressed further here (see Borsley 2005 and Cormack and Smith 2005 for a detailed discussion on the formal approaches to coordination). The aim of this chapter is to provide an overview of the main semantic and morphosyntactic features characterizing coordinating constructions from a functional–typological approach, with a focus on the variation attested to in the world’s languages. In this section the notion of coordination will be defined and analyzed. In Section 2 we will examine the semantic and morphosyntactic properties characterizing coordination in the world’s languages, and Section 3 will focus on some relevant typological patterns. Coordination is defined as a conceptual situation in which two entities, properties, or states of affairs are linked and conceived as functionally parallel (cf. Dik 1968: 25–9). However, such functional parallelism may occur at different levels. A number of scholars have focused on the semantic parallelism underlying coordination (e.g. Schachter 1977a: 89; Haspelmath 2004b: 34). Haspelmath defines as coordinate construction any syntactic construction “in which two or more units of the same type are combined into a larger unit and still have the same semantic relations with the other surrounding elements.” The identification of coordination on the basis of a semantic parallelism is especially adequate when the coordinated entities are not states of affairs (henceforth SoA), unless they are subordinated. If two non-subordinated SoAs are coordinated to each other, their functional parallelism is better described in terms other than the semantic relations they have with the surrounding elements. Haiman (1985b: 99) and Langacker (1987: 483) focus their attention on the conceptual parallelism underlying coordination, rather than on the semantic one. According to Langacker (ibid.), “the essence of coordination is the conceptual juxtaposition of co-equal structures” and all the main properties of coordination follow from this conceptual symmetry. Given two joined elements, they may have autonomous profiles or they may show an asymmetry, so that the profile of one of them is overridden by that of the other. The former (parallel) configuration corresponds to a coordination relation, the latter (non-parallel) to a dependency relation, in which the element with an autonomous profile is denoted by the head or, in case of clausal subordination, by the main clause (ibid.). The conceptual parallelism as described by Langacker subsumes any type of coordination established between any type of elements (entities, SoAs, properties), but it is difficult to find independent evidence for 293

9781441124609_Ch17_Final_txt_print.indd 293

11/20/2012 9:25:02 AM

The Bloomsbury Companion to Syntax

the identification of co-equal autonomous profiles regardless of the morphosyntactic level. There is indeed one more level at which two coordinated elements may be identified as functionally parallel, namely the level of pragmatics, at which information is inserted into the net of communicative intentions and assumptions of speaker and hearer, and may thus be topical, focal, asserted, asked, or ordered. As Cristofaro (2003: 29–50) argues in her typological survey of subordination, it is possible to establish a correspondence between the conceptual (non-)parallelism of two given SoAs and the assertive (non-)parallelism of the clauses expressing them. Let us briefly take into account the main issues that have to be dealt with when analyzing coordination from the point of view of pragmatics. First, there are two possible configurations in which two linked SoAs are pragmatically parallel: it is possible to coordinate two subordinate clauses, which are by definition non-asserted, or two independent clauses, which are instead both asserted. Second, coordination may be established between declarative clauses, requests, orders, wishes, and so on. In other words, two coordinated independent clauses may have illocutionary forces other than assertion, like interrogative or imperative illocutionary force. Finally, coordination may be established between entities or properties whose pragmatic dimension is not that of illocutionary force, but rather of topic and focus. The equivalence between conceptual and pragmatic dimension can be thus framed as follows for coordination. Two coordinated entities, properties, or SoAs require the same informational status, because any asymmetric communicative organization could not co-exist with a symmetric conceptual one. Therefore, two SoAs standing in a coordination relation will be uttered in clauses having the same illocutionary force (cf. Mauri 2008a: 37–44), while two coordinated properties or entities will occur in phrases being part of the same topical or focal part of the sentence. The pragmatic criterion provides us with concrete tests for identifying coordinating constructions across languages, independently of their morphosyntactic properties. It is indeed possible to pinpoint (i) the presence of illocutionary force by means of tests such as tag questions, whereby two coordinate clauses should either result as both at issue or both presupposed (Mauri 2008a: 37–44), (ii) the topical or focal behavior by means of tests such as dislocations and cleft sentences, whereby if two elements are linked by a coordination relation, they should not be able to be dislocated or focalized separately (cf. Coordinate Structure Constraint by Ross 1967 and the discussion in Schachter 1977a and Deane 1991). The functional parallelism inherent to coordination is thus the result of a close interrelation between the semantic and conceptual symmetry of the linked elements and the symmetry of their communicative status. A coordination relation can be thus defined at the functional level as a relation established between 294

9781441124609_Ch17_Final_txt_print.indd 294

11/20/2012 9:25:02 AM

Coordination

functionally equivalent entities, properties, or SoAs, that is, elements having the same semantic function, autonomous cognitive profiles, and the same informational status. Every construction expressing a coordination relation will be considered a coordinating construction and will be analyzed on the basis of its morphosyntactic properties.

2 Semantic and Morphosyntactic Features of Coordination 2.1 Main Types of Coordination Relation: Combination, Contrast, and Alternative Two elements linked in a coordination relation may stand in different semantic types of relations. The three basic relations that have been traditionally identified as coordinating ones are combination ((1), conjunction), contrast ((2), adversativity), and alternative ((3), disjunction). (1) It is summer and everybody goes on holidays / I love apples and pears. (2) The summer ends but many people are still on holiday / I don’t want tea, but coffee. (3) Are you coming to the cinema tonight or do you relax at home?/ You may have tea or coffee. Combination, contrast, and alternative do not exhaust the list of possible coordination relations. Dik (1968: 277–9), for instance, includes certain types of causal relations within the set (the ones conveyed by constructions with for in English, car in French, and denn in German).1 However, the three relations of combination, contrast, and alternative can arguably be described as the three basic types of symmetric links that may be established between two independent SoAs (see Mauri 2008a: 46–8), entities, or properties. Within the literature on coordination, a number of scholars have developed fine-grained analyses of the major semantic subtypes of coordination (Haspelmath 2004b and 2007, Payne 1985 and Lang 1984, among others). The classifications made by Payne (1985) and Lang (1984) are based on formal abstractions and are basically formulated in logical terms. Haspelmath (2004b) identifies the various subtypes of coordination on the basis of the lexicalized constructions attested across languages. Mauri (2008) uses the three parameters of temporality, conflict, and aim discussing coordination relations among clauses. Let us now briefly go through the main subtypes of combination, contrast, and alternative that have been recognized as typologically relevant, both for clausal and nominal coordination. 295

9781441124609_Ch17_Final_txt_print.indd 295

11/20/2012 9:25:02 AM

The Bloomsbury Companion to Syntax

Combination of SoAs may be equential (“asymmetric and” according to Lakoff 1971: 126, for example, He opened the window and jumped), characterized by the location of the linked SoAs along the same time axis at successive points, so that they are interconnected as parts of the same overall sequence of events. Simultaneous combination (overlap, Longacre 1985: 243, for example, She was dancing and clapping her hands), on the contrary, is characterized by the absence of a reciprocal order in which the SoAs occur, since they occupy the same point on the time axis. In case the location of two combined SoAs along the time axis is simply not relevant to the combination itself, the relation is said to be atemporal (Longacre 1985: 241 calls it coupling or non-temporal underlying “and” relation, for example, Birds have wings and fishes have fins, cf. additive vs. nonadditive conjunction, Halliday and Hasan 1976). Combination of entities, on the other hand, is characterized by a major distinction between natural and accidental combination (Haspelmath 2007, Wälchli 2005). In natural combination, the entities “habitually go together and can be said to form some conventionalized whole or conceptual unit” (Mithun 1988: 332). Typical examples of natural conjunction are “mother and father” and “husband and wife.” In case some formal distinction is made between natural and accidental combination, this often consists of the lack of an overt coordinator or of an intonation break in natural conjunction. Elements linked in natural conjunction are formally very close to each other, to the point that they may occur as parts of a compound word (cf. co-compound, Wälchli 2005). As pointed out by Haspelmath (2007), this can be explained in terms of economy: since the conjuncts in natural conjunction occur together very frequently, the relation between them is quite predictable and overt marking is redundant. A cidental combination, on the other hand, involves entities that are not expected to cooccur, but rather their combination is motivated by specific circumstances of the speech act. Contrast may only be established between SoAs and properties, and the conflicting relation can be due to a semantic opposition, a correction, or the denial of some expectation (cf. Haspelmath 2007; Lang 1984; Rudolph 1996). If two SoAs or properties are set against each other because they somehow show antithetic features, the relation is said to be oppositive (e.g. Paul is tall whereas John is small). Lakoff (1971: 131–6) talks about this relation in terms of “semantic opposition” and describes it as characterized by a component of symmetric “and” (i.e. atemporal combination) together with a presupposition of difference in meaning. The order in which the two SoAs are presented may be reversed without affecting the general meaning of the assertion. If the conflict is determined by the substitution of an explicitely denied state of affairs with a new one, the contrast is said to be corrective. This notion is quite uncontroversial in the literature on coordination, because this type of contrast is clearly distinguishable from all others and can be easily identified across languages. A number of 296

9781441124609_Ch17_Final_txt_print.indd 296

11/20/2012 9:25:02 AM

Coordination

languages, especially within Europe, have dedicated markers coding this type of contrast, like German sondern or Hungarian hanem. Finally, if the contrast is generated by the denial of some expectation it is said to be countere pectative. As Lang (2000: 245) puts it, the assertion of the second SoA is in contrast to “an assumption that may be either read or inferred from previous information” (ibid.: 246). Regarding alternative, the crucial distinction is that between simple and choice-aimed alternative (see Mauri 2008a: ch. 5, Mauri 2008b), which may hold between SoAs, entities, or properties. Speakers may establish an alternative relation in order to assert or elicit information about the block of linked elements (i.e. the whole set of options that the relation delimits) or in order to receive information about the individual elements. In the first case, the speaker’s intention is to talk about the set of options, and this type of alternative is called simple alternative: the speaker does not elicit any choice, but s/he simply depicts a set of possibilities. In the second case, the speaker’s aim is to ask for information about the individual elements, and this type is called choiceaimed alternative, because the speaker asks the hearer to make a choice. The two types of alternative relations are frequently coded by means of different constructions across languages, as illustrated by the Somali example in (4) where amá is used to express simple alternative (4a), while misé is used for the choice-aimed one (4b). (4) Somali, Cushitic, Afro-Asiatic (a) Amá wuu kéeni doonaa amá wuu sóo.díri doonaa. coord 3sg bring that coord 3sg send that “Either he will bring it or he will send it.” (Saeed 1993: 275) (b) ma tégaysaa misé waad  jóogaysaa? int go:2sg coord here stay:2sg “Are you going or are you staying?” (ibid.) The relations of combination and alternative may be negated, generating what Payne (1985: 4) calls rejection. It is frequent across languages to find specific constructions to express rejection, such as, for example, English neither . . . nor . . . , Latin neque . . . neque . . . , German weder . . . noch . . . These constructions are normally closely linked to the strategies that languages use to express negation (cf. Bernini and Ramat 1996), and may originate from both basic conjunctive and disjunctive constructions (as instantiated by Latin and English, respectively). As Haspelmath (2007) points out, this is connected with the logical equivalence between disjunction with wide scope negation “not (X or Y)” and conjunction with narrow scope negation “not X and not Y,” and languages seem to mirror this functional equivalence at the morphosyntactic level.

297

9781441124609_Ch17_Final_txt_print.indd 297

11/20/2012 9:25:02 AM

The Bloomsbury Companion to Syntax

2.2 Syntactic Domain of Coordinating Constructions As already pointed out, coordination relations may be established between different types of elements, ranging from SoAs to entities, and at the morphosyntactic level this translates into different linguistic units that may be linked in a coordinating construction (clauses, verb phrases, adjectival phrases, noun phrases, etc.). The set of syntactic types that may be linked in a given coordinating construction constitutes its syntactic domain. It is frequent to find languages using different coordinating constructions for different types of coordination relations (conjunctive, disjunctive, adversative constructions), but languages may also show different coordinating constructions for different syntactic levels. In other words, we may find conjunctive or disjunctive constructions in different syntactic domains. As highlighted by Haspemath (2005), it is rather frequent to find distinct constructions for combination between entities (NP conjunction) and combination between events (i.e. VP or clause conjunction). A similar distinction is attested for disjunction, though more rarely, while it is of course not possible for adversative constructions, which basically only link states of affairs and properties. An example from Korean is provided in (5), where different constructions are attested for NPs (5b & d) and clauses (5a & c) both for conjunction and disjunction. (5) Korean, isolate (Ho-min Sohn 1994: 118–25) ko Minca-nun mikwuk-ey ka-ss-ta. (a) na-nun ilpon-ey ka-(ss)-ko I-ctop Japan-to go-(pst)-and Minca-ctop America-to go-pst-decl “I went to Japan and Minca to America.” wa/hako/lang Yongho-nun (b) Minca-wa/hako/lang Minca-and Yongho-ctop “Minca and Yongho enjoy music.”

umak-ul culki-n-ta. music-acc enjoy-ind-decl

kena Minca-ka wa-ss-ta. (c) Yongho-ka wa-ss-kena Yongho-nom come-pst-or Minca-nom come-pst-decl “Either Yongho or Minca came.” ina cwuk-ul mek-keyss-ta. (d) na-nun pap-ina I-ctop rice-or gruel-acc eat-fut-decl “I will eat rice or gruel.” There are also languages distinguishing between different types of clauses, using dedicated coordinating constructions for relative and main clauses (cf. Yoruba and Rowlands 1969: 201–3). Payne (1985: 5) proposes an implicational cline that constrains the possible ranges of coordinators: S—VP—AP—PP—NP. The prediction is that individual constructions are restricted to cover contiguous categories, for example, S and VP, or AP, PP, and NP. 298

9781441124609_Ch17_Final_txt_print.indd 298

11/20/2012 9:25:02 AM

Coordination

2.3 Morphosyntactic Properties 2.3.1 Explicit Coding of the Relation Coordinate constructions may be characterized by an overt coordinator, that is, an explicit connective encoding the coordination relation at issue (syndetic coordination), or they may lack such an explicit linking device and simply consist in juxtaposition (asyndetic coordination, cf. Mithun 1988 for a detailed exemplification and discussion on the role of intonation). In case of syndetic coordination, coordinators can be analyzed on the basis of their morphosyntactic properties (distributional features, morphophonological complexity) or on the basis of their semantic properties.2 As far as distributional properties are concerned, coordinating constructions may either have a single coordinator (monosyndetic) or two coordinators (bisyndetic), which may be prepositive (preceding the coordinand) or postpositive (following the coordinand). In case of coordination among more than two elements, we may find the obligatory omission of the connective in all but the last element or its obligatory repetition in each coordinated element. A further dimension of variation of coordinators concerns their morphophonological complexity. The morphophonological complexity of the attested markers is measured on the basis of the following parameters: syntactic bondedness, number of syllables, and number of morphemes, distinguishing respectively between free vs. bound marker, mono vs. polysyllabic marker, and mono vs. polymorphemic marker. As exemplified in Table 17.1, the complexity of every marker consists of the sum of these parameters. Finally, coordinators may be analyzed on the basis of their semantic domain, that is, the set of coordination relations that they may express. On the basis of this parameter, coordinators may be classified as dedicated (monofunctional) or general (multifunctional). The concept of dedicated marker is absolute, since it applies to those cases where a given construction can be used only in one specific situation. On the contrary, the concept of general marker is scalar since the degree of generality of the markers may vary depending on the number of

Table 17.1 Morphophonological complexity of the attested coordinating markers. + = presence of the given feature; – = absence of the given feature Hebrew Italian German French

–ve o sondern tandis que

Free

Polysyllabic

Polymorphemic

– + + +

– – + +

– – – +

299

9781441124609_Ch17_Final_txt_print.indd 299

11/20/2012 9:25:02 AM

The Bloomsbury Companion to Syntax

coordination relations that they cover. An example of dedicated markers has already been provided in (4), from Somali, where the two specific coordinators amá and misé are employed for the expression of simple and choice-aimed alternative, respectively. The translation provided in (4) shows that the correspondent English disjunctive marker or is instead general, since it may cover both types of alternative. German provides a further case, with the adversative connective sondern being used only for corrective contrast, and aber being used only for counterexpectative relations. On the other hand, Italian ma, French mais, English but are general because they may be used for both corrective and counterexpectative contrast. The semantic domains of coordinators are best described by means of semantic maps and conceptual spaces (cf. Haspelmath 2003; Malchukov 2004; Mauri, 2010) because they identify multifunctional (general) and monofunctional (dedicated) markers and allow to describe the attested polysemy patterns.3 A discussion in terms of semantic maps is provided in Section 3.1.

2.3.2 (A)symmetry of the Construction Although traditional definitions of coordination define it on the basis of its syntactic parallelism, the analysis of cross-linguistic variation reveals that coordinating constructions are frequently characterized by some asymmetry. In this chapter we will focus on the parallelism of the forms encoding the linked entities or SoAs, taking into account possible differences concerning (i) verb forms for clause coordination and (ii) inflectional properties for nominals. In clause coordination, the syntactic parallelism of the construction depends on whether the linked SoAs are coded by means of the same verb form (parallel construction) or by means of different verb forms (non-parallel construction, Mauri 2008: 65–9). Following Stassen (1985) and Cristofaro (2003), we distinguish balancing and deranking verb forms (Stassen 1985: 76–83). A verb form is defined as balanced when it may occur as it is also in independent declarative clauses taken in isolation, whereas it is analyzed as deranked if it cannot be found in such clauses, because it lacks certain distinctions (such as tense, mood, aspect, or person agreement) or it is a special form not allowed in independent clauses (Cristofaro 2003: 50–60). At the clause level, a syntactically parallel coordinating construction is characterized by the same coding strategy for both SoAs, either balancing or deranking, while a syntactically non-parallel coordinating construction is characterized by different coding strategies for the two SoAs, one balancing and the other deranking (see (5a) from Korean, where the verb forms suffixed by -ko and -kena cannot receive modal specifications and depend on the second verb). Traditionally, non-parallel clausal coordinations have been described for conjunctive constructions (see Johannessen 1998; Haspelmath 2004b). However, 300

9781441124609_Ch17_Final_txt_print.indd 300

11/20/2012 9:25:02 AM

Coordination

such asymmetric strategies are also attested to express contrast and alternative relations, even though less frequently. Example (5c) instantiates a case of nonparallel clausal disjunction in Korean. Clausal asymmetric coordinations have been widely discussed in the literature, under a number of different perspectives. Lehmann (1988: 184–5) describes a continuum going from the highest degree of elaboration (i.e. coordination as traditionally defined) to the highest degree of compression, exemplified by embedded nominalized clauses without explicit linking. Asymmetric constructions such as (5a and c) find a position in the continuum on the basis of the values they show for each single parameter. Van Valin (2006) suggests a different solution and, instead of positing a continuum, establishes a new category which specifically identifies cases like (5a), namely co-subordination, characterized by syntactic dependency but no embedding (cf. Longacre 1985: 263–84, chaining structures). Some typological patterns concerning the nonparallelism of conjunctive, disjunctive, and adversative constructions will be discussed in Section 3.3. Coming now to noun phrase coordination, asymmetric constructions are characterized by nominals inflected in different cases or by some dependency strategy, typically a comitative one. Asymmetric coordination between noun phrases is only attested for combination relations. Stassen (2001) distinguishes between “and-languages”, showing different strategies for NP combination and accompaniment relations (frequent in northern and western Eurasia, India, northern Africa, New Guinea, Australia, and Meso-America), and “withlanguages”, employing the same asymmetric strategy for accompaniment and conjunctive relations (frequent in sub-Saharan Africa, East Asia, Southeast Asia, and the Pacific Islands, as well as in northern North America and lowland South America). An example of with-language is provided in (6). (6) Nkore-Kiga, Niger-Congo, Bantoid (Taylor 1985: 58) na Mugasho. n-ka-za-yo 1sg-rec.pst-go-there and/with Mugasho “Mugasho and I went there./ I went there with Mugasho.” The morphosyntactic non-parallelism shown by this type of construction is in most cases a residue of the recurrent diachronic path developing conjunctive coordinators from originally comitative strategies. The diachronic link between the two functions is clearly motivated by the functional contiguity between the two conceptual situations. As a result of the diachronic process, a number of morphosyntactic parallel features tend to arise, such as plural agreement on verbs, and the possibility to link more than two entities (whereas comitative strategies may only involve two, see Haspelmath 2007: 29–33 for a detailed discussion). 301

9781441124609_Ch17_Final_txt_print.indd 301

11/20/2012 9:25:03 AM

The Bloomsbury Companion to Syntax

3 Coordination across Languages: Some Typological Patterns After discussing the main morphosyntactic and semantic features characterizing coordination across languages, let us now focus on coordination between clauses and examine three typological issues. In Section 3.1 some implicational patterns concerning the morphosyntax and semantics of overt coordinators will be presented and in Section 3.2 the coding map of coordination relations will be described, highlighting the close connection between semantic domain and morphophonological complexity of coordinators. Section 3.3 deals with the implicational patterns attested in the analysis of non-parallel coordinating constructions, showing that certain coordination relations are more likely to be coded by means of asymmetric constructions than others. Finally, Section 3.4. identifies the “And-But-Or” language type, to which English, Italian, and French (among others) belong, and it will be argued that it is a rather infrequent type in the world’s languages, with a particular concentration in Western Europe.

3.1 Implicational Patterns in the Coding of Coordination Relations A typological survey based on 74 languages (Mauri 2008a) has shown that some coordination relations are more likely to be expressed without any overt markers as a result of their being more easily inferable from the context. In particular, if in a given language a contrast relation generated by the denial of an expectation is expressed by simple juxtaposition, this strategy will also be available for contrast relations generated by opposition and correction as in (7a). (7) (a) The combination-contrast coding implication: Syndesis for sequential, simultaneous, atemporal combination, oppositive contrast, corrective contrast → Syndesis for counterexpectative contrast. In alternative relations, on the other hand, if no marker is used in the expression of a simple alternative, then no marker will be used for an alternative where a choice is required as in (7b). (7) (b) The combination-alternative coding implication: Asyndesis for simple alternative → asyndesis for temporal and atemporal combination, asyndesis for choice-aimed alternative.

302

9781441124609_Ch17_Final_txt_print.indd 302

11/20/2012 9:25:03 AM

Coordination

Implications (7a) and (7b) may be explained with reference to the principle of syntagmatic economy, according to which what is already inferable from the context needs no further specification (cf. Haiman 1985b: 159). The coding of the various types of combination, contrast and alternative is connected to the degree to which every relation can be inferred from the context. Specifically, the more a relation is easy to infer, the less it needs to be overtly marked (see Mauri 2008a: chapter 6 for a detailed discussion and exemplification). Furthermore, cross-linguistic data also show that the degree of semantic specificity and semantic basicness of a coordinating construction is closely connected to the morphophonological complexity of the coordinating marker used, as shown by hierachies (7c) and (7d). The hierarchy in (7c) states that dedicated markers encoding a specific contrast relation (counterexpectative, oppositive, or corrective contrast) are at least as complex as the general markers used for contrast relations (but not for combination relation), that is, markers employed for corrective and counterexpectative contrast, such as English but or French mais. These general contrast markers are in turn at least as complex as dedicated and general markers used to express at least one combination relation. (7) (c) The combination–contrast coding complexity hierarchy: Dedicated and general marker for combination relations > general marker expressing contrastrelations > dedicated marker for a contrast relation. The implication in (7d) states that overt markers used to express alternative relations, either general or dedicated, are at least as morphophonologically complex as the markers used to express at least one combination relation. (7) (d) The combination–alternative coding complexity implication: Marker used for at least one alternative relation → marker used for at least one combination relation. Implications (7c) and (7d) may be explained on the basis of the economic principle of form-function asymmetry, according to which the more general a connective is (i.e. the more relations it may express), the lower is its degree of morphophonological complexity (cf. Kortmann 1997: 123–36; Zipf 1949: 66–133). On the one hand, frequency of use is a consequence of multifunctionality. The higher the number of relations a marker may express (i.e. the more general it is), the higher is the number of contexts where it may occur and, consequently, the more frequent it will be in discourse. Its phonological substance will thus be eroded, leading to morpho-phonologically simple forms. Therefore general markers, 303

9781441124609_Ch17_Final_txt_print.indd 303

11/20/2012 9:25:03 AM

The Bloomsbury Companion to Syntax

expressing more than one coordination relation, tend to be structurally simpler than dedicated ones (cf. implication (7c)). On the other hand, frequency in discourse may also be the consequence of a basic semantics. The more basic and semantically unspecified a conceptual relation is, the more it tends to correlate with high frequency of use. Since combination is the simplest coordination relation, it is the most frequently attested to in discourse (Ohori 2004: 61), and this is why markers used to express at least one combination relation, either general or dedicated, tend to be simpler than markers used to express contrast and alternative (cf. implications (7c) and (7d)). Finally, the representation of the attested semantic domains by means of semantic maps allows for a thorough analysis of how the three coordination relations at issue are related to each other. Conjunctive and adversative coordinators show recurrent overlapping polysemy patterns across languages, pointing to the combination–contrast conceptual space exemplified in Figure 17.1. The top of Figure 17.1 shows the order in which the different relations follow each other in the conceptual space. Below, some of the attested semantic domains are shown.

Conceptual Space:

seq. comb.

simult. comb.

atemp. comb.

Mangarayi

oppositive contrast

corrective contrast

counterexp. contrast gana

--

Tuvaluan

--

Hdi

lá

kae

àmá

-Maori

aa

hoki --

engari kuma

Hausa Russian

amman

i no

a Srb.-Croatian

i pa

Polish

i ale

a German

Italian

ali

vec

a

und während

sondern

mentre

bensì

e

aber ma però

Figure 17.1 Combination-contrast conceptual space and individual semantic maps. -- = no overt marker. 304

9781441124609_Ch17_Final_txt_print.indd 304

11/20/2012 9:25:03 AM

Coordination

If a coordinating marker is used to express more than one combination or contrast relation, it will convey relations that are contiguous on the conceptual space. A detailed discussion of the reasons underlying the respective order in which the relations are located on the conceptual space in Figure 17.1 is provided in Mauri (2008b: chapters 4 and 6). It suffices here to point out that the closeness of two relations on the space is due to their functional proximity. The functional proximity of two relations depends (i) on whether they share some conceptual features, (ii) on the frequency with which they are associated in discourse, and (iii) on the degree to which they can be easily inferred from each other. Combination and alternative relations, on the other hand, tend to be coded by means of completely different markers, thus showing a reduced semantic overlap. This is basically due to the fact that combination and contrast relations imply the cooccurrence of two SoAs 4 while an alternative relation implies the non-cooccurrence of the linked SoAs, which are instead presented as replaceable possibilities. Therefore combination and alternative relations are functionally very distant from each other. Yet, in languages with no overt equivalent to or, combination and alternative are expressed by means of the same construction, namely alternative is systematically conveyed through the combination of possibilities. In such cases, the potential (rather than actual) status of each combined SoA is obligatorily marked by means of some irrealis markers (like the conditional marker mo in (8a) or the dubitative adverb ‘am “perhaps” in (8b)). (8)

Wari’, Chapacura-Wanam (Everett and Kern 1997: 162)  hwam ca, mo ta (a) mo ta pa’ ta’ cond realis.future kill 1sg:realis.future fish 3sg.m cond realis.future pa’ ta’ carawa ca. kill 1sg:realis.future animal 3sg.m “Either he will fish or he will hunt.” (lit. “if he (says) ‘I will kill fish’, if he (says) ‘I will kill animals’.”)  

(b) ‘am ‘e’ ca  am mi’ pin ca. perhaps live 3sg.m perhaps  give complete 3sg.m “Either he will live or he will die.” (lit.”perhaps he will live, perhaps he will die”) In other words, in order for an alternative relation to be conveyed, either a connective coding the alternative relation or some overt irrealis marker is necessary (see Mauri 2008b for further discussion). If no overt connective of alternative is used, each SoA must display an irrealis marker and is therefore presented as possible, rather than occurring or realized, and the relation of alternative is inferred from the combination of two irrealis SoAs. 305

9781441124609_Ch17_Final_txt_print.indd 305

11/20/2012 9:25:03 AM

The Bloomsbury Companion to Syntax

3.2 The Hierarchical Coding Map of Coordination Relations A hierarchical map summarizing the results described in the preceding section is proposed in Figure 17.2, based on (i) the functional proximity between the various coordination relations, as manifested in the attested multifunctionality patterns, and (ii) their different degrees of semantic specificity and inferrability, as manifested in the attested implicational patterns of coding complexity. The coding map (see Mauri, 2010, for a comparison between semantic maps and coding maps) is structured along two perpendicular axes of increasing semantic specificity having their origin in the combination relation. Contrast and alternative are represented as further semantic specifications of the basic relation of combination. Along the horizontal axis, a combination of SoAs may be specified in terms of some discontinuity (Givón 2001b: 849) producing a contrast, or it may be specified in terms of the irreality of the SoAs it links, producing a set of alternative possibilities. The vertical axis, on the other hand, is meant to show the specifications internal to each coordination relation (i.e. the sub-types of combination, contrast, and alternative at issue). The further away from both the vertical and the horizontal axis a relation is located in the figure, the more semantically specified it is—along two hypothetical diagonals going from the origin of the axes toward the bottom right and the top right corners of the figure. The more semantically specified a relation is, the less it is easy to infer from simple juxtaposition. The coding map in Figure 17.2 predicts a number of phenomena. First, the order in which coordination relations occur from left to right mirrors the attested multifunctionality patterns. Therefore, it predicts that if a construction is used for more than one coordination relation, it will be used for relations that are contiguous along the horizontal axis of the space. Second, based on the increasing degree of semantic specificity, Figure 17.2 predicts that, other things being equal, the closer a relation is to the bottom right corner or to the top right corner of the space, the more difficult it will be to infer and, as a consequence, it will be more likely to be expressed by means of overt markers. Conversely, the closer a relation is to the origin of the axes, the easier it will be to infer and, as a result, it will also be more likely to be expressed by means of asyndetic constructions (cf. implications (7a) and (7b)). Third, the more basic and semantically unspecified a relation is, the more it correlates with a high frequency of use, and the markers coding it tend to undergo phonological erosion. Therefore, the closer a relation is to the origin of the axes in Figure 17.2, the simpler will be the morphophonology of the markers coding it; the farther it is from both the vertical and the horizontal axis, the more complex will be dedicated markers coding it (cf. implications (7c) and (7d)).

306

9781441124609_Ch17_Final_txt_print.indd 306

11/20/2012 9:25:03 AM

Coordination

INCREASING SEMANTIC SPECIFICITY

seq

sim

atemporal >

choice-aimed >

simple

ALTERNATIVE irreality COMBINATION

INCREASING SEMANTIC SPECIFICITY

INCREASING SEMANTIC SPECIFICITY

COMBINATION discontinuity CONTRAST

seq sim

atemporal> opposition

correction > denial expect

Figure 17.2 The twofold hierarchical coding map of coordination relations.

3.3 Implication of Coordination Parallelism Coordination relations show a strong tendency to be coded by means of parallel constructions. Syntactic (non-)parallelism is traditionally explained with respect to two functional principles: iconicity of independence and syntagmatic economy (cf. Cristofaro 2003: chapter 9 and Haiman 1985b: chapters 2 307

9781441124609_Ch17_Final_txt_print.indd 307

11/20/2012 9:25:03 AM

The Bloomsbury Companion to Syntax

and 4). According to the principle of iconicity of independence, the less independent two concepts are the less independent are the expressions coding them (Cristofaro 2003: chapter 9). According to the principle of syntagmatic economy, on the other hand, information that is available in the context does not need to be further specified, as shown/argued in the preceding section. The presence of a deranked verb form may depend on the fact that the relation established between the two SoAs predetermines their relative temporal location, so that tense, aspect, and mood specifications can be made explicit for just one SoA. The cross-linguistic tendency to code coordination relations by means of parallel constructions is therefore due to the fact that combination, alternative and contrast do not predetermine any semantic properties of the linked SoAs and are established between conceptually independent SoAs, which are not interconnected and not presented one in the profile of the other. However, there are remarkable exceptions, and it is possible to identify the implicational pattern (7e), constraining the attested non-parallel cases. (7) (e) The coordination parallelism implication: Non-parallel construction for non-sequential combination, contrast relation, alternative relation → non-parallel construction for sequential combination. In a given language, if a non-parallel construction is used to express a contrast relation, an alternative relation, or a non-sequential combination (simultaneous or atemporal), a non-parallel construction is also available to express sequential combination. Therefore, there seems to be one type of coordination relation that behaves differently from the others, namely sequential combination, and such differentiation can be explained in terms of syntagmatic economy since the relation of sequential combination indeed predetermines the respective temporal location of the SoAs. All languages showing a non-parallel construction only for sequential combination use an overt marker encoding the sequential temporal meaning (e.g. -Ip in Turkish (9)). The presence of an overt marker coding the temporal relation makes the successive temporal location of the SoAs explicit. Therefore, if one of the SoAs is located in time, that is, coded by means of a balanced verb form, the temporal location of the other SoA is recoverable from the context and may therefore be coded by means of a deranked form (further discussion is provided in Mauri 2008a and c). (9) Turkish, Altaic (Kornfilt 1997: 110) Hasan iş-e gid-ip ev-e dön-dü. Hasan work-dat go-Ip home-dat come-pst “Hasan went to work and came back home.” 308

9781441124609_Ch17_Final_txt_print.indd 308

11/20/2012 9:25:04 AM

Coordination

3.4 And-But-Or Languages In this final section a further generalization will be made, identifying what will be called “And-But-Or” language type (see Mauri 2008a: 289–93). “And-But-Or” languages are characterized by the following set of features: (10) (a) syndetic constructions for the three basic coordination relations; (b) internally parallel constructions for the three basic coordination relations; (c) free markers coding combination, contrast, and alternative; (d) a general marker used for both temporal and atemporal combination (“And”), a general marker used for both choice-aimed and simple alternative (“Or”), a general marker used for counterexpectative and corrective contrast (“But”), and no general markers used only for atemporal combination and opposition. “And-But-Or” languages are particularly frequent in Europe. To this language type belong most Romance languages except for Rumanian, Spanish, and Catalan (i.e. Italian, Sardinian, Portuguese, and French), Irish, all Germanic languages except for German, Icelandic, and Swedish (i.e. Norwegian, Danish, Dutch, Luxembourgish, and English), only one Slavic language (Czech), and Greek. Outside Europe, the set of features characterizing “And-But-Or” languages is not common. Of course, many languages show individual coding patterns that are also attested in “And-But-Or” languages, like the use of parallel constructions or the use of free coordinating markers. Yet, the co-existence of all features in (10) is not widespread and its occurrence in a number of adjacent languages of different families and sub-families in Europe is thus an interesting phenomenon, which might point to some areal considerations. The fact that many European “And-But-Or” languages are genetically related to each other suggests that this coding pattern may also derive from the patterns attested in the ancestor languages. However, more research needs to be done in this direction.

Notes 1. The conclusive relation (I was tired, so I decided to go to bed) and the correlative relation (The more I travel, the more I want to do it) could also be argued to be coordination relations. 2. For a diachronic survey on the grammaticalization of coordinating connectives, see Giacalone Ramat & Mauri 2010.

309

9781441124609_Ch17_Final_txt_print.indd 309

11/20/2012 9:25:04 AM

The Bloomsbury Companion to Syntax 3. Croft (2003: 144–52) makes a distinction between semantic map, which represents the multifunctionality of a given construction in a given language, and conceptual space, the overall representation of which conceptual situations may be expressed by the same construction across languages. The organization of the functions on a conceptual space represents universal relations among constructions coding these functions, thus allowing for the identification of restrictions on the cross-linguistic variation. 4. See Mauri (2008a: chapter 3) for a detailed discussion on the contraposition of the two dimensions of cooccurrence and non-cooccurrence, as associated respectively to combination and alternative relations.

310

9781441124609_Ch17_Final_txt_print.indd 310

11/20/2012 9:25:04 AM

18

Wh-Movement Luis López

Chapter Overview Introduction Wh-movement Constructions Syntax and Semantics of the Wh-phrase and the Wh-clause Wh-in-situ Multiple Wh-movement Partial Wh-movement Summary

311 312 316 319 322 323 324

1 Introduction The phenomenon referred to as wh-movement is exemplified in (1): (1) What have you bought? The interrogative word “what” appears at the beginning of the sentence. Additionally, it is the direct object of the verb “bought” and semantically it takes the same semantic role as the thing one would buy. “What” is a dependent of the verb although it does not appear anywhere near the verb. In order to express the intuition that a constituent can have a dependency with a distant constituent, linguists in the generative tradition use the technical term movement. The position that the constituent has moved from is indicated by means of a t (meaning trace). In order to track the connection between the displaced constituent and the trace linguists sometimes use a subscripted index: (2) Whati have you bought ti? 311

9781441124609_Ch18_Final_txt_print.indd 311

11/20/2012 9:23:47 AM

The Bloomsbury Companion to Syntax

The fronted phrase is referred to as wh-phrase because in English this sort of phrase often begins with [wh-]—“who,” “what,” “where,” and “when” are whphrases, but so is “how” despite the fact that orthographically it is not. In fact, we use the term wh-phrase even when talking about languages in which no [wh-] segments exist. When a wh-phrase appears in a distant dependency, then we call it wh-movement, as exemplified in (2). Wh-movement is one of the topics in generative grammar that has inspired the most literature. After literally hundreds of articles, theses, and books, we have learned many facts about its syntactic and semantic properties and cross-linguistic variation. However, there is great controversy as to how to put everything together in a comprehensive theory—and therefore, whmovement continues to be an attractive subject for every new generation of linguists, see Parodi and Quicoli, this volume, as an example of that. This chapter is organized as follows. Section 2 presents a brief description of the constructions that instantiate wh-movement in English. Section 3 provides an introduction to the syntax and semantics of a clause that includes wh-movement, again basing ourselves on English. Sections 4, 5, and 6 discuss typological phenomena. Section 7 presents a summary.

2 Wh-movement Constructions Wh-movement appears in English in at least three types of constructions: interrogative sentences, exclamative sentences, and relative clauses. Let us start by considering three examples of matrix interrogative sentences: (3) What have you bought t? (4) Who t has bought the rug? (5) How have you t bought the rug? Example (3) involves a fronted direct object, (4) a subject, and (5) an adjunct. Notice that object and adjunct wh-movement trigger inversion between the auxiliary and the subject. A subject wh-phrase triggers no inversion. If the clause does not include an auxiliary, the grammar of English forces the insertion of the “dummy” auxiliary “do” in the inversion construction. I discuss do-insertion in more detail in Section 3. (6) What did you buy t? (7) Who t bought the rug? (8) How did you t buy the rug?

312

9781441124609_Ch18_Final_txt_print.indd 312

11/20/2012 9:23:48 AM

Wh-Movement

The wh-phrase of a matrix interrogative sentence may originate in a subordinate clause: (9) What did you say [that Susana bought t]? In fact, there can be multiple—infinite—levels of embedding between the trace and the wh-phrase without a rupture of the dependency: (10) What did you say [that Susana thought [that Mary suggested [that John bought t]]]? However, since the work of Ross (1967) we have known that long-distance wh-movement is in fact very restricted. Notice that the embedding of clauses in (10) involves only complement clauses. That is: “that John bought” is the complement of “suggested,” “that Mary suggested that John bought” is the complement of “thought,” and “that Susana thought that Mary suggested that John bought” is the complement of “say.” Thus, as long as we extract from complement clauses, the result is acceptable. However, when we extract out of an adjunct clause, the result is highly ungrammatical: (11) *What did Susan buy the BMW [before John saw t]? [Compare: Susan bought the BMW before John saw the Benz]. The trace of the wh-phrase is included in an adjunct clause, showing that the wh-phrase moved out of the adjunct clause in order to reach the front position. Adjunct clauses are what linguists call islands (a term coined by Ross), constituents out of which it is impossible to move anything. For classic discussions of islands, see Merchant (2000) and Rizzi (1990), among many others. We are going to see islands again throughout this chapter. Wh-movement may target a subordinate clause. In this case, we may have a subordinate interrogative clause, as shown in the following examples. Notice that a subordinate interrogative clause is selected by a handful of interrogative verbs, such as “ask” and “wonder.” Orthographically, subordinate interrogative sentences do not exhibit a question mark. However, this should not confuse us as to their interrogative nature: (12) Mary asked [what her daughter bought t]. (13) Mary asked [who t bought the rug]. (14) Mary asked [how her daughter bought the rug t]. Notice a difference between matrix and subordinate interrogatives: the subordinate interrogatives trigger no inversion (unless the speaker is from Belfast, see 313

9781441124609_Ch18_Final_txt_print.indd 313

11/20/2012 9:23:48 AM

The Bloomsbury Companion to Syntax

Henry 1995). In fact, matrix interrogatives are the only wh-movement clauses that trigger inversion in English. The association of inversion with wh-movement is fairly common cross-linguistically but by no means general. Subordinate interrogative wh-movement can also cross an unlimited number of clausal boundaries: (15) Mary asked [what Susana said [that Mary thought [that John suggested . . . [that her daughter bought t]]]]. Exclamative clauses in English also involve wh-movement (see Castroviejo 2006 and Villalba 2008 for recent discussions of exclamatives): (16) What a beautiful house you bought t! (17) How efficiently he works t! (18) What an interesting man t is dating Chris! Given the nature of the speech act involved in exclamation, subordination and long-distance exclamatives sound somewhat unnatural. Nonetheless, the following are more or less acceptable: (19) #John exclaimed what a scoundrel Peter was t! (20) #(I’m surprised) How intelligent they thought that Mary would turn out to be t! The third type of wh-movement that can be exemplified in English is relative clauses: (21) This is the man [who t bought the rug]. (22) This is the rug [which the man bought t]. (23) This is the money [with which the man bought a rug t]. Relative clauses introduce additional challenges for the linguist. In many languages relative clauses—or at least clauses that look like relative clauses and function like relative clauses—do not instantiate wh-movement. Even in English, one could doubt that wh-movement is a necessary ingredient of relative clauses. In sentence (24a) one could replace the wh-phrase “who” with the complementizer “that,” obtaining (24b) and in (25) one could do the same with “which.” In fact, (25c) and (26b) show that a relative clause is possible without any apparent wh-phrase: (24) (a) This is the man [who bought the rug]. (b) This is the man [that bought the rug]. 314

9781441124609_Ch18_Final_txt_print.indd 314

11/20/2012 9:23:48 AM

Wh-Movement

(25) (a) This is the rug [which the man bought]. (b) This is the rug [that the man bought]. (c) This is the rug [the man bought]. (26) (a) This is the money [with which the man bought a rug]. (b) This is the money [the man bought a rug with]. However, linguists have assumed since Chomsky (1977) that there are some instances of wh-movement (and therefore wh-phrases) that are not visible but can still be detected empirically by using certain tests. In fact, the syntactic representation of (25c) looks approximately as in (27), where [wh] represents an invisible wh-phrase (but see Kayne 1994 and Bianchi 1999 for an influential alternative): (27) This is the rug [[wh] the man bought t]. Why should we assume an invisible wh-phrase and invisible wh-movement? It is worthwhile to linger a little on this topic because it gives the reader a good idea of how syntactic argumentation is built up. Let’s go step-by-step. First, notice that wh-movement in relative clauses can also be long-distance, as shown in (28a). In (28b) I show a relative clause without an overt wh-phrase: (28) (a) This is the man [who John thinks [that Peter claims [t bought the rug]]]. wh] John thinks [that Peter claims [t bought the rug]]]. (b) This is the man [[wh However, notice that the following sentences are ungrammatical: (29) (a) *This is the man who John bought the book [before Mary saw t]. wh] John bought the book [before Mary saw t]. (b) *This is the man [wh As for (29a), we have a ready explanation: the wh-phrase has been extracted out of an adjunct clause. Adjunct clauses are islands for movement. What about (29b)? If we assume that there is a “shadowy” wh-phrase and therefore “shadowy” wh-movement, then the reason why (29b) is ungrammatical is the same reason why (29a) is ungrammatical, namely, the presence of a movement island. If we do not acknowledge invisible categories and invisible movement, then we do not have an explanation for the ungrammaticality of (29b).

315

9781441124609_Ch18_Final_txt_print.indd 315

11/20/2012 9:23:48 AM

The Bloomsbury Companion to Syntax

Is it a peculiarity of linguistics or syntax to posit invisible categories on the basis of their effects? It seems to me that it is a fairly regular strategy in the natural sciences: gravitation and magnetism are just two examples of “invisible forces.” Thus, if the natural sciences have been enormously enriched by positing invisible forces, it seems that there should be no obstacle for the cognitive sciences to do the same. I conclude that relative clauses in English always involve wh-movement even when the wh-phrase is not apparent. There are other constructions that do not involve a wh-phrase but have many or all of the properties of wh-movement. One such case is neg-preposing. Neg-preposing consists of placing a phrase headed by a negative word at the beginning of the clause. Unlike wh-movement, neg-movement is optional: (30) Never in my life would I say such a thing t. (But also: I would never in my life say such a thing.) Notice that neg-preposing triggers subject–auxiliary inversion, just like matrix wh-movement of interrogative sentences. Focus-movement also shares many properties with wh-movement. Focusmovement takes a constituent that bears contrastive focus and fronts it without inversion. Like neg-movement, focus-movement appears to be optional. Unlike neg-movement, auxiliary inversion is ungrammatical: (31) (a) RUGS I bought t. (Not rags). (But also: I bought RUGS. (Not rags).) (b) *RUGS did I buy. In some languages, focus-movement and wh-movement are identical. The cases of Hungarian (Horvath 1985; Brody 1990) and Basque (Ortiz de Urbina 1987; Arregi 2002) are the best studied.

3 Syntax and Semantics of the Wh-phrase and the Wh-clause A wh-phrase can consist of a single wh-word or might have a more complex structure: (32) What (33) Which book (34) Which of the books

316

9781441124609_Ch18_Final_txt_print.indd 316

11/20/2012 9:23:48 AM

Wh-Movement

The wh-word is regarded as a determiner. Following Abney’s (1987) hypothesis we consider determiners to be heads that take as complements noun phrases or prepositional phrases headed by “of”: (35) [D What] (36) [DP Which [NP book]] (37) [D Which [PP of the books]] Let’s assume a clause structure that includes a TP and a CP, as in Chomsky (1986) and much subsequent work. Within such a theory of clausal structure, we can take the wh-phrase to surface in Spec,C. The subject–auxiliary inversion is the outcome of T-to-C movement. Since T-to-C movement is preceded by raising of the auxiliary into T, the auxiliary ends up dragged into C: (38) [CP which book [C C+T+has] [TP John tT thas read twh]]. The question now is, why is there wh-movement? There are in fact two “why” questions. Consider the question “why is the dog drinking?” One possible answer is: because drinking is necessary to restore the appropriate level of moisture in the body. Another possible answer is: because it is thirsty. Being thirsty is the proximate cause, an evolutionary mechanism that arose and stayed because it satisfied the distant cause, maintaining a minimum of water in the body. Turning back to wh-movement, there are also two possible “why” questions, one attending to proximate causes, the other to distant causes. The distant cause will be related to the functionality of language within the cognitive systems with which it interfaces. The proximate cause will be the mechanisms set up in the language faculty to bring about the distant goal. I start with the proximate cause. In the following I present a simplified model based on key features of Chomsky (2000, 2001) and Pesetsky and Torrego (2001). It is certainly not the only one in the market. For instance, Rizzi (1996) and Simpson (2000) are close relatives while Ginzburg and Sag (2001) is a more distant one. An overview of current research can be found in Cheng and Corver (2006). Consider the following structure, which represents a wh-clause before whmovement or T-to-C movement has taken place. In (39), the subject is in Spec,T and the auxiliary is adjoined to T: (39) [CP C[uWh][Q=Int] [TP John [T+has] [VP thas [VP read what[uQ]]]]]. The structure in (39) hypothesizes a number of features that together bring about movement. I include an unvalued [wh] feature (a [uWh] for short) as part of the structure of C. The [uWh] indicates that C needs to be in a syntactic 317

9781441124609_Ch18_Final_txt_print.indd 317

11/20/2012 9:23:48 AM

The Bloomsbury Companion to Syntax

dependency with a [Wh] determiner. Likewise, the wh-word also has an unvalued feature, which here I represent as [uQ]. The [uQ] feature can be valued as “interrogation,” “exclamation,” or “modification.” In this case, notice that C has a valued [Q] feature, [Q=Int], which turns the clause into an interrogative clause. Let’s take unvalued features to be the triggers of syntactic dependencies. Thus, the [uWh] turns C into a probe, bringing about a search in the c-command domain of C for a [Wh] feature. The contact between the two feature structures gives rise to an operation called Agree, which values all the unvalued features. If we posit two extra features on [uWh] we motivate the movement, the Attraction, of the wh-phrase to Spec,C as well as T-to-C:1 (40) [CP what[Q=Int] T+C[Wh][Q=int] [TP John tT(t+has) [VP thas [VP read twh]]]]. Let’s now turn to do-support. As we saw above, if there is no auxiliary in the structure then a dummy “do” is inserted: (41) What did John read? “Do-support” is one of the most famous linguistic phenomena since Chomsky (1955), the first analysis. A variety of distinct accounts of do-support have been proposed over the years but they all share a general outline that can be summarized as follows. The first assumption is that T is a bound morpheme that needs to adjoin to a verbal lexeme. The second assumption is that C attracts T, as we have seen above. Now, if T has been attracted to C, it finds itself too far from any verbal lexeme (in fact, it is taken as standard in current theory that head-tohead dependencies can only take place if there are no other heads in between, see Borer 1984). If T in C cannot attach to the lexical verb, the resulting syntactic structure cannot spell out. Thus, the dummy lexical item “do” is inserted in C and T can spell out as an affix of “do.” Let’s now turn to the other “ ‘why’ question”—the distant reason for whmovement. The existence of wh-movement is related to the interpretation of the clause. A regular clause is a proposition, like “John bought War and Peace,” which can be true or false. Wh-movement radically alters the semantic structure of the sentence. A clause that includes a wh-phrase is not a proposition but a predicate: its denotation is a set of possible propositions. For instance, the denotation of an interrogative clause is the set of possible answers (Hamblin 1973): (42) q=Which book has John read? Denotation of q= {John has read Formal Syntax, John has read Core Syntax, John has read Minimalist Syntax . . . }

318

9781441124609_Ch18_Final_txt_print.indd 318

11/20/2012 9:23:48 AM

Wh-Movement

I surmise that wh-movement forms part of the language faculty in order to bring about this semantic operation. Within this framework, the denotation of “which book” should be something like: “there is something that has the property of being a book and some other property (for instance, the property of being read by John).” Formally: (43) Denotation of “which book” = λP.∃x [P(x) ∧ book (x)] Let us now focus on the semantics of the complementizer C[wh] and the entire clause. The semantic function of C[wh] is to shift a proposition-type thing like [John read something] into a Hamblin-style set of propositions. The result is as follows (see Dayal 2006 for more details): (44) Which book did John read? CP = λP. ූx [P(x)] ∧ book (x)] (p= John read (x)) → which book λP.ූx [P(x) ∧ book (x)]

C λq[p=q]

→ ූx [p= John read (x) ∧ book (x)]] → λp. ූx [p= John read (x) ∧ book (x)]] C’ = λq[p=q] (John read (x)) → p= John read (x) TP = John read (x)

In the following sections I discuss some [Wh] phenomena that have aroused considerable interest in the last 30 years or so. They are referred to as wh-insitu (Section 4), multiple wh-movement (Section 5), and partial wh-movement (Sections 6).

4 Wh-in-situ The label wh-in-situ includes two distinct phenomena. First we have the fact that many languages in the world do not exhibit superficially any form of whmovement. The best studied among these languages are Japanese and Chinese. The following is a Chinese example taken from Huang (1995): (45) Chinese ni xiang-zhidao wo mai shenme. you ask I buy what “You ask what I bought.” 319

9781441124609_Ch18_Final_txt_print.indd 319

11/20/2012 9:23:48 AM

The Bloomsbury Companion to Syntax

Notice that the Chinese sentence in (45) is a genuine question. In English (and possibly every language) it is possible to leave a wh-in-situ in a context referred to as echo question: (46) -I just bought a $60,000 sports car. -You bought WHAT? Notice that an echo question is not a genuine question, but a way of expressing surprise or anger. They are interesting in their own right but in this short chapter I will not discuss them in detail. The second wh-in-situ phenomenon involves clauses with multiple whphrases. In English and many other languages when there is more than one wh-phrase in a sentence only one of them moves: (47) (Sorry, I didn’t understand the end of the story). Who said what? [Possible answers: John said that p, Mary said that q, Charles said that r . . . )] Notice that the fact that the answer to the question may involve a set of pairs of people and propositions they could say indicates that the in-situ “what” is a true interrogative word and the resulting sentence is a double question (a set of sets of possible answers). Let’s begin with Chinese-style wh-in-situ. Huang (1982) studied the clauses that include wh-in-situ and reached a surprising conclusion: there is in fact wh-movement in these clauses. His reasoning, though complex, is logically straightforward. Clauses that contain overt wh-movement in English and other languages have been shown to have certain properties. Clauses in Chinese that do not have overt wh-movement have many of the same properties. Thus, we should conclude that there is wh-movement in Chinese, even if it is not overt. Let’s see how the reasoning works. Recall that when we place a trace in an adjunct clause the result is sharply ungrammatical. Adjuncts are islands: (48) * What did John visit Mary [before he purchased t]? Huang realized that in Chinese you cannot have a wh-phrase in an adjunct clause: (49) Chinese *ta zai Lisi He with Lisi

weishenme mai shu yihou shengqi le? why buy book after angry art

320

9781441124609_Ch18_Final_txt_print.indd 320

11/20/2012 9:23:49 AM

Wh-Movement

The English equivalent of the previous sentence would be something like: “*Why is he angry with Lisi after he bought a book t?” (notice how different word order in Chinese is from English). When reading this sentence it is important to interpret “why” as an adjunct of “bought,” not of “angry.” The crucial question that Huang asked is: if there is no wh-movement in Chinese, why is (49) ungrammatical? Huang’s careful argumentation convinced the community of linguists that there is something like wh-movement in Chinese, even if phonetically inaudible. This proposal integrated very well with a grammatical model in which syntactic operations take place sequentially and in which some of these operations take place after the derivation has spelled out (see in particular May 1985). Since Huang’s original proposal there have been numerous discussions and alternatives (see for instance Nishigauchi 1990 and Watanabe 1992b, among many others). To my mind, Pesetsky (2000) presents conclusive evidence that there is a syntactic dependency between a wh-phrase in situ and the C[wh] head. See also Nissenbaum (2000) for an extended discussion of covert movement. One question that has often been asked is why some languages move their wh-phrases overtly while others do not. Recently, Richards (2010) has made a proposal in terms of a phonological requirement. Assume that there is a phonological requirement that the wh-phrase and C[wh] must be contained within the same prosodic phrase. Languages without overt wh-movement are those that are able to construct a big prosodic phrase that embraces the wh-phrase in situ and C[wh]. Languages that are not able to build a prosodic phrase that includes the wh-phrase and C[wh] will need to move the wh-phrase to a position adjacent to C[wh], where there will not be any prosodic barriers between them. This seems to me to be the most promising proposal currently in the market. Let us consider now the second type of wh-in-situ. Recall that in this second type of wh-in-situ we have a sentence with multiple wh-phrases but only one of them moves to the front. The earlier example is repeated here: (50) (Sorry, I didn’t understand the end of the story). Who said what? [Possible answer: John said that p, Mary said that q, Charles said that r . . . )] We have seen that Chinese-style wh-in-situ moves covertly. Can we establish if this sort of in situ wh-phrase also moves covertly? In order to test if this is the case, we follow the same methodology proposed by Huang (1982): place the wh-phrase in an island context. Consider the following example: (51) Which student suffered a headache after reading which book? [Possible answers: John suffered a headache after reading War and Peace, Chris suffered a headache after reading World and Peas, Pat . . . ] 321

9781441124609_Ch18_Final_txt_print.indd 321

11/20/2012 9:23:49 AM

The Bloomsbury Companion to Syntax

This example can be interpreted as a double question, as shown by the possibility of the pair-list reading. However, “which book” is embedded within an adjunct clause. But adjunct clauses are islands for movement. This entails that “which book” does not move out of the adjunct clause and therefore it must be connected with the matrix C[wh] by means other than syntactic movement (see Reinhart 1997 for a semantic approach to in-situ wh-phrases). Thus, when there are several wh-phrases in the sentence, only one of them moves. Which one? Can any of them move? The answer is no. As you can see in (52), if the direct object and the indirect object are both wh-phrases, only the direct objet can raise: (52) (a) What did Chris give t to whom? (b) *To whom did Chris give what t? (c) *Whom did Chris give what to t? In fact, it turns out that when there is more than one wh-phrase in a sentence, the one higher in the c-command chain is the only one that can move—a whphrase cannot raise “jumping over” another wh-phrase. This is referred to as Superiority Effect (see Chomsky 1973 for the original description, also Pesetsky 1987 for a substantial refinement). If we adopt the Attract theory that I presented above, we can understand Superiority as “attraction of the closest”: the C[uWh] attracts the closest [Wh] determiner that it finds.

5 Multiple Wh-movement We have seen that there are some languages that allow no wh-movement, while others allow for wh-movement of one wh-phrase leaving the others in situ. There is of course a third logical possibility: a language in which all wh-phrases are displaced. This property holds of several languages spoken in Central Europe: the Slavic languages as well as non-Slavic Hungarian and Romanian. The following is a Bulgarian example taken from Rudin (1988): (53) Bulgarian Koj kogo vižda?   (* Koj vižda kogo?) who whom sees Rudin’s work, together with Richards (2001), Bošković (2002), and many others has shown that multiple wh-movement is not cross-linguistically uniform.

322

9781441124609_Ch18_Final_txt_print.indd 322

11/20/2012 9:23:49 AM

Wh-Movement

Thus, a set of properties join Romanian and Bulgarian together in contrast to Serbo-Croatian, Polish, and Czech. For instance, this second group of languages allows for constituents between the first and the second wh-phrase while the first group of languages does not. In the following pair of examples, I use the perfective auxiliary e of Bulgarian, je of Serbo-Croatian, both roughly equivalent to “has.” We can see the contrasting grammaticality judgments when the auxiliary is placed between the first and second wh-phrases: (54) Bulgarian Koj (*e) kakvo (*e) who what

na kogo e dal? to whom has given

(55) Serbo-Croatian Ko je što (*je) kome (*je) dao? who has what to-whom given “Who has given what to whom?” Another interesting contrast between the Bulgarian set and the Serbo-Croatian set concerns Superiority. Bulgarian respects superiority strictly and the different wh-phrases are stacked in a fixed order. In Serbo-Croatian the subject whphrase needs to go first but the other wh-phrases may come in any order.

6 Partial Wh-movement In some languages it looks as if long wh-movement divides the wh-phrase in two parts. In the matrix Spec,C, in the position where one would expect to find the entire wh-phrase, one can see a neutralized form of a wh-word—a form that has no properties other than being [wh]. Simultaneously, a true wh-phrase appears in the Spec,C of the subordinate sentence. The following is a German example: [CP wen Sabine liebt]]? (56) [CP Was  glaubst  du  [wh] believe you whom Sabine loves “Who do you think Sabine loves?” Fanselow (2006) presents an overview of partial wh-movement and shows that it is a fairly heterogeneous phenomenon. The most common sort of analysis takes the higher wh-word to be an expletive, that is, a constituent without semantic content that is present in the structure only to satisfy a formal requirement— such as the “it” in “it is a pity that Sabine doesn’t love Golo.”

323

9781441124609_Ch18_Final_txt_print.indd 323

11/20/2012 9:23:49 AM

The Bloomsbury Companion to Syntax

7 Summary In this chapter I have presented an introduction to wh-phrases and wh-movement. I have presented a sketch of a formal analysis within the framework developed in Chomsky (2000) and Pesetsky and Torrego (2001) and presented some of the cross-linguistic phenomena that have given rise to considerable interest among linguists.

Note 1. This sketch leaves out some important details and empirical facts. For instance, I am leaving aside the lack of T-to-C with subject Wh-phrases and in subordinate clauses. For a thorough discussion, see Pesetsky and Torrego (2001).

324

9781441124609_Ch18_Final_txt_print.indd 324

11/20/2012 9:23:49 AM

19

Complementation Claudia Parodi and A. Carlos Quicoli

Chapter Overview Complements and Adjuncts Phrasal and Sentential Complements Rule Interaction in Complex Complement Structures Conclusions

325 327 336 340

1 Complements and Adjuncts In a sentence there are some constituents that play an essential syntactic and semantic role that serve to complete the meaning of categories such as verbs, nouns, adjectives, or prepositions. Such constituents are called complements. Complements are normally in contrast to adjuncts, which are constituents that play no obligatory syntactic or semantic role in the sentence. Consider for example the facts below: (1) (a) John kissed Mary. (b) *John kissed. (2) John kissed Mary in the convent. From the examples above we see that with a verb like “kiss” the complement “Mary” (traditionally called a “direct object complement”) must be obligatorily present for the sentence to be grammatical, as evidenced by the grammaticality of (1a). If the complement is not present the sentence is ungrammatical, as shown by the ungrammaticality of (1b). On the other hand, the expression “in the convent” is an adjunct, and can be absent without affecting the completeness of the sentence. 325

9781441124609_Ch19_Final_txt_print.indd 325

11/20/2012 9:25:20 AM

The Bloomsbury Companion to Syntax

There are some verbs like “offer” and many others that require two complements—a direct object complement and an indirect object complement. These are sometimes called “ditransitive” of “bitransitive” verbs. As shown by the examples in (3) the two complements must occur or the resulting sentence is ungrammatical: (3) (a) (b) (c) (d)

John offered a ring to Mary. *John offered a ring. *John offered to Mary. *John offered.

Not all verbs can occur with complements. Such verbs are called “intransitives.” For example, the verb “sleep” (an intransitive verb) cannot occur with a complement at all: (4) (a) John slept. (b) *John slept the dog. There are also some verbs such as “look” as in sentences like “John looked at the baby” that must occur with a prepositional phrase (PP) as a complement. In some formal theories of grammar, such as generative grammar, these distinctions are accounted for by the concept of “subcategorization,” which specifies the syntactic context in which the verb may be introduced in the sentence. Thus the verbs that occur in the sentences above have the subcategorization frames given in (5): (5) (a) (b) (c) (d)

sleep : kiss : offer : look :

+ [____] (intransitive) + [____ NP] (transitive) + [____ NP PP] (ditransitive) + [____ PP] (prepositional verbs)

As shown by the examples previously discussed, normally the subcategorization rule for verbs cannot be violated.1 Adjuncts, on the other hand, are accessory elements. They contribute an extra idea to the sentence, but are not obligatory. Compare: (6) (a) John kissed Mary in the convent. (b) *John kissed in the convent. (c) John kissed Mary. 326

9781441124609_Ch19_Final_txt_print.indd 326

11/20/2012 9:25:21 AM

Complementation

The phrase “in the convent” in (6a) just adds a detail—the location where the kissing occurred (but remember “the devil is in the details”). However, this information, although interesting, is not necessary to complete the meaning of “kiss.” If the adjunct is kept, but the complement is omitted, the result is an ungrammatical sentence, as shown in (6b). However, if the adjunct is removed and the complement is kept, the sentence is grammatical as shown in (6c). Because adjuncts and complements o en display different syntactic behaviors, in some formal theories, complements and adjuncts are represented at different levels in the sentence structure (“trees”). According to X-bar Theory of generative grammar, the structural representation of the VP (verb phrase) of a sentence like (6) would be as shown in (7). For newer proposals see Nunes, this volume; Chomsky 1995; Adger 2003; Hornstein et al. 2005; Luraghi and Parodi 2008. (7)

VP V’

PP in the convent

V’ V

NP

kiss

Mary

In the tree in (7) the close complement relation holding between “Mary” and “kiss” is represented by the fact that they are placed in the same level in the tree—both are “sisters” under the V’ node (pronounced V-bar node). By contrast, the more distant adjunct PP (“in the convent”) is placed at a different level from the verb. It is placed as a sister of the higher V’ node immediately under the VP node, which is the adjunct position (For an introduction of these concepts, see Haegeman 1991; Freidin 1992; Luraghi and Parodi 2008). This formalization captures the more distant relationship between a verb and its adjunct.

2 Phrasal and Sentential Complements According to their syntactic structure, complements may be classified in two basic types: (i) phrasal complements and (ii) sentential complements. These two types of complements can accompany phrasal categories such as Noun 327

9781441124609_Ch19_Final_txt_print.indd 327

11/20/2012 9:25:21 AM

The Bloomsbury Companion to Syntax

Phrases (NP), Verb Phrases (VP), and Adjective Phrases (AP). In what follows we will illustrate each case (for details, see Quicoli 1982; Bošković 1997; Luraghi and Parodi 2008, Helasvuo, this volume).

2.1 Phrasal Complements The set of examples in (8) illustrates different types of phrasal complements that may accompany VPs. (8) Phrasal Complements with VPs: (a) John hit [NP the ball]. (b) John gave [NP the ring] [PP to the girl]. (c) John put [NP the book] [PP on the table]. (d) John looked [PP at the picture]. Typically, VPs occurring in simple sentences may be complemented by a NP, a PP or both. As mentioned above, subcategorization provides the criterion used to determine whether the phrases accompanying the verb are complements or adjuncts. As noted before, all the complements that appear in the examples in (8) are obligatory. If a complement is removed in such examples the sentences will become ungrammatical. Adjectives may also be subcategorized for phrasal complements, normally a PP, as shown in (9): (9) Phrasal Complements with APs: (a) John is faithful [PP to his ideals]. (b) John is proud [PP of his son]. The PPs that occur in these examples have the status of complements. So parallel to verbs, such adjectives may be classified as transitive adjectives. Also like verbs, not all adjectives may occur with a complement. For example, the adjective “old” cannot occur with a complement as shown by the ungrammaticality of (10b). (10) (a) The book is old. (b) *The book is old to Mary. Thus, adjectives resemble verbs. They may be classified as transitive and intransitive adjectives depending on whether or not they subcategorize for a complement. The following examples illustrate phrasal complements that accompany NPs. 328

9781441124609_Ch19_Final_txt_print.indd 328

11/20/2012 9:25:21 AM

Complementation

(11) Phrasal Complements with NPs: (a) The destruction [PP of the city] pleased Nero. (b) The student [PP of physics] won the prize. Notice that the NP complements in (11) are PPs. While complements that appear in the subcategorization of VPs are normally obligatory (as those in (5)), the complements accompanying NPs can be removed without triggering ungrammaticality, as shown in (12): (12) (a) The destruction pleased Nero. (b) The student won the prize. However, the PPs in (11) may still be classified as complements because they are selected by the Noun. Unlike subcategorization, which is a strictly syntactic relation, selection is a more abstract semantic/syntactic relation holding between a lexical item and its complements. For example, in “*Mary annoys the computer” the subcategorization of the verb “annoy” is met, since the verb is accompanied by a NP complement “the computer.” However, the sentence is ungrammatical because “annoy” selects an animate object, which “the computer” is not. The sentence can be well-formed by inserting an animate complement: “Mary annoys the teacher.” Based on selection relations, it is possible to detect the differences between complements and adjuncts involving PPs as shown by the following examples: (13) (a) The student of physics with long hair won the prize. (b) *The student with long hair of physics won the prize. In (13a) the PP “of physics” is a complement of “the student” while the PP “with long hair” bears the relation of an adjunct, as shown in (14). NP

(14) N

PP with long hair

N’ N

PP

student

of physics 329

9781441124609_Ch19_Final_txt_print.indd 329

11/20/2012 9:25:21 AM

The Bloomsbury Companion to Syntax

The ungrammaticality of (13b) can be explained by the fact that the PP “with long hair” is in the position of the complement violating the selection requirement of the noun “student” as shown in the following tree: (15)

*NP N’

PP of physics

N’ N

PP

student

with long hair

In the case of nouns, as in the case of verbs, the selection requirement must be satisfied. Thus (13a) is grammatical for the same reason that sentences like “John studies physics” are grammatical. In both cases the selection requirements are satisfied by the insertion of the correct complement (“physics”). By contrast, (13b) is ungrammatical for the same reason that sentences like “*John studies with long hair” is ungrammatical. The phrase “long hair” cannot be selected as a complement of the verb “study” or the noun “student”.

2.2 Sentential Complements Sentential Complements are full sentences embedded as complements of a lexical category such as a Verb, an Adjective, or a Noun. As such, sentential complements can be embedded under a VP, an AP, or a NP. In addition, sentential complements can be finite or infinitival depending on whether or not they contain an inflected verb.

2.2.1 Finite Sentential Complements Finite Sentential Complements are embedded or subordinate clauses containing an inflected or finite verb. The following set illustrates examples of sentential complements embedded under VPs: (16) Finite Sentential Complements with VPs (a) John believes [CP that Mary likes caipirinha]. (b) Mary convinced [NP John] [CP that he should see a psychiatrist]. (c) John said [PP to Mary [CP that he will travel to Spain]. 330

9781441124609_Ch19_Final_txt_print.indd 330

11/20/2012 9:25:21 AM

Complementation

(d) John asked [CP if/whether Mary finished the job]. (e) John wondered [CP why Mary le the room]. All the sentences above contain a finite sentential complement (CP). In generative grammar sentential complements are phrases—complementizer phrases (CPs)— because they are usually introduced by a complementizer such as “that,” “whether,” or “if.” Since they are complements of verbs, they enter into subcategorization and selection relations with the verb. This explains why sometimes there is more than one complement. In (16a) there is only one sentential complement (CP), because “believe” is subcategorized for a CP complement. On the other hand in (16b), there is a phrasal complement NP and a sentential complement CP because the verb “convince” subcategorizes for both an NP and a CP complement. In (16c) there are two complements as well: a PP complement and a CP complement, since the verb “say” subcategorizes for both. In (16d) the verb subcategorizes for a CP and selects the sentential complementizer “if,” or “whether,” (a wh-word). Finally, verbs such as “ask” and “wonder” allow indirect questions to occur as their sentential complements, such as in (16e). The verb “wonder” subcategorizes for a CP sentential complement and selects the complementizer “why” (or any wh-word such as “what,” “when,” and so on). In addition, it may also select the complementizer “whether” or the complementizer “if”. APs may also occur with a sentential complement, as shown in (17): (17) Finite Sentential Complements with APs (a) It is likely [CP that John will win the election]. (b) John is happy [CP that Mary won the prize]. (c) It is unclear [CP whether John will come]. In (17a) the adjective “likely” subcategorizes for a CP complement, which is obligatory. In (17b), however, the sentential complement may be omitted. It is correct to say “John is happy”. In (17c) The adjective “unclear” may select a wh-complementizer “whether,” or any other interrogative word (wh-word), which may be omitted. Finally, NPs may also occur with a sentential complement, as shown in (18): (18) Finite Sentential Complements with NPs (a) The fact [CP that Mary likes caipirinha] bothers John. (b) The question [CP whether/why the principal should resign] became an issue. (c) Hago cuenta [CP que he hallado un tesoro]. (Cervantes) I make count that I have found a treasure “I came to the realization that I have found a treasure.” 331

9781441124609_Ch19_Final_txt_print.indd 331

11/20/2012 9:25:21 AM

The Bloomsbury Companion to Syntax

Abstract NPs such as “the fact,” “the question” and “the idea” may select sentential complements. Evidence that these CPs are complements comes from two facts. First they may select a complementizer. In (18a) the NP “the fact” selects the complementizer “that”, while in (18b) the NP “the question” selects a wh-complementizer such as “whether,” “if,” or any wh-word. If we reverse the complementizers in (18a) and (18b), the result would be ungrammatical. Compare: *The fact [CP whether/why Mary likes caipirinha] bothers John and “*The question [CP that the principal should resign] became an issue”. Second, as we will show below, such CPs are sentential complements that behave quite differently from relative clauses, which are adjuncts. Finally we can see that Spanish nominal expressions such as hacer cuenta (“come to realize”) may also select a finite sentential Complement. An interesting contrast can be observed in sentences containing CPs that are sentential complements vs. CPs that are adjuncts, such as relative clauses. These distinctions are illustrated by the examples in (19): (19) (a) The fact [CP that John discovered the secret] is amazing. (b) The fact [CP that John discovered_ ] is amazing. In (19a) the CP is a sentential complement. Notice that the subordinate sentence is complete (“John discovered the secret”). The word “that” is a complementizer and cannot be replaced by a relative pronoun such as “which” (“*The fact [CP which John discovered the secret] is amazing”). By contrast, in (19b), the CP is an adjunt. It is a relative clause. Notice that in this case the embedded sentence contains a gap (“John discovered __”), which corresponds to the relative pronoun “that”, placed in the front of the clause. Notice that in (19b) the word that can be replaced by a relative pronoun such as “which” (“The fact which John discovered is amazing”). In generative grammar these different properties are captured by assuming different positions for the complement CP and the adjunct CP. The relevant portion of the structure of (19a) that contains a sentential complement is shown in the diagram in (20): (20)

NP ...

N’ N fact

CP [that John discovered the secret]

332

9781441124609_Ch19_Final_txt_print.indd 332

11/20/2012 9:25:21 AM

Complementation

In (20) the CP is in complement position, since it is sister to the Noun “fact” under the N-bar (N’) node. By contrast, (19b) contains an adjunct CP (a relative clause). This CP is in adjunct position as shown in the partial tree representation (21). (21)

NP N’ N’ N

CP [that John discovered __] ...

fact As can be seen in the diagram, in (21) the CP is a sister to the N-bar node (N’), which is in the adjunct position inside NP.

2.3 Infinitival Sentential Complements In most languages infinitival sentential complements are CPs, o en distinguished by the fact that they contain a non inflected verb. Parallel to finite Sentential Complements, infinitival sentential complements occur with VPs, APs, and NPs. We shall now illustrate each type. Although infinitival sentential Complements may contain a phonetically realized subject (e.g. “Mary,” “the doctor,” and so on), in most cases they do not contain a phonetically realized subject. In generative grammar the subject of an infinitive is assumed to be an abstract element PRO, which is semantically interpreted as coreferential with an NP in the main clause. Examples of Infinitival Sentential Complements occurring inside VPs are given below. (22) Infinitival Sentential Complements within VP (a) John expects [CP Mary to win the race]. (b) John tried [CP PRO to cheat in the exam]. (c) John persuaded Mary [CP PRO to see a specialist]. (d) John promised Mary [CP PRO to wash the car]. (e) I prefer very much [CP for Bill to leave]. In (22a) the infinitival sentential complement contains a lexical subject “Mary”. However, the most common cases are represented by (22b–d), in which the subject of the infinitive is the abstract element PRO. In (22b) PRO is interpreted as 333

9781441124609_Ch19_Final_txt_print.indd 333

11/20/2012 9:25:21 AM

The Bloomsbury Companion to Syntax

coreferential with the main subject “John”. That is, PRO is “controlled” by the subject of the main clause. In (22c), on the other hand, PRO is understood as coreferential with, or as controlled by, the object of the main clause “Mary”. In (22d), even though the verb “promise” contains an object “Mary,” the PRO is controlled by the subject of the main clause “John.” Notice also that in the examples (22a–d) there is no overt complementizer in CP. In these cases we assume that there is an abstract complementizer that is selected by the main verb. However, in (22e) we see that the embedded clause contains the overt complementizer “for”, which shows that such complements are CPs. Some verbs allow the phenomenon of split infinitives or raising in generative terms. This phenomenon can be observed with infinitives embedded under verbs such as “seem,” “happen,” among others. This is shown in (23): (23) (a) John seems [CP tnp to adore Madonna]. (b) John happens [CP tnp to know the president]. Notice that the subject of “to adore” and to “know” is “John” in both cases. But it appears split or dislocated (i.e. “raised” in generative terms) from its original position (indicated by the NP trace tNP) to occupy the subject position of the main verb in the clause (see Stowell in this volume). We can see that “John” is not the original subject of “seem” and “happen” by replacing the infinitive with a finite verb, as shown in (24): (24) (a) It seems [CP that John adores Madonna]. (b) It happens [CP that John knows the president]. The examples above illustrate three types of infinitival constructions intensively studied in generative grammar. (22a) is known as Exceptional Case Marking Construction (ECM construction), meaning that the subject of the infinitive exceptionally receives Accusative Case from the main verb “expect.” (22b–d) are known as structures of Control, while (23c) is a case of “for . . . to” construction, in which the prepositional complementizer “for” gives Accusative Case to the subject of the infinitive. Finally, the examples in (24) are referred to as raising constructions. We will now discuss examples in which an Infinitival sentential complement is embedded under an AP. The following examples are typical. (25) Infinitival Sentential Complements with AP (a) It is important [CP for John to win the election]. (b) John is very proud [CP PRO to study at UCLA]. (c) It is useful [CP PRO to know Latin]. 334

9781441124609_Ch19_Final_txt_print.indd 334

11/20/2012 9:25:22 AM

Complementation

In (25a) the infinitival sentential complement is introduced by the prepositional complementizer “for”, which allows the infinitive verb to have a phonetically realized subject (“John”). It is an example equivalent of the “for . . . to” construction that occurs with verbs, as we have already shown above. In (25b), the infinitival sentential complement contains a PRO subject, which is controlled by “John”, the subject of the main clause containing the adjective “proud”. Finally, in (25c) the infinitival sentential complement contains a PRO subject as well. However, since in this case the subject of the main clause containing the predicate be useful is an impersonal element “it” (which has no reference) PRO in this case has impersonal or arbitrary meaning (e.g. “someone,” “anyone,” “people,” and so on). Parallel to verbs, some predicate adjectives (see Matsumoto, this volume) may also allow the phenomenon of split infinitives or raising as shown in (26): (26) (a) John is likely to win the race. (b) John is certain to win the race. As in the examples with verbs in (24), the subject of “to win” is “John,” which has been raised to the position of subject of the main clause. That “John” is actually the subject of “win” can be shown by replacing the infinitive by a finite verb as in (27): (27) (a) It is likely [CP that John will win the race]. (b) It is certain [CP that John will win the race]. We will now turn to examples of infinitival sentential complements with NPs. In English there are no examples of such constructions. However in languages such as Spanish and Portuguese they are quite common. Some core examples of these constructions in these languages are given in (28). (28) Infinitival Sentential Complements with NP (a) El hecho [CP de PRO haber aceptado tu propuesta] hizo rico a Pepe. The fact to have accepted your proposal made Pepe rich “The fact that he has accepted your proposal made Pepe rich.” (b) O fato [CP de as meninas saberem latin] impressionou o diretor. The fact of the girls to know-3rd-pl. Latin impressed the director “The fact that the girls know Latin impressed the director.” In the Spanish example in (28a) the infinitival sentential complement contains a PRO subject coreferential with “Pepe”.2 In the Portuguese example (28b) the infinitival sentential complement contains a phonetically realized subject as

335

9781441124609_Ch19_Final_txt_print.indd 335

11/20/2012 9:25:22 AM

The Bloomsbury Companion to Syntax

meninas “the girls”(see Quicoli 1982). This is possible because Portuguese has the phenomenon of personal or inflected infinitive that allows the subject to agree with the infinitive verb. In (28b) the verb saberem “to know” is in the third-person plural in agreement with as meninas. The fact that the Portuguese infinitive may have inflectional morphology for person and number shows that infinitives can also be inflected even though they do not have tense.

3 Rule Interaction in Complex Complement Structures The study of structures of complementation has played and continues to play an important role in the development of syntactic theory because it allows us to examine the functioning of syntactic rules across multiple clauses. This more complex context permits the observation of how syntactic rules interact with one another, and may lead to the discovery and formulation of general principles governing such rules. To illustrate this point, consider initially the following raising structures: (29) (a) It is likely [CP that Mark will be arrested tnp (by the FBI)]. (b) It is likely [CP that the FBI will arrest Mark]. In (29a) NP movement has applied to the embedded clause of a structure essentially like (29b), moving the NP mark from the object position to the subject position of the (passivized) verb “to arrest” in the embedded clause. Since be “likely” is a predicate adjective that allows raising, NP movement may apply to the structure roughly corresponding to (29b) to also produce the following more complicated examples: (30) (a) The FBI is likely [CP tnp to arrest Mark]. (b) Mark is likely [CP tnp to be arrested tnp (by the FBI)]. In the derivation of (30a), NP movement has applied once and moved the embedded subject of the infinitive “the FBI” to the subject position under the main clause. However, in the derivation of (30b) the NP “Mark” that starts out in the direct object position of “arrest” in the embedded clause ends up as the subject of the main clause due to two consecutive applications of NP movement—once in the embedded clause, moving “Mark” from object to subject position of “arrest”, and once again, moving “Mark” from the subject position of the embedded infinitive to the subject position in the main clause.

336

9781441124609_Ch19_Final_txt_print.indd 336

11/20/2012 9:25:22 AM

Complementation

The derivation of sentences like (30a) provides evidence that syntactic operations such as NP movement are regulated by the Cyclic Principle (a principle related to Recursion). The principle of cyclic application, allows multiple applications of the same rule (e.g. NP movement) to increasingly larger cyclic domains, as illustrated by the application of NP movement in the derivation of (30a) above. In fact, the principle of cyclic application allows NP movement to produce much more complex structures, as illustrated by (31a), which is derived by NP movement in the manner indicated in (31b): (31) (a) Mark is expected to appear to be likely to be arrested. (b) [Mark is expected [tnp to appear [tnp to be likely [tnp to be arrested tnp]]]]. As we can see from (31b), the derivation of sentences such as (31a) involves three cyclic applications of NP movement. These cyclic applications of the rule had the effect of moving the NP “Mark” that originated in the direct object position of the most embedded verb “arrest”—cycle by cycle—all the way up to the subject position of “expect”, which is three clauses away. Structures of complementation also allow us to observe the complex behavior of the rule of Wh-movement (see. López in this volume). To illustrate, consider the examples in (32) that are derived by Wh-movement from a structure essentially like (33): (32) (a) [CP Bill said [CP what the teacher did twh]]. (b) [CP What did Bill say [CP twh the teacher did twh]]. (33) [CP Bill (did) say [CP the teacher did what]]. Under some analyses such examples are taken as evidence that the rule of Wh-movement must apply cyclically. In the derivation of (32a) from an initial structure corresponding to (33) Wh-movement has applied once moving the whphrase “what” from the object position of “did” to the CP structure under the embedded clause. The trace twh indicates the position from which the wh-phrase was moved. However, in the derivation of (32b) from structure (33) Wh-movement must apply twice—first moving “what” to the CP under the embedded clause, and once again from the embedded CP to the CP under the main clause. But this is possible only if we assume that Wh-movement applies cyclically. Further interesting evidence to show that Wh-movement must apply cyclically is provided by complex complement structures like (34), which is derived by Wh-movement from a structure essentially like (35):

337

9781441124609_Ch19_Final_txt_print.indd 337

11/20/2012 9:25:22 AM

The Bloomsbury Companion to Syntax

(34) Which picture of himself 1 did Mary say that Bill 1 discovered that Helen bought? (35) [CP3 Mary did say [CP2 that Bill discovered [CP1 that Helen bought [which picture of himself]]]]. In (34) the wh-phase “which picture of himself” contains the anaphoric element “himself”. Anaphors in general must be associated, or “bound” by a preceding NP/DP within the same clause, as shown by the contrasts below: (36) (a) John expects [CP Bill1 to contradict himself1]. (b) *John expects [CP Mary to contradict himself]. In (36a) the anaphor “himself” can only refer to “Bill”, which precedes it in the same clause. The anaphor cannot refer to “John”, which also precedes it, but is not in the same clause. In (36b) the sentence is ill-formed. The anaphor cannot refer to “Mary” inside the same clause due to lack of gender agreement, nor to “John” because this NP/DP is not in the same clause as the anaphor (see Chomsky 1986b, 1995). With the above considerations in mind, let us now return to the analysis of (34), which derives in the manner shown in (35). What is interesting in this example is that the anaphoric element “himself” inside the moved wh-phrase “which picture of himself” can only refer to “Bill”, which is the subject of the intermediary clause (CP2). Notice that the wh-phrase containing the anaphor started out in the object position of “bought” in the most embedded clause (CP1), and ended up under the CP in the top clause (CP3). However, the anaphor “himself” cannot refer to the subject of the most embedded clause “Helen”, nor to the subject of the top clause “Mary” due to lack of gender agreement. The facts in (34) can be plausibly explained, however, if we assume that Wh-movement applies cyclically, clause by clause, as indicated in (37). (37) [CP3 [which picture of himself 1] did Mary say [CP2 twh that Bill 1 discovered [CP1 twh that Helen bought twh]]]. As Wh-movement reaches the intermediary clause (CP2) in (37), it places the wh-phrase containing the anaphor within the purview of the intermediary clause (CP2), creating an intermediary structure essentially like [CP2 Bill discovered [which picture of himself]], see Lopez, this volume. At this stage of the cyclic derivation the conditions for the anaphor to be bound to an antecedent are satisfied, and “himself” is bound to “Bill” (for a detailed analysis, see Quicoli 2008). Since the anaphor is bound to an antecedent at this point of the cyclic derivation, subsequent cyclic applications of Wh-movement, moving the whphrase toward the CP node of the top clause (CP3) have no effect. The result 338

9781441124609_Ch19_Final_txt_print.indd 338

11/20/2012 9:25:22 AM

Complementation

is the well-formed (34) with the anaphor “himself” correctly interpreted as coreferential with “Bill”, as desired. Complement structures also allow us to study the way syntactic rules interact with one another. Thus, consider the behavior of NP movement and Wh-movement in the example given in (38), which is derived in the manner indicated in (39): (38) Who did you say is likely to be arrested? (39) [CP who did you say [twh is likely [tnp to be arrested tnp]]]. Notice that “who” is an interrogative pronoun. Hence, it is both an NP (like all pronominal elements) and a wh-word, so that it may undergo both NP movement and Wh-movement. As shown by (39), the derivation of (38) involves the application of both these rules. As can be seen in (39), NP movement applies twice in the structure. A first application of NP movement moves “who” (as an NP) from the object position to the subject position of “to be arrested” in the embedded clause, while a second application of NP movement raises “who“ from the subject position of the embedded infinitive to the subject position of the main clause, as indicated by the NP traces tnp in (39). Next, Wh-movement applies to “who” (as a wh-word), moving it to the CP node in the main clause, as indicated by the wh-trace twh in (39), so as to derive the well-formed sentence (38). What is crucial in the example above is that application of Wh-movement must wait for the two cyclic applications of NP Movement to place “who” in the subject position of “be likely” before it can apply. If Wh-movement moves “who“ directly from the embedded subject position to the CP node in the main clause, bypassing NP movement into the subject position of “be likely”, the resulting sentence is ill-formed, as shown by the ungrammaticality of (40), which is incorrectly derived in the manner indicated in (41): (40) *Who did you say it is likely to be arrested? (41) [CP who did you say [it is likely [twh to be arrested tnp]]]. As shown by (41), in the derivation of (40) the presence of “it” in the subject position of “be likely” prevents NP movement to move “who” into this position. This would force Wh-movement to move “who” from the subject position directly to the CP in the main clause, without allowing it to be placed first into the subject position of “be likely”. Facts such as these are interesting because they offer the potential for the discovery and formulation of general principles that regulate the functioning of grammatical rules. In the example above, a plausible explanation can be given by a combination of the Cyclic Principle and the Principles of Case Theory (Chomsky 1995). Misapplication of a cyclic rule in the example above has the effect of preventing the moved NP “who” to be 339

9781441124609_Ch19_Final_txt_print.indd 339

11/20/2012 9:25:22 AM

The Bloomsbury Companion to Syntax

placed in a position where it could receive Abstract Case in the course of the derivation. The structure would then be ruled out by some version of the Case Filter, which requires lexical NPs such as “who” to have Case.

4 Conclusions In this chapter we have examined different types of complementation. We have contrasted some basic properties of complements that distinguish them from adjuncts. Essentially, complements enter into subcategorization and selection restrictions with major lexical categories such as verbs, adjectives, and nouns, while adjuncts are exempt from such restrictions. In our discussion we have illustrated some of the core cases of complement structures inside VPs, APs, and NPs. We have focused our discussion on complements inside of these major categories, and we have shown how complements and adjunts can be formally represented in sentence structures. In addition, we have examined the interaction of syntactic rules in complement structures and the principles that are involved in such interactions.

Notes 1. Some verbs may have variable subcategorization in the sense that complements may be optional. For example, a verb like “write” may occur with or without complements: (a) John writes poems to Mary. (b) John writes poems. (c) John writes. This can be formalized by assuming the following subcategorization: write: +[____ (NP) (PP)]. The items in parenthesis are optional. 2. For a traditional discussion of Spanish infinitives, see Gili Gaya (1961). For descriptive approaches, see Bosque and Demonte (1999), Di Tulio (2005), and Real Academia Española (2009). For a Generative approach, see Zagona (2002).

340

9781441124609_Ch19_Final_txt_print.indd 340

11/20/2012 9:25:22 AM

20

Adverbial Subordination Holger Diessel

Chapter Overview Introduction Clause Order and Pragmatic Function Syntactic Structure Semantic Link Conclusion

341 342 346 349 353

1 Introduction This chapter is concerned with a particular type of subordinate clause, the adverbial clause, that occurs in complex sentences expressing a temporal or logical relationship between two events. For instance, the sentence The party started when Jack arrived denotes a temporal relationship between two states of affairs, that is, the beginning of the party and Jack’s arrival, that are expressed in two clauses combined by the conjunction when. In the syntactic literature, adverbial clauses are commonly seen as embedded clauses functioning as constituents of a superordinate clause. In this view adverbial clauses are considered adjuncts (or adverbials) of the main clause (or main clause predicate) serving the same syntactic function as adverbial prepositional phrases. One piece of evidence supporting this analysis comes from the fact that adverbial clauses can often be replaced by non-clausal adverbial constituents. For instance, in the sentence The party started when Jack arrived the when-clause can be replaced by the prepositional phrase upon Jack’s arrival without any significant changes in meaning. However, a number of studies have argued that the traditional analysis of adverbial clauses as adjuncts (or adverbials) is inadequate to characterize their

341

9781441124609_Ch20_Final_txt_print.indd 341

11/20/2012 9:26:00 AM

The Bloomsbury Companion to Syntax

syntactic function (cf. Matthiessen and Thompson 1988; Givón 2001b: chapter 19, and Halliday 1994: 242–8). According to Matthiessen and Thompson (1988: 280–1), adverbial clauses are dependent but non-embedded structures, which, in contrast to other types of subordinate clauses, do not serve as syntactic constituents of a superordinate clause. Since (sentential) adjuncts do not fill an obligatory slot in the semantically associated clause, there is no cogent evidence that adverbial clauses are syntactically embedded. What is more, although the notion of adverbial subordination is commonly used for a particular clause type, it must be emphasized that adverbial clauses subsume a wide range of constructions with varying syntactic, semantic, and pragmatic properties that often overlap with the corresponding properties of coordinate sentences and other types of subordinate constructions (cf. Haiman and Thompson 1984; Lehmann 1988; see also the paper by Mauri in this volume). This chapter considers the structural variation of adverbial clauses in the light of their semantic and pragmatic features. It shows that the morphosyntactic features of adverbial clauses vary with their position and function, and that different semantic types of adverbial clauses can have radically different structural properties. In accordance with much previous work, the chapter argues that adverbial clauses constitute a heterogeneous class of subordinate clauses with fuzzy boundaries to coordinate sentences and other types of clause-linkage constructions.

2 Clause Order and Pragmatic Function One feature that distinguishes adverbial clauses from non-subordinate clauses is their linear arrangement (cf. Haspelmath 1995). While coordinate clauses and paratactic sentences are generally linked to the previous sentence, adverbial clauses can occur both before and after the associated (main) clause, as illustrated by the following examples from Persian. (1) Persian (Mahootian 1997: 40 ) (a) Cun/cunke dir bud mund-im xune. since late was stayed-1pl home “Since it was late we stayed home.” (b) Ba otobus ræft-im cunke mašin næ-dar-im. with bus went-1pl because car neg-have-1pl “We went by bus because we don’t have a car.” If we look at the cross-linguistic distribution of adverbial clauses more systematically, we find two dominant patterns (cf. Diessel 2001). In the majority of 342

9781441124609_Ch20_Final_txt_print.indd 342

11/20/2012 9:26:01 AM

Adverbial Subordination

the world’s languages, adverbial clauses are common in both positions, that is, before and after the (main) clause; but in some (rigid) OV-languages adverbial clauses are generally placed at the beginning of the sentence. In Japanese, for instance, adverbial clauses consistently precede the main clause (or main clause predicate) unless they are extraposed, which is a relatively rare phenomenon restricted to certain types of adverbial clauses in conversational discourse (cf. Ford and Mori 1994). The linear arrangement of main and adverbial clauses is closely related to their pragmatic function. It is well-known that preposed adverbial clauses serve particular organizing functions in discourse (cf. Chafe 1984, Thompson 1985, Ford 1993). In their basic use initial adverbial clauses function to present information that is pragmatically presupposed providing a thematic ground for new information asserted in subsequent clauses (cf. Lambrecht 1994). Consider for instance the following example from Time Magazine. (2) About 45 minutes later, Teresa Lewis called the police to report that her husband and stepson had been killed. But when the police arrived, Julian Lewis was still alive. Among his last words was an ominous accusation: “My wife knows who done this to me.” She did. [Time Magazine, Friday, Sept 10, 2010] When the reader of a journalistic article is told that somebody called the police, as in the first sentence of this example, he or she has good reasons to assume that the article will continue with information about what happened “when the police arrived.” The when-clause, thus, connects the complex sentence to the previous discourse; it creates a thematic ground for the ensuing (main) clause(s) based on information from the preceding sentence. The same discourse-organizing function of initial adverbial clauses has been observed in other languages (cf. Marchese 1987; Givón 2001b: chapter 4; Thompson et al. 2007). Consider for instance the following example from Supyire (Gur, Africa), in which two preposed adverbial clauses of time are thematically related to the previous discourse providing a temporal setting for new information in the sections that follow the adverbial clauses (see Carlson 1994: 588–90 for discussion). (3) Supyire (Carlson 1994: 589–90) Mu màha . . . cyìnŋikíí taanna you ab . . . sticks.def line.up yire these

fááyi rocks.def

maá cìì and.seq indf

ɲ ùŋì head

ì, at

maá and.se

a tòrò sc pass ŋ´ -kw´, ip-finish

taanna fááyi niŋ-kwuuyí ɲ ùŋì line.up rocks.def adj-surround.def head

ì at

343

9781441124609_Ch20_Final_txt_print.indd 343

11/20/2012 9:26:01 AM

The Bloomsbury Companion to Syntax

na p og Mu you

ma na  n-tare niŋké wògé ɲùŋì ì. come.impv prog ip-set.down.impv middle.def poss.def head at ahá círé yála à taanna, cond them do.well sc line.up

mu arì you ab.seq maá and.seq

cí them

maá lí and.seq it La it

ŋ` -kwɔ` ip-finish márà, cover.with.layer.of.adobe

yáha let

lá it

à perf

waha. dry

há wáha, cond dry

mu you

arì pwooré hab.seq adobe.def

ná you

fiinzígíré fonio.stem.def

maríi and.seq.prog

tà ind

cwɔˋnhɔˋ mix

e, with

kùŋikíí balls.def

ɲìnì roll

na prog

n-tare. ip-set.down.impv

“You . . . line up the sticks along on top of these rocks, and finish and then line up some of them on top of the circle of rocks setting (the other end of each stick) on the one in the middle. When you have lined them up well, you stop doing that (lit. you finish) and cover them with a layer of adobe, and let it dry. When it has dried, you mix some adobe with the fonio stems and then roll balls (of adobe) and set them down (i.e. to make a wall).” Note that the adverbial clauses in (2) and (3) exceed the confines of an individual sentence. They are used to “bridge” the boundaries between two paragraphs, that is, two thematic sections, enhancing discourse coherence (cf. Givón 2001b: chapter 19). However, at the level of the individual sentence, an initial adverbial clause can be seen as a “scene setting topic” (Lambrecht 1994: 125), which in some languages is explicitly indicated by the occurrence of a topic marker. For instance, the adverbial clause in (4a) from Isthumus Zapotec (Tehuantepec, Mexico) occurs with the same morphological marker for topichood, that is, la, as the topicalized noun phrase at the beginning of the sentence in (4b). (4) Isthumus Zapotec (Thompson et al. 2007: 294) (a) Kumu wara be la, naa nyuaa´. since sick he top went I “Since he was sick, I went.” 344

9781441124609_Ch20_Final_txt_print.indd 344

11/20/2012 9:26:01 AM

Adverbial Subordination

(b) Ngiiu-ke la, bigapa ba´du-ke. man-that top hit child-that “That man, he hit the child.” In addition to their function as scene-setting topics, preposed adverbial clauses can serve a particular speech-act function expressing politeness or coordinating the interaction between the speech participants. Consider for instance the adverbial clauses in (5) and (6). (5) Uhm well before we get into the detailed discussion of all of this, have you got something else Mary? [International Corpus of English] (6) And if I may say so Mr Speaker . . . they possibly derived some benefit from the presence of the Chancellor . . . . [International Corpus of English] In both examples the adverbial clause concerns a different level of information than the associated (main) clause. The before-clause in (5) functions to inform the hearer about the order of topics in the following discourse, and the if-clause in (6) is used to indicate that the speaker is not really in the position to state the subsequent assertion. Rather than indicating a semantic link between two events, the adverbial clauses serve a pragmatic function at the interactive level of the conversation (cf. Sweetser 1990: 76–112). In contrast to initial adverbial clauses, final adverbial clauses have usually a local semantic scope elaborating the content of the preceding (main) clause (cf. Thompson 1985; Thompson et al. 2007; Verstraete 2007). However, final adverbial clauses can convey very different types of information. First, final adverbial clauses are often used as minor additions (or afterthoughts) to the preceding main clause, as in (7), in which the when-clause spells out information that is pragmatically presupposed by the prior prepositional phrase on Friday. (7) I forgot to mention it to you on Friday, when I saw you. [International Corpus of English] Second, final adverbial clauses can present focal information following a thematic main clause that serves a similar grounding function as an adverbial clause at the beginning of a complex sentence (cf. Lambrecht 1994). For instance, the if-clause in (8) is grounded by the previous main clause, which includes two anaphoric elements, the demonstrative that and the verb happen, providing a thematic foundation for the information in the conditional clause. (8) That will happen only if the Government manages to replace the poll tax with a more acceptable alternative. [International Corpus of English] 345

9781441124609_Ch20_Final_txt_print.indd 345

11/20/2012 9:26:01 AM

The Bloomsbury Companion to Syntax

Third, adverbial clauses can function as independent sentences (or independent speech acts) resembling coordinate sentences that are syntactically and pragmatically separate from the preceding clause (cf. Green 1976; Lakoff 1984). For instance, the because-clause in (9) is intonationally unbound and includes a tag-question indicating that the adverbial clause has its own illocutionary force (see Section 4). (9) And that’s a picture frame . . . because that’s got glass in it, hasn’t it? [International Corpus of English] Finally, like initial adverbial clauses, final adverbial clauses can be used at the level of the communicative interaction between speaker and addressee. The ifclause in (10), for instance, does not express a condition for the content of the preceding (main) clause but functions instead as a pragmatic marker of politeness (cf. Sweetser 1990: 76–112). (10) I will take the big one, . . . if you don’t mind. [International Corpus of English] To sum up the discussion in this section, we have seen that adverbial clauses can serve a wide range of pragmatic functions that vary with their position relative to the associated (main) clause. In the following section we will see that initial and final adverbial clauses do not only differ with regard to their pragmatic functions but also as to their syntactic properties.

3 Syntactic Structure Complex sentences are commonly analyzed as biclausal constructions; but in language use the individual clauses are often planned and processed as separate units. The processing properties of complex sentences are crucially determined by the position of the subordinate clause (cf. Diessel 2005). An adverbial clause that precedes the main clause can be seen as a syntactic projector creating an anticipatory link to upcoming clauses that are immediately integrated in the unfolding sentence. However, if the adverbial clause follows the main clause it may only be added to the previous structure after the main clause has been completed. In contrast to complex sentences with initial adverbial clauses, complex sentences with final adverbial clauses can be planned and processed successively, that is, one clause at a time, suggesting that final adverbial clauses are potentially more independent of the (main) clause than preposed adverbial clauses (cf. ibid).

346

9781441124609_Ch20_Final_txt_print.indd 346

11/20/2012 9:26:01 AM

Adverbial Subordination

In accordance with this analysis, initial adverbial clauses are generally bound to the subsequent main clause by intonation whereas final adverbial clauses are often intonationally separated from the preceding (main) clause (cf. Chafe 1984; Ford 1993). Examining pre- and postposed adverbial clauses in conversational English, Ford (1993) found that about 40 percent of all final adverbial clauses are added to a structure that is intonationally marked as a complete sentence, whereas initial adverbial clauses are generally linked to the subsequent clause. Together with the processing analysis of complex sentences, the different intonational properties of pre- and postposed adverbial clauses seem to suggest that initial adverbial clauses are more tightly integrated in a complex sentence than final adverbial clauses (cf. Diessel 2004: chapter 3). However, as we have seen in Section 2, initial adverbial clauses are commonly used at a global discourse level, whereas final adverbial clauses are semantically linked to the preceding clause, suggesting that complex sentences are semantically more tightly organized if the adverbial clause follows the associated clause. Commensurate with this hypothesis Verstraete (2007) argued that final adverbial clauses are commonly interpreted in the light of certain epistemic and speech act features of the preceding (main) clause, whereas initial adverbial clauses are usually not affected by these features (see also Verstraete 2004). Compare for instance the two while-clauses in (11) and (12) (adopted from Verstrate 2007: 248). (11) Then you turn and run into the main lounge. He’s there, still curled up and still secured by the tape, but he must have wriggled his way through to here while you were down in the cellar. [British National Corpus] (12) While you were down in the cellar, he must have wriggled his way through to here. According to Verstraete, the final while-clause in (11) can be interpreted as the focus of the modal verb must in the main clause, whereas the initial while-clause in (12) is not amenable to such an interpretation. In other words, while the sentence in (11) can mean “that must have been the time when he wriggled his way through to here,” the sentence in (12) restricts the semantic scope of must to the main clause excluding an interpretation in which the while-clause is in the focus of the modal verb. Similarly, while a final adverbial clause can be interpreted as being part of a question, an initial adverbial clause is usually not included in the scope of an interrogative (main) clause. For instance, while the adverbial clause in (13) can be analyzed as the focus of an interrogative speech act meaning “Was that the time when you talked to her?” the adverbial clause in (14) is pragmatically presupposed and hence outside of the scope of the following question; that is, in contrast to the sentence in (13), sentence (14) cannot mean “Was that the 347

9781441124609_Ch20_Final_txt_print.indd 347

11/20/2012 9:26:01 AM

The Bloomsbury Companion to Syntax

time when you talked to her?” (but see Verstraete 2007: 249–50 for some rare exceptions). (13) Did you talk to her while she was still in her office? (14) While she was still in her office, did you talk to her? Thus, while initial adverbial clauses are structurally and intonationally incomplete without the following (main) clause, they tend to be semantically more independent of the associated clause than adverbial clauses at the end of a complex sentence, where the subordinate clause often functions as an integral semantic component of the preceding (main) clause. However, final adverbial clauses are not generally included in the scope of semantic operators in the prior (main) clause. If the adverbial clause is added to an intonationally complete sentence it often behaves like an independent assertion exhibiting a range of “main clause phenomena” (Green 1976), which are usually banned from a subordinate clause. In English, these main clause phenomena include, inter alia, negative inversion, locative inversion, and tag questions (cf. Green 1976; Lakoff 1984). Although all of these phenomena are supposed to be restricted to main clauses, it is well known that they also occur in certain (semantic) types of subordinate clauses, including certain (semantic) types of adverbial clauses, but only if the adverbial clause follows the associated (main) clause (cf. (15a–c)). Since preposed adverbial clauses are pragmatically presupposed, they are incompatible with syntactic phenomena of independent main clauses (cf. (16a–c)): (15) (a) They were stunned, because never before had there been anything quite like it. (b) I will stay, because here comes my friend John. (c) I think we should not go, because it’s raining, isn’t it? (16) (a) ?Because never before had there been anything quite like it, they were stunned. (b) ?Because here comes my friend John, I will stay. (c) *I think because it’s raining, isn’t it, we should not go. That the occurrence of main clause phenomena is restricted to final adverbial clauses has also been observed in other languages (cf. Günthner 1996; De Haan 2001). In colloquial German, for instance, adverbial clauses are often used with main clause word order, but only if the adverbial clause follows the semantically associated clause. As can be seen in (17a–b), if a causal (or concessive)

348

9781441124609_Ch20_Final_txt_print.indd 348

11/20/2012 9:26:01 AM

Adverbial Subordination

adverbial clause follows the (main) clause it can include the finite verb in second position (cf. 17a) parallel to a main clause; but if the adverbial clause precedes the main clause, the finite verb has to occur at the end of the adverbial clause (cf. 17b) suggesting that while final adverbial clauses can function as independent assertions initial adverbial clauses are always pragmatically presupposed. (17) German (a) Ich möchte gehen, weil es ist schon spät. I would.like to.go because it is already late “I would like to go because it is quite late.” (b) *Weil es ist schon spät, möchte because it is already late would.like “Because it is quite late I would like to go.”

ich I

gehen. to.go

In this section we have seen that complex sentences with initial and final adverbial clauses can have very different syntactic and semantic properties. While complex sentences with initial adverbial clauses are generally processed as a single unit, complex sentences with final adverbial clauses can either be interpreted as two independent sentences or as a biclausal construction in which main and adverbial clauses are intonationally combined and semantically dependent on each other. The following section will show that the morphosyntactic properties of adverbial clauses are not only determined by their position and pragmatic function but also by the semantic link they encode.

4 Semantic Link Adverbial clauses are commonly divided into several semantic subtypes expressing temporal, conditional, causal, purposive, and other semantic relationships (see Thompson et al. 2007 for some discussion). The semantic link between main and adverbial clauses correlates with their linear arrangement (cf. Diessel 2001). Disregarding languages in which adverbial clauses generally precede the main clause (see above), the positional tendencies of conditional, temporal, causal, and purposive clauses can be described as in (18): (18) Conditional > Temporal > Causal > Purposive[Result]

preposed

postposed

349

9781441124609_Ch20_Final_txt_print.indd 349

11/20/2012 9:26:01 AM

The Bloomsbury Companion to Syntax

Conditional clauses tend to precede the main clause, temporal clauses exhibit a mixed pattern of pre- and postposing, and causal and purposive clauses are usually placed at the end of a complex sentence. Assuming that the positional tendencies of adverbial clauses correlate with their pragmatic functions, we may hypothesize that the meaning of adverbial clauses is an important determinant of their position and pragmatic use. This hypothesis is supported by several recent studies. To begin with, the cross-linguistic tendency for conditional clauses to precede the main clause is eventually motivated by their meaning. A conditional clause describes a fictive situation providing a conceptual framework for the interpretation of subsequent clauses. Consider for instance the if-clause in (19), in which four successive clauses are semantically contiguous on the hypothetical scenario encoded in the conditional clause. (19) If all goes well, it gets very hot (up to 70 degrees) and the mixture begins to rot very quickly, and in the process weed seeds, pathagens and even perennial weed roots are killed. [International Corpus of English] Since conditional clauses are commonly used to establish a conceptual frame or “mental model” (cf. Johnson-Laird and Byran 2002) for the interpretation of related clauses, they tend to occur at the beginning of a complex sentence (functioning as a particular type of topic; cf. Haiman 1978). Apart from conditional clauses, temporal clauses are commonly used for discourse-organizing functions at the beginning of a complex sentence. However, unlike conditional clauses, temporal clauses do not immediately affect the semantic interpretation of the associated clause; rather, temporal clauses are typical background clauses that speakers use to process the transition between episodes in narrative discourse and related genres. As pointed out above (cf. (4) & (5)), preposed temporal clauses create a temporal setting for foreground information in subsequent clauses based on information from the preceding discourse, whereas final adverbial clauses complete or narrow the meaning of the preceding (main) clause (see Hopper 1979 and Ford 1993: 68–73 for some discussion). A second factor that influences the positioning of temporal adverbial clauses is iconicity of sequence. There is evidence that temporal clauses of posteriority usually follow the main clause whereas temporal clauses of anteriority typically precede it (cf. Diessel 2008). In contrast to conditional and temporal clauses, causal and purposive clauses are only rarely used for discourse-organizing functions, serving instead a more local function in the context of the preceding (main) clause. However, although both causal and purposive clauses typically occur at the end of a complex sentence they tend to serve very different pragmatic functions.

350

9781441124609_Ch20_Final_txt_print.indd 350

11/20/2012 9:26:01 AM

Adverbial Subordination

Final purposive clauses denote the goal or motivation of the activity described in the associated (main) clause, which is often at the heart of the information jointly conveyed by main and subordinate clauses (cf. Thompson 1985). Consider for instance the complex sentence in (20), in which the main clause takes up a theme from the previous discourse (i.e. the importance of concentration for playing baseball) that is elaborated by the purposive clause. Note that the main clause is semantically incomplete without the purposive clause: a sentence that begins with He must concentrate calls for an explanation that is provided by the goal or motivation expressed in the adverbial clause. (20) It helps players to develop their concentration. It is not enough for a player to know what he should be doing. He must concentrate in order to do it effectively and to do it whenever it is required. [International Corpus of English] The close semantic bond between main and purposive clauses is reflected in their morpho-syntactic structure. There are two conspicuous structural properties of purposive clauses that distinguish them from other semantic types of adverbial clauses. First, purposive clauses are morphologically more reduced than temporal, conditional, and causal clauses. As shown in a recent study by Schmidtke-Bode (2009: 151–7), across languages purposive events are commonly expressed by morphologically deprived clauses that contain an uninflected verb form and often lack an overt subject. English provides a case in point. While conditional, temporal, and causal clauses are commonly realized by fully developed clauses, purposive clauses are primarily expressed by infinitives that lack the inflectional properties of other verb forms and are usually controlled by the main clause subject (see also Cristofaro 2003: 155–94). Second, purposive clauses tend to be syntactically more closely integrated in a complex sentence than other types of adverbial clauses. Although purposive clauses are commonly analyzed as adjuncts, they often resemble complement clauses (cf. Parodi and Quicoli in this volume). Of the 80 languages in Schmidtke-Bode’s sample, 62 languages had at least one purposive clause that shared important properties with complement clauses (cf. Schmidtke-Bode 2009: 158). Consider, for instance, the purposive clause in (21a) from Imonda (Papuan, New Guinea), in which the dominant type of purposive clause includes a nominalized verb form marked by the same case suffix as the complement clause in (21b), whereas conditional, temporal, and causal relationships are expressed by fully developed clauses that do not take a case suffix but often include a topic marker, which is not permissible in purposive clauses (cf. Seiler 1985: 203).

351

9781441124609_Ch20_Final_txt_print.indd 351

11/20/2012 9:26:01 AM

The Bloomsbury Companion to Syntax

(21) Imonda (Seiler 1985: 84, 190) (a) Põl nëhe-l-m uagl-f. fence construct-nml -goal go-prs “I am going to build a fence.” (b) Ièf sabla nibia-ual-l-m õ-f. house two build-du-nmlz-goal say-prs “I want to build two houses.” Like purposive clauses, causal clauses tend to provide new and focal information; but in contrast to purposive clauses, causal clauses are often only loosely integrated in a complex sentence. In Ford’s study of adverbial clauses in conversational English, 53 percent of all causal because-clauses are added to a (main) clause with closing intonation suggesting that causal clauses are commonly used as independent assertions (cf. Ford 1993: 102–30). This hypothesis is supported by the fact that the above-mentioned main clause phenomena in adverbial clauses are largely restricted to causal (and concessive) clauses and that causal relationships are commonly expressed by constructions that resemble independent sentences. For instance, in languages in which adverbial clauses generally precede the main clause (see above) causal relationships are often expressed by coordinate sentences that follow the semantically associated main clause (cf. Diessel and Hetterle 2011). The particular properties of causal clauses are motivated by their pragmatic function. Especially in spoken discourse, causal clauses are commonly used as independent speech acts functioning to back up a previous statement that the hearer is unlikely to accept without additional information. Consider for instance the following example. (22) A: We could spend a lot of our life trying to contradict that. B: Why? A: Well, because . . . it may be a very bad chemical bath. [International Corpus of English] Although causal clauses are semantically dependent on the associated (main) clause, in conversational discourse they are often prompted by a hearer signal, as in (22), in which the because-clause is produced in response to a causal whyquestion. As Diessel and Hetterle (2011) have argued, across languages causal (adverbial) clauses are associated with a particular discourse pattern consisting of (i) an assertion that the hearer does not accept or understand, (ii) the hearer’s reaction (optional), and (iii) the speaker’s justification or explanation of the controversial statement. Interestingly, in some languages causal clauses are marked by frozen questions indicating that causal clauses are commonly used

352

9781441124609_Ch20_Final_txt_print.indd 352

11/20/2012 9:26:02 AM

Adverbial Subordination

in a particular discourse pattern that involves a causal question (e.g. English that’s why, Italian perché, Supyire ŋàhá ná yε “through what”; see Diessel and Hetterle 2011).

5 Conclusion In conclusion, this chapter has shown that adverbial clause can serve various pragmatic functions that are crucially determined by their position relative to the associated (main) clause. The particular pragmatic functions of pre- and postposed adverbial clauses are reflected in their morphosyntactic properties and are related to aspects of their meaning and communicative use. Although adverbial clauses are commonly defined as a consistent grammatical category, it must be emphasized that the notion of adverbial subordination refers to a very diverse set of constructions that exhibit an enormous amount of variation within and across languages.

353

9781441124609_Ch20_Final_txt_print.indd 353

11/20/2012 9:26:02 AM

21

Negation Liliane Haegeman1 and Terje Lohndal

Chapter Overview Introduction Negative Concord in West Flemish Negative Concord is not Multiple Agree Conclusion

354 358 358 364

1 Introduction 1.1 Variation in the Expression of Sentential Negation The syntactic expression of sentential negation has given rise to a lot of research in most, if not all, linguistic frameworks. A wide range of issues arise when one considers the encoding of sentential negation ranging from morphological to syntactic, semantic, and pragmatic issues. Within generative grammar, research on negation centers on a range of topics. We illustrate some of these in this introduction and then we will zoom in on one particular area of interest in the main discussion of the paper. Consider the data in (1)–(3). (1) French Jean ne lit pas Le Monde. Jean ne reads not Le Monde “Jean doesn’t read Le Monde.” (2) Italian Gianni non legge La Repubblica. Gianni non reads La Repubblica “Gianni does not read La Repubblica.” 354

9781441124609_Ch21_Final_txt_print.indd 354

11/20/2012 9:23:38 AM

Negation

(3) English John doesn’t read The Guardian. French sentential negation is expressed by means of a combination of a particle ne on the finite verb and a free morpheme pas; in Italian the particle non prefixed to the finite verb expresses sentential negation. Unlike Italian non, French ne cannot express negation all by itself. In English the free morpheme not expresses sentential negation. Though these examples suffice to show that negation is not uniform cross-linguistically, these and additional data also suggest that sentential negation may be expressed in a designated position in the clause. Thus, one issue that has given rise to a lot of discussion is the question of how precisely sentential negation is encoded cross-linguistically. Kayne (1989) and Pollock (1989) sparked a lot of cross-linguistic work when they suggested that the functional structure of the clause includes a designated projection “Negation Phrase” (NegP), which encodes sentential negation. The bipartite expression of sentential negation in French (1) would then illustrate a case in which the head of NegP is realized as ne and its specifier as pas. In Italian, on the other hand, only the head is spelt out as non, and in English not spells out the specifier. Comparative work on the nature of the realization of NegP has led to questions such as (i) is there is a NegP in all languages? (ii) if so, is there a unique projection that is commi ed exclusively to the expression of sentential negation? (iii) if there is such a unique projection, does it have a unique position in the clause?, (iv) is it possible to have several NegPs? (see Zanu ini 1997 and Lindstad 2007 for recent discussion), and (v) is the expression of sentential negation stable over time? (e.g. van Gelderen 2008). These questions have gained prominence as cross-linguistic work on negation has progressed, in particular on Romance languages (Zanu ini 1997, 2001), but also on a number of other and unrelated languages (e.g. Moscati 2006; Lindstad 2007; for a survey of the literature). In addition to having one designated element, the marker of sentential negation, languages o en allow for the expression of negation through other negative expressions, which, apart from encoding negation, also contribute independently to the interpretation of the clause. Again looking at the three languages above, negative adverbs such as jamais “never,” mai “never,” and never may also contribute to the expression of sentential negation. As seen in the examples, there is cross-linguistic variation: while in French (4) and Italian (5) the preverbal marker of sentential negation is also present, in English never is the sole expression of negation. Adding not to (6) cancels the sentential negation expressed by never: (4) French Jean ne lit   jamais Le Monde. Jean ne reads never Le Monde “Jean never reads Le Monde.” 355

9781441124609_Ch21_Final_txt_print.indd 355

11/20/2012 9:23:39 AM

The Bloomsbury Companion to Syntax

(5) Italian Gianni non legge mai La Repubblica. Gianni non reads never La Repubblica “Gianni never reads La Repubblica.” (6) English John has never (#not) read The Guardian. Similarly, negative arguments such as nothing and its equivalents may be implicated in the expression of sentential negation; in French (7) and in Italian (8) they co-occur with the negative particle on the finite verb, in English nothing expresses negation by itself: (7) French Jean ne lit rien. Jean ne reads nothing “Jean doesn’t read anything.” (8) Italian Gianni non legge niente. Gianni non reads nothing “Gianni does not read anything.” (9) English John has (#not) read nothing. These data already show that in some languages, various expressions of negation may co-occur, and this is also shown in (10) and (11), where the equivalents of nothing and never co-occur. Once again if nothing and never co-occur in English this gives rise to a double negation: the negation on nothing cancels the negation on never. The pa ern in which two negative expressions, such as jamais “never” and rien “nothing,” which each individually may contribute to the expression of sentential negation, co-occur and express a single sentential negation is referred to as “negative concord.” Based on the observed contrasts, it is proposed that a distinction be made between negative concord languages such as French and Italian and non-negative concord languages such as English (see, for example, Zeijlstra 2004 for a specific proposal as to how this can be captured theoretically). (10) French Jean ne lit jamais rien. Jean ne reads never nothing “Jean never reads anything.”

356

9781441124609_Ch21_Final_txt_print.indd 356

11/20/2012 9:23:39 AM

Negation

(11) Italian Gianni non legge mai niente. Gianni non reads never nothing “Gianni never reads anything.” (12) English #John never reads nothing. The data in (1)–(12), and other similar data from a wide range of languages, have led to a lot of research on the syntax and semantics of negative expressions (or “n-words,” following Laka 1990, or “n-constituents”) such as never, nothing, and also nominal expressions with a negative quantifier (no N) (cf. Giannakidou 2006).

1.2 Negative Concord and Agree Recently, there has been a revived interest in negative concord and the question it raises for how syntax and semantics connect (Zeijlstra 2004; Giannakidou 2006; Haegeman and Lohndal 2010). One point that has become the focus of a ention is that in negative concord a negative feature seems to be instantiated on a number of constituents in the clause. For instance, in Italian (11) non, mai, and niente are arguably all negative in an intuitive sense and could be said to “agree” in their negative value. This intuition is particularly relevant for research in the Minimalist Program, in which the formal relation Agree plays a prominent part. Agree involves a feature agreement dependency between a Probe and a Goal, as, for instance, between the verb inflection and the subject. In English (3) the third person ending –s can be said to match the features of the subject John. In (3) Agree is a binary relation: there is one Probe and one Goal. In addition to binary Agree, it has been proposed that the system must allow for “Multiple Agree,” by which a single Probe can establish an Agree relation with Multiple Goals (Hiraiwa 2001). It would seem that negative concord lends itself rather naturally to an analysis in terms of Multiple Agree. In (5) non could be said to Agree with mai, in (8) non could be said to Agree with niente, and (11) would be an instance of Multiple Agree where non Agrees with both mai and with niente (cf. Brown 1999 for Russian). If it can be substantiated that negative concord is derived by Multiple Agree, then this in itself would offer support for postulating Multiple Agree as part of the theory. In this chapter, we will discuss a number of empirical problems for deriving negative concord by way of Multiple Agree. Obviously our chapter will give the reader only a snapshot of the rich ongoing work on negation and of its place in the Minimalist Program, and that he or she can start exploring the rich literature by consulting the references in the present paper. 357

9781441124609_Ch21_Final_txt_print.indd 357

11/20/2012 9:23:39 AM

The Bloomsbury Companion to Syntax

The chapter is organized as follows. Section 2 introduces the phenomenon of negative concord in West Flemish, the language that we will be focusing on. In Section 3, we look at a Multiple Agree analysis of negative concord in West Flemish (Brown 1999; Zeijlstra 2004) and point out problems with this analysis. Section 4 is a conclusion.

2 Negative Concord in West Flemish The expression of sentential negation in West Flemish has been the subject of much research (see Haegeman and Zanu ini 1991, 1996; Haegeman 1995, 2002). (13) illustrates negative concord in this language. (13) West Flemish K’(en)-een doa nooit niets niet gezien. I en have there never nothing not seen “I have never seen anything there.” The morpheme –en cliticizes onto the finite verb. It is optional, but, like French ne, it cannot express negation all by itself, it must co-occur with a negative constituent. We set –en aside in this paper (see Haegeman 2002). Each of the n-constituents nooit “never,” niets “nothing,” and niet “not” in (13) can express a single (sentential) negation on its own, but the n-constituents jointly express one sentential negation. To capture negative concord, Zeijlstra (2004) advocates Multiple Agree (see Penka 2007 for a similar analysis). We will show that for West Flemish empirical problems that arise for such an analysis because the language displays constraints on which negative elements can enter into negative concord cannot be captured by the Multiple Agree analysis. For reasons of space, we will not be able to present a new analysis. Instead we refer the reader to Haegeman and Lohndal (2010) for an analysis that covers the West Flemish facts discussed here.

3 Negative Concord is not Multiple Agree Working within the Minimalist Program, Zeijlstra (2004) proposes that n-constituents that enter into negative concord are not themselves the expression of sentential negation. Rather, such expressions are semantically non-negative indefinites. This intuition is captured in terms of features. Features typically come in two flavors: interpretable/valued or uninterpretable/unvalued. As the labels suggest, features correspond in some way to semantic interpretation. Zeijlstra (2004: 245) suggests that the non-negative indefinites are associated 358

9781441124609_Ch21_Final_txt_print.indd 358

11/20/2012 9:23:39 AM

Negation

with a [u eg] feature, that is, an uninterpretable neg feature. The overt marker of sentential negation, for example, West Flemish niet, also has a [uneg] feature. Zeijlstra’s thesis is that NegP is not instantiated in all languages: it is the presence of [uneg] features in a language that triggers the projection of NegP. Sentential negation as such is then encoded by a covert negative operator OP¬ in SpecNegP, which is associated with the feature [ineg]. In Zeijlstra’s terms “OP¬ (i) introduces a negation at LF, and (ii) unselectively binds all free variables under existential closure” (247). Negative concord languages, which display [uneg] features and hence have NegP, are said to display “syntactic negation.” See Zeijlstra’s own work for discussion and motivation of the distinction between such languages and non-negative concord languages with “semantic negation.” For Zeijlstra, negative concord is derived by the operation Agree. Op¬ [ineg] in SpecNegP c-commands the (multiple) [uneg] n-constituents on the vP edge; hence an Agree relation can be established between [ineg] and [uneg]. In negative concord, multiple negative constituents jointly express one single negation. Zeijlstra (244–5) proposes that this is the result of Multiple Agree (Hiraiwa 2001): the (multiple) uninterpretable features on the negative marker and on the n-constituents are all checked by the unique interpretable feature on OP¬.

3.1 Application of the Multiple Agree Analysis On the basis of Czech (6a), we illustrate Zeijlstra’s derivation of negative concord readings. The specifier of NegP hosts a covert operator with an interpretable feature, [ineg]. In (14a), by virtue of the negative morpheme ne the finite verb vidi “see” has the [uneg] feature, and so does the n-word nikoho “no one.” Through Multiple Agree, the features [uneg] get checked and are deleted (14b). (14) Czech (Zeijlstra 2004: 250) (a) Milan nevidi nikoho. Milan  neg-sees no.one (b) [NegP OP¬ [ineg] [vP nikoho [uneg] [vP Milan nevidi [uneg]]]] In (15) and (16) Zeijlstra’s analysis is applied to West Flemish. (15a) is an example containing a negative marker niet “not” and the preverbal negative morpheme en. In terms of Zeijlstra’s analysis, niet and en both carry [uneg]; the two [uneg] features get checked by the [ineg] feature on the negative operator in SpecNegP. (15b) is a representation. In (16a), sentential negation is conveyed by means of the n-word, niemand “no one,” which may be accompanied by niet as well as by en. Negative concord is derived as in (16b): 359

9781441124609_Ch21_Final_txt_print.indd 359

11/20/2012 9:23:39 AM

The Bloomsbury Companion to Syntax

(15) West Flemish (Zeijlstra 2004: 255) (a) da Valère  niet (en) klaapt. that Valère not (en) talks “that Valère doesn’t talk.” (b) [NegP OP ¬ [ineg] [vP niet [uneg] Valère [v’ en-klaapt [uneg]]]] (16) West Flemish (ibid.) (a) da Valère tegen niemand (niet) en klaapt. that Valère against no.one  (not) en talks “that Valère doesn’t talk to anyone.” (b) [NegP OP ¬ [ineg] [[PP tegen niemand [uneg]] [vP (niet [uneg]) [vP Valère [v’ en-klaapt [uneg]]]]]]

3.2 Problems for a Multiple Agree Analysis Schematically, Zeijlstra’s implementation of multiple Agree for negative concord can be presented as in (17). In (17a) the unique [ineg] feature of the abstract negative operator establishes an Agree relation with the [uneg] feature of the two negative constituents in the clause, that is, niemand “no one” and niet “not.” The application of Multiple Agree is represented in (17b). Observe crucially that the relation between the [ineg] feature and the [uneg] on niet is not local because it crosses the [uneg] feature on niemand. In (18) we apply Zeijlstra’s approach to an example with three n-constituents nooit “never,” niemand “no one,” and niet vele “not much” entering into a negative concord relation. (17) West Flemish (a) dat er niemand niet gewerkt eet. that there no.one  not worked has (b) dat er [NegP OP ¬ [ineg] niemand [uneg] niet [uneg] gewerkt eet]

(18) West Flemish (a) dat er nooit niemand niet vele gewerkt that there no.one not much worked eet. has (b) dat er [NegP OP ¬ [ineg] nooit [uneg] niemand [uneg] niet vele [uneg] gewerkt eet] 360

9781441124609_Ch21_Final_txt_print.indd 360

11/20/2012 9:23:39 AM

Negation

On the Multiple Agree account, negative concord is a one-to-many relation in which the unique negative operator agrees with each n-word and the n-words enter into negative concord by virtue of their relation to the negative operator. Crucially, there is no relation between the individual n-words. It turns out that by this very absence of the locality restriction on Agree, an empirical problem arises for the Multiple Agree account of West Flemish negative concord. Indeed, Haegeman and Zanu ini (1996) have already shown that in West Flemish local relations between the negative elements play a role in determining the availability of negative concord. The across-the-board application of Multiple Agree to derive negative concord leads to the wrong predictions for negative concord relations among n-constituents. See Haegeman and Lohndal (2010) for further discussion. Consider (19). In (19a) niemand “no one” enters into a negative concord relation with niet “not,” in (19b) niemand enters into a negative concord relation with niet lange “not long,” and in (19c) niet lange, niemand, and niet enter into negative concord. (19) West Flemish (a) dat er doa niemand niet gewerkt eet. that there there no.one not worked has “that no one has worked there.” (b) dat er doa niet lange niemand gewerkt eet. that there there not long no.one worked has “that no one has worked there for a long time.” (c) dat er doa niet lange niemand niet gewerkt eet. that there there not long no.one not worked has “that no one has worked there for a long time.” In terms of Zeijlstra’s approach, these and similar data imply that niet lange “not long,” niemand “no one,” and the marker of sentential negation niet “not” all carry a [uneg] feature, which is checked by the [ineg] feature on the sentential negative operator. Note crucially that niet lange and niet are in a negative concord relation (19c). One would thus expect that (19d), containing niet lange and niet, will also be acceptable with a negative concord reading, but this prediction is not correct. (19d) can only (marginally) have a double negation reading. When niet is replaced by niet meer “no more” the resulting (19e) has a negative concord reading. (19) West Flemish (d) *dat Valère doa niet lange niet gewerkt eet. that Valère there not long not worked has

361

9781441124609_Ch21_Final_txt_print.indd 361

11/20/2012 9:23:39 AM

The Bloomsbury Companion to Syntax

(e) dat Valère doa niet lange niet meer gewerkt eet. that Valère there not long no more worked has “that Valère hasn’t worked there long any more.” The unacceptability of the negative concord reading in (19d) cannot be due to a ban on the co-occurrence of niet lange with niet since (19c) also contains these two items and is acceptable with the desired negative concord reading. It is also not simply due to there being an anti-adjacency requirement on niet lange and nie: in (19f) niet lange and niet are separated by the PP in dat us “in that house,” but this in itself is not sufficient to rescue the sentence. What is needed is that niet lange be separated from niet by niemand ((19g), cf. also (19c)): (19) West Flemish (f) *dat ter niemand niet lange in dat us niet gewerkt eet. that there no.one not long in that house not worked has (g) dat der niet lange niemand in dat us niet gewerkt eet. that there not long no.one in that house not worked has “that no one has worked long in that house.” Data such as those in (19) reveal that though “complex” n-constituents such as niet lange “not long” can participate in negative concord readings with niet; they can only do so provided they are separated from niet by a “simple” n-constituent such as niemand. No such constraint applies to niemand “no one” or to the other “simple” n-words such as nooit “never,” niets “nothing,” or nieverst “nowhere,” as shown in (20). (20) West Flemish (a) da Valère doa nooit  niet gewerkt eet. that Valère there never not worked has “that Valère has never worked there.” (b) da Valère doa niets niet over gezeid eet. that Valère there nothing not about said has “that Valère has not said anything about that.” (c) da Valère nieverst niet over geklaapt eet. that Valère nowhere not about talked has “that Valère has never talked about anything.” For completeness’ sake, note that, as shown in (19g), there is no adjacency requirement between the “simple” n-constituent and niet, which means that it is not the case that, contrary to what one might think, such simple n-constituents and niet form a single constituent. 362

9781441124609_Ch21_Final_txt_print.indd 362

11/20/2012 9:23:39 AM

Negation

We conclude, with Haegeman and Zanu ini (1996), that both the type of n-constituents and their positions relative to each other play a role in determining negative concord in West Flemish. Because in this language all n-constituents (niemand, niet lange, niet, niet meer, geen-NP etc.) that can express sentential negation on their own appear to be able to enter into negative concord in some combinations, Zeijlstra’s (2004) Multiple Agree analysis would lead us to expect that they can always all enter into an Agree relation with the relevant negative operator. The question arises how the application of Multiple Agree as formulated as a one time across-the-board procedure can set apart combinations that allow negative concord from those that do not. There is no way that this can be done in terms of the implementation of Multiple Agree proposed by Zeijlstra. In order to describe the co-occurrence restrictions on negative concord in some detail Haegeman and Zanu ini (1996: 143) classify West Flemish n-constituents in terms of their internal syntax and their featural make-up. Table 21.1 summarizes their classification of the n-constituents and their associated features. [q] is a quantificational feature; “bare” quantifiers such as niemand and niets correspond to our “simple” n-words. Simple n-constituents such as niemand “no one,” nieverst “nowhere,” nooit “never,” and niets “nothing” seem to be ambivalent in that they enter into negative concord with niet, as well as with composite constituents (niet + X or geen-NPs). On the other hand, composite constituents (niet + X or geen-NPs) cannot enter into negative concord with niet. Moreover, it looks as if negative concord readings are built up on the basis of stepwise local pairings: in (19a), for instance, niemand and niet can enter into negative concord, in (19d) niet lange cannot enter into a “direct” negative concord relation with niet, but as seen in (19c), niet lange can enter into an “indirect” negative concord relation with niet provided the la er is separated from the composite constituent niet lange by the simple n-constituent niemand. Recall also that the composite n-constituent niet

Table 21.1 Head features on negative elements and co-occurrence restrictions (Haegeman & Zanuttini 1996: 145) Bare Q [NEG, Q]

Geen-NP [Q]

Nie [NEG]

Bare Q [NEG, Q]

Yes niemand niets

yes niemand geen geld

yes niemand nie

Geen-NP [Q]

Yes geen mensen niemand

yes geen mensen geen tyd

no *geen mensen nie

Niet meer [Q]

Yes niemand niet meer

yes geen mensen niet meer

no *niet meer nie

363

9781441124609_Ch21_Final_txt_print.indd 363

11/20/2012 9:23:39 AM

The Bloomsbury Companion to Syntax

lange can enter into a negative concord relation with the composite constituent n-constituent niet meer (19e). We add that these West Flemish data are not only relevant for the evaluation of the Multiple Agree approach to negative concord, the data constitute a challenge to any theory that involves an across-the-board operation to associate the negative expressions in one way or another. For instance, an approach that derives negative concord readings by unselective binding of the n-constituents by one negative operator (cf. for example, Ladusaw 1992; Acquaviva 1993; Piñar 1996; Giannakidou 1997) also is not able to derive the data observed here without additional machinery. For reasons of space, we cannot here elaborate on an account that derives the West Flemish pa erns. In Haegeman and Lohndal (2010) we adopt the feature system elaborated in Zeijlstra (2004), and we pursue an approach based on binary Agree only. Our analysis is a theoretical implementation of the idea that what is going on in these cases is a pairwise matching of n-constituents and crucially relies on intervention to block the ungrammatical pa erns.

4 Conclusion This paper has provided a window onto one of the most debated issues in the recent generative literature on negation, namely how to analyze negative concord. It has offered a critique of accounts that argue for a Multiple Agree analysis and it has shown that this analysis does not work for negative concord in West Flemish. Instead a binary Agree analysis seems more feasible in order to account for the data.

Note 1. Liliane Haegeman’s research is funded by FWO as part of project 2009-OdysseusHaegeman-G091409.

364

9781441124609_Ch21_Final_txt_print.indd 364

11/20/2012 9:23:40 AM

22

Syntactic Change Thórhallur Eythórsson

Chapter Overview Introduction The Directionality of Morphosyntactic Change: Case Marking in Insular Scandinavian Reanalysis of Ambiguous Structures: The New Passive in Icelandic The Inertia Hypothesis Syntactic Reconstruction Conclusion

365 367 373 378 380 388

1 Introduction The study of syntactic change has a long history within linguistics; in fact, this history goes further back and boasts more impressive achievements than is sometimes claimed (for a critical overview, see Harris and Campbell 1995: 14–47). Work on the comparative syntax of Indo-European by Delbrück (1893– 1900) and others is full of keen observations on syntactic change, and the same goes for scholars working on individual languages, for instance Behaghel (1909, 1923–32) on the history of German or Jespersen (1909–49) on the history of English. Moreover, Wackernagel’s Law on the placement of clitics in second position (Wackernagel 1892) is still intensively discussed, both in IndoEuropean syntax and in general, most famously perhaps in work on Warlpiri (e.g. Hale 1983). Finally, in the wake of advances in linguistic typology and theoretical syntax in the 1960s, a renewed interest in historical syntax emerged, manifesting itself most conspicuously in the fierce debate between Lehmann (1974), Watkins (1976), and others on real and imagined problems in syntactic reconstruction. 365

9781441124609_Ch22_Final_txt_print.indd 365

11/20/2012 9:24:50 AM

The Bloomsbury Companion to Syntax

Despite this venerable tradition, it is no exaggeration to state that the study of syntactic change was put into a completely new perspective by Lightfoot (1979) and numerous subsequent publications (e.g. Lightfoot 1999, 2006). Viewing language change and first-language acquisition as intimately related, Lightfoot argues that language change is “contingent and fluky” and takes place in a sequence of bursts. Accordingly, language change is claimed by Lightfoot to involve the cumulative effect of changes in individual grammars embodying the language faculty in the sense of Chomsky’s I-language. While the general line of research initiated by Lightfoot has proven extremely fruitful, there is still no universal agreement on core issues, including the exact nature of syntactic change, its link to first-language acquisition, or the viability of syntactic reconstruction (of which Lightfoot is skeptical). For a comprehensive and insightful discussion of these and related matters, from the generativist perspective, see in particular Roberts (2007), himself an original thinker on syntactic change. Further important contributions, focusing on the connection between linguistic variation and change have been made by Kroch (1989, 2001) and his associates. Very briefly, Kroch sees word order change as the diachronic consequence of competing grammars in an individual (for example the competition of OV and VO orders in the history of English, resulting in the loss of the former). In this chapter I concentrate on four main issues which have figured prominently in work on syntactic change and to which I have contributed. The idea behind this procedure is that it is more profitable to the reader to be subjected to selected case studies of important problems than an enumeration of a great variety of topics. I hasten to add that the sections below are not intended to offer an exhaustive treatment of the relevant issues. It is hoped, however, that the discussion is sufficiently detailed to be understandable to the uninitiated reader. The case study presented in Section 2 of this chapter focuses on a phenomenon which has frequently been attributed to the loss of morphological case endings. This is “personalization,” or Nominative Substitution, which consists in the replacement of oblique case marking of subjects (or subject-like NPs) by nominative case. Given that it occurs in the Insular Scandinavian languages (Icelandic and Faroese), which have largely preserved their case endings intact, in contrast to Mainland Scandinavian, this change is argued to be an instance of a generalization of nominative as the unmarked case par excellence. This type of change involves morphosyntactic leveling, and it is not confined to Nominative Substitution. Another instance of such leveling is traditionally labeled “Dative Sickness,” consisting in the generalization of dative as the unmarked lexical case with subjects. It is also shown to operate in the languages in question, competing with Nominative Substitution. One of the most important types of syntactic change is reanalysis, by which new constructions emerge from existing structures. Reanalysis is “a mechanism 366

9781441124609_Ch22_Final_txt_print.indd 366

11/20/2012 9:24:51 AM

Syntactic Change

which changes the underlying structure of a syntactic pattern and which does not involve any modification of its surface manifestation” (Harris and Campbell 1995: 50, 61). A well-researched case involves the New Passive in Icelandic, which has emerged from a reanalysis of the canonical existential passive with a postverbal NP. It is argued in Section 3 that the locus of the reanalysis involves structures where the canonical passive and the New Passive cannot be distinguished morphologically. It is universally agreed that syntactic change can be caused by phonological or lexical change, or by extra-linguistic factors like contact (either in the form of direct borrowing or more indirect influence). Morphological change is also traditionally assumed to be a trigger for syntactic change, although, as mentioned above, its importance has tended to be exaggerated considerably. Another question is if syntax can change endogenously. In fact, Keenan (1994, 2002, 2009) has proposed the Inertia Hypothesis, according to which syntax, by itself, is diachronically completely inert, and only changes under the influence of “external” factors such as phonology and semantics. This hypothesis has found widespread recognition in the literature on theoretical diachronic linguistics. It is, however, not entirely obvious how to falsify this hypothesis as it is extremely difficult to exclude some external influence as being the ultimate trigger of any particular change. In Section 4 below I discuss a case which is a potential challenge to the Inertia Hypothesis—the loss of OV order in the VP in the history of Icelandic. The final section offers a discussion of syntactic reconstruction, which has been a neglected area in historical and comparative linguistics. It has even been asserted that reconstruction of syntax is not really possible given the special nature of syntax vis-à-vis other domains of grammar. In Section 5 I present arguments against this view, trying to show that syntactic reconstruction is not only a viable enterprise, but in fact a necessary part of historical syntax so that syntactic change can be modeled in a precise way. In particular, it is proposed that the tried and tested tools of the Comparative Method, in combination with the insights provided by the novel framework of Construction Grammar, can be applied successfully in syntactic reconstruction.

2 The Directionality of Morphosyntactic Change: Case Marking in Insular Scandinavian 2.1 Introduction The change from “impersonal” to personal constructions has been observed in a variety of languages, including Germanic and Romance. It is traditionally assumed that this change is connected to loss of morphological case endings, 367

9781441124609_Ch22_Final_txt_print.indd 367

11/20/2012 9:24:51 AM

The Bloomsbury Companion to Syntax

in particular in studies based on English (see Jespersen 1909–49, Lightfoot 1979 among many others). This account predicts that the change from impersonal to personal constructions should not happen in languages where morphological case is preserved. In fact, however, this type of change is attested to in such languages, including Insular Scandinavian (Icelandic and Faroese), German, and older Germanic languages. Therefore, an alternative hypothesis is called for, according to which Nominative Substitution (“personalization”) is regarded as a type of analogical change, more exactly as morphosyntactic leveling, in favor of nominative as the unmarked subject case. A “laboratory” case for testing this hypothesis, involving contemporary languages, is found in Icelandic and Faroese, which have been extensively investigated in this respect (e.g. Eythórsson 2002). Other changes in case marking are also attested in these languages, despite the preservation of case morphology, including the substitution of dative for accusative with experiencer subjects (“Dative Sickness”). The aim of this section is to show how these changes can co-exist and what their relative domains reveal about the directionality of morphosyntactic change.

2.2 Case Marking with Arguments in Insular Scandinavian The subject status of non-agentive subject-like oblique NPs has been recognized as a feature of Insular Scandinavian since the groundbreaking studies by Andrews (1976) on Icelandic and Barnes (1986) on Faroese. Thus, an oblique NP mig “me” with vanta “lack, need” in Icelandic is argued to be subject. (1) Icelandic (Eythórsson 2009) Mig vantar peninga. me-acc needs money-acc.pl “I need money.” The reason why such oblique arguments are analyzed as subjects is that they pass various tests for subjecthood in the relevant languages. Among the subject tests considered most reliable in general are PRO infinitives of the kind shown for Icelandic in (2). (2) Icelandic (Eythórsson 2009) Ég þoli ekki að ___vanta peninga. I tolerate not to need money-acc.pl “I can’t stand needing money.” This test shows that the oblique NP with verbs like vanta is the subject as it must be omitted in control infinitives. Nominative case and agreement, on the other 368

9781441124609_Ch22_Final_txt_print.indd 368

11/20/2012 9:24:51 AM

Syntactic Change

hand, are not subject tests, at least in Insular Scandinavian. If the oblique NP were an object, it might for instance be expected to be overt in such infinitive clauses, contrary to fact. Moreover, while subjects can be in oblique case in Insular Scandinavian, objects can be in the nominative. Nominative objects are mainly found with verbs taking dative subjects (DAT-NOM verbs), for example, leiðast “be bored” in Icelandic. The verb can either agree with the nominative NP or be in the (default) third person singular. (3) Icelandic (Eythórsson 2009) Mér leiðast/leiðist apar. me- at bored-by-3pl/3sg apes-nom.pl “I’m bored by apes.” The number of verbs taking oblique subjects is different in the two Insular Scandinavian languages. Icelandic is the more conservative one, with the estimated number of verbs taking oblique subjects being almost 400. In Faroese, such verbs are a lot fewer, only about 50. Some verbs taking oblique subjects listed in handbooks are uncommon in the modern languages, and are largely restricted to fixed (literary or archaic) expressions. The study of older Faroese texts (ballads) shows that oblique subjects were more numerous in Old Faroese than in the modern language.

2.3 Changes in Subject Case Marking In the history of the Insular Scandinavian languages the following two types of change in the case marking of oblique subjects have been observed: Nominative Substitution (“personalization”) and “Dative Sickness.” With Nominative Substitution accusative and dative case with subjects is replaced by nominative, as in the Icelandic examples in (4–5). (4) Icelandic (Eythórsson 2002) (a) Bátinn rak að landi. boat-def.acc drifted to shore (b) Báturinn rak að landi. boat-def.nom drifted to shore “The boat drifted to the shore.” (5) Icelandic (Eythórsson 2002) (a) Leiknum lauk með jafntefli. match-def.dat finished with draw 369

9781441124609_Ch22_Final_txt_print.indd 369

11/20/2012 9:24:51 AM

The Bloomsbury Companion to Syntax

(b) Leikurinn lauk með jafntefli. match-def.nom finished with draw “The match ended in a draw.” “Dative Sickness” consists in the substitution of dative case for accusative case with subjects, as in (6), as opposed to the original accusative exemplified in (1). (6)

Icelandic (Eythórsson 2009) Mér vantar peninga. me-dat needs money-acc.pl “I need/lack money.”

Nominative Substitution is very common in Faroese, but in Icelandic it mostly affects theme subjects as in (4) and (5). “Dative Sickness” exclusively affects accusative experiencer subjects and is common both in Icelandic and in Faroese. In fact, in Faroese the accusative has become virtually obsolete as subject case, having been largely ousted by the nominative, although dative subjects are still found to a limited degree with experiencer verbs.

2.4 Structural Case and (Thematic and Idiosyncratic) Lexical Case Yip, Maling and Jackendoff (1987) set forth a hypothesis on the assignment of case to arguments, positing a fundamental distinction between structural case, on the one hand, and lexical (inherent) case on the other hand. According to this proposal, structural case is a predictable syntactic case, but lexical case is limited to certain verbs or verb classes. Lexical case, in turn, is divided into regular (thematic) case and idiosyncratic (or “quirky”) case. Thematic case is predictable from the lexical semantics of the verb, but idiosyncratic case is completely unpredictable. According to this threefold division, case in Icelandic can be classified as follows: First, nominative is structural case with subjects, while accusative is structural case with objects; in other words, nominative and accusative are the unmarked syntactic cases and not restricted to particular verbs. Nominative is also structural case with objects of verbs taking oblique subjects, that is, DATNOM verbs like leiðast “be bored” in (3). Second, dative in Icelandic is a regular (thematic) lexical case with subjects that have the thematic role of experiencer (e.g. with leiðast) or recipient (e.g. with áskotnast “obtain”). The arguments for this analysis include the productivity of dative case, which manifests itself, among other things, in the strong

370

9781441124609_Ch22_Final_txt_print.indd 370

11/20/2012 9:24:51 AM

Syntactic Change

tendency to generalize dative as subject case at the expense of accusative with experiencer verbs (“Dative Sickness”). Moreover, dative is regular lexical case with the objects of many verbs (e.g. hjálpa “help”). Third, both accusative and dative case with theme subjects, as in (4) and (5), are unpredictable in Icelandic, and are thus idiosyncratic. The diminished productivity of verbs taking oblique theme subjects is manifested by their tendency to acquire nominative case with subjects (Nominative Substitution). Finally, it may be mentioned that genitive is always considered idiosyncratic case, both with subjects and objects. The genitive with arguments of verbs is still found in Modern Icelandic, but has been lost in Faroese.

2.5 Changes in Case Marking of Arguments On the basis of the threefold division of case sketched above for Icelandic, a hypothesis on the directionality of changes in case marking of arguments can be set forth, as follows (Eythórsson 2002): (7) Case Directionality Hypothesis (a) idiosyncratic case → structural case (b) idiosyncratic case → regular lexical (thematic) case According to this hypothesis, structural case replaces (idiosyncratic) lexical case whereas idiosyncratic case yields to thematic case and structural case. The manifestations of these changes in Icelandic include the two types of change described above: Nominative Substitution, whereby idiosyncratic case is replaced by structural case, and “Dative Sickness,” whereby idiosyncratic case is replaced by thematic case. Thus, these changes are in accordance with the hypothesis in (7) on the directionality in case marking. The validity of this hypothesis has been further supported by the findings of extensive empirical evidence gathered from surveys, conducted in recent years, of the case system of Icelandic. The motivation for the changes is most plausibly seen in connection with the productivity and frequency of the relevant case patterns. Nominative is the unmarked case in Icelandic par excellence and predominant with subjects; it is therefore unsurprising that it has a strong tendency to be generalized at the expense of oblique case with subjects (which may have become idiosyncratic due to the fact that its relation to a particular lexical semantics has diminished). Dative, on the other hand, became the unmarked oblique subject case with experiencer and recipient subjects; this is clearly manifested by the strong tendency for dative to be substituted for accusatives by “Dative Sickness.” In view of these facts, it is proposed that Nominative Substitution

371

9781441124609_Ch22_Final_txt_print.indd 371

11/20/2012 9:24:52 AM

The Bloomsbury Companion to Syntax

(“personalization”) is a type of analogical change which involves morphosyntactic leveling in favor of nominative as the unmarked case. By the same token, “Dative Sickness” can be seen as leveling in favor of dative at the expense of accusative as the unmarked oblique subject case. Thus, there are two types of change affecting the case marking of subject arguments: the generalization of nominative as the unmarked case at the expense of all other case patterns with subjects, and the generalization of dative as a regular lexical (thematic) case with subjects. A highly interesting result from recent surveys of Faroese is the fact that the development in this language independently confirms the Case Directionality Hypothesis in (7), which was originally proposed on the basis of Icelandic. By implication, the Faroese findings also support the hypothesis of the triple division of (Yip et al. 1987). In Faroese accusative has become almost obsolete as subject case, having been ousted by dative with experiencers and, more generally, by nominative. In fact, the comparison with Icelandic clearly suggests that Nominative Substitution has drastically reduced the occurrences of oblique case with subjects in Faroese; this includes dative case, which would seem to have become idiosyncratic even with experiencer subjects. The premise for this conclusion, of course, is that older Faroese was similar to Icelandic and had a lot more oblique subjects than Modern Faroese. The historical and comparative study of the Scandinavian languages provides strong arguments in favor of that assumption.

2.6 Conclusion The particular value of the investigation of the changes in case marking in Insular Scandinavian has wider implications. Similar changes to those studied above have taken place in other languages, including some of the older Germanic languages. In many of these languages oblique subjects (or subjectlike obliques) have been ousted by the kind of morphosyntactic leveling termed Nominative Substitution. Changes comparable to “Dative Sickness,” another type of morphosyntactic leveling, are also known from other languages, for example, German, where, however, the facts are not as transparent. Thus, careful study of the changes in case marking in modern Insular Scandinavian has resulted in a deeper understanding of the mechanisms of morphosyntactic change. This is why the contemporary Insular Scandinavian languages can be regarded as providing a “laboratory” case for testing hypotheses originally proposed for languages such as Old and Middle English, which are only attested in written sources.

372

9781441124609_Ch22_Final_txt_print.indd 372

11/20/2012 9:24:52 AM

Syntactic Change

3 Reanalysis of Ambiguous Structures: The New Passive in Icelandic 3.1 Introduction Reanalysis is one of the most important types of syntactic change, by which new constructions emerge on the basis of existing structures. According to Harris and Campbell (1995: 50, 61), reanalysis is “a mechanism which changes the underlying structure of a syntactic pattern and which does not involve any modification of its surface manifestation”; they add that surface manifestations can appear after the reanalysis of the underlying structure has taken place, in the form of morphological change or word order (cf. also Roberts 2007: 122–3). A well-researched case involves the New Passive in Icelandic, which has emerged from a reanalysis of the canonical existential passive with a postverbal NP (Eythórsson 2008 contra Maling and Sigurjónsdóttir 2002). The locus of the reanalysis involves structures where the canonical existential passive and the New Passive cannot be distinguished morphologically.

3.2 The New Passive vs. the Canonical Passive In the past few decades a new construction has been gaining ground in Icelandic, which has the morphosyntactic characteristics of passive but with the postverbal NP corresponding to the object of the active clause exhibiting accusative case, and not nominative. The accusative NP is demonstrably an object. This construction is known as the New Passive. (8) Icelandic (Eythórsson 2008) mig. (a) Það var barið it was hit-n.sg me-acc “I was hit.” lítinn (b) Það var barið it was hit-n.sg little “A little boy was hit.”

strák. boy-acc

litla strákinn. (c) Það var barið it was hit-n.sg little-def.acc boy-def.acc “The little boy was hit.” The canonical passive in Icelandic is formed in a similar way as in other Germanic languages. An NP corresponding to the accusative object of an active verb appears in nominative case, which the participle agrees with. The NP

373

9781441124609_Ch22_Final_txt_print.indd 373

11/20/2012 9:24:52 AM

The Bloomsbury Companion to Syntax

occurs in the canonical subject position and can either be definite or indefinite, as in (9). (9) Icelandic (Eythórsson 2009) tölvan var keypt í gær. (a) Nýja new-def.nom computer-def.nom was bought-nom yesterday “The new computer was bought yesterday.” tölva var keypt í gær. (b) Ný new-nom computer-nom was bought-nom yesterday “A new computer was bought yesterday.” Alternatively, the NP stands further to the right, in clauses which, in the absence of any other clause-initial phrase, are introduced by the expletive það “it.” This is the expletive passive, in which the NP is subject to the Definiteness Effect (DE), and occurs either preverbally (to the left of the participle) or postverbally (to the right of the participle), as shown in (10). (10) Icelandic (Eythórsson 2008) ný  tölva (a) Það var kypt it was bought-nom new-nom computer-nom (*nýja  tölvan). (*new-def.nom computer-def.nom) “A (*the) new computer was bought.” (*nýja tölvan). (b) Það var ný tölva it was new computer-nom (*new-def.nom computer-def.nom) keypt. bought-nom “A (*the) new computer was bought.” Moreover, lexical case (dative or genitive) with NPs is preserved both in the canonical passive, as in (11), and in the New Passive, as in (12b). The participle appears in the default neuter singular form. Note in particular the formal ambiguity between canonical passive and the New Passive with the postverbal indefinite dative NP in (11a). The example in (11b) can only be the canonical expletive passive, as shown by the occurrence of the NP to the left of the participle. Likewise, (12b) is out both as a canonical passive and as a New Passive. The preverbal placement of a definite NP is ungrammatical in the canonical passive and so is the preverbal placement of any NP in the New Passive. (11) Icelandic (Eythórsson 2008) stelpu í skólanum. (a) Það var hrint it was pushed-n.sg girl-dat in school-def “A girl was pushed in school.” 374

9781441124609_Ch22_Final_txt_print.indd 374

11/20/2012 9:24:52 AM

Syntactic Change

í skólanum. (b) Það var stelpu hrint it was girl-dat pushed-n.sg in school-def “A girl was pushed in school.” (12) Icelandic (Eythórsson 2008) stelpunni í skólanum. (a) Það var hrint it was pushed-n.sg girl-def.dat in school-def “The girl was pushed in school.” hrint í skólanum. (b) *Það var stelpunni it was girl-def.dat pushed-n.sg in school-def The New Passive in Icelandic is most common among children and adolescents, but it does occur among adults as well. The oldest users on record were born around 1950 (Maling and Sigurjónsdóttir 2002; Eythórsson 2008). It is important to note that the New Passive is used alongside the canonical passive; there are no speakers on record who only have the New Passive and not the canonical passive. Moreover, it is unclear that there is any semantic difference between the New Passive and the canonical passive; this is comprehensible as it can be assumed that the former has emerged from the latter. To recapitulate, the following characteristics of the New Passive are crucial for our purposes: z An NP complement of a past participle is in accusative case (unless it

bears lexical case). z The NP can be definite (i.e. it does not obey the DE). z The NP is an object. z The NP cannot undergo NP-movement. z The participle does not agree with the NP but appears in the (default)

neuter singular form. Finally, what is particularly intriguing about the New Passive is the fact that this construction contradicts the generalization that an accusative object NP does not occur in a clause unless it also contains a nominative NP (see Section 2 above).

3.3 The Emergence of the New Passive Bearing in mind these characteristics, the emergence of the New Passive in Icelandic can be sketched as follows. The point of departure is the claim in Maling and Sigurjónsdóttir (2002: 112) that the original locus of the change was with dative forms because the New Passive appears to be more common with

375

9781441124609_Ch22_Final_txt_print.indd 375

11/20/2012 9:24:52 AM

The Bloomsbury Companion to Syntax

verbs taking dative objects than with those taking accusative objects. In other words, the change may have begun with those forms where the morphological evidence of non-agreement is least obvious; this can therefore be interpreted as “defective” agreement. For example, post-verbal oblique NPs (datives and genitives) and neuter singular NPs involve a formal ambiguity between the New Passive and the canonical expletive passive, as shown in (13) and (14). (13) Icelandic (Eythórsson 2008) (a) Það var skammað lítið barn. it was scolded-n.sg little-nom/acc child-nom/acc “A little child was scolded.” (b) Það var hrint litlum strák. it was pushed-n.sg little-dat boy-dat “A little boy was pushed.” The ambiguity is even more robust in cases where there is a violation (“leakage”) of the DE with postverbal NPs in the canonical expletive passive, as occasionally happens in standard Icelandic. (14) Icelandic (Eythórsson 2008) (a) Það var skammað litla barnið. it was scolded-n.sg little-def.nom/acc child-def.nom/acc “The little child was scolded.” (b) Það var hrint litla stráknum. it was pushed little-def.dat boy-def.dat “The little boy was pushed.” Furthermore, even though it may have nominative case marking, the postverbal NP in the canonical expletive passive is arguably object-like, at least for some speakers, as it does not pass the standard subject tests, for example, Conjunction Reduction, as easily as NPs in other positions in passive sentences. As shown in (15a) the preverbal NP is fine in an expletive passive clause followed by a conjoined clause, but the postverbal NP is much less felicitous in such structures, as indicated by the question marks in (15b). (15) Icelandic (Eythórsson 2008) laminn og fór að gráta. (a) Það var lítill strákur it was little boy-nom.sg beaten-nom.sg and began to cry “A little boy was beaten and began to cry.” lítill strákur og fór að gráta. (b) ??Það var laminn it was beaten-nom.sg a little boy-nom.sg and began to cry “A little boy was beaten and began to cry.’ ” 376

9781441124609_Ch22_Final_txt_print.indd 376

11/20/2012 9:24:52 AM

Syntactic Change

Thus, the preconditions for a change from nominative to accusative in passive may already have been latent in Icelandic for a considerable period.

3.4 The Nature of the Change The New Passive is very plausibly analyzed as a change that originates in child language acquisition (see Benediktsdóttir 2008). In order for structures with postverbal NPs to signal to the acquirer that the construction in question is the canonical existential passive, either or both of the following two conditions on the postverbal NP must be met: (i) the NP must be unambiguously nominative, and/or (ii) it must be indefinite. In instances involving a formal ambiguity in morphological case forms and a “leakage” in the DE a reanalysis of the postverbal NP can take place. Due to this reanalysis, the postverbal NP in “object position” is assigned object case, that is, accusative rather than nominative. The premise for this analysis is that while nominative is the unmarked case par excellence in Icelandic, accusative is the unmarked object case. Thus, if nominative is not assigned to the external argument, it can be assigned to the internal argument. By assumption, then, this is the reason for nominative case assignment to the internal argument in the canonical passive to transitive verbs, where there is no external argument (see Section 2). If, on the other hand, the evidence for nominative is somehow weakened, for example, as a result of shifts in the input available to the acquirer, accusative can be assigned to the internal argument instead; hence the accusative case assignment in the New Passive of Icelandic. On the account outlined above, a child faced with ambiguous evidence regarding the morphological case of the postverbal NP concludes that the case of the NP in object position should be accusative and not nominative. The ambiguous structures give rise to a change in case assignment in favor of accusative case—the unmarked case with objects. Once the postverbal NP has been reanalyzed as an object which is assigned accusative case, the New Passive emerges. Presumably, the reanalysis takes place as a result of a drop in the frequency of the evidence that the acquirers have for nominative case with the postverbal NP below a certain threshold; exactly what that threshold is remains to be specified (for some general speculations, see Lightfoot 1999, 2006; for instance). As the New Passive does not involve nominative case assignment to the object, the absence of participial agreement is an inevitable part and parcel of the change. Recall that the assumption is that the New Passive emerges in structures where the participial agreement could be interpreted as “defective.” It is reasonable to assume that this reanalysis is favored by lack of agreement in other types of passives that are ubiquitous as a “depersonalizing device” in Icelandic: the impersonal passive of intransitives, including unergatives, 377

9781441124609_Ch22_Final_txt_print.indd 377

11/20/2012 9:24:52 AM

The Bloomsbury Companion to Syntax

reflexives, and unaccusatives. In all of these structures the participle is in the default neuter singular form.

3.5 Conclusion In conclusion, the main motivation for the emergence of the New Passive can be identified. Due to instances of formal and structural ambiguity, the postverbal NP is reanalyzed as an object that is assigned accusative case rather than nominative. The other properties of the New Passive—the lack of NP-movement and the DE—follow from this reanalysis. It should be stressed, however, that while the above scenario of the emergence of the New Passive is plausible and can be backed up by considerable evidence, we are not in a position at this time to solve the “actuation problem” of this change: why did the change happen in Icelandic when it did, and not at some other time? Although this question cannot be given an answer at present, the fact remains that there do not seem to be any unambiguous examples of the New Passive containing either full NPs in the accusative or pronouns until the mid-twentieth century (Maling and Sigurjónsdóttir 2002: 129). Therefore, there is no reason to think that the change qua the reanalysis of the canonical expletive passive as New Passive began much earlier than that, given the copious written documentation of Icelandic.

4 The Inertia Hypothesis 4.1 Introduction The type of syntactic change involved in the New Passive in Icelandic is arguably triggered by structural and formal ambiguity. In fact, it is universally agreed that syntactic change can be caused by changes to Primary Linguistic Data (PLD) arising from phonological, morphological, or lexical change, or extra-linguistic factors like contact. According to Keenan (1994, 2002, 2009), however, syntax itself cannot change endogenously; in other words, syntactic change is subject to “Inertia.” The Inertia Hypothesis can be stated in a general form as follows: (16) Syntactic change does not arise unless caused by external factors. This hypothesis essentially claims that syntax, by itself, is diachronically completely inert, and, more generally, that linguistic change may only originate as an interface phenomenon, involving “external” factors such as phonology and semantics. This hypothesis has found widespread recognition in the literature on 378

9781441124609_Ch22_Final_txt_print.indd 378

11/20/2012 9:24:52 AM

Syntactic Change

theoretical diachronic linguistics (e.g. Longobardi 2001b; Lightfoot 2002; Crisma and Longobardi 2009; Biberauer and Roberts 2009; among others). A weaker form of the hypothesis may be taken to suggest that such domains as pragmatics, stylistics, and (genre-conditioned) language use must also be regarded as “external” factors. A number of problematic questions arise in this connection, for example, if it can ever be excluded that syntactic change is ultimately caused by external factors. In other words, it is not immediately obvious that the Inertia Hypothesis is falsifiable and therefore meets accepted scientific standards. A number of publications have addressed this issue. For example, Biberauer and Roberts (2008) propose on the basis of Middle and Early Modern English that a syntax-internal cause for syntactic change can be identified; this situation arises when an initial, extra-syntactically induced change creates a system which leads to cascades of changes over several centuries, leading to a typological shift. Another case that potentially supports the assumption of “pure” syntactic change might be seen in the loss of OV order in the VP in the history of Icelandic.

4.2 The Loss of OV Order in the VP in Icelandic In older Icelandic, the VP exhibited both Verb-Object (VO) order, as in (17a), and Object-Verb (OV) order, as in (17b); this was the state of affairs from the earliest recorded history of Icelandic until the first half of the nineteenth century, when the OV orders were lost (e.g. Rögnvaldsson 1994/5: 35). (17) (a) . . . at   sárit mundi hafa  grandat honum. that wound-def would  have killed-n.sg. him “ . . . that the wound would have killed him.” (b) . . . og kvaðsk enginn maðr  þetta gera  vilja. and said no   man  this do-inf want-inf “ . . . and no one said that he wanted to do this.” The VP word order variation has been mapped with considerable precision in Old Icelandic prose by Rögnvaldsson (e.g. 1994, 1995, 1996), using a computerized concordance. Hróarsdóttir (2000) has also studied this variation in the subsequent period of Icelandic (ca. 1400–1850), based on her selection from a corpus of written documents of various kinds. In spite of this extensive documentation, none of the past studies have succeeded in pinpointing the possible mechanisms behind the OV/VO variation and the loss of OV order, neither in terms of syntax, information structure, stylistics, or text genre. True, parallel word order changes also happened in related languages, Mainland Scandinavian (Delsing 379

9781441124609_Ch22_Final_txt_print.indd 379

11/20/2012 9:24:52 AM

The Bloomsbury Companion to Syntax

2000) and Middle English (Pintzuk 1991), some centuries prior to the OV loss in Icelandic, and those changes have standardly been connected to external factors such as the loss of inflectional morphology, or to language contact, or both. However, neither of these factors can be held responsible for the OV loss in Icelandic. As the inflectional system was preserved virtually intact in Icelandic, morphological reduction cannot have been the trigger of this syntactic change. Language contact, in particular with Danish or other Mainland Scandinavian languages, is also unlikely to have been a direct cause for the loss of OV, given the geographical isolation of Iceland.

4.3 Conclusion It has not be possible so far to identify any factor that would be a necessary precondition for the sudden loss of OV orders in the early nineteenth century. Therefore, until such a factor can be identified, it would seem that the OV loss in Icelandic is in fact due to a syntax-internal cause: challenging Inertia. A lot more empirically founded research is needed to test the Inertia Hypothesis, with the aim to show that when all potential external factors have been filtered out there indeed remain cases of “pure” syntactic change.

5 Syntactic Reconstruction 5.1 Introduction It is sometimes said that no language has changed so much in the past decades as the reconstructed Indo-European proto-language. Contrary to the intention of this joke, this situation actually bears witness to the vitality of the field of historical and comparative Indo-European linguistics. The principal goal of linguistic reconstruction is to model the grammar of prehistoric stages from which the attested languages can be derived; the reconstructions reflect the state of the art in the field of historical and comparative linguistics at any given time. But while the reconstruction of phonological and morphological systems and lexical items is a well-integrated domain in historical linguistics, syntactic reconstruction has long been a neglected area. There are several reasons for this neglect, including the preoccupation with potential problems considered to be inherent in syntax and to be virtually insurmountable for such an undertaking (see already Watkins 1976, for an early skeptical view). In this section I present arguments against this prevailing view, trying to show that reconstruction is not only feasible in syntax, but in fact a necessary part of historical syntax so that syntactic change can be modeled in a precise way. 380

9781441124609_Ch22_Final_txt_print.indd 380

11/20/2012 9:24:53 AM

Syntactic Change

5.2 Syntactic Reconstruction: Pros and Cons The reasons for lingering skepticism regarding the viability of syntactic reconstruction include the following points. First, syntactic structures are not atomic like lexical items, morphemes, and phonemes, but complex and schematic. Second, there is no agreement on the diachronic transmission of syntactic or grammatical structures; while lexical items can be transmitted from one generation of speakers to the next, sentences are created afresh with every new generation of acquirers (Lightfoot 1979 and later work). Third, syntactic change is clearly not regular like the types of phonological change subject to “sound laws” (Lightfoot 2002; Pires and Thomason 2008). Finally, syntax is presumed to lack the degree of arbitrariness in form–meaning mappings that is generally found in the lexicon (Harrison 2003). All of these points have been taken to show that the reconstruction of syntax is not possible in the same way as the reconstruction of phonological systems and lexical items. Nevertheless, several researchers have presented arguments in support of the viability of syntactic reconstruction (see Eythórsson and Barðdal 2011; Barðdal and Eythórsson 2012a; 2012b, for a detailed discussion and further references, in particular to Kikusawa 2003; Guardiano and Longobardi 2005; Roberts 2007; and Harris 2008). In response to this skepticism it must be stated that the alleged obstacles to syntactic reconstruction seem in many cases to be theory-dependent. Although it is evident a priori that sentences have a more abstract structure than lexical items, it is unclear that the difference is as fundamental as often assumed. Thus, according to the findings of current theoretical linguistics, the structure of words and morphemes is more abstract than traditionally thought; conversely, the structure of phrases and clauses may be analyzable in a more concrete way than usually done within certain dominant frameworks, for example, generative syntax. In short, the difference in the degree of abstractness between words, on the one hand, and sentences, on the other hand, may have been exaggerated in previous scholarship. In this light, Construction Grammar appears a promising framework for reconstructing syntax, as it takes the basic unit of language, the construction, to be a form–meaning pairing (Barðdal and Eythórsson 2012a). The tried and tested method in traditional historical–comparative linguistics, known as the Comparative Method, is also based on correspondence sets of form and meaning, to which constructions provide exactly the right kind of input. In other words, there is a natural leap from the synchronic form–meaning pairings posited by Construction Grammar to those required for historical reconstruction in terms of the Comparative Method. Given the premise of Construction Grammar that syntax involves constructions similar to the lexicon, the difference between syntactic reconstruction and lexical and phonological reconstruction becomes minimal. 381

9781441124609_Ch22_Final_txt_print.indd 381

11/20/2012 9:24:53 AM

The Bloomsbury Companion to Syntax

5.3 Reconstructing Case Frames On the basis of the evidence from the Germanic languages, the predicate *etana“eat” can be reconstructed for Proto-Germanic (cf. Gothic itan, Old English and Old Saxon etan, Old High German ezzan, Old Icelandic eta; Modern English eat, German essen, Modern Icelandic eta (éta), etc.). Furthermore, since this verb takes a nominative subject and an accusative object in all the daughter languages, the reconstruction of the form *etana- “eat” with the corresponding case frame in Proto-Germanic is completely unproblematic. However, due to the ubiquity and high productivity of the nominative–accusative case pattern in Germanic and other Indo-European (as well as non-Indo-European) languages, the degree of arbitrariness in the reconstruction of canonical subject and object case marking is bound to be low. Therefore, although certainly demanded by the facts, this reconstruction appears trivial and thus not particularly interesting. Non-canonical case subject marking involves a higher degree of arbitrariness than canonical subject marking, and therefore such patterns have a higher validity for reconstruction. In this connection, it should be expiclitly stated that is assumed here that subject-like oblique NPs in Old Germanic were in fact subjects, parallel to the situation in modern languages like Icelandic, Faroese, and German, discussed in Section 2 (for arguments, see, for example, Eythórsson and Barðdal 2005). On this view, the leftmost argument of the argument structure generally behaves syntactically as a subject. As will be discussed presently, the relevant oblique NPs pass the tests for subjecthood applicable in Old Germanic. As an example, the lexical entry for the Proto-Germanic predicate “hunger” may be reconstructed as *hungrijan- (some of the languages, like Old Icelandic, point to *hungrōn-, which however, must be secondary). This reconstruction is made on the basis of attested forms, such as huggrjan (Gothic), hungiren (Old High German), hyngran (Old English), gihungrjan (Old Saxon), hungera (Old Frisian), and hungra (Old Icelandic). All of these verbs mean “hunger” and all of them occur with an accusative subject, as in the following examples (Barðdal and Eythórsson 2012a): (18) Gothic (Barðdal and Eythórsson 2012a) þana gaggandan du mis ni huggreiþ. this-one.acc going to me not hungers “The one who comes to me will not starve.” (St John 6: 35) (19) Old High German (Barðdal and Eythórsson 2012a) Mih hungrita hungrita, inti ir   gabut mir ezzan. me.acc hungered  and you gave me eat.inf “I was hungry and you fed me.” (Tatian 152: 3)

382

9781441124609_Ch22_Final_txt_print.indd 382

11/20/2012 9:24:53 AM

Syntactic Change

(20) Old English (cf. Barðdal and Eythórsson 2012a) þone þe to me cymð. Ne hingrað not hungered the-one.acc that to me came “The one who comes to me will not starve.” (Lindisfarne Gospels 1, St John 6: 35) (21) Old Icelandic (Barðdal and Eythórsson 2012a) mann hungrar þá til líkamligra krása. man.acc hungers then to bodily delicacies “A man starts having cravings for fleshly delicacies.” (Leif. 48 14. 18. 20) The form and meaning correspondence seen in the individual languages exemplified above clearly suggest inheritance from an earlier stage, yielding the traditional reconstruction *hungrijan- (*hungrōn-) “hunger.” Although usually not explicitly stated in traditional historical–comparative work, it is perfectly reasonable to reconstruct this particular verb together with an accusative subject. This reconstruction is in accordance with the Comparative Method, demanding the consideration of the evidence from the earliest and most archaic daughter languages. In addition to Old Germanic, the accusative with this verb is still found in Modern Icelandic and German, which are more conservative in this respect than, say, Modern English, which has nominative (I hunger). The occurrence of nominative subject is in accordance with the directionality manifested by Nominative Substitution, the strong tendency to replace oblique case by nominative with subjects, as discussed in Section 2 (cf. Eythórsson 2002). In summary, the reconstruction of the verb *hungrijan- (*hungrōn-) “hunger” with an accusative subject is uncontroversial given the traditional premises of the Comparative Method. This combination of form, including the specific subject case marking pattern, and meaning yields the input for the correspondence set for the accusative subject construction with this particular lexical item in Proto-Germanic. The same kind of argument can be used mutatis mutandis for other predicates taking oblique subjects in Germanic and cross-linguistically.

5.4 Reconstructing Grammatical Relations 5.4.1 Introduction Given that syntax is a legitimate object of the Comparative Method, the question arises whether grammatical relations, such as the subject and the object relation, are also reconstructable. Barðdal and Eythórsson (2012a, 2012b) argue that this is indeed the case, and that it is sufficient to reconstruct the constructions that distinguish subjects from objects in order to capture the relevant grammatical relations for a particular proto-stage. This method can be illustrated

383

9781441124609_Ch22_Final_txt_print.indd 383

11/20/2012 9:24:53 AM

The Bloomsbury Companion to Syntax

by a comparison of the behavioral subject properties found in the Germanic languages. At least four behavioral properties in Old Germanic clearly single out the subject relation, as opposed the object relation. z Position (Subject-Verb Inversion) z Subject-to-Object Raising z Subject-to-Subject Raising z Control Infinitive

The relevance of these properties in Old Germanic will be illustrated in the following.

5.4.2 Position (Subject-Verb Inversion) There is a very clear contrast between subject-initial orders and other word order patterns in Germanic, both with nominative subjects and oblique subjects. Subject-initial clauses constitute the neutral or unmarked word order pattern, whereas all other word order patterns are derived. The patterns relevant as subject properties include the clause-initial position of the finite verb in declarative clauses (the so-called Narrative Inversion) and fronting (“topicalization”) of non-subjects. An early example of fronting (presumably for focus), with the finite verb in second position preceding an enclitic subject pronoun, is found already in an inscription written in the runic alphabet from around AD 500. (22) Early Runic (Eythórsson 2001, 2011) Hariuha haiti-ka farauisa. Hariuha am-called.-I.nom knowing-danger.nom “I am called Hariuha, the one who knows danger.” In the unmarked case the NP following the finite verb in inversion patterns is the subject.

5.4.3 Subject-to-Object Raising Raising-to-Object constructions are found already in Old Germanic, including Gothic. (23) Gothic (Barðdal and Eythórsson 2012a) saljan at izwis. [ik] wenja mik . . . I.nom hope myself.acc stay.inf with you “I hope to stay with you.” (1 Cor 16.7, from Harbert 2007: 263) It should be pointed out that the Greek original of the Gothic example in (30) has the verb elpizō “hope,” which is a control verb. This in turn shows 384

9781441124609_Ch22_Final_txt_print.indd 384

11/20/2012 9:24:53 AM

Syntactic Change

that Raising-to-Object truly existed in Gothic and is not a calque from Greek. Raising-to-Object can therefore be reconstructed as a subject property for Proto-Germanic. Oblique subject predicates are also found in Raising-to-Object constructions in both North and West Germanic, for example, in Old Icelandic and Old Saxon. (24) Old Icelandic (Barðdal and Eythórsson 2012a)  fátt um finnast við Þorstein um veturinn. Einar lét sér Einar.nom let self.dat little about think with Þorstein of winter.def “Einar showed indifference towards Þorstein during the winter.” (Þorsteins Þáttur Síðu-Hallssonar eftir Morkinskinnu, chapter 3) (25) Old Saxon (Barðdal and Eythórsson 2012a) ina thô gehungrean lêt. Sô he since he.nom self.acc then hunger.inf let “Since he then let himself hunger.” (Heliand 1059) The lack of any relevant evidence from Gothic is presumably due to the small size of the Gothic corpus rather than any systematic absence.

5.4.4 Subject-to-Subject Raising In a parallel manner, Raising-to-Subject may also be reconstructed for ProtoGermanic, for example, by means of verbs meaning “begin” (*-ginnan, found in both East and West-Germanic with different prefixes). (26) Gothic (Barðdal and Eythórsson 2012a) allai þai gasaihvandans duginnaina bilaikan ina all.nom who seeing begin mock.inf him “all, who see (it), begin to mock him.” (Lk. 14.29) The subject allai “all” of the lower verb bilaikan “mock” is in the nominative case and occurs here as the subject of duginnan “begin.” Since only the subject of the lower verb, and not its objects, behave in this way, Subject-to-Subject raising is generally regarded as a subject test in the syntactic literature. Again, examples involving oblique subject predicates are found in both North and West-Germanic, for instance in Old Saxon: (27) Old Saxon (Barðdal and Eythórsson 2012a) than is  uuerk tregan, an  is hugi hreuuen. ôðo beginnad imu easy begin him.dat then his work regret at his mind rue “He will find it easy to begin to regret what he did, to rue it in his mind.” (Heliand 3234)

385

9781441124609_Ch22_Final_txt_print.indd 385

11/20/2012 9:24:53 AM

The Bloomsbury Companion to Syntax

Thus, not only can Raising-to-Subject be reconstructed for Proto-Germanic, but the subject behavior of the oblique argument of the lower verb in examples like (27) can be reconstructed as well.

5.4.5 Control Infinitive Finally, Control Infinitives are generally taken to be the most conclusive subject test cross-linguistically. Such cases are well attested in Old Germanic. Examples of control infinitives involving oblique subjects from Old Norse and Early Middle English are given below. (28) Old Icelandic (Barðdal and Eythórsson 2012a) Indriði  kveðst eigi ___  svo á lítast. Indriði.nom says not pro.dat so on seem.inf “Indriði says he does not like that.” (Þorsteins þáttur Síðu-Hallssonar from Flateyjarbók, chapter 2) (29) Early Middle English (Barðdal and Eythórsson 2012a) dremen of win. good is, quaþ Joseph, to ___ good is, said Joseph to pro.o l dream.inf of wine “It is good, said Joseph, to dream about wine.” (Gen. & Exod. 2067, ca. 1250, here cited from Seefranz-Montag 1983: 134) For our purposes, it is sufficient to be able to document the occurrence of control constructions in the different Old Germanic sub-branches in order to argue that such constructions must be reconstructed for the Germanic protolanguage. This also includes control constructions with oblique subject predicates, although the relevant examples are missing in Gothic. (See, however, Section 5.5 for a potential argument for the occurrence of an oblique subject in a control infinitive in Gothic.)

5.4.6 Summary At least the following subject properties can be safely reconstructed for Proto-Germanic: Position (Subject-Verb Inversion), Raising-to-Object, Raising-to-Subject, and Control. More interestingly, Raising-to-Object and Raising-to-Subject show distinctly that in North and West Germanic oblique subjects behave syntactically like nominative subjects and not like objects. Position and (possibly) Control Infinitives support that analysis. Thus, it is clear that the constructions that distinguish subjects (including oblique subjects), from objects, and thereby also the relevant grammatical relations, can be reconstructed in accordance with the principles of the Comparative Method.

386

9781441124609_Ch22_Final_txt_print.indd 386

11/20/2012 9:24:53 AM

Syntactic Change

5.5 Missing Evidence and the Predictive Power of the Comparative Method A problem facing the historical syntactician, like any other linguist working on languages only documented in historical texts, involves missing evidence. The linguist must decide if the absence of the evidence is accidental or systematic. An example illustrating this dilemma with respect to the reconstruction of grammatical relations comes from Gothic. In this language there is one example of the verb luston “lust for.” The twist is that this single occurrence of luston is in a control infinitive, with the subject being omitted on identity with the nominative subject in the matrix clause while the genitive object is overt. (30) Gothic (Barðdal and Eythórsson 2012a) luston izos. hvazuh saei saihviþ qinon du ___ whoever who sees woman.acc to pro.acc lust her.gen “whoever looks on a woman in order to lust for her.” (Mt. 5.28) As this is the only instance of luston “lust for” in Gothic, it is impossible to verify that this verb had a Accusative–Genitive case frame in this language. On the basis of the comparative evidence from the other Old Germanic languages, however, it is reasonable to assume that it did (and of course it manifestly preserved the genitive object in Gothic). The verb *luston- “lust, desire” occurs with the Accusative–Genitive case frame in North and West Germanic, as shown in the following examples, from Old Icelandic and Old English: (31) Old Icelandic (Barðdal and Eythórsson 2012a) þig lysti þessa. er when you.acc desired that.gen “When you desire that.” (Ljósvetningasaga, chapter 19) (32) Old English (Barðdal and Eythórsson 2012a) Hine nānes þinges ne lyste on ðisse worulde. him.acc no.gen thing.gen not desire on this world “He desired nothing in this world.” (Boethius Cons.Phil. 35–6) In this light, it is highly plausible that oblique subjects could be omitted in Control Infinitives in Gothic, as in other subbranches of Germanic, and are thus reconstructable for Proto-Germanic. To be sure, it cannot be excluded that the accusative argument had been replaced by nominative with the verb “lust” (by Nominative Substitution, cf. 2 above), but given the archaic character of Gothic and the retention of oblique case with other “impersonal” verbs, for example, huggrjan “hunger” discussed in Section 2; but this possibility must be considered less likely. 387

9781441124609_Ch22_Final_txt_print.indd 387

11/20/2012 9:24:54 AM

The Bloomsbury Companion to Syntax

5.6 Conclusion On the basis of work employing the model of Construction Grammar (Barðdal and Eythórsson 2012a, 2012b), it has been shown that the reconstruction of grammatical relations is much better motivated than traditionally assumed. The first step in this procedure is reconstructing lexical predicates and their argument structure constructions, and as a next step, the relevant constructions involving subject behavior can be reconstructed. On a successful reconstruction of these constructions, the grammatical relations of the arguments fall out. This exercise shows how grammatical relations can be reconstructed in a methodologically rigorous way for prehistoric periods of languages. It is reasonable to assume that the application of the Comparative Method in combination with Construction Grammar can be extended to other abstract domains within syntax. This is among the tasks of future research in the domain of syntactic reconstruction. The goal of this section was to argue that reconstruction is not only feasible in historical and comparative syntax, just as in other domains of grammar and in the lexicon, but in fact a necessary part of any serious theory of syntactic change. Given that the principal goal of linguistic reconstruction is to model the grammar of the prehistoric stages from which the attested languages can be derived, the omission of syntax from such undertaking will inevitably leave us with a very incomplete picture. The preceding discussion should have made clear that there is no reason to content oneself with that state of affairs.

6 Conclusion The four case studies presented in this chapter have focused on different key issues in syntactic change. Section 2 dealt with the nature of the morphology– syntax connection and the question of the impact of the loss of morphology on syntactic change. It was concluded that the relevance of morphology-driven syntactic change tends to be exaggerated and that, at least in many cases, alternative explanations for the changes in question must be sought. These explanations include leveling conditioned by frequency and other usage-based factors. Section 3 dealt with reanalysis, perhaps the most important type of syntactic change, presenting arguments from a well-investigated case involving structurally ambiguous constructions as the locus for change. Section 4 addressed the problem of the Inertia Hypothesis, making a case for a potential challenge to this much-discussed hypothesis, from the history of Icelandic. Finally, Section 5 argued that, contrary to a widely held view, syntactic reconstruction is a viable enterprise and a necessary part of historical syntax.

388

9781441124609_Ch22_Final_txt_print.indd 388

11/20/2012 9:24:54 AM

23

Acquisition of Syntax 1 Maria Teresa Guasti

Chapter Overview Introduction Acquisition of Word Order Acquisition of Wh-movement Acquisition of Pronominal Reference Conclusion

389 390 391 398 402

1 Introduction This chapter is concerned with the acquisition of syntax. In order to illustrate how this happens, I have chosen three distinctive properties of human language: word order, displacement, and locality. Order of words is a fundamental property of human language, which is acquired since the first word combinations. Children speaking a variety of different languages put words together respecting the order found in their language of exposure and rarely are mistaken in doing this. There is evidence that even before they start speaking, children are sensitive to word order and use it to understand sentences. But how do children figure out the relevant order at only 18 months or earlier? Another distinctive feature of human language is displacement or movement. Wh-elements are often moved to a sentence initial position and they have to be reconstructed in their base position (where a copy is left) to be interpreted. Thus, a relation has to be established between an overt element and a silent one. A similar operation is at stake in the formation of relative clauses. Although children form questions very early and do not have particular problems with the movement operation, the full mastery of constructions involving such operation is full of tricks. Relations hold between adjacent

389

9781441124609_Ch23_Final_txt_print.indd 389

11/20/2012 9:26:17 AM

The Bloomsbury Companion to Syntax

words, but they also hold between distant elements, of which movement is a special case. In general, all relations connecting two syntactic positions are subject to locality constraints. Dependencies between two lexical elements are exemplified by a pronoun and a linguistically specified antecedent. Children use pronouns and understand sentences containing them, but challenges come from specific formal properties of pronouns. We are going to examine the acquisition of these key properties of human languages in a cross-linguistic perspective. In the course of acquisition both universal and language specific features emerge, whose study may shed light on the nature of language in general.

2 Acquisition of Word Order At an average age of 24 months, children start to combine words by respecting the word order of their language. English learners use the SVO order, while their Turkish, Dutch, and German peers employ the SOV order. For Dutch and German this is observed in sentences including infinitive verbs, as in (1a). In finite sentences, children respect the V2 constraint and place the finite verb in second sentence position, as in (1b). (1) Dutch (a) Hij op kussens slap-en. (Hein, 2; 6) He on cushions sleep-inf “He sleeps on the cushions.” (b) Hij doet ‘t niet. (Hein, 2; 4) He makes it not “He does not make it.” Before this age, at around 17 months, one-word stage children use word order in comprehending sentences. Using the intermodal preferential looking paradigm, Hirsh-Pasek and Golinkoff (1986) established that children rely on word order to comprehend active reversible sentences. In this experiment, toddlers are seated in front of two monitors and hear a sentence, such as Big Bird is tickling Cookie Monster. One screen displays Big Bird tickling Cookie Monster and the other Cookie Monster tickling Big Bird. Toddlers look longer at the screen matching the sentence they have just heard, thus showing that they have made a connection between the sentences and the visual scene. As sentences in this experiment are reversible, it is only by paying attention to word order that children could establish the association sentence-scene (beyond word order, we assume that children at this age know nouns and verbs, otherwise they would not be able to understand sentences). How comes that children at 17 390

9781441124609_Ch23_Final_txt_print.indd 390

11/20/2012 9:26:18 AM

Acquisition of Syntax

months of age are already comfortable with word order? Based on the large body of evidence showing that infants are sensitive to prosodic properties since birth (see Guasti 2002 for a review), Nespor et al. (1996) have hypothezised that infants figure out the word order between heads and complements in their language at the pre-lexical stage based on the main prominence within the phonological phrase (see also Mazuka 1996). Nespor et al. (2008) have further refined this proposal based on the observation that main prominence is acoustically realized in different ways within branching phonological phrases depending on whether it is stress-initial or stress-final. Precisely, main prominence is realized through higher pitch and intensity in the former case and through increased duration in the latter. Prominence is stress-initial in OV languages (or complement-head languages) and stress-final in VO languages (or head-complement languages). Thus, infants can exploit prominence type to establish whether the complement is initial (OV) or final (VO), respectively. If prominence is realized through pitch and intensity, the complement is initial (OV); if it is realized through duration, the complement is final (VO). Experimental evidence from Christophe et al. (2003) supports these views: infants at 2 months of age can discriminate between two languages (French, VO and Turkish, OV) on the basis of phonological phrase prominence, that is, they can exploit a phonological property to bootstrap the syntactic organization within phrases (Phonological bootstrapping hypothesis). Thus, it is not surprising that when toddlers put words together they already know how to order them. For languages with mixed word order, like Dutch or German, prominence will be realized through pitch plus intensity or duration depending on the phrase: NPs and APs are head-initial and VPs are head final; PPs can be both. This means that infants exposed to these languages will have to determine word order phonological phrase by phonological phrase. Bion et al. (2007) have shown that 14 month old German infants use prosodic prominence within phonological phrases to discriminate differences in word order within VPs, adding further evidence to the validity of the phonological bootstrapping hypothesis.

3 Acquisition of Wh-movement In many languages, wh-questions display movement of wh-elements to sentence initial position. Children exposed to these languages employ wh-movement since their first questions (Dutch, German, Italian, Swedish). For example, Guasti (2000) examined the transcripts of four English-speaking children between 1–6 and 5–1 years. Of the 2809 wh-questions found, only 1 percent had an in-situ wh-element and some of these were probably echo-questions. The same holds for other early languages (Clahsen et al. 1995 for German; 391

9781441124609_Ch23_Final_txt_print.indd 391

11/20/2012 9:26:18 AM

The Bloomsbury Companion to Syntax

Guasti 1996 for Italian; Santelmann 1998 for Swedish). Some examples are in (2) and (3). (2) English (Brown 1973) Where dis goes? (Adam, 2; 8) (3) French (Hamann 2000) Où il est canard? (Marie, 2; 1) Where he is duck In those languages where wh-in-situ and wh-movement are both possible, whin-situ takes priority. This is evident in French. Hamann (2006) established that initially children highly prefer wh-in-situ and start to employ wh-movement only later. Examples from wh-in-situ questions are in (4). (4) French (Hamann 2000) (a) Est où maman. (Augustin, 2; 6) Is where mommy? (b) Le mouton i(l) fait comment: bee. (Augustin, 2; 6) The sheep he makes how: beh Hamann explained preference for wh-in-situ by proposing that children opt for the most economical option and adopt wh-movement only later. In addition, she pointed out that the first questions featuring wh-movement did not show inversion of the verb (see also Hamann 2000, Hulk and Zuckermann 2000, Plunkett 2004), an option that is valid in French, although inversion of the verb is also possible in adult French, as shown in (5). (5) French Où est-elle? Where is she? So far, we have concentrated on one aspect of question formation: wh-movement or lack thereof to sentence initial position (or Spec CP). Another aspect concerns movement of the verb or of the auxiliary to a position preceding the subject, that is, to the head of CP. It might not have gone unnoticed that the English wh-question in (2) feature wh-movement, but not auxiliary inversion. In fact, (2) would have to be correctly expressed by (6): (6) English Where does this go?

392

9781441124609_Ch23_Final_txt_print.indd 392

11/20/2012 9:26:18 AM

Acquisition of Syntax

Since the 1960s, an asymmetry in the acquisition of English questions has been observed: English speaking children move the wh-element to the sentence initial position, but have trouble with Subject Auxiliary Inversion (SAI). In Bellugi’s (1971) seminal work, it was proposed that acquisition of SAI proceeds in four stages, which have been investigated since then in several studies (see Stromswold 1990 for an extensive discussion of these stages). (7)

Failure to perform SAI SAI is first applied in positive yes/no questions only. SAI is also applied in positive wh-questions. SAI is applied in both positive and negative questions.

Although the clear cut division in (7) was not always found, SAI challenges English-speaking children that sometimes may perform SAI in some questions more than in others (see Guasti et al. 1995 for evidence concerning the third stage; see Hiramatsu and Lillo-Martin 1998 for criticism). In addition, difficulties with SAI are evident not only through failure to invert the auxiliary, but through failure to express it, as shown in aux-less questions in (8) (see Guasti and Rizzi 1996): (8)

English (Brown, 1973) What that train doing? (Adam, 2; 4)

Inversion of the verb/auxiliary (or movement to C) is problematic in early English, is not preferred in French, but is not neglected in early languages like early German, Dutch, Swedish (Clahsen et al. 1995; Haegeman 1995; Santelmann 1998). An example of how inversion looks like in these languages is provided in (9) from German. (9)

German Wohin faehrst du mit dem Zug? Where go- rs-2sg you with the train? “Where do you go with the train?”

In these languages fronting of the verb to the second position (V2) is not only found in wh-questions, but is generalized to all matrix sentences (also declaratives) and, unlike in English, it is not restricted to the set of auxiliary verbs (do, can, must, be, etc.), but is generalized to all verbs, auxiliaries, and lexical verbs. One or both of these two aspects of the syntax of these languages must be responsible for the fact that inversion of the verb/auxiliary is not troublesome for their acquisition. Evidence from Italian suggests that it is the presence

393

9781441124609_Ch23_Final_txt_print.indd 393

11/20/2012 9:26:18 AM

The Bloomsbury Companion to Syntax

of a special class of auxiliaries that disturbs English learners. In Italian whquestions the subject cannot stay between the wh-element and the verb; it has to stay in a left or in a right peripheral position, as in (10) or be null. Rizzi (1996) interpreted the adjacency restriction between the verb and the wh-element by claiming that in questions the verb moves to the head of CP and the wh-element moves to the Spec of CP, as in English. Interestingly, differently from what happens in English, both auxiliaries and lexical verbs move to C, that is, auxiliaries and lexical verbs behave in the same way. (10) Italian (Il bambino), cosa comprerà (il bambino)? (The child), what  buy-f t-prs-3sg (the child)? “What will the child buy?” (Il bambino), cosa ha comprato (il bambino)? (the child), what has bought (the child)? “What has the child bought?” Italian learners produce questions as in (10) (Guasti 1996) or with null subjects, both with lexical verbs and with auxiliaries. Although Italian is not a V2 language, movement of the verb to C goes smoothly in child language. Hence, what is difficult for English children must be the existence in their language of a special class of auxiliaries with irregular properties with respect to other verbs: only auxiliaries invert in questions and precede negation. On the basis of the data reviewed, we can conclude that if wh-movement is compulsory in the adult language, it is observed in the child language since the earliest productions. If it is just an option, wh-in-situ is initially chosen as the most economic solution. In languages that also feature movement of the verb/auxiliary to C, this is observed since the earliest production, along with wh-movement, unless its application is hindered by independent aspects of the language, as in English. Thus, children acquire smoothly those properties that are regular or always applied (wh-movement), opt for the most economical option if they have a choice, and fail to apply some operations for some time if independent properties make this application difficult. Another structure that features wh-movement of some sort (either of a relative (empty/lexical) operator, or of a relative head; see Haegeman 1994 for the first type of analysis and Vergnaud 1974 and Kayne 1994 for the second) is constituted by relative clauses in (11). (11) English (a) The boy who is pushing the fairy. (b) The boy who the fairy is pushing.

394

9781441124609_Ch23_Final_txt_print.indd 394

11/20/2012 9:26:18 AM

Acquisition of Syntax

Production of relative clauses is observed from 3 years, in particular of subject relative clauses, as in (11a) or of even simpler structures as in (12). (12) English Here is the boy that took the cake. Relative clauses relativizing grammatical functions other than the subject are learned slowly and their full acquisition takes years. One problem observed in several languages is the failure to use relative pronouns and the generalized use of a complementizer, like that in English, to relativize all grammatical functions. An exception are subject and object relatives in French, which are introduced by different complementizers qui and que, respectively. Children use qui when they relativize the subject and que when they relativize the object since 3 years (Labelle 1990) (we return later again to the production of object questions). Like children speaking other languages, they use que “that” to form nonsubject relative clauses. The failure to use relative pronouns in relative clauses relativizing grammatical functions other than subject and object is sometimes coupled with the use of resumptive pronouns (Labelle (1990) for French; Guasti and Cardinaletti (2003) for French and Italian; Pérez-Leroux (1995) for Spanish; Goodluck and Stonjanovic (1996) for Serbo-Croatian; Varlokosta and ArmonLotem (2002) for Modern Greek). Some examples with resumptive pronouns from French and Italian are in (13): (13) French (Guasti and Cardinaletti 2003) (a) L’ ȃne que le cowboy lui donne à boire. The donkey that the cow-boy to-him gives to drink “The donkey to whom the cow-boy gives some water.” Italian (ibid.) (b) Tocca la panca che ci sono seduti i due bambini. Touch the bench that there are sitting the two children “Touch the bench on which the two children are sitting.”

(7; 1)

(7; 2)

Resumptive pronouns are less frequent in early subject and object relatives than in relatives on other grammatical functions. Thus, children produce relative clauses both with gaps and with resumptive pronouns, depending on which function is being relativized. It is worth pointing out that French, Italian, and Spanish (see references above) resumptive relative clauses are common in the sub-standard varieties of the language (see also Fiorentino 1999; Berruto 1980; Sornicola 1981; Blanche-Benveniste 1997). Thus, the use of these kinds of relative clauses by children is not surprising, as it is attested in the input. According to Guasti and Cardinaletti (2003), relative pronouns are learned during the

395

9781441124609_Ch23_Final_txt_print.indd 395

11/20/2012 9:26:18 AM

The Bloomsbury Companion to Syntax

school years, even through explicit teaching. When this happens, resumptive pronouns are dismissed, at least when the standard variety of Italian is spoken. Some examples of the use of relative pronouns are below: (14) Italian (Guasti and Cardinaletti 2003) (a) Tocca la panca su cui il pompiere è disteso. “Touch the bench on which the fireman lies down.” (b) Tocca la scatola in cui si è nascosta la giraffa. Touch the box in which itself is hidden the giraffe “Touch the box where the giraffe has hidden.”

(9; 6) (9; 8)

On the basis of these results and in spite of the fact that children use resumptive relative clauses, one can argue that children produce relative clauses via whmovement; this is evident in subject and object relative clauses, where children use gaps. It is also evident in French where the complementizers used in the two relative clauses are distinct and children choose the appropriate form. As the complementizer qui, present in subject relative clauses, is often taken as a sign that movement has applied (see Rizzi 1990), we can safely conclude that children use the movement strategy to form subject relatives. This proposal can be extend to relative clauses relativizing other grammatical functions if some other assumptions are added. Children’s lexicon does not include relative pronouns, which are learned through explicit teaching during the school years; therefore, the use of the complementizer that is extended to all grammatical functions (French subject relative clauses aside). In so doing, the information about which function is relativized is lost, hence the use of resumptive pronouns, that can be seen as partial copies of the element moved (an operator or the relative head). Thus, we can conclude that children use movement to form relative clauses. Another aspect of children’s formation of relative clauses is the tendency to transform relative clauses relativizing the various grammatical functions into subject relative clauses by exploiting different strategies (French: Labelle 1990; English: McDaniel et al. 1998; French and Italian: Guasti and Cardinaletti 2003). This is most evident for object relatives, especially when the arguments of the verb in the relative clause are both animate. In this case, children older than 6 years use passivization, as illustrated in (15): (15) Italian (ibid.) (a) Tocca il cammello che è stato comprato dal bambino. Touch the camel that is been bought by+the child “Touch the camel that has been bought by the child.”

(9; 3)

396

9781441124609_Ch23_Final_txt_print.indd 396

11/20/2012 9:26:18 AM

Acquisition of Syntax

(b) Touche le cochon que le monsieur est en train de laver. Touch the pig that the man is washing. (c) Touche le cochon qui était lavé. Touch the pig that was washed.

(7; 1)

The same tendency is evident also in genitive relatives, where two subject relative clauses are used. For the target in (16a), children produced the sentence in (16b). (16) Italian (ibid.) (a) Target: Tocca la signora il cui figlio disegna. Touch the lady the whose son draws “Touch the lady whose son is drawing.” (b) Tocca la signora che ha il figlio che disegna. Touch the lady that has the son that draws

(5; 8)

The tendency to transform object relatives into subject relatives has been discussed extensively in Utzeri (2007) and in Belletti and Contemori (2010). In a study on the production of subject and object relatives in Italian, they note that this tendency is not only evident in children but also in adults, although it is much more pronounced in the former case than in the latter. In addition, while adults (and older children) use passivization almost exclusively, as in (15), young children use passivization less and opt for other strategies. For the target in (17a), they produce a structure with passivizing si, as in (17b) or change the verb in a way to make a subject relative possible, as in (18b). (17) Italian (ibid.) (a) Target Il bambino che il re pettina. The child that the king combs (b) Production Il bambino che si fa pettinare dal re. The child that himself makes comb by+the king “The child that makes himself comb by the king.” (18) (a) Target Il bambino che la mamma bacia. The child that the mother kisses (b) Production Il bambino che riceve un bacio dalla mamma. The child that receives a kiss by+the mother 397

9781441124609_Ch23_Final_txt_print.indd 397

11/20/2012 9:26:19 AM

The Bloomsbury Companion to Syntax

Interestingly, when passive starts to be part of the children’s repertoire, which happens between age 5 and 6, passivization of the kind in (15) is also found in the child’s production. Thus, we observe a shift toward subject relatives, mostly studied in the case of object relatives, but present also when relative clauses on other grammatical functions are elicited. A second phenomenon we observe is the development in the structures employed to operate this shift: change of verb or use of Passivizing si in younger children and periphrastic passive in older children and adults. This shift in the formation of relative clauses is likely to reflect Keenan and Comrie’s (1977) accessibility hierarchy and is likely to express the fact that subject relative clauses are easier or simpler than other relative clauses. Friedemann et al. (2009) have proposed that object relative clauses are more difficult than subject relative clauses to acquire because they feature the intervention of an NP—the embedded subject in (19a) between the head of the relative (the boy) and the position in which this head must be interpreted—indicated between angle brackets. This is not so in subject relative clauses, where no element intervenes between the head of the relative and the position of interpretation, as illustrated in (19b). (19) English (a) The boy that the clown is following . (b) The boy that is following the clown. In other words, connecting the head to its position of interpretation violates some sort of locality constraint and the relation in (19b) is more local than in (19a). What we learn from this is that children are able to move element in the sentence and do so, if necessary, but prefer shorter movements and are disturbed by the intervention of other elements in the dependency. To some extent the same holds true for adults.

4 Acquisition of Pronominal Reference A lively topic in the acquisition of pronominal reference is the interpretation of pronouns. Based on studies in a wide variety of languages, it has been established that children from about 3 years do not experience difficulties in interpreting reflexives pronouns, that is, their pairing a sentence like in (20) with a situation in which the bear correctly is scratching himself and not someone else (Chien and Wexler 1990). (20) English The bear is scratching himself. 398

9781441124609_Ch23_Final_txt_print.indd 398

11/20/2012 9:26:19 AM

Acquisition of Syntax

This finding has been replicated in many languages (French, Hebrew, Icelandic, Italian, and Spanish; Jakubowicz 1984; Samper-Padilla 1990; Mckee 1992; Sigurjónsdóttir and Hyams 1992; Grodzinsky and Kave 1993) and it is valid regardless of the form of reflexive pronouns: weak pronouns occurring in the position in which an NP occurs, as in the English example in (20) or clitics that precede the verb, as in the Italian translation of (20) given in (21): (21) Italian L’  orso  si gratta. The bear himself  scratches “The bear scratches himself.” Children not only comprehend sentences including reflexives pronouns correctly, but they produce correct sentences including them. Bloom et al. (1994) established that 2–5 year old English learners use myself when the antecedent is I to express an action directed toward the speaker. Thus, principle A of the binding theory, which governs the legal interpretation of reflexive pronouns, does not seem to cause problems to children. Nonreflexive pronouns (pronouns from now on), ruled by principle B of the binding theory, present a different picture. In several studies on various languages it has been established that children interpret pronouns in a nonadult way up to age 5 or 6 (let us call these nonadult interpretations “principle B errors” for convenience): a sentence like (22) is often paired with a situation in which the bear is scratching himself. (22) English The bear is scratching him. In other words, it seems that (22) is interpreted as (20), with the pronoun referring to the same individual denoted by the nominal expression (anaphoric interpretation), rather than referring to a character not mentioned in the sentence, but present in the extralinguistic context (deictic interpretation). This finding was first discussed in Jakubowicz (1984), then in Chien and Wexler (1990) for English (see also Grimshaw and Rosen 1990; McKee 1992; Kiguchi and Thornton 2004), but it was discovered in other languages as well, besides English: Dutch (Baauw 1999) and Icelandic (Sigurjónsdóttir and Hyams 1992). Adult controls were present in some studies, but did not misinterpret pronouns. Depending on the study, the nonadult interpretation varies from 29 percent to 82 percent and, according to Conroy et al. (2009), part of this variability is due to the different tasks used. Another source of variability may be the age range of the children who participated in the studies 2; 6 to 6; 6 years. Since there is development in this age range, it is possible that higher percentages of errors 399

9781441124609_Ch23_Final_txt_print.indd 399

11/20/2012 9:26:19 AM

The Bloomsbury Companion to Syntax

are found in those studies in which younger children took part. This proviso notwithstanding, in several studies principle B errors are not found when the antecedent is a quantificational element or a wh-element, as in (23): (23) English Every bear is scratching him. I know who scratched him. In these cases, children took the pronoun to refer to the same salient character in the extralinguistic context, that is, they assign a deictic interpretation to the pronoun. The nominal–quantificational asymmetry (NQA) was found in some studies on English (Chien and Wexler 1990, Avrutin and Thorton 1994) and Dutch (Philip and Coopmans 1996). Other studies also on English did not report this asymmetry (Kaufman, 1988; Grolla 2005). In sum, pronouns are often interpreted as reflexive pronouns, but an asymmetry is often found depending on the nature of the antecedent (referential, as in (22) or quantificational, as in (23)). The NQA was questioned in Elbourne (2005) and was experimentally investigated in Conroy et al. (2009). These authors claimed that the asymmetry is largely due to experimental shortcomings, one being the fact that the scenarios used to test sentences including referential antecedents (22) and quantificational ones (23) were not the same as in previous studies. This difference in the set up impacts on the availability of the antecedent for a pronoun; it turns out that in the quantificational context used in previous studies, the deictic interpretation of a pronoun is favored with respect to the referential context. In a series of three experiments, Conroy et al. (2009) tested the interpretation of pronouns taking care of this flaw. In order to do this, they used the same stories for sentences including a referential and for sentences including a quantificational antecedent and had the main character be the intended deictic referent in both conditions. With these changes and others that we will not discuss, children made only about 10 percent principle B errors in both the referential and the quantificational contexts, a percentage expected in experimental tasks. Thus, the NQA is no longer evident when the context and thus the accessibility of the antecedent is controlled for. Conroy et al. (2009) also showed that when they made their contexts more similar to those found in previous studies, the contrast between nominal and quantificational contexts was evident again, with 56 percent acceptance of the anaphoric interpretation of pronoun in nominal contexts and only 16 percent in quantificational ones. Thus, children in controlled experimental settings interpret pronouns correctly and do not take them to be anaphoric to the local antecedent. These results are so discrepant with a large body of evidence showing the contrary that the question why English learners (like their peers speaking Dutch or Icelandic) are highly susceptible to errors in the interpretation of pronouns cannot be dismissed by simply appealing to methodological flaws. 400

9781441124609_Ch23_Final_txt_print.indd 400

11/20/2012 9:26:19 AM

Acquisition of Syntax

This concern finds its rationale in three observations. First, children do not display any problem in the interpretation of reflexives pronouns, as we saw above, regardless of the methodology employed in the experiments. Second, children do not err in the interpretation of pronouns in contexts ruled by principle C, as in (24), that is, they do not take he to be anaphoric with the clown (English: Crain and McKee 1985; Italian: Guasti and Chierchia 1999/2000; Kazanina and Phillips 2001; Leddon and Lidz 2005). (24) English He was playing guitar, while the clown was dancing. Third, not all kind of pronouns give rise to principle B errors. Cardinaletti and Starke (1999) discuss a typology of pronouns and show that pronouns are of three types: clitics, weak, and strong pronouns. Clitics are the most deficient elements, they are heads and need to move to a position adjacent to the verb. They cannot occur in the position in which NPs are found. Weak pronouns are less deficient than clitics, as they are XPs and can occur in positions occupied by NPs, and finally strong pronouns are not deficient. Interestingly, in English and similar languages, pronouns are ambiguous between being strong or weak. As strong pronouns, they can be focalized and used deictically. This is not possible with clitics. Now, it is interesting to note that principle B errors are found in those languages in which pronouns are ambiguous between being weak or strong pronouns, but not in clitic languages, even when the contexts is not taken care of. For example, in French (Hamann et al. 1997), Spanish (Baauw et al. 1998), Italian (McKee 1992), and Greek (Varlokosta 2000) children make principle B errors about 10 percent of the time. In clitic languages, no NQA is observed to the extent that both referential and quantificational antecedents were tested and this was done in French and Spanish. Finally, McKee tested Italian and English speaking children with the same stories and found that only English children made principle B errors. Taking these observations together, we are led to the conclusion that the experimental set up matters, as suggested by Conroy et al. (2009), but also the formal status of pronouns matters. It is likely that the ambiguous status of pronouns impacts on their interpretation and on the way pronoun resolution is solved (as suggested by Conroy et al. 2009). In online studies on adults’ interpretation of pronouns, it has been established that comprehenders temporarily consider illicit antecedents for pronouns (Badecker and Straub 2002) and as the sentence unfolds they abandon the ungrammatical candidates. Children are less prone to abandon an initial interpretation (Trueswell et al. 1998) and thus are led to make principle B errors, that is, unlike adults they sometimes do not abandon ungrammatical candidates. Since children speaking clitic languages do not err, we must conclude that the processing of clitics is somehow different from that of weak pronouns 401

9781441124609_Ch23_Final_txt_print.indd 401

11/20/2012 9:26:19 AM

The Bloomsbury Companion to Syntax

or that illicit antecedents are not considered in these languages, a conjecture that has never been tested so far.

5 Conclusion We examined the acquisition of word order, displacement, and locality. Children obey word order since their first multi-word combinations, as they exploit phonological properties to figure out whether the language is head-complement or complement-head. Children move elements, for example, in forming questions and do so if this is the only option, but they opt for the non-movement strategy whenever available. Some aspects of the language may interfere with the acquisition of movement, as the presence of an idiosyncratic class of auxiliaries in English, which is responsible for difficulties in subject auxiliary inversion. Although movement is applied in relative clauses, the use of relative pronouns takes time, as a more economical option exists, which consists in the generalized use of the complementizer. Also relativizing grammatical functions other than subjects is costly, as the subject in the relative clause may interfere in the relation between the relative head and its copy in the position of interpretation. Finally, the interpretation of weak pronouns is hindered by their ambiguity and their correct interpretation is initially evident only in supportive contexts.

Note 1. I would like to thank Carlo Cecchetto, Francesca Foppolo, and Angeliki Zachou for comments on previous versions of this article. The preparation of this article has been supported by PRIN 2007 “Grammatical and Semantic Features in Acquisition.”

402

9781441124609_Ch23_Final_txt_print.indd 402

11/20/2012 9:26:19 AM

24

Usage and Syntax Paul J. Hopper

Chapter Overview Usage and the Study of Syntax On-line Grammar Example of TCUs Syntax and Speaking Practices Conclusion

405 407 408 409 420

1 Usage and the Study of Syntax The introduction of usage into a consideration of syntax moves the starting point away from the internal, cognitive, representational view of sentence structure (i.e. competence) to the existential facts of interactive discourse (performance). This revised perspective means facing the challenge of giving a formal account of syntax while retaining a focus on the real-time unfolding of discourse. Usage-based approaches to linguistic structure have, in a certain sense, been around for a long time. Earlier work, such as the articles in Li (1974, 1976) and Givón (1979), consisted basically of sentence-level studies of the pragmatics of word order in the simple sentence and of the relationship between the subject and the topic in isolated sentences. Often, manufactured discourses were used in order to illustrate contextual features of grammar. A number of scholars associated with the American missionary organization Summer Institute of Linguistics had already in the 1960s and 1970s ventured into discourse and grammar as a tool of Bible translation (Longacre 1976; Grimes 1975). When technologies such as optical scanning, digital recording, and high-speed search software became available in the 1990s, it was possible to create large corpora of spoken, interactive languages, and the face of discourse and grammar studies

405

9781441124609_Ch24_Final_txt_print.indd 405

11/20/2012 9:24:57 AM

The Bloomsbury Companion to Syntax

changed radically. The present chapter will deal exclusively with recent studies of the relationship between grammar and spoken conversational discourse. A useful survey is that of Ochs et al. (1996). The study of syntax from the point of view of conversational interactions diverges from “textbook” syntax based on imaginary sentences. While our graphic representation of syntax necessarily depicts utterances as completed entities, recent usage-based approaches depend crucially on temporal sequencing such that the exact course of an utterance is not precisely known in advance of its production. It is not just that the hearer is presented with a novel utterance, but rather both speaker and hearer are embarking on new territory. The speaker does not formulate an utterance in advance, but rather unfolds it in real time. The bird’s eye view of a sentence, which permits the beginning, middle, and end to be simultaneously accessible, is not available to either the speaker or the hearer. The speaker is guided in his or her production by sedimented routines of talking, and the hearer is similarly alerted in advance to the likely course of an utterance by cues in the speaker’s utterance. These routines are not uniformly in the nature of “grammar” (syntax), but consist largely of short prefabricated phrases. Consequently the tasks of syntax are framed in terms of identifying and describing linguistic resources for managing the flow of speech and giving and receiving cues for interactions among speakers. A second feature of usage-based syntax is that usage can only be studied through actual examples of speech. Mental (i.e. made-up/fictional) sentences are excluded from consideration as data. Consequently, questions of data sources and techniques of transcription assume major prominence. Syntax from the perspective of usage includes not merely the arrangements of conventional words common to written and mental sources, but markers of hesitation, reaction, laughter, and many other forms that are relevant to an interaction. Some of these markers index low-level phonetic phenomena such as glottal stops, drawls, and exhalations, any of which could be a crucial clue to how the turn at talk is to be interpreted. Finally, while speech can be monologic, dialogue is a far more frequent and characteristic setting for human language. Dialogue takes place through orderly alternations among speakers known as turns at talk, or simply turns. The identification of places in a conversation where turns change has resulted in an elaborate theory of interaction, known as interactional linguistics (Couper-Kuhlen and Selting 2001). The turn constructional unit (TCU), the short complete utterance type characteristic of active conversational dialogue, is the fundamental building block of spoken discourse. Syntax is subservient to the TCU and is framed by it. The TCU begins at the point where the speaker has gained the floor and ends when the speaker’s right to the floor has expired, unless this right is renewed by the speaker. The TCU is held to be a universal feature of

406

9781441124609_Ch24_Final_txt_print.indd 406

11/20/2012 9:24:57 AM

Usage and Syntax

human dialogic interaction, which takes place through the orderly assignment of turns at talk. Linell (2010) is a recent argument for the foundational nature of dialogic discourse to all human thought. Syntax and interaction are inextricably linked: “Grammar is not only a resource for interaction and not only an outcome of interaction, it is part of the essence of interaction itself. Or, to put it another way, grammar is inherently interactional” (Ochs et al. 1996: 38). Current theory postulates that the study of grammatical structure in language begins with generic problems of interaction (Sidnell 2009: 16–23). From this point of view, syntactic structure is a secondary by-product of kinds of interaction, and to begin with the autonomous calculus of grammatical forms is to doom oneself to a vicious circle of formalisms. Usage-based grammar requires that we first attend to the types of interactions that occur between speakers and study the possible forms that such interactions take. These forms will constitute the grammars of the interactions, and the grammar of the language is the aggregate of such interactional grammars.

2 On-line Grammar As the name suggests, “on-line” grammar replaces synchronic syntax with a perspective of the unfolding of structure as a spoken interaction progresses through time. Such developing structure is known as sequencing (Schegloff 2007). A brief example will be considered. Like other examples of spoken language in this article, the data are from the Santa Barbara Corpus of Spoken American English (Du Bois et al. 2001–4). In the following sequence, the speaker has told how the air was so thick during a dust storm that the headlights of cars were green: (1) (a) (b) (c) (d) (e)

. . . Yeah. they just looked green. . . . It was a wei=rd. . . . ugly. . . . ugly day.

She sums up her story with three comments (lines c–e). These are marked with utterance-final intonation (transcribed with a period/full stop). In a standard syntactic analysis the intonations and the turn-completions would be edited out and the sequence would be analyzed as something like: It was a weird, ugly, ugly day. The predicate noun phrase would be said to have a structure approximately:

407

9781441124609_Ch24_Final_txt_print.indd 407

11/20/2012 9:24:57 AM

The Bloomsbury Companion to Syntax

NP

Article

Modifier Phrase

Adj

a weird

Noun

Adj

Adj

ugly

ugly day

But to analyze the noun phrase in this way would be a mistake. For ugly in line d escalates weird in line c, and the repetition of ugly in line e confirms the substitution. The two adjectives look the same lexically, but they do different work. They enter at different times and with different pragmatic functions. Yet the indefinite article a in line c projects a noun day that comes in at line e; the interlocutor, on hearing a, now anticipates a noun that will resolve the projection and fulfill the formula a^MODIFIER^day. Are we entitled to say, then, that there is a Noun Phrase (NP) [a weird ugly ugly day]? Yes, but it is an emergent NP, one that exists as a structural constituent only after the fact of its fulfilled utterance, and that retrospectively creates an ad hoc formula [a^MODIFIERi^MODIFIERj^ MODIFIERj^day]. Such formulae are not novel, that is, they are not thought up de novo in every instance. Instead they are modeled analogically on phrases that are actually remembered in a form that is identical to or very similar to previously used and previously heard phrases, such as a weird day and an ugly day.

3 Example of TCUs Consider the following exchange: (2) (a) (b) (c) (d) (e) (f)

ALICE:

. . . Y- % – [Did you take] the Tylenol, ANNETTE: [XX] ALICE: like I told you to yester[2day2]? ANNETTE: [2Yeah2]. I took some this morning then [3too3].

408

9781441124609_Ch24_Final_txt_print.indd 408

11/20/2012 9:24:57 AM

Usage and Syntax

(g) ALICE: [3Okay3]. (h) ANNETTE: Just to make sure. Some of the transcriptional conventions of the corpus are present here: the percent sign (%) is a glottal stop; square brackets, which may be identified by numbers, indicate overlapping speech; XX is a sequence that could not be clearly heard by the transcriber. Pauses and breaks of various lengths are written with hyphens (-). Commas, question marks, and periods (full stops) mark different kinds of intonation. Each line is a turn constructional unit (TCU), that is, each line comprises a turn at talk. Often, the TCU consists of a token such as “yeah” (line e), “okay” (line g), or even a failed turn “Y-%” (line a). However, TCUs may resemble the clauses and even full sentences of written language (for example line b and line f), a point to which we will return later. From the point of view of sentence syntax, many TCUs have the appearance of being incomplete, like lines e and h. The units of significant meaning include the many functional signals of a completed turn. In line 1, for example, Alice begins (presumably) to say “yes,” then starts again getting no further than a phonetic closure (%). At this point Annette is entitled to her own turn.

4 Syntax and Speaking Practices TCUs are identified through a combination of prosody, pragmatics, and syntax (Ford and Thompson 1996; Selting 2005). They exist by virtue of their potential for offering interlocutors orderly opportunities to present their own turns. The syntax of TCUs accommodates a number of apparently universal strategies for creating dialogical discourse, the most important of which will now be discussed.

4.1 TCUs and Clauses TCUs often resemble the clauses and sentences familiar from theoretical syntax. In the example given above, it would be easy to derive normatized complex sentences consisting of a main and a subordinate clause: (3) (a) Did you take the Tylenol like I told you to yesterday? (b) I took some this morning then too, just to make sure. Such a procedure would reverse the priority of the spoken versus written versions and claim a secondary status for the spoken version. While there is some 409

9781441124609_Ch24_Final_txt_print.indd 409

11/20/2012 9:24:57 AM

The Bloomsbury Companion to Syntax

reality for sentences in this sense, speakers do not formulate complete sentences in advance of their production, but are instead guided by the interlocutor in the task of integrating shorter utterances into TCUs and, eventually, longer, syntactically more complex forms. Thus in the extract given, the “subordinate clauses” like I told you to yesterday and just to make sure are prompted by Annette’s “XX,” whose (undecipherable) content opens a turn space for Alice, and Alice’s “okay.” But the term “subordinate” is here a misnomer, since (i) there is no special grammatical status for the two TCUs line e and line h, and (ii) lines e and h are as fully informative as the presumed “main clauses” lines b and f. That is, while lines e and h are in a sequence, they are not in a hierarchical relationship to one another. To take usage-based syntax seriously, then, is to understand the starting point for an analysis to be the conversational interchanges themselves, rather than extrapolated syntactic content matched against putative written counterparts.

4.2 “Subordination” in Usage-based Syntax The types of subordination that are grouped together in sentence syntax are sometimes revealed to be quite separate when their discourse contexts are studied. Two instances of this are complementation and relativization.

4.2.1 Complementation Syntacticians recognize several different types of constructions that are grouped under the heading of complements. These include nonfinite clauses such as infinitives and gerunds, as well as full clauses that resemble sentences but function as subjects and objects of predications. It is this latter type that we will be considering here. The most common kind of complement clause that occurs in spoken English is that in which the complement clause serves as a direct object, as in “John declared that his refrigerator was being bugged.” Sentence syntax analyzes such constructions as consisting of a main clause (“John declared”) and a direct object clause “that his refrigerator was being bugged.” As argued in Thompson (2002), however, there is actually little reason even at the sentence level to analyze the subordinate clause as a direct object. The substitutability of the subordinate clause by a pronoun such as “it,” which is sometimes cited (“John knows that they are discussing him”/”John knows it”), is not possible with a wide range of main clause verbs such as “figure” and “declare.” Other standard tests for direct object status such as Passive likewise often fail to apply to complement clauses. When considered from the point of view of spoken usage, there is even less evidence for direct object complements. Many cases are of the following kind: 410

9781441124609_Ch24_Final_txt_print.indd 410

11/20/2012 9:24:57 AM

Usage and Syntax

(4) (H) I phoned her and said, . . . I don’t know whether I can sit in the chair at a=ll or not. Here, the main clause would at first seem to be “I don’t know” and the complement clause “whether I can sit in the chair at all or not.” However, on closer consideration it is clear that the speaker is not asserting a state of ignorance about her ability to sit in the chair, but is rather expressing uncertainty. Moreover, the main point she is conveying to the listener is about the chair and whether she can sit in it rather than about her own knowledge. Very many such main clauses are of this type. In the next example, two people are discussing a sleeping dog: (5) ALICE: I wonder if we should wake her up? MARY: . . . No, . . . she’ll get scared and want to go outside. Mary’s negative response is directed at the question of waking up the dog, not at Alice’s mental state of uncertainty reflected in her “wonder.” We may conclude then that the clauses following verbs like “know” and “wonder” are neither objects nor are they really subordinate, since the focus of information is normally on the “subordinate,” not the “main,” cause. This conclusion is reinforced by the observation that the number of different verbs that occur in transcripts in the “main clauses” of such constructions is quite limited. Almost always these verbs present an evaluative, epistemic, or evidential stance (Thompson 2002), and this fact suggests that what at first sight seemed to be a main clause+complement clause construction is better analyzed as a fusion of an introductory phrase such as “I wonder” or “I think” with a new sequence. Invariably, such first-person phrases are in the same TCU as the following “complement.” They are in the nature of frozen idioms rather than of syntactically analyzable subject+verb phrases. They are in fact not separate clauses but rather modal prefixes that precede the principal statement (the “complement clause”), and project their scope in terms of degree of certainty or evaluation.

4.2.2 Relativization A similar fusing of elements is found in relative clauses when these are examined in natural conversation. Object relative clauses such as: (6) (a) There’s like one lemon left on this tree that I can reach. (b) and they keep calling me on the— (H) their stupid intercom system that you can hardly understand X. 411

9781441124609_Ch24_Final_txt_print.indd 411

11/20/2012 9:24:57 AM

The Bloomsbury Companion to Syntax

(c) Are you sure we’re doing work6 from the same class that I’m— that I’m— that I go to every night. (d) (H) . . . I’ve also made them car loans, that they’ve never missed a payment on, in which “that” is the object of the relative clause verb (6a & b) or is a prepositional complement (6c & d) are found both with, as in these examples, and without an overt pronoun, as in the following: (7) (a) it sounds like the tape I made last night (b) Do what you want with the time you have (c) This party I went to Friday night, where Joy was jamming on that harmonica? Relative clauses appear in conversation both with and without a relative pronoun. Relative clauses introduced by a pronoun are distinctly separate clauses: (8) (a) The force of the water turns the turbine, which turns the rotor in the generator, (H) and generates power. (b) So we have this frozen horse hoof, that we have to start out on, cause you don’t want to . . . cripple up a . . . (H) really good horse, In each of the examples in (7), a relative pronoun—“which” in (a), “that” in (b)—introduces a clause that fills a TCU. The prosody, with the relative clauses set off by comma intonation, supports the analysis of the relative clauses here as distinct grammatical entities, with head nouns (“turbine,” “hoof”) and relative clauses in separate sequenced TCUs. In other examples of relative clauses, however, it is less obvious that the head noun and the relative clause are separately distinguishable. This is especially true of object and prepositional-complement relatives, where the pronoun is optional: (9) (a) Everything’ll mutate into something you can eat eventually. (b) I’m really happy with the stuff= . . . I got now, 412

9781441124609_Ch24_Final_txt_print.indd 412

11/20/2012 9:24:58 AM

Usage and Syntax

Here, “something you can eat” is less clearly a construct of a head noun and relative clause, and “I got now” forms a close unit with “the stuff.” As noted in Fox and Thompson (2007), from the point of view of usage it is evident that: (1)

(2)

(3) (4)

There are several kinds of object relative clauses that are not distinguished in sentence syntax. These include not only the well-known distinction between restrictive and nonrestrictive clauses, but other types also. The most common type of relative clause does not function as a descriptive modifier of a head noun, but serves as a limiting or identifying phrase attached to a very general word that is usually a pronoun or quantifier such as “one,” “all,” “something,” etc. The “relative pronoun” in object relative clauses is almost always absent. Relative clauses should be seen as manifesting a continuum of biclausality, at one end of which is a tight integration into a more or less formulaic phrase, at the other a loosely attached incremental phrase that elaborates an already complete nominal phrase. The absence or presence of a relative pronoun, and also the restrictive/nonrestrictive distinction, reflect the relative clause’s location at the more-integrated vs. the less-integrated ends of the continuum respectively.

4.2.3 Biclausal, Constructions Relative clauses and complement clauses are part of the general grammar of biclausal constructions, ones that consist of a main clause and a subordinate clause. Among the syntactic constructions that are biclausal, besides relative and complement clauses, are: conditionals, adverbials of various kinds, and cleft and pseudoclefts (also known as wh-clefts). As with relative and complement clauses, an examination of usage shows that there are greater and lesser degrees of integration between the two clauses. In German this difference shows up in the word order. At the most integrated level, spoken German follows the same convention as the written language, that is, the subordinate clause has verb-final word order and the main clause verb second. If the subordinate clause comes first, “verb second” means that the finite verb of the main clause (often an auxiliary) must come before the subject. Take a German conditional sentence as an example (Günthner 1999): (10) German Wenn ich den sehen würde, würde ich ihn direkt darauf ansprechen. if I him see were:to, would  I him directly about:it speak:to “If I were to see him, I would ask him about it right away.”

413

9781441124609_Ch24_Final_txt_print.indd 413

11/20/2012 9:24:58 AM

The Bloomsbury Companion to Syntax

The finite auxiliary (in bold face) is in the second place, preceded by the ifclause, with the subject ich “I” following. In such utterance sequences, the protasis (if-clause) sets a truth-value condition on the apodosis (then-clause). However, in spoken German it often happens that the word order of a protasis resembles that of an independent main clause, that is, the finite verb follows the subject noun phrase: (11) German Wenn du if you

luscht hasch und=zeit desire have and time

en kindergottesdienst wir machen morgen we make tomorrow a children’s:service in der lutherkirche. in the Lutheran:church “If you’d like to and have time, tomorrow we’re having a children’s service in the Lutheran church.” The then-clause (apodosis) “ignores” the if-clause (protasis) and presents the normal main clause subject-verb word order. Here the relationship protasis– apodosis is looser. The protasis does not set a truth condition on the apodosis (the children’s service is not logically dependent on the addressee’s assent). Instead, as Günthner notes, the protasis “presents the conditions under which the following utterance is relevant to the interlocutor” (Günthner 1999: 5). The scope of the protasis often goes well past the following clause. Indeed, sometimes the directly following clause must be skipped over in order to ascertain a relevant sequence: (12) (a) JENNIFER: . . . if you take all the points. (b) All thirteen hearts, (c) and the [queen of diamonds, (d) DAN: [So that’s twenty-six points]. (e) JENNIFER: (H) it’s called], (f) right. (g) It’s called shooting the moon, (h) and you get – (i) . . . you give everyone else .. twenty-six points, (j) . . . or else you’re minus twenty-six points. The protasis in lines a–c is separated from the apodosis in lines g–j by the “parenthetical” sequence in lines e–f. What in written syntax is an integrated two-clause sequence, with an if-clause setting up truth value conditions for 414

9781441124609_Ch24_Final_txt_print.indd 414

11/20/2012 9:24:58 AM

Usage and Syntax

a then-clause, is seen in natural spoken syntax to be a functionally disjunct sequence with one utterance being loosely relevant to the interpretation of a subsequent, not necessarily adjacent utterance. Consider now another biclausal construction of English, the so-called pseudocleft or wh-cleft (Quirk et al. 1985: 1387–9), illustrated through sentences like: (13) What they wanted was a house of their own. Pseudocleft sentences consist of an initial clause, called the wh-clause, and a second clause introduced by is or was. Such sentences are seen as an alternative version of a single-clause transitive sentence: (14) They wanted a house of their own. In the pseudocleft version the verb (here, “wanted”) receives a strong secondary focus, and there is a primary focus on the direct object (“a house of their own”): (15) What they wanted was a house of their OWN. The discourse factors conducive to this pattern of stresses have been studied by Prince (1978) for written and edited spoken texts. However, an examination of interactive discourse suggests a different picture of the pseudocleft. The most striking difference is that in its most typical form, the construction is by no means biclausal. Instead, the wh-clause appears alone, carrying out a variety of interactive tasks (Hopper 2001). These include taking up a new theme that follows on from a previous sequence, claiming a longer turn, and drawing attention to an upcoming significant sequence. The “second clause” of the written and formal registers (here: “. . . was a house of their own”) is not necessarily a clause at all but is simply the continuation of the discourse opened by the whpiece (“what they wanted”). In the following example, Sharon, a grade-school teacher, is talking about ways of dealing with large classes of children, in particular mixed third- and fourth-graders (8- to 9-year olds): (16) (a) SHARON: (b) (c) (d) (e)

well, what you do with those third-graders, you know, is you just like, (H) take them, 415

9781441124609_Ch24_Final_txt_print.indd 415

11/20/2012 9:24:58 AM

The Bloomsbury Companion to Syntax

(f) and put them, (g) you know, (h) with one of the smarter fourth-graders, (i) who’s very [ver]bal, (j) CAROLYN: [uh]. (k) SHARON: and (-) and well-beha=ved. In line c, Sharon uses a wh-clause to reintroduce the theme of Coping with Third Graders, and then suggests a solution. The follow-up, instead of the single clause found in the model pseudocleft of grammatical descriptions, is an elaborate discourse segment that continues for several turns. There is no single focused element; instead, Sharon uses the wh-clause “what you do with those third-graders” in a way that resumes the topic and launches a recommendation. There are many examples of this sort of thing in Hopper (2001) and Günthner and Hopper (2010). As is often the case with “biclausal” constructions, the single clause follow-up is simply one of a variety of possible continuations from the wh-clause. In the following example: (17) But what they did bury in that freshly poured concrete, was one-inch steel water pipe. the pseudocleft is similar in form to the model pseudocleft, that is, there is a double focus on the verb (did bury) and the direct object noun phrase (one-inch steel water pipe). The speaker, the tour guide at a dam, here addresses the rumor that the bodies of dead workmen were buried in the concrete. Planned discourses of this kind have made the noun phrase complement pseudocleft the prototype for the construction in syntactic studies. In spoken language, however, it is quite rare, and is simply an unprivileged possibility. At the other end of the scale of integration, the pseudocleft is little more than a prefix to a TCU, as in line c of the following: (18) (a) (b) (c) (d)

. . . The rotor and the . . . (H) runner, . . . have come out of our A-Seven generator, what happened we lost a water seal down there. (H) So we had to pull those out to fix that water seal,

Here, “what happened” serves to introduce the explanation in lines c–d for a fact mentioned in the previous TCUs (lines a–b).

4.2.4 Incrementation Speakers frequently expand an already complete TCU by adding something further. In the following example, 416

9781441124609_Ch24_Final_txt_print.indd 416

11/20/2012 9:24:58 AM

Usage and Syntax

(19) (a) (b) (c) (d)

. . . the wind was blowing real hot when I first got up. . . . And it was early. . . . Real early, like seven o’clock in the morning,

the speaker adds to her statement in line b two further expansions (lines c & d). Such expansions of a TCU are known as increments (for recent discussion of increments, and relevant literature, see Ono and Couper-Kuhlen 2007; Auer 2007). Typically, increments occur after a TCU has been completed and has not been taken up by an interlocutor (Ford et al. 2002: 19). Increments are said to be retractive in that they depend on and modify a preceding TCU (Auer 2007). They often take the form of a single noun phrase or prepositional phrase, and indeed in Ford et al. (2002) increments are defined as nonclausal expansions. Other researchers have however seen no reason to limit increments in this way. In the following dialogue, for example: (20) (a) (b) (c) (d)

. . . You know, but you can only actually keep them over, for like five minutes. Cause they have like buses to catch and stuff,

line d can be characterized as an increment, that is, as a retractive expansion of a previous TCU. Here, a teacher is explaining that unruly pupils can only be detained after school for short periods. Line d is added as an expansion (in this case, as often, an explanation) for line c. The period (full stop) punctuation of line c, which is the transcriptional sign for a prosodic fall, indicates that the TCU has closed, so the speaker’s addition of line d constitutes a continued turn. Increments often give evidence of emergent constituentization, cf. Helasvuo 2011 (this volume). Consider Sharon’s discourse about dealing with third- and fourth-graders, in which she says: (21) (a) (b) (c) (d) (e) (f)

and put them, you know, with one of the smarter fourth-graders, who’s very [ver]bal, CAROLYN: [uh]. SHARON: and (-) and well-beha=ved.

After introducing the noun phrase “one of the smarter fourth-graders” in line c, Sharon adds an elaboration “who’s very verbal” in line d, and increments 417

9781441124609_Ch24_Final_txt_print.indd 417

11/20/2012 9:24:58 AM

The Bloomsbury Companion to Syntax

this, after Carolyn’s reactive token in line e, with a further elaboration (line f) “and– well-behaved.” The end product would be seen in standard syntax as a restrictive relative clause: (22) One of the smarter fourth graders who’s very verbal and well-behaved. Yet the relative clause has been delivered piece by piece, supported by Carolyn’s “uh,” with comma intonation after each TCU and a pause in the middle of line f. Since Sharon does not presumably have a specific child in mind in line c, the relative clause would have to be understood as limiting the domain of the head noun to the set of verbal and well-behaved children, that is, as a restrictive relative clause. This consideration brings us back to the point made previously: that the relationship of a relative clause to its antecedent is not dichotomized into restrictive and nonrestrictive, but is on a continuum of looser to tighter integration with its head.

4.2.5 Projection An earlier part of a TCU projects a later part or a subsequent TCU. Projection— the anticipation and thus prestructuring of a segment of discourse—is an important methodological concept in the analysis of spoken grammar (Liddicoat 2004; Auer 2005; Hopper and Thompson 2008). In the following extract, (23) (a) (H) When we sell Traveler’s Checks, (b) . . . we put that money in a clearing account, line a with its “when” clause projects line b, and in line b the speaker’s use of “put” projects a directional adverb or adverbial expression, here “in a clearing account.” The interlocutor is prompted by the introductory “when” of line a to expect a follow-up like that of line b; the interlocutor does not expect a request for the time of day, and indeed would understand such a request as a deferral (nonresolution) of the projection. The interlocutor would further, on hearing “we put,” anticipate a direct object (resolved by “that money”) and a directional adverbial (resolved by “in a clearing account”). The speaker similarly, knowing the conventions for constructing coherent sequences, builds his/her turn on the basis of this knowledge, confident that the interlocutor will make the same projections in comprehending. The term projection in conversation analysis originally referred to the combination of semantic, syntactic, and prosodic resources that alerts listeners to the end of the current speaker’s turn at talk. In recent years, however, the term has been used in a more general sense, the strategies for foreshadowing upcoming discourse (Auer 2007). Projection is closely associated with the grammar 418

9781441124609_Ch24_Final_txt_print.indd 418

11/20/2012 9:24:58 AM

Usage and Syntax

of spoken language, in fact the grammar of a language could be said to comprise the conventions for signaling projection, since these conventions are what enable the construction and comprehension of sequences to take place. Without it, utterances would either be detached from a communicative event entirely, or utterance sequences would be predetermined and devoid of communicative value (Auer 2005). Projection is what makes verbal communication an open and collaborative affair. As participants develop a sense of where the discourse is going, they can tacitly mold it, allow it to continue, harmonize with the speaker’s goals, interrupt it with their own contribution, offer supportive tokens of various kinds, and predict when their turn will come. The scope of a projection can be local, that is, short range, or more extensive. Short range projections are quite exact, and are made possible by idiomatization and by conventional phrasing. An indefinite article in English projects a noun, sometimes across an adjective or noun modifier. Examples are “an oldfashioned name” or “an out of town team,” where the indefinite article projects a noun modified by a fixed expression such as “old-fashioned” or “out of town” that goes with the noun. Formulaicity is an important part of projection. Speakers’ familiarity with a store of expressions makes it easier to anticipate and therefore to understand the course of an utterance. Another example: (24) KRISTIN:

. . . (H) Well let’s . . . kind of figure out a cla=ssic, u=m, %= diet, uh, . . . quota, % % @for @you.

The speaker works out her noun phrase from a starting point of an indefinite article and an adjective (“a cla=ssic”) and finally settles on the noun she wants; the interlocutor is prompted to listen for an upcoming noun, and waits for the speaker to signal that she has made her choice. Projection also works on wider stretches of discourse. The pseudocleft that was examined earlier presents a good example. The wh-clause that forms its starting point could, in written syntax, introduce any verb, such as “what they recovered . . . ,” “what they decided . . . ,” and so on, and such verbs do occur in spoken English. However, an overwhelming majority of recorded pseudoclefts contain one of the two verbs “do” and “happen” (Hopper 2001). The verb “do” projects a single action by a human agent or agents as the theme of the subsequent discourse: (25) . . . (H) What I have to do is, submit all of this to our accounting department, 419

9781441124609_Ch24_Final_txt_print.indd 419

11/20/2012 9:24:58 AM

The Bloomsbury Companion to Syntax

“Happen,” on the other hand, projects a longer and more complex sequence, one that could be characterized as an event: (26) RICKIE: REBECCA: RICKIE: REBECCA: RICKIE:

REBECCA: RICKIE:

. . . W- what will happen is, . . . you’ll walk into the courtroom here. . . . [Mhm]. [There are door]s right here. [2Yeah2]. [2(H)2] There are seats right here, . . . u=m, . . . that are for the audience, but normally there’s nobody in the audience but, [Okay]. [your] husband will be there, (H)= u=m, . . . (TSK) a=nd=, . . . then, . . . there are two tables right here. . . . I’ll be seated at this table.

Here, as typically, there are more participants, and more things happening, intertwined with states. Speakers use “do” for shorter range projections and “happen” for projections that either involve nonhuman agents or ones that are more complex and demand more attention and which moreover warn the interlocutor that extended turns are needed.

5 Conclusion It is fitting that a chapter on usage should be included in a general handbook of syntax, which rightly encompasses every level from morphology to the semantic and pragmatics of sentences. It is important, however, to maintain sight of the special perspective furnished by usage. It is not the function of usage to supply a layer of contextual facts that supplement standard linguistic structures. Rather usage shifts the starting point for the study of syntax and reverses the directionality that gives priority to pre-interpreted structure. It places linguistic structure in a temporal interactive framework, to which linear hierarchical syntax, and the sentences of written language, are secondary. While it may simplify the analysis of sequences to detach syntax from usage, it must not be forgotten that the price of such separation is the exclusion from all consideration of the real environments in which language is at home.

420

9781441124609_Ch24_Final_txt_print.indd 420

11/20/2012 9:24:58 AM

25

Construction Grammar and the Syntax-Semantics Interface Laura A. Michaelis

Chapter Overview Introduction Foundations Formalism Facts Conclusion

421 422 424 429 435

1 Introduction A foundational assumption of traditional generative grammar is that a grammar is organized in the mind of the speaker as a number of hermetically sealed modules, which in the course of a derivation hand off data one to the other.1 Sign-Based Construction Grammar (SBCG) assumes that the grammar contains no such modules, but rather that grammar is an inventory of signs, complexes of linguistic constraints that contain information about form, meaning, and use, and that constructions are the means by which simpler signs are combined into more complex signs. The notion of construction, on this view, is a formalization, in a constraint-based architecture, of the notion of construction in traditional grammar. A simple illustration of a construction is the subjectless tagged sentence shown in (1): (1) Looks nice, doesn’t it?

421

9781441124609_Ch25_Final_txt_print.indd 421

11/20/2012 9:24:01 AM

The Bloomsbury Companion to Syntax

In a sentence like (1) the missing subject of the main clause can only be interpreted on the basis of the reference of the tag’s subject. In particular, the addressee(s) of (1) must determine what it refers to in order to reconstruct the missing first argument of the main clause. While idiosyncratic, the biclausal construction that licenses (1) shares properties with more general constructions, including that which licenses a question tag that is of opposite polarity to the main clause and pronounced with rising intonation, for example, That’s okay with you, isn’t it? The particular combination, arrangement, and interpretation of these inherit ited, construction-licensed signs is, however, peculiar to sentences of this form: a main clause missing a subject, followed by a question tag whose pronominal subject provides the reference of the missing subject of the main clause (Kay 2002). Generalizations about constructions are captured through the interaction of a hierarchical classification of types and the type-based inheritance of grammatical constraints (Sag 2010). The approach described here is distinct from traditional generative grammar, as it is constraint-based rather than derivational and assumes neither underlying structure nor empty categories. In addition, semantic constraints and use conditions are directly associated with the phrase-structure rules that define constructions, rather than being a “read off ” syntactic representation or relegated to a “pragmatic component.” The constructionist program does, however, fall within generative grammar in the historically broader sense of aiming to provide a fully explicit account of the sentences of each language under study. In this chapter, I will lay out the case for construction-based grammar, and in particular SBCG, by describing the theory’s foundations, the theory’s formal implementation, and facts that support a construction-based view of grammar. Section 2 will be devoted to the foundations of construction-based syntax. In Section 3, I will describe the formal foundations of SBCG. Section 4, devoted to the facts, will describe two major lines of evidence for construction-based syntax. Section 5 will provide concluding remarks.

2 Foundations To some linguists, Construction Grammar seems to make an obvious point. Why would anyone think that syntax isn’t based on constructions? A er all, the category of construction has been a part of grammatical discourse since ancient times. But while we find continuations of that tradition in pedagogical and field grammars, as far as syntacticians are concerned, construction-based analysis stopped making theoretical sense when grammar was redefined, as per transformationalist assumptions, as a mechanism for assembling symbols into phrases. According to the transformationalist tradition, syntactic rules do only one thing: 422

9781441124609_Ch25_Final_txt_print.indd 422

11/20/2012 9:24:01 AM

Construction Grammar and Syntax-Semantics

determine what symbol sequences function as units for syntactic purposes. They cannot add conceptual content to that contributed by the words. If sentence meaning does not come from “construction meaning,” there seems little point in positing constructions. In fact, in the “rule-free” conception of grammar promoted by Chomsky (1989, 1995), grammatical constructions are “taxonomic epiphenomena” whose properties are predictable from the interaction of fixed principles with language-particular parameter settings (Chomsky 1989: 43). The difference between a construction-based approach to grammar and one based on interacting universal principles can be viewed in part as a distinction between a positive licensing strategy—ruling certain structures in—and a negative suppression-based strategy—ruling certain structures out (Zwicky 1994; Malouf 2003). The constraints in a suppression-based theory like Government and Binding (GB) theory include the case filter, the binding principles and restrictions on long-distance dependencies, for example, subjacency. The theory is deemed successful if each of the ill-formed sentences of the language under study violates at least one constraint. By contrast, according to the licensingbased view of grammar adopted by SBCG: [a]n expression is syntactically well-formed if its phonological form is paired with its semantics as an instance of some syntactic construction. It follows that an expression is ungrammatical only because there is no combination of constructions that license it, not because there is some cross-constructional filter that rules it out. (Zwicky 1994: 614) As observed in the above passage, constructions interact in the licensing of language objects. That is, a linguistic expression can instantiate multiple types at once. For example, the fronting pattern exemplified by (2) is an instance of both the Topicalization construction and the Filler-Gap construction: (2) That I’m not so sure about. These kinds of interactions are described by type hierarchies, as discussed by Sag (2010). Why would anyone prefer a licensing-based model? A er all, suppressionbased syntactic theories offer constraints of potentially universal significance, and they are inarguably more economical than licensing-based models, since there are far fewer general constraints than there are constructions. Construction grammarians prefer the licensing model not because it is more elegant, but because it provides descriptive precision that suppression-based approaches cannot. Construction Grammar retains descriptive goals that generative-transformational grammar long ago exchanged for the promise of bright-line tests that 423

9781441124609_Ch25_Final_txt_print.indd 423

11/20/2012 9:24:01 AM

The Bloomsbury Companion to Syntax

would separate the relevant (“core”) grammatical phenomena from the irrelevant (“peripheral”) ones. If one takes Chomsky’s claims seriously, the loss of descriptive coverage that resulted from this move was a sign of progress in the “search for explanatory adequacy” (Chomsky 1995: 435). But, as Sag (2010), observes, the generative-transformational tradition finesses the core phenomena too: when one considers that tradition’s signature phenomenon, the English filler-gap dependency, one finds that it is silent concerning salient parameters of variation among the extraction constructions, including the syntactic categories of the filler and head daughters, the type of wh-element in the filler daughter (if any), and the use of the auxiliary-initial pattern in the head daughter. This in turn should lead us to ask how a theory that takes cross-linguistic parametric variation seriously can overlook intra-linguistic variation of a similar nature.

3 Formalism For many years, the only formal reference work available to construction grammarians was an unpublished (but widely circulated) course reader, Fillmore and Kay (1995). It outlines a model that has come to be known as Berkeley Construction Grammar (BCG). This work contains a compelling treatment of interactions between argument-structure constructions (e.g. passive and ditransitive) and shows that one can use the mechanism of lexeme-construction unification to describe English nominal and verbal syntax without recourse to the unary-branching phrases and “inaudible” determiners of X’-based approaches. However, this work also uses a cumbersome nested-box notation for constructions that permits an undefined degree of recursion, and an open-ended and loosely organized repertoire of features. In addition, while Fillmore and Kay (1995) persuasively argue that formal and semantic commonalities among constructions can be captured by means of inheritance relations (rather than, say, transformations), the work does not provide any means of notating such taxonomic relationships other than scattershot notations in construction diagrams. Construction grammarians seeking a more comprehensive and principled system of formal representation were inclined to look to an allied declarative model, Head-Driven Phrase-Structure Grammar (HPSG; Pollard and Sag 1987; 1994). Like BCG, HPSG treats words and phrasal patterns as similar kinds of form-meaning pairings, uses feature structures to model semantic and syntactic classes of grammar objects, and assumes a hierarchical classification of grammatical structures. SBCG, described by Sag (2010, forthcoming), Kay and Sag (forthcoming), and Michaelis (2009), among others, is an attempt to “expand the empirical coverage of HPSG, while at the same time putting BCG on a firmer theoretical footing” (Sag forthcoming). SBCG is a theory of constructional meaning because 424

9781441124609_Ch25_Final_txt_print.indd 424

11/20/2012 9:24:01 AM

Construction Grammar and Syntax-Semantics

it assumes that rules of syntactic combination (descriptions of local trees) are directly associated with interpretive and use conditions, expressed by semantic and pragmatic features that attach to the mother or daughter nodes in these descriptions (Kay and Michaelis forthcoming, Sag forthcoming). This amounts to the claim that syntactic rules have meanings. This claim sets Construction Grammar apart from prevailing models of meaning composition. Such theories are based on a principle that Jackendoff (1997: 48) describes as the “doctrine of syntactically transparent composition.” According to this doctrine, “[a]ll elements of content in the meaning of a sentence are found in the lexical conceptual structures [ . . . ] of the lexical items composing the sentence” and “pragmatics plays no role in determining how [lexical conceptual structures] are combined.” To propose a construction-based model of semantic composition like SBCG is not, however, to deny the existence of syntactically transparent composition. It is instead to treat it, in accordance with Jackendoff (1997: 49), as a “default in a wider array of options.” That is, whenever a class of expressions can be viewed as licensed by a context-free phrase-structure rule accompanied by a rule composing the semantics of the mother from the semantics of the daughter, a construction-based approach would propose a construction that is functionally equivalent to such a rule-to-rule pair. But the constructional approach also enables us represent linguistic structures in which the semantics of the mother does not follow entirely from the semantics of the daughters. In this section, we will discuss two features that make SBCG a useful formalism for constructionbased syntax. The features are locality and variable-grain description.

3.1 Locality In SBCG, the phrase types in the target language are described by means of combinatory constructions. Combinatory constructions describe constructs— signs that are built from one or more distinct signs. Constructions in SBCG take the form of type constraints. A type constraint is a conditional statement that tells what properties a construct will have if it is an instance of the type in question. Intuitively, constructs are local trees (mother–daughter configurations) with feature structures (specifically, signs) at the nodes. Constructions can describe only such mother–daughter dependencies and not, for example, mother–granddaughter dependencies (Sag 2007, 2010). A construct is modeled in SBCG as a feature structure that contains a mother (mtr) feature and a daughters (dtrs) feature. The value of the mtr feature is a sign and the value of the dtrs feature a list of one or more signs. What then is a sign? A sign, as in the Saussurean tradition, is a form–meaning pairing. A sign is modeled as a type of feature structure, or attribute–value matrix. A sign specifies values for six features (Sag 2010, forthcoming): 425

9781441124609_Ch25_Final_txt_print.indd 425

11/20/2012 9:24:01 AM

The Bloomsbury Companion to Syntax z PHON z FORM: a list of formatives comprising the expression z ARG-ST: a ranked list of a lexical expression’s arguments z SYN: CAT and VAL(ENCE) z SEM: INDEX and FRAMES z CNTXT: TOPIC and FOCUS

By treating phrases as feature structures, SBCG captures properties common to lexemes and phrase types in a way that BCG did not. As mentioned, according to the BCG vision, the grammar is an inventory of trees (nested boxes) with an indefinite depth of recursion. By contrast, argument-structure constructions like the Transitive construction are represented by feature structures, as in Figure 25.1. The construction shown in Figure 25.1 expresses a constraint on transitive lexemes: each such lexeme assigns the grammatical relation (rel) object to one argument in its valence (val) set, provided that this argument is not the highest ranking or “distinguished” argument (hence the negative value assigned to the feature DA, or “distinguished argument”). The Transitive construction presumably represents a class of lexemes (those that take direct objects), but it is unclear why a lexeme description like that in Figure 25.1 should qualify as a construction, as it does not contain nested boxes. SBCG, by contrast, proposes two types of constructions: the aforementioned combinatory constructions, which describe properties of phrase types, and lexical-class constructions, which in turn describe properties shared by classes of lexemes (like devour) and words (like devoured). The only difference between lexical-class constructions and combinatory constructions is the type name in the antecedent of the type constraint. Because both words and phrases are signs, the two can be described

syn [voice active] gf obj val rel ⎣ ⎣ DA − ⎦

Figure 25.1 The Transitive construction as per BCG (Fillmore and Kay 1995)

trans - verb - lexeme

applicative lexeme ⇒

ARG -ST saturation - frame ACTOR x SEM | FRAME THEME y SURFACE z

Figure 25.2 The Applicative lexical-class construction 426

9781441124609_Ch25_Final_txt_print.indd 426

11/20/2012 9:24:02 AM

Construction Grammar and Syntax-Semantics

phrase SYN [VAL < > MTR MRKG M : unmk subject - pred - cxt ⇒ DTRS < X , H > CAT [ VF fin] HD - DTR : H VAL < X > MRKG M : unmk

Figure 25.3 The Subject-Predicate combinatory construction

in the same way. This is shown by Figures 25.2–25.3, which illustrate, respectively, a lexical-class construction and a combinatory construction. The Applicative construction, shown in Figure 25.2, describes the lexeme class to which the verbs fill and cover belong (as in, for example, She filled the bathtub with champagne and They covered the wall with a sheet): this lexeme class is a subtype of the transitive-lexeme (trans-verb-lexeme) class, as indicated by the typing of the feature structure to the right of the arrow. As shown by the arg-st list, verbs of this lexeme class express the theme argument as a PP headed by with. The semantic constraints associated with this lexeme class are, as indicated by the frame, labeled saturation-frames in the frames list. This frame is intended to capture the resultant-state entailment that the theme occupies a critical mass of points in a planar region (Michaelis and Ruppenhofer 2001). The Applicative construction also describes one of the classes to which the verbs spray and load belong: the lexical entries of these verbs lack an arg-st list, making them compatible with both the Applicative and Oblique-Goal lexical-class constructions. The Subject-Predicate construction, shown in Figure 25.3, licenses basic declarative clauses. This construction contains two features not previously discussed: hd-dtr and marking (mrkg). The value of the hd-dtr feature is a syntactic feature structure that describes the head sign in a phrase. The index value of this feature structure is identical to that of the second sign in the dtrs list. The value of the mrkg feature is the type of the grammatical marker that accompanies a sign (e.g. that in the case of phrasal verbal sign whose dtrs are the complementizer that and a clause). In the case of the Subject–Predicate construction, the value unmk indicates that neither the verbal head nor its mtr has a grammatical marker. As described in Figure 25.3, a subject–predicate construct consists of two daughter signs, the second of which is a finite verbal sign that selects for the first sign by means of its val feature. As shown in this figure, the mtr of a subject–predicate construct has an empty val set, indicating that it is a complete predication. Thus, SBCG captures properties common to lexical items and phrases by describing both as feature structures. It is true that combinatory constructions describe sign configurations (via the mtr and dtrs features), while lexicalclass constructions describe single signs. But signs and sign configurations are 427

9781441124609_Ch25_Final_txt_print.indd 427

11/20/2012 9:24:02 AM

The Bloomsbury Companion to Syntax

the same thing as far as the licensing mechanism is concerned. The principle that governs the licensing of language objects in SBCG is the Sign Principle. According to the Sign Principle of SBCG (Sag forthcoming), a sign is lexically licensed if it satisfies a lexical entry and constructionally licensed if it is the mother sign of some construct type.2 This means that one can verify the grammaticality of a phrase based only on the properties of its topmost (mtr) feature structure, since these properties include identifying information about that node’s daughters (e.g. the frames on the mtr’s frames list).

3.2 Variable Granularity As is widely recognized by proponents of Construction Grammar and exemplar-based approaches (e.g. Bybee 2007), many grammatical generalizations are not very general. Thus, grammar must contain constraints of varying grains. Independent-clause (IC) exclamatives (e.g. What fools they are!) provide an example of a fine-grained constraint (Sag forthcoming): (3) God, *(I can’t believe) who they hired/where they went! As (3) shows, IC and subordinate-clause exclamatives differ with regard to the syntactic category of the filler daughter: who and where are not possible filler daughters of IC exclamatives in English, although they are in some other languages (Michaelis 2001). A grammar that provides no mechanism for imposing category restrictions will overgenerate. How does SBCG avoid overgeneration? It treats nodes, and in particular the mtr nodes of constructs, as feature structures—not category labels. A description of a feature structure is a set of properties. As property sets, feature-structure descriptions follow the logic of set inclusion: the more properties in the description, the smaller the class of language objects that description picks out. For example, the feature set that describes an IC exclamative includes that which defines the filler-head construction, shown in Figure 25.4.

MTR filler-hd cxt ⇒

SYN [VAL L1] GAP L2

DTR < [SYN X], H > Phrase CAT verbal HD-DTR H: SYN VAL L1 GAP < [SYN X] >

Figure 25.4 The Filler-Head construction 428

9781441124609_Ch25_Final_txt_print.indd 428

11/20/2012 9:24:02 AM

Construction Grammar and Syntax-Semantics

Inclusion relations among feature-structure descriptions allow us to model constructs at each step along the idiomaticity continuum, with an array of constructions of correspondingly graded generality.

4 Facts The descriptive goals of modern syntacticians are broad, and concern general phenomena like the discourse-syntax interface, meaning composition and argument realization. Thus, the challenge for proponents of construction-based syntax is to demonstrate that constructions figure in the most basic functions that grammar performs. In this section, I will discuss ways to confront this challenge, highlighting two fundamental grammar functions: semantic composition and complement licensing. In Section 4.1, I will argue that a sensible model of semantic composition requires recourse to constructional meaning. In Section 4.2, I will give evidence that constructions are licensors of complements.

4.1 Composition Requires Constructions Let us assume, following the tradition of Generalized Phrase Structure Grammar (Gazdar et al. 1985), that for some class of expressions it is licensed by a phrasestructure rule and that this phrase-structure rule is paired with an interpretive rule that composes the semantics of the mother from the semantics of the daughters. In such cases, Construction Grammar would propose a construction that does the same work that such a rule-to-rule pair does. Obviously, we have not validated a constructionist approach by simply showing that it replicates what other theories do without constructions. But Construction Grammar can also describe linguistic structures in which the mother of a given local tree may not follow from the semantics of the daughters. Can a syntactic theory based on strict composition do the same thing? It appears that the answer is no, at least if we use the following definition of compositionality, taken from the Szabó (2007): “If a language is compositional, it cannot contain a pair of non-synonymous complex expressions with identical structure and pairwise synonymous constituents.” The problem with this understanding of meaning composition is that it yields a counterintuitive result for syntactically regular idioms of the type described by Fillmore et al. (1988). Two examples of such idioms are given in (4) and (5), respectively: (4) Pseudoimperative: Now watch me be somehow unable to make it out there. (5) Pseudoconditional: If you’re Obama, you might not like the idea of the Clintons in the White House. 429

9781441124609_Ch25_Final_txt_print.indd 429

11/20/2012 9:24:03 AM

The Bloomsbury Companion to Syntax

Both the Pseudoimperative and the Pseudoconditional are syntactically regular: the former has a syntactic form indistinguishable from that of a regular imperative sentence and the latter has the syntactic form of an ordinary hypothetical conditional. But (4) does not exhort the hearer to watch the speaker and (5) does not predict a future outcome based on the hearer’s ability to shi identity. If we are to maintain Szabó’s definition of compositionality, we must conclude from these interpretive affordances either that English is not compositional or that Pseudoimperatives and Pseudoconditionals have distinct hierarchical representations from their transparently interpreted analogs. Neither appears to be an acceptable conclusion. Thus, a conception of composition based exclusively on X’-syntax leads to a nonsensical result for examples like (4–5). By contrast, a constructional approach allows a single phrase-structure analysis for both the idiomatic and transparent readings of (4–5) and, for the idiomatic readings, posits constructions that attach semantic interpretations directly to complex syntactic objects. For example, under a constructionist analysis, the Pseudoconditional and the regular conditional have a common supertype that is used to define the properties that the two subtypes have in common. In short, constructional approaches recognize as instances of compositionality cases in which two different meanings for the same syntactic form are licensed by two different collections of form-meaning licensers, that is, by two different collections of constructions.

4.2 Constructions License Complements Here we will consider two lines of evidence supporting the contention that constructions license complements: valence variation and “weird sisterhood.”

4.2.1 Valence Variation According to Rappaport Hovav and Levin (1998), henceforth RHL, valence augmentation and other contextual effects on verb meaning are the products of lexical derivations that build up complex event structures from simpler ones. Unlike the construction-based model of argument structure proposed by Goldberg (1995), the RHL model is based on lexical projection; as they put it: “Many aspects of the syntactic structure of a sentence—in particular, the syntactic realization of arguments—are projected from the lexical properties of the verbs” (RHL: 97). Each of a verb’s syntactic frames is associated with a distinct verb meaning, although every verb has one basic class membership. The more aspectual representations a verb has the more syntactic variation it will display, and vice versa. To represent verb meaning and semantic operations on verb meaning, RHL propose (i) a set of Aktionsart-based templates and (ii) an operation that augments one such schema up to another one. The 430

9781441124609_Ch25_Final_txt_print.indd 430

11/20/2012 9:24:03 AM

Construction Grammar and Syntax-Semantics

RHL model assumes that verbs unify with event-structure templates based on Aktionsart class. According to this model, activity verbs like sweep are lexically intransitive, although such verbs can gain a second argument by combining with an accomplishment template, as in, for example, She swept the floor. Arguments supplied by templates are referred to as nonstructural arguments. The RHL model makes three predictions about null complements (Ruppenhofer 2004; Goldberg 2005, 2006): z As nonstructural arguments, the second arguments of bivalent state,

achievement and activity verbs should always be omissible, for example: Have you eaten? z Nonstructural participants are subject only to a recoverability condition based on prototypicality (p. 115); therefore all null complements should have existential (indefinite) interpretations, as in, for example: She reads. z As structural arguments, patient arguments of accomplishment verbs, for example, kill and break, should never be omissible. However, each of these predictions proves false. First, the second argument of a bivalent state, achievement, or activity verb is not always omissible, as shown by (6–8), respectively: (6) She enjoys *(things). (7) She found *(something). (8) We discussed *(issues). Second, null instantiated second arguments need not have indefinite interpretations; they may instead be interpreted as zero anaphors, as in (9)–(11): (9) I remember (that). (10) I prepared (for that event) for weeks. (11) She arrived (there). Third, as shown by Goldberg (2005), accomplishment verbs do allow null instantiated patient arguments. For example, verbs of emission and ingestion license existential null complements, as in (12)–(13): (12) He cried (lachrymal fluid) into his beer. (13) He swallowed (saliva) nervously. In addition, as observed by Ruppenhofer (2004), almost any accomplishment verb in an iterated-event context allows an existential null complement, as in (14)–(16): 431

9781441124609_Ch25_Final_txt_print.indd 431

11/20/2012 9:24:03 AM

The Bloomsbury Companion to Syntax

(14) The police only arrest if there is a high-profile situation. (15) You just take and take. (16) She has never failed to impress. We will now explore how a construction-based model of null complementation, proposed by Kay (2004), circumvents the problems described above. Instead of defining a class of “structurally intransitive” verbs, the constructionist model posits a null-complementation construction. This construction licenses a lexeme whose arg-st list contains one more sign than its val set does. That is, one of the arguments on the arg-st list lacks a corresponding valence member. Among the frames in the frames set of this lexeme is a quantifier frame, which represents the construal (existential or anaphoric) of the missing valence member. The null-complementation construction is a type of derivational construction. A derivational construction builds a new lexeme from one or more lexical signs: the mother of a derivational construct is of the type lexeme; its daughters are lexemes. The null-complementation construction builds a lexeme with a covert valence member from a lexeme with an optionally covert valence member. A construct licensed by the null-complementation construction is shown in Figure 25.5.

lexeme FORM < eat > ARG - ST NP i, NP x overt MTR SYN | VAL NP INST i eat − fr timed − meal − fr exist − fr SEM | FRAMES EATER i , , INST x BV x FOOD x lexeme

DTRS

FORM < eat > ARG - ST NP i, NP x overt (ini) SYN | VAL NP , NP INST i INST x eat − fr SEM | FRAMES EATER i FOOD x

Figure 25.5 A null-complementation construct 432

9781441124609_Ch25_Final_txt_print.indd 432

11/20/2012 9:24:03 AM

Construction Grammar and Syntax-Semantics

In the construct shown in Figure 25.5, the dtr lexeme is the verb eat, which licenses two arguments, an eater and a food. The sign types of its valence members are, respectively, overt and (ini). According to the hierarchy of sign types laid out by Kay (2004), this (ini) sign type may resolve to either an overt sign or an ini sign. In the construct shown in Figure 25.5, it has resolved to ini. The reason that we do not see this ini sign in the valence set of the mtr lexeme is that an ini sign is a type of covert sign; as such, it is subject to the constraint in (17): (17) covert ⇒ sign & [FORM < >] The constraint in (17) ensures that the type ini has no form value. A sign with no form value will not appear on a lexeme’s val list, since the val value is a list of overt signs. As a subtype of the type null-comp, the ini sign type obeys the constraint in (18): ⎡ ⎡ ⎢ ⎢ ⎢SEM ⎢ INDEX x ⎢ ⎢FRAMES (18) null-comp ⇒ covert & ⎢ ⎢ ⎢⎣ ⎢⎣

⎡fram e ⎤ ⎢ ⎥ ⎢ BV x ⎥ ⎣ ⎦

⎤⎤ ⎥⎥ ⎥⎥ ⎥⎥ ⎥⎥ ⎥⎦ ⎥⎦

The constraint in (18) ensures that if there is a sign of the type null-comp, the lexeme that licenses this sign will have a quantifier frame on its frames list. This quantifier frame takes a bound variable (bv) as its argument, and this bound variable shares its index with the covert sign. The constraint in (19) ensures that the quantifier frame is in particular an existential-quantifier frame: (19) ini ⇒ null-comp & [SEM [FRAMES ]] The above constraint captures the existential interpretation of a missing argument in sentences like (20): (20) I’ve eaten. Sentence (20) means something like “I’ve eaten some food at a canonical meal time” rather than “I’ve eaten that food at a canonical meal time.” Described in procedural terms, the null-complementation construction that licenses the construct in Figure 25.5 builds an eat lexeme in which the food argument is missing from the verb’s val set but remains part of its frames set, where it is the bound variable of an existential quantifier. It is important to note, however, that while the mother and daughter lexemes have distinct val sets, they have the same arg-st set: the food participant appears on the arg-st list of the mtr lexeme even though it does not appear on the mtr lexeme’s val list. Additional null-complementation constructions are required to account for the fact that null-complementation restrictions on verbs can be overridden in

433

9781441124609_Ch25_Final_txt_print.indd 433

11/20/2012 9:24:03 AM

The Bloomsbury Companion to Syntax

certain contexts. As mentioned above, patient arguments of accomplishment verbs are not generally subject to null instantiation. For example, sentences like (21)–(23) sound awkward: (21) The police arrested *(someone) last night. (22) Sam took *(something) without permission. (23) I just impressed *(someone)! However, as observed by Goldberg (2005), accomplishment verbs that do not allow unexpressed patient arguments in episodic contexts invariably allow them in habitual-generic predications and existential-perfect predications, as shown by (24)–(26), repeated from (14)–(16) above: (24) The police only arrest if there is a high-profile situation. (25) You just take and take. (26) She has never failed to impress. The contexts illustrated here map neatly to constructions—the habitual present-tense inflectional construction in (24) and (25) and the perfect-participle derivational construction in (26). Let us presume that the mtr lexemes of these constructions resemble the lexeme daughter eat in the derivational construct shown in Figure 25.5, in that its second valence member has the type (ini). This means that constructions like the habitual-present tense inflectional construction can “feed” the indefinite null-instantiation construction, thus licensing null-object arrest and null-object take in (24)–(25), respectively.

4.2.2 Weird Sisterhood Many verb frames specify sisterhood relations that are not predicted by the general-purpose constituency rules that combine heads and complements and heads and specifiers. Many of these patterns have specialized communicative functions. A look at these phenomena suggests that highly detailed constructions, rather than non-category-specific phrase-structure rules, pair predicates and their complements. In this section, we will look at just one case of weird sisterhood: Nominal Extraposition. In Nominal Extraposition (Michaelis and Lambrecht 1996), an exclamatory adjective, for example, amazing, licenses an NP complement: (27) I know it’s just it’s unbelievable the different things that are happening in America today. (28) I’ll date myself a little bit but it it’s remarkable the number of those things they need. (29) I know. I love that game. It’s amazing the words they come up with. 434

9781441124609_Ch25_Final_txt_print.indd 434

11/20/2012 9:24:24 AM

Construction Grammar and Syntax-Semantics

The pattern exemplified in (27)–(29) is idiosyncratic in two respects. First, adjectives are not case assigners and should not therefore license non-oblique NP complements. Second, this NP complement is interpreted as denoting a scalar degree (Michaelis and Lambrecht 1996). In (29), for example, the NP the words they come up stands in for a scalar expression like “the number of words they come up with”; that is, it is not the words themselves but their numerousness that is deemed remarkable. The fact that the complement of amazing in (29) has a scalar interpretation follows from the fact that (29) is an exclamation, but the pairing of an exclamatory adjective with an NP sister that denotes a degree, metonymically or otherwise, requires a construction that provides for this syntax and this meaning.

5 Conclusion What makes construction-based syntax a radical departure from formal syntactic theory as we know it? While transformational-generative syntacticians have sought to import syntactic generalizations into the lexicon (as illustrated by the approaches of Baker 1996; Hale and Keyser 1998; and Borer 2001), construction grammarians have moved in the opposite direction: it is the lexicon that provides a model for the syntax-semantics interface in Construction Grammar. Lexical-class constructions, which define classes of lexemes or words, and combinatory constructions, which define classes of phrases, are both constraints on feature structures. In phrasal constructions, a list-valued feature of the mother is used to represent the property of having the daughters it does. Thus, rather than seeing syntax, semantics, and lexicon as independent modules, with the lexicon characterized as a bag of idiosyncratic form-meaning pairings, SBCG proposes a lexicon structured by hierarchically organized lexical classes and extends this model to relations among classes of phrases.

Notes 1. Portions of this chapter appear as Michaelis (forthcoming). 2. According to the Sign Principle, a lexical sign can be constructionally licensed, if it corresponds to the mtr sign of a derivational or inflectional construction. In fact, the only lexical signs that are licensed by lexical entries are those that are not “produced” by derivational or inflectional constructions.

435

9781441124609_Ch25_Final_txt_print.indd 435

11/20/2012 9:24:24 AM

26

Topic, Focus, and the Cartography of the Left Periphery Luigi Rizzi

Chapter Overview Introduction The Left Periphery of the Clause Topic and Focus Illustrated by Clitic Left Dislocation and Contrastive Focus Movement The Interface with Semantics/Pragmatics: Association with Old/New Information Some Differences between Topic and Focus A Structural Approach to Topic and Focus Configurations Topic and Mod Conclusion

436 437 439 440 444 445 449 450

1 Introduction The initial periphery of the clause is the structural zone dedicated to the expression of “scope-discourse” semantics: the scope of different kinds of operators and the discourse-related articulations of Topic—Comment—and (left-peripheral) Focus—Presupposition. Till the mid–1990s, a widespread assumption was that a single X-bar projection, the complementizer phrase, or CP, would suffice to structurally express such properties, much as a single IP projection had previously been assumed to express the internal structure of the clause. The cartographic projects (Rizzi 1997, 2000, (ed.) 2004; Cinque 1999, (ed.) 436

9781441124609_Ch26_Final_txt_print.indd 436

11/27/2012 11:07:21 AM

Topic, Focus, and the Cartography of the Left Periphery

2002; Belletti (ed.) 2004, 2009; among others) drew much attention to the study of the precise structural shape of these zones, showing the necessity of assuming richer representations of syntactic structures. In this chapter I will illustrate some fundamental interface properties of Topic and Focus, and discuss their structural position in a cartography of the left periphery of the clause. The illustration will be largely based on data from Italian, a language displaying a particularly rich, diversified, and thoroughly studied use of the initial periphery, with more occasional reference to data from cross-linguistic comparison.

2 The Left Periphery of the Clause The traditional approach to the complementizer system assumes a single head, C, and its X-bar projection, the CP, to constitute the structure of the initial periphery of the clause. This conception was shown to be oversimplified around the mid-1990s, when various kinds of empirical evidence were produced in favour of a more articulated approach. A very straightforward type of evidence is provided by the fact that different kinds of complementizers occur in different orders with respect to a given element. Consider for instance three types of complementizers in Italian: che “that,” introducing embedded declaratives, se “if” introducing indirect questions, and di (homophonous to preposition di, “of”; see Rizzi 1982, Kayne 1984 for arguments supporting its complementizer status), introducing control infinitives, as in (1a, b and c). (1) (a) Penso che partiro’. “I think that I will leave.” (b) Penso di partire. “I think of to leave.” (c) Non so se partirò. “I don’t know if I will leave.” These elements occur in distinct positions with respect to a Topic expressed in the Clitic Left Dislocation construction (whose properties are discussed below); che must precede the Topic (at least in restrictive varieties of Italian), if can both precede and follow the Topic; di is necessarily preceded by the Topic: (2) (a) *Penso, a Gianni, che gli dovrei parlare. “I think, to Gianni, that I should speak to him.” (b) Penso che, a Gianni, gli dovrei parlare. “I think that, to Gianni, I should speak to him.” 437

9781441124609_Ch26_Final_txt_print.indd 437

11/27/2012 11:07:22 AM

The Bloomsbury Companion to Syntax

(3) (a) Penso, a Gianni, di dovergli parlare. “I think, to Gianni, “of” to have to speak to him.” (b) *Penso di, a Gianni, dovergli parlare. “I think ‘of’, to Gianni, to have to speak to him.” (4) (a) Non so, a Gianni, se gli potremo parlare. “I don’t know, to Gianni, if we could speak to him.” (b) Non so se, a Gianni, gli potremo parlare. “I don’t know if, to Gianni, we could speak to him (?).” This simple distributional evidence suggests that different complementizer particles occupy different positions and a more complex space must be assumed to constitute the initial periphery of the clause. In the proposal in Rizzi (1997) it is assumed that the complementizer space is delimited by two types of heads, with other kinds of heads allowed to occur in the space thus defined. The highest head, expressed by che in Italian, that in English, etc. is called Force in Rizzi (op. cit.); the lowest head, filled by di in Italian, for in English, etc. is called Fin(iteness). Fin agrees in finiteness with the clause it introduces. In some languages, both delimiting positions can be simultaneously expressed, as illustrated by particles mai and a in Welsh which, according to Roberts’ (2004) analysis, surround the topic-focus field in examples like the following: (5) Welsh Dywedais i [mai ‘r dynion  fel arfer a [werthith  y ci]]. “Said I C the men as usual C  will-sell the dog.” (Roberts 1999) Other heads like se (dubbed Int(errogative) in Rizzi 2001) can occur in the space delimited by Force and Fin. If Int expresses interrogative force, perhaps the highest head should be restricted to declarative force, or be redefined (e.g. as simple subordinator, as in Rizzi 1997: 328, fn. 6 based on Bhatt and Yoon 1991; see Haegeman 2006; Radford 2010); nevertheless, I will stick here to the original label, which has come to designate the highest C head in much standard practice. Complementizers like that and if can co-occur in a fixed order in certain special indirect questions in Spanish, in which a previously asked question is understood as being reported by the speaker (Plann 1982; Suñer 1994): (6) Suñer (1994: 349): Me preguntaron que si tus amigos ya te visitaron en Granada. “They asked me that if your friends had already visited you in Granada.” 438

9781441124609_Ch26_Final_txt_print.indd 438

11/27/2012 11:07:22 AM

Topic, Focus, and the Cartography of the Left Periphery

Saito (2010) shows that in Japanese a similar kind of “report” marker to can co-occur in a fixed order with the interrogative particle ka, and with the particle no, that he analyzes as a filler of the Fin position; given the head-final order of the language, the three particles occur sentence finally and in an order which mirrors the hierarchical order: (7) Taroo-wa [CP kare-no imooto-ga soko-ni ita (no) ka (to)] minna-ni tazuneta. T.-TOP he-GEN sister-NOM there-in was no ka to all-DAT inquired “Taroo asked everyone if his sister was there.”   (ibid.: 41) So, module headedness, the hierarchical order Decl/report—Interrogative— Finiteness observed in the Romance languages is reproduced in Japanese.

3 Topic and Focus Illustrated by Clitic Left Dislocation and Contrastive Focus Movement The space delimited by Force and Fin contains positions dedicated to express Scope-discourse semantic properties, in Chomsky’s (2004) sense: different kinds of operators taking scope over the clause, and positions pertaining to the way in which information is “packaged” (Vallduvi 1992, Samek-Lodovici, this volume.) and discourse is organized: in primis, Topic and (left-peripheral) Focus. I will illustrate some fundamental properties of Topic and Focus, or, more precisely of the Topic—Comment and Focus—Presupposition articulations, through certain Italian/Romance constructions whose syntactic and interpretive properties have been thoroughly investigated: the Clitic Left Dislocation, in which an initial Topic is construed with a Comment containing a resumptive clitic pronoun, as in (8a); and the contrastive Focus, in which a prosodically marked initial constituent binds a gap in the following clause, the Presupposition, and an explicit negative tag making the contrast is strongly invited, as in (8b):1 (8) (a) Il suo libro, lo dovresti leggere. (Topic—Comment) “His book, you should read it.” (b) IL SUO LIBRO dovresti leggere, non il mio. (Focus—Presupposition) “HIS BOOK you should read, not mine.” The two articulations have in common the critical use of the left periphery of the clause; they also sharply differ in their syntactic properties and in their properties on both interfaces with sound and meaning. On the sound side, the topic-comment and the focus-presupposition articulations are marked by clearly distinct prosodic contours, which make the two constructions easily identifiable already in the speech signal (Frascarelli 2000).

439

9781441124609_Ch26_Final_txt_print.indd 439

11/27/2012 11:07:22 AM

The Bloomsbury Companion to Syntax

The following figures, taken from Bocci’s (2009) experimental study, illustrate the two articulations in Sienese Italian: (9) Pitch contour of Topic—Comment A Michelangelo (Top), Germanico vorrebbe presentare Pierangela

“To Michelangelo (Top), Germanico would want to introduce Pierangela.” (10) Pitch contour of Focus—Presupposition A MICHELANGELO (Foc) Germanico vorrebbe presentare Pierangela 550 500

Pitch (Hz)

400 300 200 100

0

3.61823 Time (s)

“TO MICHELANGELO (Foc) Germanico would want to introduce Pierangela.” The contour of contrastive focus, Bocci (2004, 2009) argues, is more prominent than the contour of topic; even sharper is the distinction between the contour of the presupposition, which is completely flattened, and the contour of the comment, which remains “hilly,” with varying prominence as in sentences without an explicit topic.

4 The Interface with Semantics/Pragmatics: Association with Old/New Information The properties of Topic and Focus at the interface with semantics/pragmatics can be highlighted by creating mini-discourse contexts and checking the 440

9781441124609_Ch26_Final_txt_print.indd 440

11/27/2012 11:07:22 AM

Topic, Focus, and the Cartography of the Left Periphery

appropriateness of the different constructions (see Roberts (2010) for detailed discussion of the interpretive properties of Topics). A Topic position cannot be used to provide the requested new information in a wh-question–answer pair: (11) Q: Che cosa hai dato a Gianni? “What did you give to Gianni?” A: Gli ho dato il tuo libro. “I gave him your book.” A’: #Il tuo libro, glielo ho dato. “Your book, I gave it to him.” A Topic can be felicitously used in a question–answer pair when it takes up a piece of information already given in the question: (12) Q: Che cosa hai fatto col mio libro? (e con quello di Piero?) “What did you do with my book? (and with Piero’s?)” A: Il tuo libro, lo ho dato a Gianni (sul libro di Piero, non so che dirti). “Your book, I gave it to Gianni (on Piero’s book I don’t know what to say).” In fact, the explicit mention of the topic in (12) A is fully natural when an element of choice is introduced (as is the case if the parenthesized part of (12) Q is produced, introducing a possible choice to be made in the background information between my book and Piero’s book); otherwise, the most natural answer simply involves the pronoun that takes up the contextually given referent, and the explicit Topic is felt as redundant. (13) The Topic-Comment articulation: an element is selected from the given, background information and made prominent; a comment is expressed about it: “Among the elements of the background, I select X(Topic) and tell you about it that Y (Comment).” The assumption that Topics are associated to given information, in the sense expressed by (13), is apparently challenged by the observation that indefinite topics are quite normally felicitous in Italian and other Romance languages. For instance: (14) Un problema di questo genere, non lo risolverò mai. “A problem of this kind, I will never solve.”

441

9781441124609_Ch26_Final_txt_print.indd 441

11/27/2012 11:07:23 AM

The Bloomsbury Companion to Syntax

But, even though the particular problem referred to by the topic in (14) may not have been introduced in previous discourse, the phrase “di questo genere” (of this kind) connects it to a set of familiar problems, and this kind of “partitivity” is sufficient to license the use of a Topic–Comment structure. The partitivity does not have to be explicit. Consider the following dialogue. A son may reply to his father’s reproach A as is B or in B’ (15) A: Questa settimana non hai fatto niente per preparare l’esame . . . “Yesterday you didn’t do anything to prepare the exam . . .” B: Beh, un libro, lo ho letto . . . “Well, a book, I have read . . .” meaning: a book relevant for the preparation of the exam. B’: Beh, ho letto un libro . . . “Well, I’ve read a book . . .” Here the book does not have to be connected to the preparation of the exam. Reply B strongly invites the conclusion that the book is part of the exam’s program, while reply B’ implies nothing on the nature of the book: it could be connected or unrelated to the exam, in which case B’ would convey the message that indeed the son did not do anything to prepare for the exam, but he nevertheless engaged in a worthwhile activity, reading an unrelated book. So, the use of the indefinite topic, as in B, invites a quasi-partitive interpretation “a book which is part of a relevant presupposed set” (in this case, the set of readings for the preparation for the exam), even in the absence of any explicit partitive structure. In such an “understood partitive” context, even a non specific indefinite may be possible as a Topic in Italian, where the non-specific interpretation is explicitly marked by the subjunctive mood in the relative clause in examples like the following: (16) Gianni sta cercando di rimettere in sesto il budget del dipartimento, ma . . . “Gianni is trying to restore the departmental budget but . . . ” . . . una segretaria che sappia tenere la contabilità, non l’ha ancora trovata. “ . . . a secretary who can(+Subj) do the accounting, he hasn’t found yet.” That is, Gianni hasn’t found yet a good candidate in the set of possible collaborators who may help him restore the Department’s budget. In conclusion, the “givenness” of a topic may be quite indirect, mediated by an explicit or implicit partitive connection to a familiar set of entities. As for Focus, in Standard Italian, the left peripheral focal position cannot correspond to simple new information focus, which is normally expressed in sentence-final position: 442

9781441124609_Ch26_Final_txt_print.indd 442

11/27/2012 11:07:23 AM

Topic, Focus, and the Cartography of the Left Periphery

(17) Q: Che cosa hai letto? “What did you read?” A: Ho letto il tuo libro. “I read your book.” A: # IL TUO LIBRO ho letto. “Your book I read.” The left peripheral focus in Italian is not only new information, but must convey a new information that contrasts with some expectation attributed to the interlocutor. A typical context is one in which contrastive focus corrects a previous statement: (18) A: So che ieri hai letto un articolo per preparare l’esame . . . “I know that yesterday you read an article to prepare the exam . . . ” B: Scherzi? UN LIBRO INTERO ho dovuto leggere (non un articolo). “You kiddin’? A WHOLE BOOK I had to read, not an article.” Contrastive focus in this sense strongly invites a negative tag, explicitly denying the interlocutor’s statement, or expectation; on the other hand, new information focus does not normally occur with an explicit negative tag excluding alternatives: for instance, adding a negative tag to (17) A like non quello di Gianni “not Gianni’s” would sound rather redundant, unless I had some reason to believe that you expected that I would read Gianni’s book. Contrastive focus is licit not only in environments of explicit correction of previous statements. It is sufficient that the new information contrasts with some “natural expectation” imputed to the interlocutor. Consider for instance the following fragment of dialogue: (19) A: So che l’esame di linguistica è piuttosto leggero . . . “I know that the linguistics exam is rather light . . . ” B: Sei matto! CINQUE LIBRI ho dovuto preparare! “You crazy! FIVE BOOKS I had to prepare! Therefore, (20) Contrastive focus introduces new information that contrast with some natural expectation imputed to the interlocutor. So, Standard Italian can only use the clause-initial focal position to convey the special, contrastive interpretation expressed by (20). There is a parametrization here: even closely related systems like the Sicilian dialects (and the regional

443

9781441124609_Ch26_Final_txt_print.indd 443

11/27/2012 11:07:23 AM

The Bloomsbury Companion to Syntax

varieties of Italian spoken in Sicily and other southern areas) use a clause-initial position also for new information focus (Cruschina 2008). The following fragment, not felicitous in standard Italian, is fine in Sicilian:2 (21) Sicilian A: Chi scrivisti? “What did you write? B: N’articulu scrissi. “An article I wrote.” As our examples show, both contrastive and new information Foci can be definite or indefinite. So one may ask: if Focus is associated to new information, how can it be definite, and referring to a contextually given entity? In fact, focus is new information in a relational sense: what is new is not (necessarily) the referent of the focussed element, but the fact that it is this referent that participates with that particular theta role in the event that is described (i.e. for (17) B, “the entity x such that I read x = your book”; for (19) B: “The entity x such that I had to prepare x = five books, and you expected otherwise,” etc. So, a focal construction assigns a value to a variable contained in the presupposition; the assigned value may be a referent already introduced in discourse, or not (on the rich literature on the semantic properties of Focus see the synthesis in Rooth 1996).

5 Some Differences between Topic and Focus A Focus configuration gives rise to a Weak Cross-Over (WCO) violation, while a Topic does not give rise to any detectable effect of this kind: (22) Giannii, suai madre loi ha sempre apprezzato. “Gianni, his mother always appeciated him.” (23) ??GIANNIi suai madre ha sempre apprezzato ti (non Piero). “GIANNI his mother always appreciated, not Piero.” Lasnik and Stowell (1991) have proposed that WCO is the “signature” of core Operator-variable constructions. If this is so, the WCO effect with Focus is a consequence of the proposed interpretative status of the construction: the focal element assigns a value to a variable. The Topic–Comment configuration, on the other hand, does not plausibly involve any operator—variable structure, is more akin to the assignment of an antecedent to a pronoun, or to a kind

444

9781441124609_Ch26_Final_txt_print.indd 444

11/27/2012 11:07:23 AM

Topic, Focus, and the Cartography of the Left Periphery

of higher subject—complex predicate relation (on similarities and differences between Topic–Comment and Subject–Predicate relations see Rizzi 2006); so, there is no reason to expect a WCO effect in this case. Consider then the compatibility with a clitic. Cinque (1990) observed that direct object topic requires a resumptive clitic: a structure with an object gap like (24b), with the intonation and pragmatics of a Topic—Comment configuration, is ungrammatical; whereas a contrastive focus is inconsistent with the clitic (as in (25a)), while a structure with an object gap as in (25b) is fine (see Bocci 2004 for the discussion of a special case of contrastive focus consistent with the clitic): (24) (a) Il tuo libro, lo ho comprato. “Your book, I bought it.” (b) *Il tuo libro, ho comprato. “Your book, I bought.” (25) (a) *IL TUO LIBRO lo ho comprato (non il suo). “YOUR BOOK I bought it (not his).” (b) IL TUO LIBRO ho comprato ___ (non il suo). “YOUR BOOK I bought (not his).” Cinque proposes to interpret this contrast in the following way. The principle banning vacuous quantification, a particular case of Chomsky’s (1986b) principle of Full Interpretation, requires variables to be bound by operators: no free variable and no vacuous operator (not binding any variable) is allowed. If syntactic variables are gaps, empty DP positions (at least in languages not having a grammaticalized “resumptive strategy”), then (24b) violates the ban against vacuous quantification in one direction: there is a variable without an operator binding it. Given the operator status of left-peripheral focus, (25b) is fine: an operator binds a variable; (25a), on the other hand, is excluded: there is an operator with no good syntactic variable (= a gap) to bind. (24a) is also fine: a non-operator (the Topic) is construed with a non-variable (the clitic pronoun), a special case of the assignment of a referent to a pronoun.

6 A Structural Approach to Topic and Focus Configurations We have shown that Topic–Comment and Focus–Presupposition configurations involve a left-peripheral position construed with a clause-internal position. We

445

9781441124609_Ch26_Final_txt_print.indd 445

11/27/2012 11:07:24 AM

The Bloomsbury Companion to Syntax

should now try to be more precise about the nature of the left peripheral positions and of the computational mechanisms engaged in the generation of such configurations. I would like to adopt here the so-called criterial approach to scope-discourse semantics: scope-discourse properties are expressed by dedicated functional heads, which assign to their dependents interpretive roles such as topicality, focus, etc. (much as thematic properties are assigned by lexical elements to their dependents). So, there is a system of functional heads, merged in particular hierarchical orders in the initial periphery of the clause, which attract topics, foci, and the like. We thus have representations like the following:3 (26) (a) This book Top you should read ___. (b) THIS BOOK Foc you should read ___, not that one. The involvement of a system of dedicated functional heads is made plausible by the fact that many languages manifest overt Topic and Focus particles connecting the Topic to the Comment and the Focus to the Presupposition, as in the following case taken from the West African language Gungbe, analyzed along the lines of Aboh (2004): (27) Gungbe (a) Ùn sè [dò [dàn ló yà [Kòfí hù ì]]]. “I heard that snake the TOP Kofi killed it.” (b) Ùn sè [dò [dàn ló wè [Kòfí hù ___]]] (Aboh 2004). “I heard that snake the FOC Kofi killed.” In short, such heads, dubbed Criterial heads in Rizzi (1996, 1997), have a dual function: the syntactic function of attracting a phrase to their specifiers, and the interface function of triggering interpretive routines in the interface systems. At the interface with semantics and pragmatics the interpretive routines can be expressed as follows, at a rough first approximation: (28)

TopP XP Top

YP

Interpret XP as Topic Interpret YP as Comment

446

9781441124609_Ch26_Final_txt_print.indd 446

11/27/2012 11:07:24 AM

Topic, Focus, and the Cartography of the Left Periphery

(29)

FocP ZP Foc

WP

Interpret ZP as Focus Interpret WP as Presupposition That is, for the Top head, “my specifier is to be interpreted as the Topic, and my complement as the Comment”; and similarly for the Foc head. The criterial heads also trigger the appropriate rules of prosodic contour assignment, to yield such representations as (9) and (10). This approach thus insists on the assumption of completely transparent interface representations, somehow “wearing on their sleeves” certain basic interpretive properties; hence on the attempt of “syntacticizing as much as possible” (Cinque and Rizzi 2010) the systems at the interpretive interfaces. One additional difference that the interpretive routines (28) and (29) can help us understand has to do with the uniqueness of (left peripheral) Focus, opposed to the possible proliferation of Topics. In Italian and other Romance languages there can be as many topics as there are topicalizable arguments: (30) (a) A Maria, il tuo libro, Gianni glielo deve dare. “To Maria, your book, Gianni must give it to her.” (b) *A MARIA (,) IL TUO LIBRO Gianni deve dare (non a Giulia, il disco). “To MARIA YOUR BOOK Gianni must give, not to Giulia the record.” The uniqueness of focus follows from the interpretive properties of the structure, according to (29): if a FocP were recursively embedded as the complement of a higher Foc head, we would have that the complement of the higher Foc, interpreted as presupposed, contains a focus position, an inconsistent interpretive property: hence, FocP cannot be recursive. No such problem arises for Top: there is no particular interpretive constraint on the Comment, except that it must specify some new information about the Topic; nothing else is excluded, in particular nothing prevents a Comment to have, in turn, a Topic—Comment structure. Therefore, there is no interpretive ban against Top recursion. Languages may or may not exploit this option, as a matter of syntactic parametrization, but there is no general ban against it. On the other hand, there is a general interpretive ban against Foc recursion, and in fact the uniqueness of left-peripheral Focus seems to hold in general. So, different languages exploit to different degrees the options of an expanded left periphery. Verb Second languages like German or the other 447

9781441124609_Ch26_Final_txt_print.indd 447

11/27/2012 11:07:24 AM

The Bloomsbury Companion to Syntax

Germanic languages (except Modern English) exploit at most one left peripheral position at a time in main clauses, as a consequence of the Verb Second constraint permitting at most one element to escape from the clause and reach the periphery (Grewendorf 2002). Gungbe permits at most one Topic and one Focus in a strict order: (31) Gungbe (Aboh 2004) dò Kòfí yà gànkpá mè wè kpònòn lé sú—ì dó. “. . . that Kofi Top PRISON IN Foc policemen Pl shut him there.” (Aboh 2004) Hungarian permits a proliferation of Topics higher than the Focus position (Kiss 1987; Brody 1990; Puskas 2000). Italian and other Romance languages permit a proliferation of TopPs to precede and follow the unique Foc position, for example: (33) Credo che a Gianni, QUESTO, domani gli dovremmo dire. C Top Foc Top IP “I believe that to Gianni, THIS, tomorrow we should say.” Which leads to a basic map of the left periphery as follows: (34) Force Top* Foc Top* Fin IP (Rizzi 1997) with Top* indicating the recursive character of the Topic position.4 This map has been refined in many important ways. For instance, Benincà and Poletto (2004), Frascarelli and Hinterhölzl (2007), and Bianchi and Frascarelli (2009) have proposed a more refined typology of topics, which is also linked to positional differences. For instance, the higher Top position admits a contrastive interpretation in that it can be naturally contrasted with another topic, while the lower, postfocal Top position does not naturally allow contrast: (35) (a) A Gianni, IL LIBRO gli dovevate dare, (e a Piero, IL DISCO, piuttosto che il contrario). “To Gianni, THE BOOK you should give, (and to Piero THE RECORD, rather than the other way around).” (b) Il LIBRO, a Gianni, gli dovevate dare (* e IL DISCO, a Piero, piuttosto che il contrario). “The BOOK to Gianni you should give, (and THE RECORD, to Piero, rather than the other way around).”

448

9781441124609_Ch26_Final_txt_print.indd 448

11/27/2012 11:07:24 AM

Topic, Focus, and the Cartography of the Left Periphery

So, we have the possibility of a contrastive Topic, as much as we have a contrastive Focus, but this is restricted to the highest (pre-Foc) Topic position: the lower Top position is a “familiarity” or “given” Topic, which may merely reiterate a given information without being able to express an explicit contrast with another Topic.5

7 Topic and Mod Adverbials can be highlighted by being preposed to the left periphery. Referential adverbials (tomorrow, next week, etc.) may well function as bona fide Topics in some structures; other kinds of adverbials, for example, the –ly adverbs are harder to interpret as topics, and still they can be highlighted by preposing: (36) Quickly, John left the room. Here clause-initial quickly does not imply, or invite any connection with a familiar or presupposed set of quick events. Moreover, contrary to topic movement, this kind of adverb preposing is clause-bound: in Quickly, Bill said that John left the room it is hard to construe the adverb with the embedded clause. In Rizzi (2004) it is proposed that the adverb targets the Spec of a specific left-peripheral head, Mod(ifier), dedicated to the highlighting of adverbials. Referential adverbials like temporal adverbials have the option of targeting Mod as well, in addition to being able to target the Top position. We can thus express certain different effects that a preposed adverbial and a regular argumental topic determine on movement. Top often determines some kind of island effect, more or less weak depending on the language; a preposed adverbial, targeting a Mod head, does not. For example, in English: (37) (a) The book which, tomorrow, John will give to Bill. (b) ?? The book which, to Bill, John will give tomorrow. A preposed adverbial does not have any adverse effect when it is crossed by a movement path, as in (37a), while an intervening Topic gives rise to a deviant structure. Given the difference in the landing site of Topic movement and adverb preposing, we can at least state that TopP gives rise to a mild island effect, while ModP does not. Not only that: in some cases, an intervening adverbial may even have a facilitating effect on extraction, as in cases like the following, in which the adverbial alleviates a that-trace effect, as originally observed by Bresnan (1977) (the 449

9781441124609_Ch26_Final_txt_print.indd 449

11/27/2012 11:07:24 AM

The Bloomsbury Companion to Syntax

adverb effect, Browning 1996, or anti-adjacency, Rizzi 1997); no such improving effect is induced by a topic: (38) (a) The man who I think that *(next year) ___ will sell his house. (b) *The man who I think that his house, ___ will sell next year. See Rizzi (2010) for an analysis of anti-adjacency capitalizing on the different position and properties of the two landing sites TopP and ModP.

8 Conclusion The discourse-related Topic—Comment and Focus—Presupposition articulations are expressed in the initial periphery of the clause and manifest welldifferentiated properties at the syntactic level, and at the interpretive interfaces. I have illustrated the main properties of these articulations, and presented a structural analysis based on results of the cartographic projects. According to this line of research, syntactic representations specify richly articulated structural zones, such as the initial periphery of the clause. These syntactic representations explicitly encode information used by the interpretive systems to express the special properties of Topic and Focus at the interfaces with sound and meaning.

Notes 1. I will not use in my discussion the term Topicalization, which has often been used in the literature on English to refer to both Topic movement and Focus movement, generating some confusion; I will rather assume two distinct processes of Topic and Focus movement, targeting distinct positions in the left periphery (the former also involving the doubling with a resumptive clitic: as shown later in the chapter). The English glosses of Clitic Left Dislocation are somewhat misleading, as English, lacking syntactic clitic pronouns, does not have the exact equivalent of this construction. Ross’s (1967) Left Dislocation has distinct properties from Clitic Left Dislocation, and is closer to the distinct left peripheral construction called “Hanging Topic,” on which see Cinque (1990), Benincà and Poletto (2004), and Belletti (2009, chapter 11). 2. Cruschina actually argues that Sicilian specifies two distinct left peripheral positions, a lower one dedicated to new information focus, less prosodically marked, and requiring inversion and a higher one dedicated to contrastive focus, prosodically more marked, and not requiring inversion; Standard Italian only specifies the latter position in the left periphery. 3. Notice that English uses the gap strategy not only for Focus, but also for Topic constructions; presumably in English the gap is licensed through the mediation of a null operator locally construed with the Topic, much as in Chomsky’s (1977) analysis.

450

9781441124609_Ch26_Final_txt_print.indd 450

11/27/2012 11:07:24 AM

Topic, Focus, and the Cartography of the Left Periphery 4. Belletti (2004, 2009) argues that a Topic and Focus can also appear in the low periphery of the clause, associated to the vP: we thus have peripheries dedicated to scopediscourse semantics associated to both phase domains (in the sense of Chomsky’s 2001 Phase Theory), the CP, and the vP. Clitic right dislocation can be analyzed as movement to a low Topic position (see also Cecchetto 1999), much as the clause final focal position expressing new information focalization may involve movement to a low Focus position. So, in (i) Gli ha parlato Gianni, a Piero “To-him spoke Gianni, to Piero,” which could be felicitously used to answer a question like Who spoke to Piero?, the subject is moved to the low new information Focus position, and the right dislocated indirect object is moved to the low Topic position, according to Belletti’s analysis. 5. Frascarelli and Hinterhölzl (2007) and Bianchi and Frascarelli (2009) further distinguish the contrastive Topic from an “aboutness shift” Topic that occupies an even higher position in the cartography of the left periphery.

451

9781441124609_Ch26_Final_txt_print.indd 451

11/27/2012 11:07:24 AM

27

Optimality Theory as a Possible Framework for the Minimalist Program1 Vieri Samek-Lodovici

Chapter Overview Introduction The Relation between the MP and OT OT as the Framework of Choice for Pursuing MP Goals Conclusions

452 452 455 462

1 Introduction This paper examines the relation between Optimality Theory (henceforth “OT,” Prince and Smolensky 1993, 2004) and the Minimalist Program (henceforth “MP,” Chomsky 1995, 2000, 2005), arguing that OT offers a useful theoretical tool for pursuing the core goals of the MP (see also Broekhuis and Dekkers 2000). I start in Section 2 showing that OT and the MP are logically independent of each other but also compatible with one another. Section 3 compares OT and non-OT analyses of Italian rightmost focus, arguing that the OT analysis best captures the MP’s core goals.

2 The Relation between the MP and OT At its core, the MP is a hypothesis about how language, and syntax in particular, is organized. As Chomsky put it, the MP emanates from the strong minimalist 452

9781441124609_Ch27_Final_txt_print.indd 452

11/20/2012 9:24:27 AM

Optimality Theory for the Minimalist Program

thesis that “language [is] an optimal solution to interface constraints that the faculty of language must satisfy,” providing “an optimal way to link sound and meaning” (Chomsky 2005: 3). If the MP thesis is correct, “Universal Grammar would be restricted to properties imposed by interface constraints” (ibid.). The MP thus provides us with a research program, not a specific theory of Universal Grammar or its syntactic component. Many different theories are potentially consistent with the strong minimalist thesis. As Chomsky recognized “the program is theory-neutral” (ibid.: 2, fn. 2). One important question for the MP concerns the relation holding between interface constraints. Are they compatible with each other, or do they conflict? This is an empirical question; we cannot know the answer a priori. It is a question worth pursuing, because it affects how we define and conceive grammaticality. If constraints at the interfaces are compatible with each other, those syntactic structures that satisfy all constraints can be considered grammatical, while any structure that fails even just one constraint is ungrammatical. If, on the other hand, constraints at the interfaces conflict with one another, then their simultaneous satisfaction becomes impossible, forcing a different theory of what it means to be grammatical. OT provides such a theory, as we will see shortly. There are theoretical and empirical reasons suggesting that constraints do conflict. First of all, if the interfaces are independent of each other, mutual consistency among interface constraints is unexpected. The null hypothesis is that at least some constraints will conflict. Second, current theories of phonology arguably provide us with the best understood model of what Chomsky calls the Sensory-Motor interface. If the MP is correct and syntactic properties follow from conditions at the interfaces, some of the constraints pertinent to the phonological interface should affect syntactic derivations (an example is provided in the next section). The last 15 years have shown that phonological constraints conflict with each other, hence even those phonological constraints relevant for syntax would be expected to conflict too. Third, many linguistic analyses presented as conflict-free on closer inspection turn out to involve constraint conflict, suggesting that the phenomena under study resist a conflict-free approach. An interesting case of this kind is provided in Chomsky (2005), where the formal foundations of phase theory include a No Tempering Condition (NTC) stating that the application of the operation “merge” to two syntactic objects X and Y leaves X and Y unchanged (ibid.: 5). Chomsky, however, also maintains that the NTC is systematically violated by tensed T heads, which inherit their tense and agreement features from C (ibid.: 10). If grammaticality required the satisfaction of all constraints, either the analysis of T heads or the NTC would have to be revised, or else any tensed sentence would be incorrectly predicted to be ungrammatical. 453

9781441124609_Ch27_Final_txt_print.indd 453

11/20/2012 9:24:27 AM

The Bloomsbury Companion to Syntax

Chomsky addresses this problem by proposing that these systematic NTC violations are imposed by constraints at the conceptual interface (C–I). More precisely, he observes that inheritance of tense and agreement features from C to T violates the NTC, but it also derives the crucial distinction between A and A’ positions. Since this distinction is required at the conceptual interface, the NTC violations satisfy an interface requirement and are therefore deemed necessary (ibid.: 10–11). Note how Chomsky’s analysis follows an optimality theoretic view of grammaticality: finite clauses are grammatical despite their systematic violations of the NTC and despite the NTC’s fundamental role in licensing structure building. Yet this does not entail that NTC violations are freely available: they are possible only when necessary to meet independent constraints at the conceptual interface. Put in OT terms, the constraints at the conceptual interface are ranked higher than the NTC, and any tensed clause that violates the NTC in order to satisfy these higher ranked constraints remains grammatical, as there is no other structure providing a be er solution for the conflict at hand. Chomsky’s analysis of tensed clauses shows that the MP can and should be pursued under an optimality theoretic view of grammaticality, at least under its original description, since constraint conflict and resolution are inherent to the role played by the NTC in the analysis of tense.

2.1 Optimality Theory OT provides a theory of constraint interaction, offering a formally rigorous definition of what structures count as grammatical in a model of grammar where constraints are allowed to conflict with one another. Under OT, linguistic constraints are universal and freely ranked, with higher ranked constraints taking precedence over lower ranked ones. Different rankings determine different resolutions of the available constraints conflicts. For example, if constraints A and B conflict and A outranks B (or, following OT conventions, “A»B”), B will be violated as much as is necessary to satisfy A.2 The opposite holds when B outranks A. Linguistic structures are grammatical with respect to a ranking of the conflicting constraints. A structure is grammatical when it performs optimally with respect to the ranked constraints, that is, when it violates all constraints minimally and always favors higher ranked constraints over lower ranked ones whenever a conflict arises. For example, if A and B are the only constraints available and A outranks B, a structure S that satisfies A and violates B once will be preferred to a competing structure that violates A and satisfies B. It will also be preferred to a structure that satisfies A and violates B twice (a more concrete example is discussed in the next section). Structure S provides the best possible 454

9781441124609_Ch27_Final_txt_print.indd 454

11/20/2012 9:24:27 AM

Optimality Theory for the Minimalist Program

solution to the conflict, violating B only as much as is necessary to satisfy the higher ranked A; as such, it provides an optimal solution to the conflict between A and B under the ranking A»B and therefore it is grammatical under this ranking. The other two structures are instead suboptimal under this ranking, and therefore ungrammatical. What structure is deemed grammatical depends on the specific constraint ranking being considered. Different rankings select different structures. Cross-linguistic variation thus follows from the different rankings available for the universal constraints governing human language, not from separate parametrized properties. Each ranking identifies a grammar, that is, a way to solve the conflicts that exist among UG constraints (detailed introductions to OT can be found in Prince and Smolensky 1993, 2004; McCarthy 2002. For OT syntax see Dekkers et al. 2000; Müller 2000; Legendre 2001; and Ackema and De Hoop 2006). OT and the MP are logically independent from one another. Neither implies the other and they may fail or succeed independently. If they are both incorrect, syntactic properties will not be derivable from interface constraints, and, in addition, constraints will turn out never to conflict. The Principles and Parameters analytical framework of the 1980s and early 1990s fits this description. Yet, it was later abandoned in favor of theories giving a central role to economy, such as Minimalism and Optimality Theory. It is therefore likely that either OT, or the MP, or both, provide a be er model. If both are correct, syntactic properties will prove to be derivable from conflicting constraints rooted in the phonological and conceptual interfaces (i.e. the Sensory-Motor and Semantic/ConceptualIntentional interfaces of Chomsky 2005).

3 OT as the Framework of Choice for Pursuing MP Goals This section considers in more concrete terms how OT can assist the pursuit of MP goals. It does so by briefly examining how different analyses provided under the OT and minimalist frameworks capture Zubizarreta’s observation that focus in Italian and other Romance languages falls clause rightmost because focus needs stress and main stress occurs clause rightmost in these languages (Zubizarreta 1994, 1998). Zubizarreta’s insight ought to be considered an important piece of evidence for the MP, as it shows that observable syntactic properties—namely the position of focused constituents—are sensitive to the location of prosodic prominence, that is, they are sensitive to a property governed by constraints pertinent to the Sensory-Motor interface rather than syntax, exactly as expected under Chomsky’s strict minimalist hypothesis. As I will show, however, only the OT analysis is able to capture the minimalist nature of Zubizarreta’s observation. 455

9781441124609_Ch27_Final_txt_print.indd 455

11/20/2012 9:24:27 AM

The Bloomsbury Companion to Syntax

3.1 Conflicts with Interface Constraints In Italian the distribution of constituents in a clause—subjects, objects, indirect objects, and adjuncts—is sensitive to their discourse status. Consider for example, the distribution of subjects. Unfocused subjects occur in specTP, hence preceding verbs whenever the la er provide new information, see (1a). Focused subjects, instead, occur in postverbal position as illustrated in (1b) (Antinucci and Cinque 1977; Calabrese 1992; Belle i and Shlonsky 1995; Samek–Lodovici 1996; Zubizarreta 1998). (1) (a) Discourse-given subjects: (b) Focused subjects:

[TP S V  ] [TP  V SF]

This distribution is more rigid than often realized. Consider, for example, a scenario where Mary has been taken to a hospital in a coma. A friend calls her husband asking the general question “Any news?” and is told that Mary has finally spoken, suggesting that the coma has ended. The sentences in (2) show three distinct syntactic structures and we want to consider which of them is grammatical under this context. They differ in the position of the subject and in the position of main stress (shown in capitals). Due to the generality of the question “Any news?”, focus applies to the TP encompassing the entire clause in all three structures (the subscript “F” marks the focused constituent). Crucially, the only grammatical structure under this context is A, showing an unstressed preverbal subject in specTP. The stressed postverbal subject in B is ungrammatical because it forces a contrastively focused interpretation of the subject that is inconsistent with the fact that Mary is the only person being discussed. The same holds of the stressed preverbal subject in C. (2)

Q: Notizie? News? “Any news?” A: [Maria ha PARLATO]F. Mary has spoken “Mary SPOKE.” B: *[Ha parlato MARIA]F. has spoken Mary “MARY spoke.” C: *[MARIA ha parlato]F. Mary has spoken “MARY spoke.”

456

9781441124609_Ch27_Final_txt_print.indd 456

11/20/2012 9:24:27 AM

Optimality Theory for the Minimalist Program

If we change scenario and replace the initial question with one focusing only the subject by inquiring who spoke at some party conference, as in (3), we get a different grammaticality pa ern. The preverbal subject in A is ungrammatical, because it is unstressed and focus requires stress. Structure B with a clause-final stressed subject is now grammatical. Answer C may at first appear grammatical too. This is unsurprising, as Italian allows for focus fronting. We know, however, that these constructions place focus in a higher position than specTP (Rizzi 1997, 2004, this volume; Samek-Lodovici 2006, 2009).3 In so far as these analyses are correct, they exclude the possibility that a subject might focus in specTP. Therefore, if leave aside focus fronting and take C to stand for the syntactic structure with a stressed subject in specTP, rather than fronted to a higher position, then C, too, is ungrammatical. (3) Q: Chi ha parlato? Who has spoken? “Who spoke?” A: *[TP MariaF ha PARLATO]. B: [TP Ha parlato MARIAF]. Has spoken Mary. “MARY spoke.” C: *[TP MARIAF ha parlato]. As mentioned, Zubizarreta (1998) proposes a particularly insightful explanation of the above distribution. Unfocused subjects raise to specTP. Focused subjects, however, focus clause-finally because focus needs stress and stress in Italian falls clause-rightmost. Zubizarreta’s proposal potentially derives the syntactic distribution of focused phrases from constraints governing prosodic prominence, thus avoiding the need to govern focus through additional constraints within the syntactic component of grammar. The issue I wish to address here concerns what framework provides the best formalization for Zubizarreta’s explanation, where “best” assesses how well the chosen framework captures its minimalist import. In particular, I will compare the OT analysis in Samek-Lodovici (2005) with the phase-theoretic account in Kahnemuyipour (2004) and the minimalist analysis in Zubizarreta (1998), which predates phase-theory. Consider the OT analysis first. At its most essential, it involves the three constraints in (3). The constraint EPP favors movement of the subject to specTP (Chomsky 1982, 1995; Grimshaw 1997).4 The constraint Right-ip-Head governs the position of prosodic prominence within the intonational phrase (or “ip”) that encompasses the entire clause (Truckenbrodt 1995). Main stress coincides with the prosodic prominence peak, therefore for our purposes the constraint 457

9781441124609_Ch27_Final_txt_print.indd 457

11/20/2012 9:24:27 AM

The Bloomsbury Companion to Syntax

simply states that stress falls rightmost in the clauses. Finally, the constraint Stress–Focus states that focused phrases must be stressed (Truckenbrodt 1995). (4)

EPP =

Overtly realize the specifier of TP.

Right-ip-Head = Align the right boundary of every intonational phrase with its head. Stress-Focus = Focused phrases are stressed. As shown in tableaux (5) when the entire clause is focused, structure A is selected as optimal because it satisfies all three constraints. Structure B, with the subject stuck in specVP, is suboptimal because it violates EPP by leaving the specTP position unrealized. Structure C is also suboptimal because its stress does not occur rightmost in the intonational phrase encompassing the clause, thus failing Right-ip-Head. As in any OT tableaux, the optimal structure is marked by the symbol “ ,” and each constraint violation is marked by the star symbol “*.” (5) Clause-wide focus

Stress-Focus

Right-ip-Head

EPP

[TP Maria ha PARLATO]F A.       Mary has spoken B.   [TP ___ ha parlato MARIA]F C.     [TP MARIA ha parlato]F

* *

When focus is restricted to the subject, the three constraints conflict with each other and each structure violates a distinct constraint. As shown in tableaux (6), structure A violates Stress-Focus, because the focused subject is not stressed. Structure B violates EPP because specTP is left unrealized. Structure C violates Right-ip-Head because stress is not rightmost. If grammaticality required satisfaction of all constraints, none of these structures would be grammatical. Under OT, instead, the grammatical structure is the one that best satisfies the grammar of the language, that is, the one that complies best with higher constraints even if this determines further violations of lower constraints. In Italian, Stress-Focus and Right-ip-Head are ranked above EPP, making structure B grammatical. This ranking is also responsible for the ungrammatical status of A and C: they violate higher ranked constraints that B satisfies and therefore they are suboptimal. (A more extensive and detailed analysis is available in Samek-Lodovici 2005. Similar analyses have been applied to the interaction of syntax and prosody in other languages; see Büring and Gutierrez-Bravo 2002; Büring 2001; Szendröi 2001; and Dehé 2005). 458

9781441124609_Ch27_Final_txt_print.indd 458

11/20/2012 9:24:27 AM

Optimality Theory for the Minimalist Program

(6) Focused subjects A.  [TP MariaF ha PARLATO]      Mary has spoken

Stress-Focus

Right-ip-Head

*

B.   [TP ___ ha parlato MARIAF] C.     [TP MARIAF ha parlato]

EPP

* *

Even in the simplified form just described, the OT analysis captures the minimalist import of Zubizarreta’s results: the syntactic position of focused subjects is driven by independent constraints whose nature is non-syntactic, such as Stress-Focus and Right-ip-Head. In particular, the constraint Right-ip-Head was first proposed on the basis of prosodic evidence alone (see Truckenbrodt 1995, 1999, based on Prince 1983; Nespor and Vogel 1986; Halle and Vergnaud 1987; Hayes 1995; and Selkirk 1995). It governs the position of prosodic prominence and therefore it belongs to the Sensory-Motor interface, not syntax. What determines the overall syntactic distribution of focused and unfocused subjects is the conflict between Right–ip-Head and EPP unleashed by the need to provide focus with stress. What OT provides is the formal framework for modeling this conflict while preserving the non-syntactic nature of some of the constraints involved. The result is an analysis where constraints at the interfaces derive syntactic properties.

3.2 Conflict-free Analyses The available alternative analyses developed in frameworks that subscribe to the MP’s goals but not to constraint conflict appear unable to capture the role of prosody in driving the distribution of focus, thus failing to derive syntactic properties from interface constraints. Let me start with the phase-based analysis in Kahnemuyipour (2004), where main stress is assigned syntactically to the highest (i.e. leftmost) item of the phase spellee at the end of each phase (ibid.: 88). For example, the final stress on “BOOK” in (7a) below is assigned as follows: first the verb raises from V to v and the object “book” raises to the specifier of the aspectual phrase AspectP to get case. Next, the complement of v, which is phasal, is spelled out: this complement is the AspectP phrase shown in (7b). Finally, main stress is assigned by the stress rule to the highest item of the spelled-out phrase, namely “book.” (7) (a) Ali bought a BOOK. (b) [Ali [vP bought [AspectP a BOOK [VP bought

a book ]]]] 459

9781441124609_Ch27_Final_txt_print.indd 459

11/20/2012 9:24:28 AM

The Bloomsbury Companion to Syntax

The analysis adheres to the classic Y-model where syntax feeds phonology, and hence also prosody, and the syntactic derivation precedes the assignment of prosodic prominence (ibid.:188). The flow of information presupposed by the Y-model, however, is at odds with the goals of the MP. An autonomous syntactic module that feeds the prosodic interface but is otherwise insulated from it by definition cannot have any of its properties determined by the prosodic interface. The effect is an inevitable dissociation of syntax from prosody, which in turn affects the analysis of Italian focus. Since the position of stress is identified syntactically, it is no longer possible to derive the position of focus from the position of stress. As mentioned, stress is assigned to the highest position available in the spellee. In order to get stressed, focus must move to such a position for independent reasons, thus separating the analysis of focus movement from that of stress assignment. Instead, focus moves in order to reach a CP-level focus projection à la Rizzi (1997, 2004, this volume), and this operation is then followed by remnant movement of the remaining unfocused phrase to a topic projection above focus. For example, a focused subject would raise to the higher focused projection as shown in (8a) but eventually appear postverbally due to the remnant movement of TP shown in (8b) (Kahnemuyipour 2004: 187–8). (8) MARIAF [TP ha parlato]] MARY  has spoken (b) [TopP [TP ha parlato] [FocP  MARIAF [TP ha parlato]]] (a) [FocP

It is worth noticing that the analysis incorrectly predicts negative subjects to be unable to focus in clause-final position. Like any other Italian postverbal negative phrase, postverbal negative subjects require licensing by a suitable c–commanding licenser— for example, a neg-marker—in the surface representation. For example, in (9a) the licensing neg-marker “non” (not) is mandatory. As (9b) shows, however, licensing is incorrectly predicted to be unavailable with postverbal focused negative subjects under the above analysis because they lie outside the TP preceding them and therefore are not c-commanded by the neg-marker in it (a similar point is made in Cardinale i 2001; see also Samek-Lodovici 2006, 2009). (9) (a) *(Non) ha parlato NESSUNOF  not has spoken anybody “NOBODY spoke.” (b) [TopP [TP non ha parlato] [FocP  NESSUNOF [TP non ha parlato]]] 460

9781441124609_Ch27_Final_txt_print.indd 460

11/20/2012 9:24:28 AM

Optimality Theory for the Minimalist Program

Similar observations hold for the minimalist analysis provided in Zubizarreta (1998) prior to the development of phase theory. Under this analysis, prosodic prominence is governed by the following two rules, simplified here for the purposes of this discussion. The Focus Prominence Rule (or “FPR”) assigns prosodic prominence to focused phrases, while the Nuclear Stress Rule (or “NSR”) places main stress on the rightmost item of the clause. Both rules apply at the final stages of the syntactic derivation (ibid.: 139– 41). When the items selected by the FPR and NSR rules do not coincide, the derivation proceeds until the discrepancy is resolved. Consider again a clause involving a focused subject. The subject is maintained to raise to the specifier of a higher focus projection, as in Kahnemuyipour’s analysis; see (10a) below. This causes a discrepancy between the two rules because the FPR would assign stress to the focused subject while the NSR would assign stress to the past participle “parlato” (spoken). In order to avoid this discrepancy, TP left-adjoins to the focus projection thus leaving the focused subject rightmost in the clause as shown in (10b) (ibid.: 138). The FPR and NSR rules fire again, but they now coincide, both assigning stress to the focus subject. The syntactic derivation thus ends, yielding a structure with a postverbal focused subject. (10) (a) [FocP MARIAF [TP ha parlato]] MARY has spoken (b) [FocP [TP ha parlato] [FocP  MARIAF [TP ha parlato]]] Like the preceding analysis, and for the same reasons, the analysis incorrectly predicts the impossibility of focused negative subjects in clause-final position. This ma er aside, the analysis differs from the previous one in that the relation between main stress and focus is exploited to determine the position of focus relative to the other constituents of the clause. The final operation moves TP further in order to match focus and stress. This may appear to comply with the aims of the MP, in that conditions on stress assignment affect syntactic properties, namely the position of focus. This, however, is not genuinely the case because the NSR is defined in syntactic terms and fully embedded within the syntactic derivation. As such, it is fully detached and independent from the constraints governing the interface with prosody (for more discussion on this point, see Samek-Lodovici 2005 and Rizzi, this volume). The syntactic module thus is self-governed and independent from the prosodic interface, rather than having its properties determined by it. In conclusion, both analyses define stress syntactically and do so in order to preserve the classic Y-model where syntax feeds prosody. This, however, insulates syntax from the prosodic interface, making it impossible to derive syntactic properties from conditions at the interface as required by the MP. 461

9781441124609_Ch27_Final_txt_print.indd 461

11/20/2012 9:24:28 AM

The Bloomsbury Companion to Syntax

4 Conclusions From a logical point of view, the MP and OT are independent of each other. Grammar may or may not be entirely reducible to interface constraints as maintained by the MP, and in both cases the constraints of grammar may or may not conflict with each other, as maintained by OT. In other words, OT and the MP concern fundamental aspects of the cognitive organization of human language that are independent from each other. Crucially, however, if constraints at the interfaces do conflict with each other, it is not possible to pursue the MP without a theory of constraint conflict such as OT. The analysis of Italian-focused subjects examined in the last section illustrates this point. A conflict-based approach to grammar allows us to derive syntactic properties from independently necessary prosodic constraints. The corresponding OT analysis, in turn, provides a formal account of exactly how prosody determines the syntactic distribution of Italian clause-final focus. Current conflict-free analyses of the same phenomenon require syntactic accounts of stress-assignment that further insulate syntax from the prosodic interface. While further research is welcome, these results suggest that the MP is best pursued under a conflict-based approach to grammar.

Notes 1. I would like to thank Hans Broekhuis, Ralf Vogel, Ellen Woolford, Hedde Zeijlstra, and the audience at the second DEAL conference at Leiden University in 2008 for useful comments and discussion on some of the concepts expressed in this paper. 2. Strictly speaking this only holds when A can be satisfied and no higher-ranked constraint conflicts with A. If another higher ranked constraint C conflicts with A, A might be violated in order to satisfy C, making it unnecessary to violate B for A’s sake. Note that in this case A is involved in two distinct conflicts, one with C and one with B. The way the conflict with C is resolved affects how the conflict with B is resolved. More generally, the complex way in which multiple constraint conflicts interact with each other makes a formally rigorous theory of constraint conflict such as OT absolutely necessary (thanks to Jane Grimshaw for useful comments about this important point). 3. The actual position of fronted focused phrases in Italian is under debate, ranging from the CP-level focus projection proposed by Rizzi (1997, 2004, this volume) to the TP adjoined position followed by a right-dislocated clause proposed in SamekLodovici (2006, 2009). Equally debated is the issue whether left-peripheral focus may apply to new-information focus or whether it is restricted to contrastive focus; see among others Brune i (2003), Belle i (2004a), and Samek-Lodovici (2005, 2006). 4. The constraint is rooted in Chomsky’s Extended Projection Principle, which requires the syntactic realization of the specTP position (Chomsky 1982, 1995).

462

9781441124609_Ch27_Final_txt_print.indd 462

11/20/2012 9:24:28 AM

Appendix—Resources in Syntax Maria Freddi and Silvia Luraghi

In this appendix, we briefly survey a number of electronic resources for corpus research on syntax. Clearly, this does not exhaust all possible types of resources: in principle, any type of written or oral text can be used for syntactic research, independent of the format in which it is available. In addition, one’s theoretical orientation also plays an important role in determining one’s need for specific resources. Thus, linguists who believe in introspection as the only safe procedure for describing a language’s syntax do not need resources for corpus research such as those listed here. The number of projects devoted to syntactic parsing has grown dramatically in the last few years, covering an increasing number of languages. The recently updated entry “Treebank” in Wikipedia lists the following languages: Arabic, Bulgarian, Catalan, Chinese, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Modern Greek, Ancient Greek, Hebrew, Hindi, Hungarian, Icelandic, Italian, Japanese, Korean, Latin, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Slovene, Spanish, Swedish, Thai, Turkish, Urdu, and Vietnamese (http://en.wikipedia.org/wiki/Treebank URL consulted on August 8, 2012). Remarkably, however, the Wikipedia article does not contain references to all syntactic parsed corpora surveyed here. Proliferation of Treebanks has led to a very large number of resources, which o en differ not only in the underlying theoretical model but also in the quantity and type of information included. As a consequence, standardization is the next challenge facing corpus research in this field. Given the number of current projects, it is impossible to give a full list of resources. In what follows, we therefore limit ourselves to a selection, which is, however, representative of state-of-theart research. The list also includes so ware developed within some of the major parsing projects. Among the pioneering parsed corpora are the Penn-Helsinki Treebanks, which makes use of constituent structure, and the Prague Dependency Treebank, based on syntactic dependencies. The Penn Treebank contains 3 million words of skeletally parsed text, over 2 million words of text parsed for predicate-argument structure (Taylor 2003). More details regarding the syntactic bracketing employed are given in Freddi (this volume). Based on the same kind of bracketing other corpora have been developed, notably the collection of historical 463

9781441124609_App_Final_txt_print.indd 463

11/20/2012 9:25:40 AM

The Bloomsbury Companion to Syntax

English corpora called the Penn-Helsinki Treebank Parsed Corpora of Historical English (www.ling.upenn.edu/hist-corpora/). These comprise the Penn-Helsinki Parsed Corpus of Middle English, (PPCME2), the Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME), and the Penn Parsed Corpus of Modern British English (PPCMBE) (Kroch et al. 2010), syntactically annotated corpora of prose text samples of English from the indicated time periods. The Old English period is covered by other interuniversity projects, which use texts sampled in the Helsinki Corpus of Diachronic English. Details on each corpus can be found in the following web pages: www-users.york.ac.uk/~lang18/pcorpus.html www-users.york.ac.uk/~lang22/YcoeHome1.htm www-users.york.ac.uk/~sp20/corpus.html Historical projects devoted to other languages that follow the same annotation standards include the Canadian Parsed Corpus of Historical French (Kroch and Santorini 2010) and the Tycho Brahe Corpus of Historical Portuguese (www.tycho. iel.unicamp.br/~tycho/corpus/en/index.html). A number of historical corpora for research on syntax of various Slavonic languages is being developed at the University of Regensburg; see http://rhssl1. uni-regensburg.de/SlavKo/korpus. Syntactic parsing is performed by means of the so ware EXMARaLDA and conforms to the standard of the Annis2 database (see below). The Prague Dependency Treebank, or PDT (http://ufal.mff.cuni.cz/pdt2.0/) is a corpus of Czech annotated for dependencies following the Prague school syntactic tradition (see Freddi this volume for more details). This includes three layers of annotation resulting in three different representations, namely the morphological (including the lemma and POS-tag), the analytical (syntactic dependency), and tectogrammatical (semantic and pragmatic layer). An example of the three-layered representation is given in Figure A1 below. Many other treebanks follow annotation standards variously derived from this. An example of a strict application of this standard is the Index Thomisticus (http://itreebank.marginalia.it/), which contains the works of Thomas Aquinas. The Ancient Greek and Latin Dependency Treebanks (http://nlp.perseus.tu s. edu/syntax/treebank/), currently under development at Tu s University, make use of the analytical layer only. The corpora include a sample of mostly Ancient Greek texts and some Latin authors. The annotated corpora are freely available online and a tutorial can be taken to familiarize with the annotation scheme. Interested users can volunteer to join the annotators team. The Proiel project (Pragmatic Resources in Old Indo-European Languages; www.hf.uio.no/ifikk/english/research/projects/proiel//) is a dependency treebank which also displays three levels of linguistic annotation. Besides syntactic annotation, Proiel provides morphological and pragmatic information; information 464

9781441124609_App_Final_txt_print.indd 464

11/20/2012 9:25:41 AM

Appendix—Resources in Syntax

atree .

rf

t-layer

jÍt PRED

les DIR3 .rf

a/lex.rf

a/aux

a/lex.rf

a/a

a-layer

AuxS

a/au

ux

.rf

x.rf

#PerPron ACT

.

šel

AuxK

Pred

Byl

by

AuxV

AuxV

do AuxP

lesa

Byl

Byl

šel

do

být

lesa

být

jit

do

les

VpYS---XR--AA-- Ve---------

w-layer

m-layer

Adv

Byl

by

. . z---------

VpYS---XR--AA-- RR---2------- NNIS2---A----

šel

dolesa

.

Figure A1 The Prague Dependency Treebank: three-layered representation

regarding zero anaphoras is incorporated into the syntactic level, unlike what is done in the PDT. An example of a Proiel dependency tree is given in Figure A2. This project started as a parallel corpus of the Greek text of the New Testament with its translations into the old Indo-European languages Latin, Gothic, Armenian, and Old Church Slavonic. It is currently being expanded to classical 465

9781441124609_App_Final_txt_print.indd 465

11/20/2012 9:25:41 AM

The Bloomsbury Companion to Syntax

VOC PRED λεψω

αμηv

OBL

COMP δΤi

üμiv

PRED καi

PRED

PRED

καi

καi PRED PRED

PRED PRED

iδαv AUX

PID εnεθuμησαv

OUK OBJ

SUB

OBJ

AUX

αKOUσαi

OUK

OBJ

XSUB

ATR Πoλλoi

SUB

OBJ α

καi

OBJ

OBJ

αKOUεΤε

SUB

βλἑΠεTε

ηκαuααv

XSUB XOBJ

SUB

XOBJ

iδεiV

δ

V

SUB

ΠρoϕῆTαi

SUB δikαioi

Figure A2 Proiel: dependency tree

Greek and Latin authors. Among the latter, Caesar’s De Bello Gallico has been ported from the Perseus Treebank by means of a specially designed conversion so ware (see Lee & Haug 2010). Among multilingual parallel corpora, SMULTRON (Stockholm MULtilingual TReebank; www.cl.uzh.ch/research/paralleltreebanks/smultron_en.html) is a 466

9781441124609_App_Final_txt_print.indd 466

11/20/2012 9:25:41 AM

Appendix—Resources in Syntax

parallel treebank containing English, German, and Swedish annotated with phrase structure trees. The trees have been aligned on sentence, phrase, and word levels. A parallel corpus based on the Prague Dependency Treebank is the Czech-English Dependency Treebank (Mikulová & Stepanek 2009). The Turku Dependency Treebank (http://bionlp.utu.fi/fintreebank.html) is a parsed corpus of Finnish which follows the Stanford Dependency Scheme. The latter derives typed dependency parses of English sentences from phrase structure parses. A typed dependencies parse represents dependencies between individual words, adding dependencies with grammatical relations labels such as subject or indirect object (de Marneffe et al. 2011; http://nlp.stanford.edu/ manning/papers/LREC_2.pdf). Arabic has been annotated following both the Penn (www.ircs.upenn.edu/ arabic/) and the Prague (http://ufal.mff.cuni.cz/padt/PADT_1.0/index.html) standards. An example of the latter Treebank is given in Figure A3. In addition, a treebank version of the Quran is available at http://corpus.quran.com/ treebank.jsp. Language-specific treebanks must take into account typological differences variously affecting some details in the annotation standards. An example is the Sinica Treebank of Chinese (http://turing.iis.sinica.edu.tw/treesearch/), which uses Information-based Case Grammar. The issue of typological diversity is addressed in Rios et al. (http://lotos.library.uu.nl/publish/articles/000262/bookpart.pdf), regarding a parallel treebank of Spanish and Quechua. The authors argue for a formalism based on Role and Reference Grammar as especially appropriate for Quechua. Another recent project concerns the development of a syntactically parsed corpus of premodern Japanese. The Oxford Corpus of Old Japanese, which is part of the project, is currently under construction (http:// vsarpj.orinst.ox.ac.uk/corpus/). Also, learner corpora have started to be annotated for syntactic relations. A recent outcome is the dependency learner corpus developed by Dickinson and Ragheb, building from the CHILDES dependency annotation scheme (http:// childes.psy.cmu.edu/grasp/). Among freely downloadable so ware for syntactic annotation are EXMARaLDA (www.exmaralda.org/index.html), and the tools (http://ufal.mff. cuni.cz/pdt2.0). An important project ANNIS2 (www.s 632.uni-potsdam.de/d1/annis/) aimed at designing so ware for the search and visualization of complex multilevel linguistic corpora with diverse types of annotation is being developed at the “Institut für deutsche Sprache und Linguistik” of the Humboldt University in Berlin. This includes PAULA, an Interchange Format for Linguistic Annotation (Potsdamer Austauschformat für Linguistische Annotation / Potsdam Interchange Format for Linguistic Annotation).

467

9781441124609_App_Final_txt_print.indd 467

11/20/2012 9:25:41 AM

The Bloomsbury Companion to Syntax

Figure A3 Arabic Treebank

468

9781441124609_App_Final_txt_print.indd 468

11/20/2012 9:25:41 AM

Glossary

A-position: A position that in the D-structure can be occupied by an argument (but not by a non-argument). Theta roles can be assigned to A-positions. In the sentence The girl reads the books the DPs the girls and the books are in A-positions. Accidental conjunction: Coordination relation of combination which involves entities that are not expected to co-occur, but whose combination is motivated by specific circumstances of the speech act. Adjunction: Structure building operation that preserves the phrasal status of its target. Adjuncts: Accessory constituents of a sentence that contribute additional information such as time, place, manner, etc. They are accessories since the sentence is complete without them. Adverbial clause: A subordinate clause that indicates a particular semantic relationship between two events. Adverbializer: A grammatical marker that is used to form adverbs, such as the English suffix -ly. Added to the adjectives quiet and fortunate, for example, it forms the adverbs quiet-ly and fortunate-ly. Agree: An operation in the syntax that creates a dependency between an interpretable and an uninterpretable feature. Alignment: The use of identical grammar for a range of conceptual or semantic categories, especially the grammar concerned with core arguments of a verb or clause. Types of alignment are Nominative–Accusative, Ergative–Absolutive, Tripartite, Neuter, and Split-S. AN: Adjective-Noun order. Anaphor: A noun phrase that must obligatorily co-refer with another noun phrase usually in the same clause for the sentence to be grammatical. Reflexive pronouns are anaphors. Anticausative: A syntactic derivation that removes the Subject (Agent) from the initial syntactic pattern. Antisymmetry: Principle by which all languages have underlying specifier-headcomplement word order in all types of phrases. Antitopic: Referring expression used to confirm the identity of a continuing topic, usually at the end of a sentence. In the English sentence He’s nice, that guy, the phrase that guy is an antitopic. Asymmetric V2: A language is defined as an asymmetric V2 language when the phenomenon is only found in main clauses, but not in embedded ones (except for a small class of clauses selected by bridge verbs). Atemporal conjunction: Coordination relation of combination in which the location of two combined States of Affairs along the time axis is not relevant to the combination itself.

469

9781441124609_Glos_Final_txt_print.indd 469

11/20/2012 9:25:25 AM

Glossary Attraction: A possible formulation of the operation Move/Internal Merge, according to which a head acts as a “magnet” for the item that is displaced. Auto-benefactive: Result of successive application of two elementary derivations: the benefactive (adding an Indirect Object to the set of arguments and the meaning “for [the sake of]” to the meaning of the base proposition) and the indirect reflexive (encoding the referential identity of the main argument of the initial structure (Subject) with Indirect Object). Bar-level: Identification of a given syntactic object as a head, an intermediate projection, or a maximal projection. Basic word order: Least marked word order for affirmative active main-clause statements with nominal rather than pronominal arguments. Behavioral phenomena: Syntactic processes such as NP deletion under co-reference, or control of the interpretation of reflexive pronouns, “floated” quantifiers, or possessors external to phrases containing possessed items. Binary branching: Property of syntactic structures in which mother nodes immediately dominate two constituents. Binding theory: The module of grammar that governs the possible links between a pronoun and a lexical noun or a quantifier. Bridge verbs: A special class of verbs which allow for V2 in their complement clause. They are usually verbs of saying or of believing. Building block model of syntax: The assumption in most syntactic theories that constructions are built up out of atomic units that exist independently of the constructions that define them. C-command: It stands for constituent command. It is a binary relation between nodes of a syntactic tree that reflects structural properties. Cartography of syntactic structures: The attempt to draw realistic maps of syntactic configurations. Case Theory: Also known as Abstract Case Theory, in generative grammar this is a theory according to which all noun phrases in a sentence must be case marked. The case of a noun phrase may be morphologically expressed as in Mary saw him, where the pronoun him has morphological accusative case. In Mary saw John, the noun phrase John has abstract accusative case. Causal clause: A subordinate clause denoting an event that is interpreted as the cause or reason of a related event. Causative construction: A linguistic expression denoting a complex situation consisting of two component events: (i) the causing event, and (ii) the caused event. Choice-aimed disjunction: Coordination relation of alternative in which the speaker’s aim is to ask for information about the individual elements: the speaker asks the hearer to make a choice. Clitic: A word that does not bear an accent and needs to lean phonologically on the preceding or following word. Clitic Left Dislocation: A common construction expressing Topic – Comment configurations in the Romance languages, involving a Topic construed with a resuptive clitic pronoun in the comment. Coding phenomena: Linguistic elements with phonological form such as case markers, pronouns, and agreement forms, as well as constituent order.

470

9781441124609_Glos_Final_txt_print.indd 470

11/20/2012 9:25:25 AM

Glossary Colligation: The repeated co-occurrence of grammatical categories in a corpus, for example, article + noun; also a colligational pattern. Collocation: The habitual co-occurrence of individual lexical items. Comparative method: a technique for studying the development of languages by performing a feature-by-feature comparison of two or more languages with common descent from a shared ancestor. Complement: A constituent of a sentence that is obligatorily required to complete the meaning of sentential elements such as verbs, nouns, adjectives, and prepositions. Complements differ from adjuncts in that complements are obligatory while adjuncts are optional. Complementizer: Element that introduces a complement or subordinate clause, such as English that in I’m thrilled that you came. In generative grammar it is abbreviated as C/CP. Complex sentence: A sentence consisting of two or more clauses, usually a main clause and one or more dependent clauses. An English example is She told me [that I should come], where the main clause is She told me, and the subordinate clause is that I should come. Computing parsimony: An unwarranted assumption in syntactic argumentation in which an analysis that minimizes processing operations is favored over other analyses without further motivation. Conditional clause: A subordinate clause denoting an event that is interpreted as the condition of a related event. Configurational languages: Languages with relatively fixed word order and hierarchic structure (e.g. English). This term is used in opposition with non-configurational languages. Configurationality: Term used to refer to whether a language has a hierarchic rather than a flat structure. Constituency: The hierarchic organization of syntactic structure within the clause. Constituent: A linguistic unit which is a component of a larger construction. Constraint conflict: The conflict that arises between the demands of distinct constraints when no syntactic or prosodic operation can create a structure that satisfies all constraints at once. Constraint ranking: An ordered set of constraints where the demands of higherordered constraints override those of lower-ranked ones whenever a conflict arises between them. Construct: In SBCG, a configuration of a mother sign and one or more daughter signs. Construct state (or Status Constructus) construction: In Semitic languages it is a nominal construction that expresses a genitival relation between the Head Noun (HN) and the Noun Phrase (arguably, a DP) without the mediation of a preposition. The HN occurs first in the DP, without article, in the construct state form, and inherits its (in)definiteness value from the complement. Construction: In SBCG, a feature structure that describes either a construct or a class of lexemes. Construction, combinatory: In SBCG, a description of a construct.

471

9781441124609_Glos_Final_txt_print.indd 471

11/20/2012 9:25:26 AM

Glossary Construction, derivational: In SBCG, a combinatory construction that describes a sign configuration in which the mother is a lexeme sign and the (sole) daughter is also a lexeme sign. Derivational constructions capture lexeme–lexeme relationships like active–passive. Construction, inflectional: In SBCG, a description of a combinatory construction that describes a sign configuration in which the mother is a word sign and the (sole) daughter is a lexeme sign. Inflection constructions include the Preterit construction, which licenses the construction of the verbal word whose form is laughed from the verbal lexeme whose form is laugh. Construction, lexical-class: In SBCG, a construction that describes properties common to a group of lexemes (e.g. proper names). Construction Grammar: A family of grammatical theories that are based on the idea that the primary unit of grammar is the grammatical construction. In Construction Grammar the grammar is held to consist of a set of lexical entries and an inventory of constructions, ranging from general phrase-structure rules to highly specific and idiomatic patterns. Containment: Type of constituenthood relation. Control: In generative grammar it is the relation holding between a noun phrase in the main clause and the missing subject of an infinitive verb (PRO). According to this theory, PRO is interpreted as identical to a noun phrase in the main clause, which is referred to as the controller of PRO. Controller: The argument that determines the interpretation of participant-coding grammar within a certain domain (e.g. determining the interpretation of reflexive and pronominal forms). Coordinate clause: A non-dependent and non-embedded sentence that can indicate a similar semantic relationship between two events. Coordinating construction: Any construction expressing a coordination relation, that is, a relation established between functionally equivalent entities, properties, or States of Affairs. Every construction expressing a coordination relation is a coordinating construction. Coordination relation: Relation established between functionally equivalent entities, properties, or States of Affairs, that is, elements having the same semantic function, autonomous cognitive profiles, and the same informational status. Corpus: Term used in linguistics to refer to a body of data. A corpus may consist of just a list of words, but more often the kinds of corpora (plural of corpus) that are of special interest to syntacticians consist of large samples of connected discourse. Corrective contrast: Coordination relation of contrast in which a conflict is determined by the substitution of an explicitly denied State of Affairs with a new one. Counterexpectative contrast: Coordination relation of contrast in which a conflict arises from the denial of some expectation generated by the first clause or by the context. CP expletives: Pronouns with no semantic import that only occur when the first position before the inflected verb is not occupied by any other constituent. Creak: Special voice quality resulting from glottalization.

472

9781441124609_Glos_Final_txt_print.indd 472

11/20/2012 9:25:26 AM

Glossary Cross-linguistic extension of language-specific analyses: An unwarranted assumption in syntactic argumentation in which an analysis of a construction in a language is extended to a similar but not identical construction in another language without further motivation. Cyclic principle: In generative grammar, a principle that allows multiple applications of the same rule to increasingly larger domains. Dative sickness (þágufallssýki): A traditional term in Icelandic for the generalization of the dative case at the expense of accusative case marking with subjects. Degrammaticalization (of middle): A diachronic process which suggests that a particular grammatical form (e.g. middle voice form), albeit physically preserved in the paradigm, loses a large part of its original grammatical functions (thus being partly lexicalized). Dependency tree: A tree diagram that shows the hierarchical relations holding between a verb and its dependent elements (usually arguments). Derivation: In generative grammar it is a representation of how rules and principles apply, step by step, to construct or generate a sentence. Descriptive adequacy: A way of evaluating the adequacy of a linguistic theory by seeing if it correctly accounts for the language being studied—for example, by seeing if the predictions of the theory match the judgments of mature native speakers as to which phrases and sentences are ill-formed, which are well-formed, and what range of meanings they have. Direct causation: A causative situation in which the caused event is temporally adjacent to the causing event, without any other event intervening between them. Direct-indirect object alignment: The use of identical grammar for the most patientlike argument of a simple transitive clause (P) and for the Theme argument of a three-argument clause (T), but different grammar for the most goal-like argument of a three-argument clause (G). Directive causation: A causative situation in which the causer draws upon a nonphysical—verbal or social—means in causing the causee to carry out an action or to undergo a change of condition or state. Directionality parameter: In X-bar theory, specification of whether the head precedes or follows its complement Discontinuous constituent: Multi-word constituent whose elements are not linearly adjacent within a sentence, but are separated from each other by intervening items. Discourse-given constituents: Constituents whose referents have been mentioned or can be inferred from the discourse context. Distributional method/analysis: A method of syntactic argumentation that uses the pattern of occurrence of syntactic units (words or phrases) in particular roles in a range of constructions in a language. Do-support: The appearance of a dummy auxiliary do in interrogative and other types of sentences. Dominance: Type of constituenthood relation in which one constituent dominates the other. In the VP constituent eats apples the verb eats dominates the direct object apples.

473

9781441124609_Glos_Final_txt_print.indd 473

11/20/2012 9:25:26 AM

Glossary DP (=Determiner Phrase): A phrase headed by a Determiner (i.e. article, demonstrative, possessive, quantifier). Elicitation: A technique of gathering linguistic data in which the researcher directly asks a native speaker whether certain linguistic examples are well-formed or not, and what they mean—often in a personal interview. The examples may be specifically designed to test particular hypotheses. Emphatic reflexive (intensifier): A linguistic expression signaling that the referent is to some degree unexpected in the discourse role or clausal role where it occurs. Empty category: In a parse tree, any element without corresponding overt phonetic forms in S-Structure (phonetically null place-holders, and hence inaudible or silent). The different types of empty categories depend on their syntactic properties. They can be base-generated (namely, pro and PRO) or derived from movement or deletion (traces). Enclitic: A backward-leaning clitic. Endocentricity: Property of syntactic objects that are internally headed. Ergative–absolutive alignment: The use of identical grammar for the core argument of an intransitive clause (S) and for the most patient-like argument of a transitive clause (P), but different grammar for the most agent-like argument of a transitive clause (A). Exceptional Case Marking Construction: A construction in which the lexical subject of an infinitival complement is assigned Case, normally accusative case, by the verb in the main clause. It corresponds essentially to the traditional accusative with infinitive construction. Explanatory adequacy: A way of evaluating the adequacy of a linguistic theory by seeing if it could plausibly be learned by a child who has access only to relatively simple naturally occurring data and (perhaps) innate knowledge applicable to any human language. Extended Projection Principle (EPP): Principle by which all senteces must have a subject. Extended Standard Theory (EST): Generative model that makes deep structures become closer to surface structures. Finite complement: A subordinate clause containing a tensed, conjugated verb. Fixed word order: (i) Word order that is the only (or about the only) possibility for a language’s sentences; (ii) Referring to a language in which a single word order is used for all or most sentences. Focus (or Focus constituent): An element expressing new information; it may be contrastive when it contrasts some natural expectation imputed to the interlocutor. Focus-movement: The positioning of a phrase that bears contrastive focus at the beginning of a clause. Force and finiteness: The specifications delimiting the complementizer system. Force expresses the clausal type, Finiteness agrees with the finite or non-finite character of the verbal morphology. Form-function isomorphism: (Weak) an unwarranted assumption in syntactic argumentation in which differences in form necessarily entail differences in underlying structure or in function; (strong) an unwarranted assumption in syntactic

474

9781441124609_Glos_Final_txt_print.indd 474

11/20/2012 9:25:26 AM

Glossary argumentation in which identity of form necessarily entails identity of underlying structure or of function. Frame: A corpus-based syntactic-semantic pairing for each word (as in the FrameNet project). Free ride principle: An unwarranted assumption in syntactic argumentation in which a theoretical construct established in one grammatical analysis is used without further motivation in some other analysis. Free word order: (i) Possiblity of having several common sentence word orders in a language; (ii) Referring to a language with multiple possible word orders. Fuzzy system: A grammatical system whose boundaries are not defined absolutely. Generality: An unwarranted assumption in syntactic argumentation in which a more general analysis is favored without further motivation. Generalized semantic roles: The broadest semantic role concepts that subsume narrower compatible semantic role types; for example, Actor may subsume Agent, Experiencier, Instrument, etc., while Undergoer may subsume Patient, Theme, Recipient, etc. Generative grammar: An approach to linguistics that uses explicit formal rules and representations in an effort to explain how speakers take a finite amount of knowledge and experience with a particular language and use it productively to create and interpret new linguistic utterances in that language. Genitive: In a broad sense the property of encoding the arguments of a Nominal Head. Global extension of construction-specific analyses: An unwarranted assumption in syntactic argumentation in which an analysis of a single construction in a language is extended to other constructions in the language without further motivation. GN: Genitive-Noun order. Harmony: Situation in which all phrasal heads are aligned with the verb, that is, in which all phrases are either head-initial or head-final. A language that displays such harmony can be said to be harmonic. Head-final: Referring to a phrase in which the head is final, or to a language with this pattern for all phrases. Head-initial: Referring to a phrase in which the head is initial, or to a language with this pattern for all phrases. Head marking and dependent marking: Refers to wherein the grammar relations among constituents are marked morphologically. For this purpose, the head of the clause is identified as the verb (predicate), and the dependents are other elements, particularly the arguments. If a language has pronominal affixes on the verb (predicate), for example, the pattern is said to be head marking: the grammatical roles of the arguments are marked on the head of the clause (see-I-him the boy). If the roles are marked on nouns with case affixes, however, the pattern is said to be dependent marking (see I-nominative boy-accusative). Also for these purposes, nouns are considered heads of noun phrases or determiner phrases, and modifiers are considered dependents. If the relation of possession is marked on the possessed noun (the head), the pattern is said to be head

475

9781441124609_Glos_Final_txt_print.indd 475

11/20/2012 9:25:26 AM

Glossary marking (boy his-shoe). If it is marked on the possessor, the pattern is said to be dependent marking (boy-genitive shoe). Host: A word to which a clitic attaches phonologically. The host and the clitic are called clitics group. Indirect causation: A causative situation in which the caused event does not immediately follow the causing event, with at least one event intervening between them. Indirect learning: Learning the complex features of a language by deducing them from simpler, more easily observable features of the language and how those interact with Universal Grammar and with one another. Inertia hypothesis: A hypothesis proposed by E. Keenan stating that syntactic change does not arise unless caused by external factors. Infinitive complement: A subordinate clause containing a tenseless, infinitive verb. Inflected infinitive: A verbal form that lacks tense, but has morphemes for person and number to indicate agreement with its subject. Inflected infinitives are also called personal infinitives. Innateness: The view that some knowledge relevant to language is automatically present in the minds of all normal humans by virtue of their intrinsic make up (e.g. genetic factors, etc.) rather than by their idiosyncratic experience and learning. Interpretable feature: A syntactic feature with a value that has a correlation on the semantic side. Also known as a valued feature. Intonational units (also intonational groups): Stretch of speech uttered under a single coherent intonational contour. At intonational unit boundaries, there is often pitch reset in the beginning of the unit, acceleration in tempo on the initial unstressed syllables, pausing, and prosodic lengthening of the final syllables. Islands: Constituents out of which it is impossible to extract anything. Label: Element that identifies the relevant grammatical properties of phrasal syntactic objects. Last Resort Condition: Condition by which, in the Minimalist Program, transformations must be driven by the need to check a feature. Left-peripheral focus: Focused constituents occurring at the left-periphery of the clause, whether clause-initially or preceding the subject, if present, or the head carrying tense and agreement inflection (i.e. the auxiliary or the finite verb). Lexeme: In SBCG, a feature structure that represents the syntactic category and meaning of a set of inflected words considered by speakers to be distinct manifestations of the same word, for example, run, runs, and ran. Lexical causative: The causative type involving suppletion, with no formal similarity between the basic verb and the causative verb. Lexicalization: Process by which a morphologically complex word or phrase becomes a recognized vocabulary item in a language. The English term lawn mower is lexicalized, but the phrase grass slicer is not. Linear restriction: Syntactic restriction according to which the inflected verb must be in second position after one constituent only in the clause.

476

9781441124609_Glos_Final_txt_print.indd 476

11/20/2012 9:25:26 AM

Glossary Listeme: A description of a word, lexeme, or phrase that is listed in the grammar; listeme is a generalization of the notion lexical entry to include not only words and lexemes but also multiword expressions of various kinds. Locality: An (often implicit) assumption in phrase-structure grammars by which phrase-structure rules are limited to describing the relationship between a mother node and one or more daughter nodes. Long distance scrambling: Movement operation that takes a phrase from an embedded finite clause to a higher one. Macro-comparative approach: A way of developing and testing linguistic hypotheses by investigating how they apply to a range of genetically and historically unrelated languages. Major constituent: Clauses are divided into two parts, the subject and the predicate, which form the major constituents of the clause. Manipulative causation: A causative situation in which the causer acts physically on the causee. Meaning Shift Unit (MSU): A Sinclairian notion central to a description of the phraseological tendency in language; the corpus-based unit of meaning identified by syntactic-semantic mapping. Merge: Structure building operation. An operation that combines lexical items and complex objects built from lexical items. It specifies the relevant properties of the resulting structure Methodological opportunism: A fallacy in syntactic argumentation in which certain constructional tests/criteria/arguments are arbitrarily selected to define a grammatical category or distinction, ignoring other constructional tests/criteria/ arguments, either within a language or across languages. Micro-comparative approach: A way of developing and testing linguistic hypotheses by comparing closely related languages and dialects, to see what varies across them in patterned ways (parametric clusters) and what does not vary. Middle (voice): A cluster of deagentivized (intransitivized) syntactic patterns encoded by the same (middle) morphology. Minimalist Program: The most recent incarnation of the goals of generative grammar, which puts particular emphasis on finding the sparest possible set of hypotheses that will account for the linguistic knowledge and capacities of humans. Morpheme order: Order of morphemes within a complex word. Morphological causative: The causative type involving a morphological process that applies to a basic verb, that is, affixation. Morphosyntactic leveling: the generalization of an unmarked variant in cases of morphosyntactic altnernation Multiple Agree: An operation in the syntax that creates a dependency between one interpretable feature and many uninterpretable features at the same time. Multiple wh-movement: A term used to describe a phenomenon in some languages whereby all wh-phrases in a sentence are moved. NA: Noun-Adjective order.

477

9781441124609_Glos_Final_txt_print.indd 477

11/20/2012 9:25:26 AM

Glossary Natural conjunction: Coordination relation of combination in which the entities habitually go together and can be said to form some conventionalized whole or conceptual unit. Neg-preposing: The positioning of a phrase headed by a negative word at the beginning of a clause. Negative concord: A phenomenon by which multiple negative expressions in the syntax jointly convey a single sentential negation in the semantics. Neutral alignment: The use of identical grammar for all three core arguments of intransitive and transitive clauses (S, A, P). New passive: a new construction in Icelandic which exhibits passive morphosyntax but preserves accusative case marking of the NP. NG: Noun-Genitive order. Node: (In tree) Any of the points of connection between branches in a tree-diagram; see parse tree. A node is said to dominate those nodes that emanate from it. (In concordance) A word or word form under scrutiny; also key-word-in-context. Nominative–accusative alignment: The use of identical grammar for the core argument of an intransitive clause (S) and the most agent-like argument of a transitive clause (A), but different grammar for the most patient-like argument of a transitive clause (P). Nominative substitution: The generalization of nominative case at the expense of oblique case marking with subjects. Non-configurational languages: Languages which present a flat, rather than hierarchic, structure. In non-configurational languages word order is governed mostly by pragmatic factors and not so much by syntactic constraints. These languages typically allow for discontinuous constituents (cf. configurational languages). Nonredundancy in representation: An unwarranted assumption in syntactic argumentation in which an analysis that is maximally nonredundant in syntactic representation is favored over other analyses, without further motivation. Examples include Storage parsimony and Syntagmatic nonredundancy. NP: Noun phrase. In word order typology, Noun-Postposition. Null complementation: The absence of a semantically licensed argument of a given predicator when that argument is contextually recoverable, as in, for example, I’ve already eaten (where an ‘ingestible’ item is presumed to exist but not mentioned). Object shift: Overt movement of pronominal objects to the left outside their VP. It might affect DPs optionally in some languages. Oblique subject: An NP exhibiting oblique (non-nominative) case marking which passes the crucial subject tests in a given language. Oppositive contrast: Coordination relation of contrast in which two States of Affairs or properties are set against each other because they show somehow antithetic features. OSV: Object Subject Verb order. OV: (i) Object preceding Verb (i.e., SOV, OVS, OSV); (ii) referring to a language with this word order. OVS: Object Verb Subject order.

478

9781441124609_Glos_Final_txt_print.indd 478

11/20/2012 9:25:26 AM

Glossary Parametric cluster: A set of properties that tend to covary across languages, appearing together, because of the way that linguistic principles interact. These are related to implicational universals and are what make indirect learning possible. Parse tree: A branching diagram that represents the syntactic structure of a sentence in which different branches connect nodes. Parsing: The segmentation of a sentence into its constituent groups or phrases; thence, generally the automatic assignment of syntactic roles by parsing programs or parsers. Part-of-speech tagging: The automatic assignment of a part-of-speech (a POS tag) to each word in a corpus by tagging programs or taggers. Partial Wh-movement: A term used to describe a phenomenon in some languages whereby a wh-phrase seems to be split in two different positions. Participant roles: Finely-grained semantic role concepts like “cutter,” “cut thing,” etc., determined by the meaning of specific predicates like cut, chop, slice, hack, etc. Phonological bootstrapping: The process whereby information from the phonological domain is used to crack the code of another linguistic domain, for example, syntax. Phrase: In generative grammar a phrase is a grammatical expression containing an obligatory nucleus or “head” that may or may not be accompanied by other elements related to the head of the phrase. In this theory a sentence is formed by a combination of phrases. Pivot: The argument that controls cross-clause grammatical phenomena concerned with participants, for example, deletion of a phrase from one clause when it designates a participant co-referent with a participant in another clause. PN: Preposition-Noun order. Possessive-reflexive: A subtype of auto-benefactive which encodes the referential identity of the main argument of the initial structure (Subject) with the possessor of another argument. Primary-secondary object alignment: The use of identical grammar for the most patient-like argument of a simple transitive clause (P) and the most goal-like argument of a three-argument clause (G), but different grammar for the Theme argument of a three-argument clause (T). Probe: A component of the syntactic operation “Agree,” which consists of searching in the c-command domain a constituent that bears features that allow the head directing the probe to valuate its own. Proclitic: A foreward leaning clitic. Purposive causative: The syntactic causative subtype involving two clauses, one representing an event carried out for the purpose of realizing another event denoted by the other. Purposive clause: A subordinate clause denoting an event that is interpreted as the goal or purpose of a related event. Raising: In generative grammar it refers to the movement of a phrase (e.g. noun phrase or verb phrase) from a subordinate clause to the main clause.

479

9781441124609_Glos_Final_txt_print.indd 479

11/20/2012 9:25:27 AM

Glossary Reanalysis: A mechanism that changes the underlying structure of a syntactic pattern and does not involve any modification of its surface manifestation. Reciprocal: A syntactic derivation that suggests the conjunction of the base proposition with its “symmetric” equivalent, where two of the arguments switch. Reflexive: A syntactic derivation that encodes the referential identity of the main argument of the initial structure (Subject) and some other argument (most often, Direct Object). Right dislocated constituents: Discourse-given constituents dislocated to a clauseexternal position at the right-periphery of the clause. Scope-discourse semantics: The scope of operators and the expression of discourserelated properties like Topic and Focus. Scrambling: Term used to describe word order variability at the sentence level (used of languages with free word order). In generative grammar, scrambling can be analyzed as a movement operation or as free order of merge in the sentence. Second position: Position after the first word or constituent phrase in a sentence (also called Wackernagel’s position). Second-position clitic: Clitic that appears in second position in a sentence (also called a Wackernagel clitic). Selection: In a generative grammar selection refers to a co-occurrence restriction regulating syntactic/semantic compatibility between constituents in a sentence. Semantic domain of a coordinating construction: The set of coordination relations that the construction may express. On the basis of this parameter, coordinating constructions may be classified as DEDICATED (monofunctional) or GENERAL (multifunctional). Semantic roles: Broad semantic concepts determined by the meaning inherent in verb types, and which subsume compatible participant roles; for example, Agent subsumes “cutter,” ”hugger”; but not ”cut thing” or “one who dies.” Sequential causative: The syntactic causative subtype in which the clause of cause strictly precedes the clause of effect, mirroring the temporal relation between cause and effect. Sequential conjunction: Coordination relation of combination characterized by the location of the linked States of Affairs along the same time axis at successive points, so that they are interconnected as parts of the same overall sequence of events. Sequentiality: The linear organization of elements, as opposed to organization based on hierarchic relations. Shallow parsing: A kind of parsing that uses less fine-grained constituent analysis than in a full parse (see also skeleton or flat parsing). Sign: In Sign-Based Construction Grammar, a feature structure representing the syntactic, pragmatic, morphological, and semantic properties of a listeme or phrase. Sign-Based Construction Grammar (SBCG): A subtype of Construction Grammar in which the grammar is taken to consist of a set of listemes and a set of constructions. Listemes and constructions license classes of linguistic objects (signs and constructs, respectively).

480

9781441124609_Glos_Final_txt_print.indd 480

11/20/2012 9:25:27 AM

Glossary Simple disjunction: Coordination relation of alternative in which the speaker’s intention is to talk about the set of options. Speakers do not elicit any choice, but they simply depict a set of possibilities. Simultaneous conjunction: Coordination relation of combination characterized by the absence of a reciprocal order in which the States of Affairs occur, since they happen to occupy the same point on the time axis. SOV: Subject Object Verb order. Specifier: Element within a constituent that precedes a head-plus-complement phrase, or that follows a complement-plus-head phrase. Split alignment/Split system: The use of multiple alignment systems in different sub-parts of the grammar of a language. Split-S alignment: The use of distinct grammar for core arguments of different intransitive verb or clause types (S’s), especially the treatment of some S’s identically to the most agent-like argument of a transitive clause (A) and others like the most patient-like argument of a transitive clause (P). Storage parsimony: An unwarranted assumption in syntactic argumentation in which an analysis that posits fewer basic categories, features, or operations is favored over other analyses without further motivation. Subcategorization: In formal approaches, it specifies the syntactic context in which a verb or another lexical item may be introduced in a sentence. Subject-Auxiliary-Inversion (SAI): The process whereby the subject is moved in front of the auxiliary verb as in English questions. Subject clitic inversion: In some languages subject inversion is restricted to clitic pronouns only. This is the case of modern Romance languages like French, but also some German dialects like Cimbrian. Subject inversion: The phenomenon according to which the subject occurs in a position immediately after the inflected verb. This is considered as a test to show that the inflected verb has moved to a position in the left periphery of the clause. Superiority: Condition on transformations. For example, when two or more whphrases coexist in the same sentence, the highest wh-phrase in the tree is moved by wh-movement. More technically, the highest wh-phrase in the c-command chain is the one that is displaced by wh-movement. SV: (i) Subject preceding Verb (i.e., SVO, SOV, VSO); (ii) referring to a language with this constituent order. SVO: Subject Verb Object order. Symmetric V2: A language is defined as a symmetric V2 language when the phenomenon is found both in main and in embedded clauses, generally after the complementizer. Symmetry: An unwarranted assumption in syntactic argumentation in which an analysis that is symmetric is favored over an asymmetric analysis without further motivation. Syntactic annotation: The encoding of syntactic information in a corpus. Syntactic argumentation: Evidence and reasoning provided to support an analysis of the grammatical structure of a sentence.

481

9781441124609_Glos_Final_txt_print.indd 481

11/20/2012 9:25:27 AM

Glossary Syntactic causative: The causative type with the expression of cause and that of effect occurring in separate clauses. Syntactic domain of a coordinating construction: The set of syntactic types (NPs, VPs, clauses) that may be linked in a given coordinating construction. Syntactic gradience: Boundary indeterminacy between syntactic categories. Syntactic pattern: A pattern of mapping of semantic arguments onto syntactic functions (grammatical relations). Syntactic reconstruction: The modeling of the syntax of prehistoric stages from which the syntax of attested languages can be derived. Syntactic tree: In generative grammar, a formal representation of the structural relations among the constituents of a sentence in the form of a tree diagram. Syntagmatic nonredundancy: An unwarranted assumption in syntactic argumentation in which an analysis of a sentence that minimizes redundant representation of grammatical information is favored over other analyses without further motivation. T-to-C movement: An analysis that seeks to account for the inversion of auxiliary and subject in interrogative questions and other types of sentences. The central notion is that the functional head T raises and adjoins to the functional head C. Temporal clause: A subordinate clause denoting an event that is temporally related to another event. “Top down” research: A “Galilean” style of research, in which new empirical hypotheses are developed not simply by observing data, but by logical deduction from the basic principles of the theory under consideration. Such hypotheses are then confirmed or disconfirmed, often by data gathered specifically to test those hypotheses (i.e. by experiments; in linguistics, often by elicitation). Topic: An element which is selected from the presupposed background, is made syntactically prominent, and about which a Comment is expressed. Trace: A place-holder for a displaced constituent. Treebank: Another term for a syntactically annotated corpus; also parsed corpus. Tree-diagram: Diagram that displays the internal hierarchic structure of clauses. The diagram is two-dimensional, its root is at the top and consists of the symbol S for sentence. S forms the topmost node, the root of the tree, from which the branches (NP, VP, etc.) emanate. In generative grammar it is nowadays common to consider CP the topmost node. Tripartite alignment: The use of distinct grammar for all three core arguments of intransitive and transitive clauses (S, A, P). Type hierarchy: In SBCG, a taxonomic classification of feature structures of the type sign. If a type B is a subtype of another type A, then feature structures of type B must satisfy all constraints that the grammar imposes on objects of type A as well as the grammatical constraints imposed on type B. Type hierarchies in SBCG take over functions formerly served by inheritance networks in construction-based grammars. Uninterpretable feature: A syntactic feature that does not have a correlation on the semantic side. Also known as an unvalued feature. Universal: Implicational statement of a correlation between several characteristics of languages, often word-order based.

482

9781441124609_Glos_Final_txt_print.indd 482

11/20/2012 9:25:27 AM

Glossary Universality of syntactic strutures: The assumption in some syntactic theories (such as generative grammar) that all languages have the same underlying syntactic structures. Universals: Properties that hold of all natural human languages, either in absolute terms (called Universal Grammar, an instance of innateness) or in conditional terms (implicational universals, related to the notion of parametric clusters). UTAH (Uniformity of Theta-Assignment Hypothesis): Hypothesis argued for at length in Baker 1988, which assumes that identical thematic relationships between predicates and their arguments are syntactically represented by identical structural relationships between those items at the level of D-structure. V2: Word order in which the verb appears after the first constituent of the sentence, or referring to a language with this word order (also known as verb second). Vacuous projection: Syntactic projection that has solely one immediate constituent. Valence: The repertoire of syntactic complements with which a predicator (verb, noun, or adjective) must combine. In SBCG, a predicator’s valence is represented as a list-valued feature, VAL(ENCE). Verb-final: Referring to a sentence word order in which the verb is final, or to a language with this word order. Verb first: This defines the position of the inflected verb, located in absolute clause initial position. Generally it is assumed that in V2 languages all cases of V1 have to be analyzed as including an empty constituent (either an operator or a null Topic) in front of the verb. Verb-initial: Referring to a sentence word order in which the verb is initial, or to a language with this word order. Verb-medial: Referring to a sentence word order in which the verb is neither initial nor final, or to a language with this word order. Verb second: See V2. Verb third: This defines the position of the inflected verb only, which is located after two constituents. Sometimes the term is also used to simply refer to a violation of the verb second constraint, although the verb might also be the fourth or the fifth constituent of the clause. VO: (i) Verb preceding Object (i.e. SVO, VSO, VOS); (ii) referring to a language with this constituent order. Voice: A regular encoding of syntactic patterns through the verbal morphology. VOS: Verb Object Subject order. VS: (i) Verb preceding Subject (i.e. VSO, VOS, OVS); (ii) referring to a language with this constituent order. VSO: Verb Subject Object order. Wackernagel clitic: Clitic that appears in second position in a sentence (also called a second-position clitic) (after Wackernagel 1892). Wackernagel’s position: Position after the first word or constituent phrase in a sentence (also called second position) (after Wackernagel 1892). Wh-in-situ: A term used to describe those wh-phrases that remain in the position where their semantic role is assigned.

483

9781441124609_Glos_Final_txt_print.indd 483

11/20/2012 9:25:27 AM

Glossary Wh-movement: A term used to describe the fact that wh-phrases in many languages appear displaced from the predicate that assigns them a semantic role. Wh-phrase: A DP headed by a wh-determiner. Word order: Order of the words or constituents in a larger phrase or construction, most often a sentence. Word order typology: Typology of languages based on word order and correlations between the word order in different types of phrases. X’: In X-bar theory, constituent consisting of a head (X) plus a complement, in either order (pronounced “X-bar”). X-bar theory: Idea that all types of phrases have the same structure, where the lexical category of the head of any phrase is represented by “X.” Within this theory, an X-phrase (XP) consists of a specifier plus a constituent X’ (read as “X-bar”), which contains the head X plus any complement: whether the head precedes or follows its complement is determined by a directionality parameter. XP: In X-bar theory, X-phrase—a phrase consisting of a specifier plus an X’ constituent.

484

9781441124609_Glos_Final_txt_print.indd 484

11/20/2012 9:25:27 AM

Bibliography

Aarts, Bas (2007), Syntactic Gradience: The Nature of Grammatical Indeterminacy, Oxford: Oxford University Press. Aarts, Bas, Joanne Close, and Sean Wallis (2012), “Choices over time: Methodological issues in current change,” in B. Aarts, J. Close, G. Leech, and S. Wallis (eds), The Verb Phrase in English: Investigating Recent Language Change with Corpora, Cambridge: Cambridge University Press. Abeillé, Anne (2003), “Introduction,” in A. Abeillé (ed.), Treebanks: Building and Using Parsed Corpora, Dordrecht: Kluwer, pp. xiii–xxvi. Abney, Steven (1987), The English Noun Phrase in Its Sentential Aspect. PhD dissertation, Massachusetts Institute of Technology, Cambridge, MA. Aboh, Enoch O. (2004), The Morphosyntax of Complement-Head Sequences, New York: Oxford University Press. Ackema, Peter and Helen De Hoop (2006), “Current issues in optimality theoretic syntax,” Linguistics, 44, 873–87. Acquaviva, Paolo (1993), The Logical Form of Negation: A Study of Operator-Variable Structures in Syntax. PhD dissertation, Scuola Normale Superiore, Pisa. Adams, James N. (1994), “Wackernagel’s law and the position of unstressed personal pronouns in Classical Latin,” TPhS, 92, 104–5. — (1996), “Interpuncts as evidence for the enclitic character of personal pronouns in latin,” Zeitschrift für Papyrologie und Epigraphik, 111, 208–10. Adger, David (2003), Core Syntax: A Minimalist Approach, Oxford: Oxford University Press. Agbayani, Brian and Chris Golston (2010), “Second-position is first-position: Wackernagel’s law and the role of clausal conjunction,” Indogermanische Forschungen, 115, 1–21. Aikhenvald, Alexandra Y. (2000), “Transitivity in Tariana,” in R. M. W. Dixon and A. Aikhenvald (eds), Changing Valency: Case Studies in Transitivity, Cambridge: Cambridge University Press, pp. 145–72. Aissen, Judith (1999), “Markedness and subject choice in optimality theory,” Natural Language and Linguistic Theory, 17, 673–711. Alexiadou, Artemis, Haegeman Liliane, and Melita Stavrou (2007), Noun Phrase in the Generative Perspective, Berlin: Mouton de Gruyter. Allan, Rutger (2003), The Middle Voice in Ancient Greek, Amsterdam: J. Gieben. Alsina, Alex and Sam Mchombo (1990), “The syntax of applicatives in Chichewa: Problems for a theta theoretic asymmetry,” Natural Language and Linguistic Theory, 8, 493–506. Andersen, Paul Kent (1991), A New Look at the Passive. Frankfurt am Main: Lang. — (1994a), Empirical Studies in Diathesis, Münster: Nodus Publikationen. — (1994b), “Remarks on Dionysios Thrax’s concept of ‘diathesis’,” Historiographia Linguistica, 21, 1–37. Anderson, Mona (1979), Noun Phrase Structure. PhD dissertation, University of Connecticut. Anderson, Stephen R. (1976), “On the notion of subject in ergative languages,” in C. N. Li (ed.), Subject and Topic, New York: Academic Press, pp. 1–24. — (1983), “A VP in a VSO language?” in G. Gazdar, E. Klein, and G. Pullum (eds), Order, Concord, and Constituency, Dordrecht: Foris Publications, pp. 9–55. — (1993), “Wackernagel’s revenge: Clitics, morphology, and the syntax of second position,” Language, 69, 68–98. — (2003), Aspects of the Theory of Clitics, Oxford: Oxford University Press.

485

9781441124609_Ref_Final_txt_print.indd 485

11/20/2012 5:54:04 PM

Bibliography Anderson, Stephen R. and Sandra Chung (1977), “On grammatical relations and clause structure in verb-initial languages,” in P. Cole and J. M. Sadock (eds), Syntax and Semantics 8: Grammatical Relations, New York: Academic Press, pp. 1–25. Andrews, Avery (1976), “The VP complement analysis in Modern Icelandic,” Proceedings of NELS, 6, 1–21. — (2007), “The major functions of the noun phrase,” in T. Shopen (ed.), Language Typology and Syntactic Description, Volume I: Clause Structure (2nd edn), Cambridge: Cambridge University Press, pp. 132–223. Antinucci, Francesco and Guglielmo Cinque (1977), “Sull’ordine delle parole in Italiano: l’emarginazione,” Studi di Grammatica Italiana, VI, 121–46. Applegate, Richard Brian (1972), Ineseño Chumash Grammar. PhD dissertation, University of California, Berkeley. Arregi, Karlos (2002), Focus in Basque. PhD dissertation, Massachussets Institute of Technology. Artiagoitia, Xabier (2010), “The DP Hypothesis in the Grammar of Basque,” MS, Universidad del País Vasco. Auer, Peter (2005), “Projection in interaction and projection in grammar,” Text, 25(1), 7–36. — (2007), “Why are increments such elusive objects? An afterthought,” Pragmatics, 17(4), 647–58. Austin, Peter and Joan Bresnan (1996), “Non-configurationality in Australian aboriginal languages,” Natural Language and Linguistic Theory, 14, 215–68. Avrutin, Sergey and Rosalind Thornton (1994), “Distributivity and binding in child grammar,” Linguistic Inquiry, 25, 265–71. Baauw, Sergio (1999), “The role of the clitic-full pronoun distinction in the acquisition of pronominal coreference,” in A. Greenhill, H. Littlefield, and C. Tano (eds), Proceedings of the 23rd Annual Boston University conference on language development. Somerville, MA: Cascadilla Press, pp. 32–43. Baauw, Sergio, Maria A. Escobar, and William Philip (1998), “A delay of Principle B effect in Spanish speaking children: The role of lexical feature acquisition,” in A. Sorace, C. Heycock, and R. Shillcock (eds), Language Acquisition: Knowledge Representation and Processing: Proceedings of GALA ‘97, Edinburgh, UK: Edinburgh University Press, pp. 16–21. Badecker, William and Kathleen Straub (2002), “The processing role of structural constraints on the interpretation of pronouns and anaphors,” Journal of Experimental Psychology: Learning, Memory and Cognition, 28, 748–69. Bailyn, John F. (2001), “On scambling: A reply to Boškovic and Takahashi,” Linguistic Inquiry, 32, 635–58. Baker, Mark (1988), Incorporation: A Theory of Grammatical Function Changing, Chicago: University of Chicago Press. — (1991), “On some subject/object non-asymmetries in Mohawk,” NLLT, 9(4), 537–76. — (1992), “Thematic conditions on syntactic structures: evidence from locative applicatives,” in I. M. Roca (ed.), Thematic Structure: Its Role in Grammar, Berlin: Foris, pp. 23–46. — (1996), The Polysynthesis Parameter, New York: Oxford University Press. — (2001), “Phrase structure as a representation of ‘primitive’ grammatical relations,” in W. D. Davies and S. Dubinsky (eds), Objects and other Subjects, Dordrecht: Kluwer, pp. 21–51. — (2002), The Atoms of Language, New York: Basic Books. — (2003), Lexical Categories: Verbs, Nouns and Adjectives, Cambridge: Cambridge University Press. — (2008), “The macroparameter in a microparametric world,” in T. Biberauer (ed.), The Limits of Syntactic Variation, Amsterdam: John Benjamins, pp. 351–74. — (2010), “Formal generative typology,” in B. Heine and H. Narrog (eds), Oxford Handbook of Linguistic Analysis, New York: Oxford University Press, pp. 285–312. Baker, Mark, Kyle Johnson, and Ian Roberts (1989), “Passive arguments raised,” Linguistic Inquiry, 20(2), 219–51.

486

9781441124609_Ref_Final_txt_print.indd 486

11/20/2012 5:54:05 PM

Bibliography Barðdal, Jóhanna and Thórhallur Eythórsson (2012a), “Reconstructing syntax: Construction grammar and the comparative method,” in H. C. Boas and I. A. Sag (eds), Sign-Based Construction Grammar, Stanford: CSLI Publications, pp. 267–321. — (2012b), “Hungering and lusting for women and fleshly delicacies: Reconstructing grammatical relations for Proto-Germanic,” Transactions of the Philological Society. 110 (3) Barnbrook, Geoff and John Sinclair (1995), “Parsing Cobuild entries,” in J. Sinclair, M. Hoelter, and C. Peters (eds), The Languages of Definition: The Formalisms of Dictionary Definitions for Natural Language Processing, Studies in Machine Translation and Natural Language Processing, Luxembourg: European Commission, pp. 13–58. Barnes, Michael (1986), “Subject, nominative and oblique case in Faroese,” Scripta Islandica, 37, 13–46. Bartonek, Antonín (2003), Handbuch des mykenischen Griechisch, Heidelberg: Winter. Bayer, Josef and Jaklin Kornfilt (1994), “Against scrambling as an instance of move-alpha2,” in N. Corver and H. van Riemsdijk (eds), Studies on Scrambling: Movement and Nonmovement Approaches to Free Word-Order Phenomena, Berlin: Mouton de Gruyter, pp. 17–60. Behaghel, Otto (1909) “Beziehungen zwischen Umfang und Reihenfolge von Satzgliedern,” Indogermanische Forschungen, 25, 110–42. — (1923–32), Deutsche Syntax: eine geschichtliche Darstellung, 4 vols, Heidelberg: Carl Winter. Belletti, Adriana (2001), “Inversion as focalization,” in A. Hulk and J. Y.Pollock (eds), Inversion in Romance, Oxford: Oxford University Press, pp. 60–90. — (2004a), “Aspects of the low IP area,” in L. Rizzi (ed.), The Structure of CP and IP, Oxford: Oxford University Press, pp. 16–51. — (ed.) (2004b), Structures and Beyond: The Cartography of Syntactic Structures, 3 vols, New York: Oxford University Press. — (2009), Structures and Strategies, London: Routledge. Belletti, Adriana and Carla Contemori (2010), “Intervention and attraction on the production of subject and object relatives by Italian (young) children and adults,” in J. Costa, A. Castro, M. Lobo, and F. Pratas (eds), Proceedings of Gala 2009, Cambridge: Cambridge Scolars Publishing. Belletti, Adriana and Luigi Rizzi (1988), “Psych-verbs and theta-theory,” Natural Language and Linguistic Theory, 6, 291–352. Belletti, Adriana and Ur Shlonsky (1995), “The order of verbal complements: A comparative study,” Natural Language and Linguistic Theory, 13, 489–526. Belloro, Valeria A. (2007), Spanish Clitic Doubling: A Study of the Syntax-Pragmatics Interface. PhD thesis, SUNY, Buffalo. Bellugi, Ursula (1971), “Simplification in children’s language,” in R. Huxley and E. Ingram (eds), Language Acquisition: Models and Methods, New York: Academic Press, pp. 95–119. Beltrami, Giulia (2010), Support Verb Constructions as Patterns of Use. A Corpus-based Study of British and American English. Unpublished Master’s thesis, University of Pavia. Benediktsdóttir, Ásbjörg (2008), Nýja þolmyndin: Fyrsta þolmyndun barna? [The New Passive: Children’s First Passive?]. BA thesis, University of Iceland. Benincà, Paola (1984), “Un’ipotesi sulla sintassi delle lingue romanze medievali,” Quaderni Patavini di Linguistica, 4, 3–19. — (1995), “Complement clitics in medieval romance: the Tobler-Mussafia law,” in A. Battye and I. Roberts (eds), Clause Structure and Language Change, Oxford: Oxford University Press, pp. 296–325. — (2006), “A detailed map of the left periphery of medieval romance,” in R. Zanuttini, H. Campos, E. Herburger, and P. Portner (eds), Negation, Tense and Clausal Architecture: Cross-Linguistics Investigations, Washington: Georgetown University Press, pp. 53–86. Benincà, Paola and Cecilia 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, 2 vols, New York/Oxford: Oxford University Press, pp. 52–75.

487

9781441124609_Ref_Final_txt_print.indd 487

11/20/2012 5:54:05 PM

Bibliography Bentzen, Kristine (2007a), Order and Structure in Embedded Clauses in Northern Norwegian. PhD dissertation, University of Tromsø. — (2007b), “The degree of verb movement in embedded clauses in three varieties of Norwegian,” in K. Bentzen and Ø. A. Vangsnes (eds), Nordlyd 34: Scandinavian Dialect Syntax, Tromsø: CASTL, pp. 127–46. Bentzen, Kristine and Westengaard Marit (2007), “The (non-) effect of input frequency on the acquisition of word order in Norwegian embedded clauses,” in I. Gülzow and N. Gagarina (eds), Frequency Effects in Language Acquisition. Defining the Limits of Frequency as an Explanatory Concept, Berlin: Mouton de Gruyter, pp. 271–306. Berinstein, Ava (1985), Evidence for Multiattachment in K’ekchi Mayan. New York: Garland. Bernstein, Judy (1991), “DPs in french and walloon: Evidence for parametric variation in nominal head movement,” Probus, 3(2), 101–26. Berretta, Monica (1985), “I pronomi clitici nell’italiano parlato,” in G. Holtus and E. Radtke (eds), Gesprochenes Italienisch in Geschichte und Gegenwart, Tübingen: Narr, pp. 185–224. Berruto, Gaetano (1980), La variabilità sociale della lingua, Torino: Loescher. Bhaskararao, P. and K. V. Subbarao (eds) (2004), Non nominative subjects, 2. vols, Amsterdam: John Benjamins. Bhatt, Rakesh M. (1999), Verb Movement and the Syntax of Kashmiri, Dordrecht: Kluwer. Bhatt, Rakesh and James Yoon (1991), “On the composition of comp and parameters of V-2,” WCCFL, 10, 41–52. Bianchi, Valentina (1999), Consequences of Antisymmetry: Headed Relative Clauses, Berlin: Mouton the Gruyter. Bianchi, Valentina and Mara Frascarelli (2009), “Is Topic a Root Phenomenon?” MS, University of Siena/University of Roma 3. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finnegan (1999), Longman Grammar of Spoken and Written English, London: Longman. Biberauer Theresa, Andrs Holmberg, and Ian Roberts (2010), A Syntactic Unviersal and its Consequences. MS, Universities of Cambridge and Newcastle. Biberauer, Theresa and Ian Roberts (2008), “Cascading parameter changes: Internallydriven change in Middle English and Early Modern English,” in Th. Eythórsson (ed.), Grammatical Change and Linguistic Theory: The Rosendal Papers, Amsterdam: Benjamins, pp. 79–113. — (2009), “The return of the subset principle,” in P. Crisma and G. Longobardi (eds), Historical Syntax and Linguistic Theory, Oxford: Oxford University Press, pp. 58–74. Bidese, Ermenegildo (2008), Die diachronische Syntax des Zimbrischen, Tübingen: Guenter Narr Verlag. Biktimir, Tuvana (1988), “Impersonal passives and the -ArAk construction in Turkish,” in Daniel Slobin and Karl Zimmer (eds), Studies in Turkish Linguistics, Amsterdam: John Benjamins, pp. 53–76. Bion, Ricardo A. H., Barbara Hoehle, and Michaela Schmitz (2007), “The role of prosody on the perception of word-order differences by 14-month-old German infants,” in Proceedings of ICPhS, Saarbrücken, Germany, pp. 1537–40. Blanche-Benveniste, Claire (1997), Approches de la langue parlée en français, Gap–Paris: Editions Ophrys. Bloom, Paul, Andrew Barrs, Janet Nicol, and Laura Conway (1994), “Children’s knowledge of binding and coreference: Evidence from spontaneous speech,” Language, 70, 53–71. Bloomfield, Leonard (1933), Language (reprinted 1984), New York: Holt. Bocci, Giuliano (2004), Contrastive focalisation on topics and preverbal subjects in italian: Syntax-free prosodic focalisation or syntactic movement to focP?” Rivista di Grammatica Generativa, 29, 3–39. — (2009), On Syntax and Prosody in Italian. PhD dissertation, University of Siena. Boeckx, Cedric (2008), Bare Syntax, Oxford: Oxford University Press. Boeder, Winfried (1969), “Über die Versionen des georgischen Verbs,” Folia Linguistica, 2, 82–152.

488

9781441124609_Ref_Final_txt_print.indd 488

11/20/2012 5:54:05 PM

Bibliography Bögel, Tina (2010), “Pashto (endo-)clitics in a parallel architecture,” Proceedings of the LFG10 Conference, Carleton University, Ottawa: CSLI Publications. Böhmová, Alena, Jan Hajič, Eva Hajičová, and Barbora Hladká (2003), “The Prague Dependency Treebank: A three-level annotation scenario,” in A. Abeillé (ed.), Treebanks: Building and Using Parsed Corpora, Dordrecht: Kluwer, pp. 103–27. Boley, Jacqueline (1989), The Sentence Particles and the Place Words in Old and Middle Hittite, Innsbruck: Universität. Borer, Hagit (1984), Parametric Syntax: Case Studies in Semitic and Romance Languages, Dordrecht: Foris Publications. — (1994), Deconstructing the Construct. MS, Amherst: University of Massachusetts. — (2003), “Exo-skeletal vs. endo-skeletal explanations: Syntactic projections and the lexicon,” in J. Moore and M. Polinsky (eds), The Nature of Explanation in Linguistic Theory, Stanford, CA: CSLI Publications, pp. 31–67. Borsley, Robert D. (2005), “Against conjP,” Lingua, 115, 461–82. Bošković, Želiko (1997), The Syntax of Nonfinite Complementation: An Economy Aproach, Cambridge, MA: Massachusetts Institute of Technology Press. — (2001), On the Nature of the Syntax-Phonology Interface: Cliticization and Related Phenomena, Amsterdam: Elsevier Science. — (2002), “On multiple Wh-fronting,” Linguistic Inquiry, 33, 351–83. — (2004), “Topicalization, focalization, lexical insertion, and scrambling,” Linguistic Inquiry, 35(4), 613–38. Bošković Željko and Daiko Takahashi (1998), “Scrambling and last resort,” Linguistic Inquiry, 29, 347–66. Bosque, Ignacio and Violeta Demonte (eds) (1999), Gramática Descriptiva de la Lengua Española, Madrid: Espasa Calpe. Bossong, Georg (1985), Empirische Universalienforschung: differentielle Objektmarkierung in den neuiranischen Sprachen, Tübingen: Gunter Narr. — (1998), “Vers une typologie des indices actanciels. Les clitiques romanes dans une perspective comparative,” in P. Ramat and E. Roma (eds), Sintassi Storica, Roma: Bulzoni, pp. 9–43. Brandi, Luciana and Patrizia Cordin (1989), “Two Italian dialects and the null subject parameter,” in O. Jaeggli and K. Safir (eds), The Null Subject Parameter, Dordrecht: Kluwer, pp. 111–42. Bresnan, Joan (1977), “Variables in the theory of transformations,” in P. Culicover, T. Wasow, and A. Akmajian (eds), Formal Syntax, New York: Academic Press, pp. 157–96. — (1978), “A realistic transformational grammar,” in M. Halle, J. Bresnan, and G. A. Miller (eds), Linguistic Theory and Psychological Reality, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 1–59. — (ed.) (1982a), The Mental Representation of Grammatical Relations, Cambridge, MA: Massachusetts Institute of Technology Press. — (1982b), “Passive in lexical theory,” in J. Bresnan (ed.), The Mental Representation of Grammatical Relations, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 3–86. — (2001), Lexical-Functional Syntax, Oxford: Blackwell. Bresnan, Joan and Jonni Kanerva (1989), “Locative inversion in Chichewa: A case study of factorization in grammar,” Linguistic Inquiry, 20, 1–50. Bresnan, Joan and Lioba Moshi (1990), “Object asymmetries in comparative Bantu syntax,” Linguistic Inquiry, 21, 147–85. Broadwell, George A. (2006), Choctaw Reference Grammar, Bloomington: Indiana University. Brody, M. (1990), “Some remarks on the focus field in Hungarian,” UCL Working Papers, 2 vols, University College of London. Broekhuis, Hans and Joost Dekkers (2000), “The minimalist program and optimality theory: Derivations and evaluations,” in J. Dekkers, F. van der Leeuw, and J. van de Weijer (eds), Optimality Theory. Phonology, Syntax, and Acquisition, Oxford: Oxford University Press, pp. 186–422.

489

9781441124609_Ref_Final_txt_print.indd 489

11/20/2012 5:54:05 PM

Bibliography Brown, Jerrold C. (1981), “The verbal art of Horace’s ode to Pyrrha,” Transactions of the American Philological Association (1974-), 111, 17–22. Brown, Roger (1973), A First Language. Cambridge, MA: Harvard University Press. Brown, Sue (1999), The Syntax of Negation in Russian, Stanford: CSLI Publications. Browning, Marguerite A. (1996), “CP recursion and that-t Effects,” Linguistic Inquiry, 27, 237–56. Brunetti, Lisa (2003), “Is there any difference between contrastive focus and information focus?” in M. Weisgerber (ed.), Proceedings of the Conference “sub7 – Sinn und Bedeutung,” Arbeitspapier Nr. 114, FB Sprachwissenschaft, Universität Konstanz, Germany. Büring, Daniel (2001), “Let’s phrase it! focus, word order, and prosodic phrasing in German double object constructions,” in G. Müller and W. Sternefeld (eds), Competition in Syntax, New York: Mouton de Gruyter, pp. 69–105. Büring, Daniel and Rodrigo Gutíerrez-Bravo (2002), “Focus-related word order variation without the NSR. A Prosody-based crosslinguistic analysis,” in S. Mac Bloscaidh (ed.), Syntax at Santa Cruz 3, Santa Cruz, CA: ICSC, pp. 41–58. Bybee, Joan (2002), “Sequentiality as the basis for constituent structure,” in T. Givón and B. Malle (eds), The Evolution of Language out of Pre-Language, Amsterdam / Philadelphia: John Benjamins, pp. 109–134. — (2007), Frequency of Use and the Organization of Language, Oxford: Oxford University Press. — (2010), Language, Usage and Cognition, Cambridge: Cambridge University Press. Bybee, Joan and Joanne Scheibman (1999), “The effect of usage on degrees of constituency: the reduction of don’t in English,” Linguistics, 37(4), 575–96. Bybee, Joan and Paul Hopper (2001), “Introduction to frequency and the emergence of linguistic structure,” in J. Bybee and P. Hopper (eds), Frequency and the Emergence of Linguistic Structure, Amsterdam/Philadelphia: Benjamins, pp. 1–24. Calabrese, Andrea (1992), “Some informal remarks on focus and logical structures in Italian,” in S. Kuno and H. Thráinsson (eds), Harvard Working Papers in Linguistics, 1, 91–127. Camacho, José (2003), The Structure of Coordination: Conjunction and Agreement Phenomena in Spanish and Other Languages, Dordrecht: Kluwer Academic Publishers. Campbell, Lyle (2000), “Valency-changing derivations in K’iche,” in R. M. W. Dixon, R. M. W., and, A. Y. Aikhenvald (eds), Changing Valency: Case Studies in Transitivity, Cambridge: Cambridge University Press, pp. 236–43. Cardinaletti, Anna (2001), “A second thought on emarginazione: destressing vs. ‘right dislocation’,” in G. Cinque and G. Salvi (eds), Current Studies in Italian Syntax: Essays Offered to Lorenzo Renzi, Amsterdam: Elsevier, pp. 117–35. Cardinaletti, Anna and Michael Starke (1999), “The typology of structural deficiency. A case study of the three classes of pronouns,” in H. van Riemsdijk (ed.), Clitics in the Languages of Europe, Berlin-New York: Mouton de Gruyter, pp. 145–232. Carlson, Robert (1994), A Grammar of Supyire, Berlin: Mouton de Gruyter. Castroviejo, Elena (2006), Wh-exclamatives in Catalan. PhD dissertation, Universitat Autònoma de Barcelona. Cecchetto, C. (1999), “A comparative analysis of left and right dislocation in romance,” Studia Linguistica, 53, 40–67. Cennamo, Michela (1993), The Reanalysis of Reflexives: A Diachronic Perspective, Napoli: Liguori. — (1998), “The loss of the voice dimension between late Latin and early Romance,” in M. Schmid, J. R. Austin, and D. Stein (eds), Historical Linguistics 1997, Amsterdam: Benjamins, pp. 77–100. Chafe, Wallace L. (1979), “The flow of thought and the flow of language,” in T. Givón (ed.), Syntax and Semantics 12: Discourse and Syntax, New York: Academic Press, pp. 159–181. — (1984), “How people use adverbial clauses,” Berkeley Linguistics Society, 10, 437–49. — (1987), “Cognitive constraints on information flow,” in R. Tomlin (ed.), Coherence and Grounding in Discourse, Amsterdam: John Benjamins, pp. 21–51.

490

9781441124609_Ref_Final_txt_print.indd 490

11/20/2012 5:54:06 PM

Bibliography — (1994), Discourse, Consciousness, and Time: The Flow and Displacement of Conscious Experience in Speaking and Writing, Chicago: University of Chicago Press. Chafe, Wallace L. (ed.) (1980), The Pear Stories: Cognitive, Cultural and Linguistic Aspects of Narrative Production, Norwood, NJ: Ablex. Chametzky, Robert A. (2000), Phrase Structure: From GB to Minimalism, Oxford: Blackwell. — (2003), “Phrase structure,” in R. Hendrick (ed.), Minimalist Syntax, Oxford: Blackwell, pp. 192–225. Cheng, Lisa and Norbert Corver (2006), Wh-movement: Moving on, Cambridge, MA: Massachusetts Institute of Technology Press. Chien, Yuchin and Kenneth Wexler (1990), “Children’s knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics,” Language Acquisition, 1, 225–95. Chomsky, Noam (1955), The Logical Structure of Linguistic Theory. PhD dissertation, University of Pennsylvania. — (1957), Syntactic Structures, The Hague: Mouton de Gruyter. — (1965), Aspects of the Theory of Syntax, Cambridge, MA: Massachusetts Institute of Technology Press. — (1970), “Remarks on nominalization,” in R. A. Jacobs and P. S. Rosenbaum (eds), Readings in English Transformational Grammar, Waltham, MA: Ginn and Company, pp. 184–221. — (1973), “Conditions on transformations,” in S. Anderson and P. Kiparsky (eds), Festschrift for Morris Halle, New York: Rinehart and Winston, pp. 232–86. — (1977) “On Wh Movement,” in A. Akmajian, P. Culicover, and T. Wasow (eds), Formal Syntax, New York: Academic Press, pp. 71–132. — (1980), Rules and Representations, New York: Columbia University Press. — (1981), Lectures on Government and Binding, Dordrecht: Foris. — (1982), Some Concepts and Consequences of the Theory of Government and Binding, Cambridge, MA: Massachusetts Institute of Technology Press. — (1986a), Barriers. Cambridge, MA: Massachusetts Institute of Technology Press. — (1986b), Knowledge of Language: its Nature, Origin, and Use, New York: Praeger. — (1989), “Some notes on economy of derivation and representation,” MIT Working Papers in Linguistics, 10, 43–74. — (1995), The Minimalist Program, Cambridge, MA: Massachusetts Institute of Technology Press. — (2000), “Minimalist inquiries: The framework,” in R. Martin, D. Michaels, and J. Uriagereka (eds), Step by Step. Essay on Minimalist Syntax in Honor of Howard Lasnik, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 89–156. — (2001), “Derivation by phase,” in M. Kenstowicz (ed.), Ken Hale: A Life in Language, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 1–52. — (2004), “Beyond explanatory adequacy,” in A. Belletti (ed.), Structures and Beyond: The Cartography of Syntactic Structures, 3 vols, New York: Oxford University Press, pp. 104–31. — (2005), On Phases. MS, Cambridge, MA: Massachusetts Institute of Technology. Christophe, Anne, Marina A. Nespor, Maria T. Guasti, and Brit Van Øyen (2003), “Prosodic structure and syntactic acquisition: the case of the head-direction parameter,” Developmental Science, 6, 211–20. Chumbow, Beban S. and Pius N. Tamanji (1994), “Bafut,” in P. Kahrel and R. van den Berg (eds), Typological Studies in Negation, Amsterdam: Benjamins, pp. 211–36. Cinque, Guglielmo (1980), “On extraction from NP in Italian,” Journal of Italian Linguistics, 5, 47–99. — (1990), Types of A’ Dependencies, Cambridge, MA: Massachusetts Institute of Technology Press. — (1999), Adverbs and Inflectional Heads. Oxford and New York: Oxford University Press. — (ed.) (2002), The Structure of CP and DP: The Cartography of Syntactic Structures, 1 vol, Oxford and New York: Oxford University Press.

491

9781441124609_Ref_Final_txt_print.indd 491

11/20/2012 5:54:06 PM

Bibliography — (2005), “Deriving Greenberg’s Universal 20 and its exceptions,” Linguistic Inquiry, 36, 315–32. — (2009), “The syntax of adjectives: Some comparative remarks,” Handout. Cinque, Guglielmo and Luigi Rizzi (2010), “The cartography of syntactic structures,” in B. Heine and H. Narrog (eds), The Oxford Handbook of Linguistic Analysis, Oxford and New York: Oxford University Press, pp. 51–65. Citko, Barbara (2005), “On the nature of merge: External merge, internal merge, and parallel merge,” Linguistic Inquiry, 36, 475–96. Clahsen, Harald, Claudia Kursawe, and Martina Penke (1995), “Introducing CP: Wh-questions and subordinate clauses in German child language,” in Essex Research Reports in Linguistics 7, Essex, England: University of Essex, Department of Linguistics, pp. 1–28. Cole, Peter (1982), “On defining bounding nodes for subjacency,” Linguistic Inquiry, 13, 139–45. — (1983), “The grammatical role of the causee in the universal grammar,” International Journal of American Linguistics, 49, 115–33. Collins, Chris (2002), “Eliminating labels,” in S. D. Epstein and T. D. Seely (eds), Derivation and Explanation in the Minimalist Program, Oxford: Blackwell, pp. 42–64. — (2004), A Smuggling Approach to the Passive in English. MS, Cornell University. Comrie, Bernard (1975), “Causatives and universal grammar,” Transactions of the Philological Society (1974), 1–32. — (1976), “The syntax of causative constructions: cross-language similarities and divergences,” in M. Shibatani (ed.), Syntax and Semantics 6: The Grammar of Causative Constructions, New York: Academic Press, pp. 261–312. — (1978), “Ergativity,” in W. Lehmann (ed.), Syntactic Typology, Austin: University of Texas Press, pp. 329–94. — (1985), “Causative verb formation and other verb-deriving morphology,” in T. Shopen (ed.), Language Typology and Syntactic Description, Vol. III, Grammatical Categories and the Lexicon, Cambridge: Cambridge University Press, pp. 309–48. Comrie, Bernard and Helma van den Berg (2006), “Experiencer constructions in Daghestanian languages,” in I. Bornkessel, M. Schlesewsky, B. Comrie, and A. Friederici (eds), Semantic Role Universals and Argument Linking: Theoretical, Typological, and Psycholinguistic Perspectives, Berlin: Mouton de Gruyter, pp. 127–54. Comrie, Bernard and Maria Polinsky (eds) (1993), Causatives and Transitivity, Studies in Language Companion Series, 23, Amsterdam: Benjamins. Conrad, Susan (2010), “What can a corpus tell us about grammar?” in A. O’Keeffe and M. McCarthy (eds), The Routledge Handbook of Corpus Linguistics, London: Routledge, pp. 227–40. Conroy, Anastasia, Eri Takahashi, Jeff Lidz, and Colin Phillips (2009), “Equal treatment for all antecedents: How children succeed with Principle B,” Linguistic Inquiry, 45, 446–86. Cooreman, Ann (1988), “The antipassive in Chamorro: variations on the theme of transitivity,” in M. Shibatani (ed.), Passive and Voice, Amsterdam: John Benjamins, pp. 561–95. Cormack, Annabel and Neil Smith (2005), “What is coordination,” Lingua, 115, 395–418. Cornilescu, Alexandra (1995), “Romanian genitive constructions,” in C. Guglielmo and G. Giusti (eds), Advances in Romanian Linguistics, Oxford: Oxford University Press, pp. 1–52. Corver, Norbert and Henk van Riemsdijk (1994), Studies on Scrambling: Movement and NonMovement Approaches to Free Word-Order phenomena, The Hague: Mouton. Couper-Kuhlen, Elizabeth and Margret Selting (2001), “Introducing interactional linguistics,” in M. Selting and E. Couper-Kuhlen (eds), Studies in Interactional Linguistics, Amsterdam: John Benjamins, pp. 1–22. Couro, Ted and Margaret Langdon (1975), Let’s Talk ‘Iipay Aa: An Introduction to the Mesa Grande Diegueño Language, Banning, CA: Malki Museum Press. Craig, Colette (1977), Jacaltec, Austin: University of Texas Press.

492

9781441124609_Ref_Final_txt_print.indd 492

11/20/2012 5:54:06 PM

Bibliography Crain, Stephen and Cecile McKee (1985), “Acquisition of structural restrictions on anaphora,” in S. Bernan, J. Choe, and J. McDonough (eds), Proceedings of NELS 15, Amherst: University of Massachusetts, GLSA, pp. 96–110. Creissels, Denis (2008), “Direct and indirect explanations of typological regularities: The case of alignment variations,” Folia Linguistica, 42, 1–38. — (2009), “Ergativity/accusativity revisited,” Paper presented at the Association of Linguistic Typology VIII, Berkeley, July 24–28, 2009. Crisma, Paola (1991), Functional Categories inside the NP: A study on the Distribution of Nominal Modifiers, Tesi di laurea, Università di Venezia. — (1999), “Nominals without the article in Germanic languages,” Rivista di Grammatica Generativa, 24, 105–25. — (2011), “On the so-called : ‘Indefinite article’,” paper presented at CGG21, University of Seville, April 7–9, 2011. Crisma, Paola and Giuseppe Longobardi (2009), “Change, relatedness and inertia in historical syntax,” in P. Crisma and G. Longobardi (eds), Historical Syntax and Linguistic Theory, Oxford: Oxford University Press, pp. 1–13. Crisma, Paola and Giuseppe Longobardi (eds) (2009), Historical Syntax and Linguistic Theory, Oxford: Oxford University Press. Cristofaro, Sonia (2003), Subordination Strategies. A Typological Study, Oxford: Oxford University Press. Croft, William (1990), Typology and Universals, Cambridge: Cambridge University Press. — (1995), “Intonation units and grammatical structure,” Linguistics, 33, 839–82. — (2001), Radical Construction Grammar: Syntactic Theory in Typological Perspective, Oxford: Oxford University Press. — (2003), Typology and Universals (2nd edn), Cambridge: Cambridge University Press. — (2009) “Methods for finding language universals in syntax,” in S. Scalise, E. Magni, and A. Bisetto (eds), Universals of Language Today, Berlin: Springer, pp. 145–64. — (2010), “Ten unwarranted assumptions in syntactic argumentation,” in K. Boye and E. Engberg-Pedersen (eds), Language Usage and Language Structure, Berlin: Mouton de Gruyter, pp. 313–50. — (To appear), “Hypothesis formation,” in S. Luraghi and C. Parodi (eds), A Companion to Syntax. Croft, William and D. Alan Cruse (2002), Cognitive Grammar, Cambridge: Cambridge University Press. Croft, William, Hava Bat-Zeev Shyldkrot, and Suzanne Kemmer (1987), “Diachronic semantic processes in the middle voice,” in P. Ramat et al. (eds) Historical Linguistics, ICHL 7 Proceedings, Amsterdam: Benjamins, pp. 179–92. Cruschina, Silvio (2008), Discourse-related Features and the Syntax of Peripheral Positions – A Comparative Study of Sicilian and other Romance Languages. Doctoral dissertation, University of Cambridge. Cruse, D. Alan (1992), “Monosemy vs. polysemy (review article on Ruhl, On monosemy),” Linguistics, 30, 577–99. Cruttenden, Alan (1986), Intonation, Cambridge: Cambridge University Press. Culicover, Peter and Ray Jackendoff (2005), Simpler Syntax, Oxford: Oxford University Press. Dahlstrom, Amy (1983), “Agent-Patient languages and split case marking languages,” Proceedings of the Ninth Annual Meeting of the Berkeley Linguistics Society, 37–46. — (1986), Plains Cree Morphosyntax. PhD dissertation, University of California, Berkeley. Davies, John (1981), Kobon, Amsterdam: North-Holland. Davies, William D. and Stanley Dubinsky (eds) (2001) Objects and Other Subjects, Dordrecht: Kluwer. Dayal, Veneeta (2006), “Multiple wh-questions,” in M. Everaert and H. van Riemsdjik (eds), The Blackwell Companion to Syntax, Oxford, UK: Blackwell, pp. 275–326. Deane, Paul (1991), “Limits to attention: A cognitive theory of island phenomena,” Cognitive Linguistics, 2, 1–63.

493

9781441124609_Ref_Final_txt_print.indd 493

11/20/2012 5:54:06 PM

Bibliography De Haan, Gerard (2001), “More is going on upstairs than downstairs: Embedded root phenomena in West Frisian,” Journal of Comparative Germanic Linguistics, 4, 3–38. Dehé, Nicole (2005), “The optimal placement of up and ab: A comparison,” Journal of Comparative Germanic Linguistics, 8, 185–224. Dekkers, Joost, Frank van der Leeuw, and Jeroen van de Weij (2000). Optimality Theory. Phonology, Syntax, and Acquisition, Oxford: Oxford University Press. DeLancey, Scott (1981), “An interpretation of split ergativity and related patterns,” Language, 57, 626–57. — (1984), “Transitivity and ergative case in Lhasa Tibetan,” Proceedings of the Tenth Annual Meeting of the Berkeley Linguistics Society, 131–40. — (2000), “The universal basis of case,” Logos and Language, 1, 1–15. — (2004), The Blue Bird of Ergativity. MS, University of Oregon Delbrück, Berthold (1878), Die altindische Wortfolge aus dem Catapathabrahmana. Halle: Waisenhauses. — (1893–1900), “Vergleichende Syntax der indogermanischen Sprachen,” in K. Brugmann and B. Delbrück, Grundriss der vergleichenden Grammatik der indogermanischen Sprachen, Strassburg: Karl J. Trübner. Delfitto, Denis and Paola Paradisi (2009), “For a diachronic theory of genitive assignment in romance,” in P. Crisma and G. Longobardi (eds), Historical Syntax and Linguistic Theory, Oxford: Oxford University Press, pp. 292–310. Delsing, Lars-Olof (2000), “From OV to VO in Swedish,” in S. Pintzuk, G. Tsoulas, and A. Warner (eds), Diachronic Syntax: Models and Mechanisms, Oxford: Oxford University Press, pp. 255–74. Den Besten, Hans (1983), “On the interaction of root transformations and lexical deletive rules,” in W. Abraham (ed.), On the Forma Syntax of Westgermania, Amsterdam: Benjamins. Déprez, Viviane (1994), “Parameters of object movement,” in N. Corver and H. van Riemsdijk (eds), Studies on Scrambling: Movement and Non-Movement Approaches to Free Word-Order Phenomena, The Hague: Mouton, pp. 101–53. Devine, Andrew M. and Laurence D. Stephens (2000), Discontinuous Syntax. Hyperbaton in Greek, Oxford/New York: Oxford University Press. Di Tullio, Ángela (2005), Manual de Gramática del Español. Buenos Aires: La Isla de la Luna. Diesing, Molly (1992), Indefinites, Cambridge, MA: Massachusetts Institute of Technology Press. — (2010), “Clitics revisited,” Procedings of Formal Approaches to Slavic Linguistics, 18, 104–24. Diesing, Molly and Eloise Jelinek (1995), “Distributing arguments,” Natural Language Semantics, 3(2), 123–76. Diessel, Holger (2001), “The ordering distribution of main and adverbial clauses: A typological study,” Language, 77, 343–65. — (2004), The Acquisition of Complex Sentences, Cambridge: Cambridge University Press. — (2005), “Competing motivations for the ordering of main and adverbial clauses,” Linguistics, 43, 449–70. — (2008), “Iconicity of sequence. A corpus-based analysis of the positioning of temporal adverbial clauses in English,” Cognitive Linguistics, 19, 457–82. Diessel, Holger and Katja Hetterle (2011), “Causal clauses: A cross-linguistic investigation of their structure, meaning, and use,” in P. Siemund (ed.), Linguistic Universals and Language Variation, Berlin: Mouton de Gruyter, pp. 21–52. Dik, Simon (1968), Coordination. Its Implications for the Theory of General Linguistic. Amsterdam: North-Holland. — (1978), Functional Grammar. Amsterdam: North Holland. Dirven, René (1973), “Emphatic and reflexive in English and Dutch,” Leuvense Bijdragen, 62, 285–99. Dixon, Robert M. W. (1972), The Dyirbal Language of North Queensland, Cambridge: Cambridge University Press. — (1977), A Grammar of Yidiny, Cambridge: Cambridge University Press.

494

9781441124609_Ref_Final_txt_print.indd 494

11/20/2012 5:54:07 PM

Bibliography — (1979), “Ergativity,” Language, 55, 59–138. — (1994), Ergativity, Oxford: Oxford University Press. — (2010), Basic Linguistic Theory, Vols 1, 2, Oxford: Oxford University Press. Dixon, Robert M. W. and Alexandra Y. Aikhenvald (eds) (2000), Changing Valency: Case Studies in Transitivity, Cambridge: Cambridge University Press. Donohue, Mark and Søren Wichmann (eds) (2008), The Typology of Semantic Alignment Systems, Oxford: Oxford University Press. Dover, Kenneth J. (1960), Greek Word Order, Cambridge: Cambridge University Press. Dryer, Matthew S. (1982), “Passive and inversion in Kannada,” in M. Macaulay and Orin D. Gensler (eds), BLS 8, pp. 508–35. — (1986), “Primary object, secondary object, and antidative,” Language, 62, 808–45. — (1992), “The Greenbergian word order correlations,” Language, 68, 81–138. — (1998), “Aspects of word order in the languages of Europe,” in A. Siewierska (ed.), Order in the Languages of Europe. European Science Foundation Language Typology series, Berlin: Mouton de Gruyter, pp. 283–319. — (2008a), “Order of subject and verb,” in M. Haspelmath, M. S. Dryer, D. Gil, and B. Comrie (eds), The World Atlas of Language Structures Online, Munich: Max Planck Digital Library, chapter 82 (online at http://wals.info/feature/82; accessed July 19, 2010). — (2008b), “Relationship between the order of object and verb and the order of adjective and noun,” in M. Haspelmath, M. S. Dryer, D. Gil, and B. Comrie (eds), The World Atlas of Language Structures Online, Munich: Max Planck Digital Library, chapter 97 (online at http://wals.info/feature/97; accessed July 19, 2010). — (2008c), “Order of subject, object and verb,” in M. Haspelmath, M. S. Dryer, D. Gil, and B. Comrie (eds), The World Atlas of Language Structures Online, Munich: Max Planck Digital Library, chapter 81 (online at http://wals.info/feature/81; accessed August 19, 2010). Dowty, David (1991), “Thematic proto-roles and argument selection,” Language, 67, 547–619. Dryer, Matthew S. (1982), “Passive and inversion in Kannada,” in M. Macaulay, et al (eds), BLS. Berkeley Linguistic Society, pp. 311–21. — (1986), “Primary objects, secondary objects, and antidative,” Language 62.4: 808–45. DuBois, John (1987), “The discourse basis of ergativity,” Language, 63, 805–55. DuBois, John and Robert Englebretson (2001–4), Santa Barbara Corpus of Spoken American English, Parts 1–3, Philadelphia: Linguistic Data Consortium. Duffield, Nigel (1993), “Irish construct state nominals and the radical pro-drop phenomenon,” in A. J. Shafer (ed.), Proceedings of NELS 23, University of Massachusetts: Amherst, GLSA, pp. 113–27. Egerland, Verner (2002), “Sull’omissione del pronome clitico oggetto in italiano antico,” Verbum, 4(2), 349–66. Elbourne, Paul (2005), “On the acquisition of principle B,” Linguistic Inquiry, 36, 333–65. Etxepare, Ricardo (2003), “Valency and arguments structure,” in J. I. Hualde and J. Ortiz de Urbina (eds), A Grammar of Basque, Berlin and New York: Mouton de Gruyter, pp. 363–426. Eythórsson, Thórhallur (2001), “The syntax of verbs in Early Runic,” Working Papers in Scandinavian Syntax, 67, 1–55. — (2002), “Changes in subject case marking in Icelandic,” in D. Lightfoot (ed.), Syntactic Effects of Morphological Change, Oxford: Oxford University Press, pp. 196–212. — (2008), “The new passive in Icelandic really is a passive,” in Th. Eythórsson (ed.), Grammatical Change and Linguistic Theory: The Rosendal Papers, Amsterdam: Benjamins, pp. 173–219. — (2009), “Stöðugleiki og breytingar í færeysku og íslensku: Beygingar og setningagerð” [Stability and change in Faroese and Icelandic: Morphology and syntax], in T. Sigurðardóttir and M. Snædal (eds), Frændafundur 6, Tórshavn: Fróðskapur, pp. 75–93. — (2011), “Variation in the syntax of the older runic inscriptions,” Futhark, 2, 27–49. Eythórsson, Thórhallur (ed.) (2008), Grammatical Change and Linguistic Theory: The Rosendal Papers, Amsterdam: Benjamins.

495

9781441124609_Ref_Final_txt_print.indd 495

11/20/2012 5:54:07 PM

Bibliography Eythórsson, Thórhallur and Jóhanna Barðdal (2005), “Oblique subjects: A common Germanic inheritance,” Language, 81, 824–81. — (2011), “Die Konstruktionsgrammatik und die komparative Methode,” in Th. Krisch and Th. Lindner (eds), Indogermanistik und Linguistik im Dialog: Akten der XIII. Fachtagung der Indogermanischen Gesellschaft vom 21. bis 27. September 2008 in Salzburg, Wiesbaden: Reichert Verlag, pp. 148–56. Faarlund, Jan Terje (1990), Syntactic Change: Toward a Theory of Historical Syntax, Berlin and New York: Mouton de Gruyter. — (2010), “Word order,” in S. Luraghi and V. Bubenik (eds), Continuum Companion to Historical Linguistics, London and New York: Continuum International Publishing Group, pp. 201–11. Fanselow, Gisbert (2001), “Features, theta-Roles, and free constituent order,” Linguistic Inquiry, 22, 405–37. — (2006), “Partial Wh-movement,” in M. Everaert and H. van Riemsdjik (eds), The Blackwell Companion to Syntax, Oxford: Blackwell, pp. 437–92. Farmer, Ann K. (1984), Modularity in Syntax. A Study of Japanese and English, Cambridge, MA: Massachusetts Institute of Technology Press. Ferraresi, Gisella and Maria Goldbach (eds) (2008), Principles of Syntactic Reconstruction, Amsterdam: John Benjamins. Fillmore, Charles (1968), “The case for case,” in E. Bach and R. T. Harms (eds), Universals in Linguistic Theory, New York: Holt, Rinehart and Winston, pp. 1–88. — (1977), “The case for case reopened,” in Peter Cole and Jerrold Sadock (eds), Syntax and Semantics 8: Grammatical Relations, New York: Academic Press, pp. 59–82. — (2009a), “FrameNet as a lexical resource: Its goals, methods and potential,” in FrameNet Masterclass, Milan, Catholic University of the Sacred Heart, December 3, 2009. — (2009b), “A valency dictionary of English,” Review Article, International Journal of Lexicography, 22(1), 55–85. Fillmore, Charles and Paul Kay (1995), Construction Grammar Coursebook. Unpublished MS, Department of Linguistics, University of California, Berkeley. Fiorentino, Giuliana (1999), Relativa debole. Sintassi, uso, storia in italiano (Weak Relative Clause. Syntax, Use, History of Italian), Franco Angeli: Milano. Fischer, Olga and Wim van der Wurff (2008), “Syntax,” in R. Hogg and D. Denison (eds), A History of the English Language, Cambridge: Cambridge University Press, pp. 109–98. Flobert, Pierre (1975), Les verbes déponents latins des origines à Charlemagne, Paris: Les belles lettres. Fontana, Josep (1993), Phrase Structure and the Syntax of Clitics in the History of Spanish. PhD dissertation, University of Pennsylvania, Philadelphia. — (1996), “Phonology and syntax in the interpretation of the Tobler-Mussafia law,” in A. L. Halpern and A. M. Zwicky (eds), Approaching Second. Second Position Clitics and Related Phenomena, Stanford: Center fort he study of Language and Information, pp. 41–83. Ford, Cecilia E. (1993), Grammar in Interaction: Adverbial Clauses in American English Conversations, Cambridge: Cambridge University Press. Ford, Cecilia E., Barbara A. Fox, and Sandra A. Thompson (2002), “Introduction,” in C. E. Ford, B. A. Fox, and S. A. Thompson (eds), The Language of Turn and Sequence, Oxford: Oxford University Press, pp. 3–13. Ford, Cecilia E. and Junko Mori (1994), “Causal markers in Japanese and English conversations: A cross-linguistic study of interactional grammar,” Pragmatics, 4, 31–61. Forrest, Linda (1999), The Role of Attention and Activation in Language Production: A Neurocognitive Linguistic Approach. PhD dissertation, University of Oregon. Fortescue, Michael (1984), West Greenlandic, London: Croom Helm. Fox, Barbara, Makoto Hayashi, and Robert Jasperson (1996), “Resources and repair: A crosslinguistic study of syntax and repair,” in E. Ochs, E. A. Schegloff, and S. A. Thompson (eds), Interaction and Grammar. Studies in Interactional Sociolinguistics, 13 vols, Cambridge: Cambridge University Press, pp. 185–237.

496

9781441124609_Ref_Final_txt_print.indd 496

11/20/2012 5:54:07 PM

Bibliography Fox, Barbara and Paul J. Hopper (eds) (1994), Voice: Form and Function, Amsterdam: Benjamins. Fox, Barbara and Robert Jasperson (1995), “A syntactic exploration of repair in English conversation,” in P. Davis (ed.), Alternative Linguistics: Descriptive and Theoretical Modes, Amsterdam and Philadelphia: John Benjamins, pp. 77–134. Fox, Barbara A. and Sandra A. Thompson (2007), “Relative clauses in English conversation: Relativizers, frequency, and the notion of construction,” Studies in Language, 31(2), 293–326. Fox, Danny and David Pesetsky (2004), “Linearization of syntactic structure,” Theoretical Linguistics, 31(1–2), 1–46. Fraenkel, Eduard (1932), “Kolon und Satz, I: Beobachtungen zur Gliederung des antiken Satzes,” Nachrichten der Göttinger Gesellschaft der Wissenschaften, Phil.-hist. Klasse, 197–213. — (1933), “Kolon und Satz, II: Beobachtungen zur Gliederung des antiken Satzes,” Nachrichten der Göttinger Gesellschaft der Wissenschaften, Phil.-hist. Klasse, 319–54. Frascarelli, Mara (2000), The Syntax-Phonology Interface in Focus and Topic Constructions in Italian, Dordrecht: Kluwer Academic Publishers. Frascarelli, Mara and Roland Hinterhölzl (2007), “Types of topics in German and Italian,” in S. Winkler and K. Schwabe (eds), On Information Structure, Meaning and Form, Amsterdam: Benjamins, pp. 87–116. Freidin, Robert (1992), Foundations of Generative Grammar, Cambridge, MA: Massachusetts Institute of Technology Press. Friedmann, Na’ama, Adriana Belletti, and Luigi Rizzi (2009), “Relativized relatives. Types of intervention in the acquisition of A-bar dependencies,” Lingua, 199, 67–88. Fuss, Erich (2008), Word Order and Language Change. On the Interface between Syntax and Morphology, Habilitationsschrift, University of Frankfurt. Garside, Roger, Geoffrey Leech, and Tamás Váradi (1992), Manual for the Lancaster Parsed Corpus (online at http://khnt.hit.uib.no/icame/manuals/LPC/LPC.PDF). Gazdar, Gerald, Ewan Klein, Geoffrey Pullum, and Ivan Sag (1985), Generalized Phrase Structure Grammar. Cambridge, MA: Harvard University Press. Gelderen, Elly Van (2008), “The negative cycle,” Linguistic Typology, 12, 195–243. Geniušienė, Emma (1987), The Typology of Reflexives, Empirical Approaches to Language Typology 2, Berlin: Mouton de Gruyter. George, Leland and Jaklin Kornfilt (1977), “Infinitival double passives in Turkish,” in J. Kegl, D. Nash, and A. Zaenen (eds), NELS, 7, pp. 65–79. Giacalone Ramat, Anna and Caterina Mauri (2011), “The grammaticalization of coordinating interclausal connectives,” in H. Narroge B. Heine (eds), Oxford Handbook of Grammaticalization, pp. 653–64, Oxford: Oxford University Press. Giannakidou, Anastasia (1997), The Landscape of Polarity Items. Doctoral dissertation, Groningen University. — (2006), “N-Words and negative concord,” in Martin Everaert and Henk van Riemsdijk (eds), The Blackwell Companion to Syntax, Volume III, Oxford: Blackwell, pp. 327–91. Gianollo, Chiara (2005), Constituent Structure and Parametric Resetting in the Latin DP: A Diachronic Study. Doctoral dissertation, Università di Pisa. Gildea, Spike (1993a), “The rigid VS order in Panare (Cariban): A historical explanation,” International Journal of American Linguistics, 59, 44–63. — (1993b), “On the evolution of a counter-universal pattern of split ergativity,” in Michael Bernstein (ed.), Proceedings of the Ninth Eastern States Conference on Linguistics, Ithaca: Cornell University, pp. 103–16. — (1994), “Semantic and pragmatic inverse: ‘Inverse alignment’ and ‘inverse voice’ in Carib of Surinam,” in T. Givón (ed.), Voice and Inversion, Amsterdam: John Benjamins, pp. 167–230. — (1998), On Reconstructing Grammar: Comparative Cariban Morphosyntax, Oxford: Oxford University Press.

497

9781441124609_Ref_Final_txt_print.indd 497

11/20/2012 5:54:07 PM

Bibliography Gildea, Spike and Flávia Alves (2010), “Nominative-absolutive: Counter-universal split ergativity in Jê and Cariban,” in S. Gildea and F. Queixalós (eds), Ergativity in Amazonia, Amsterdam: John Benjamins, pp. 159–200. Gildersleeve, Basil L. and Gonzalez Lodge (1913), Latin Grammar (3rd edn), New York: Macmillan. Gili Gaya, Samuel (1961), Curso Superior de Sintaxis Española, Barcelona: Bibliograf. Gramática RAE = Nueva gramática de la lengua española. Ginzburg, Jonathan and Ivan Sag (2001), Interrogative Investigations: The Form, Meaning and Use of English Interrogatives, Stanford, CA: CSLI publications. Giorgi, Alessandra (1983), “Toward a theory of long-distance anaphors,” The Linguistic Review, 3, 307–61. Giorgi, Alessandra and Giuseppe Longobardi (1991), The Syntax of Noun Phrases. Cambridge: Cambridge University Press. Givón, Talmy (1979), On Understanding Grammar, San Diego: Academic Press. — (1984), Syntax: A Functional-Typological Introduction, Volume 1, Amsterdam: John Benjamins. — (2001a), Syntax: An Introduction, Volume 1 (2nd edn), Amsterdam: John Benjamins. — (2001b), Syntax: An Introduction, Volume 2 (2nd edn), Amsterdam: John Benjamins. Gleason, Henry Allan Jr. (1955/1961), An Introduction to Descriptive Linguistics, New York: Holt, Rinehart & Winston. Goldberg, Adele (1995), Constructions: A Construction Grammar Approach to Argument Structure, Chicago: University of Chicago Press. — (2005), “Constructions, lexical semantics and the correspondence principle: Accounting for generalizations and subregularities in the realization of arguments,” in N. ErteschikShir and T. Rapoport (eds), The Syntax of Aspect, Oxford: Oxford University Press, pp. 212–36. — (2006), Constructions at Work, Oxford: Oxford University Press. Goldstein, David M. (2010), Wackernagel’s Law in Fifth-Century Greek. PhD thesis, University of California at Berkeley. Gonda, Jan (1979), The Medium in the Rigveda, Leiden: Brill. Goodluck, H. & Stonjanovic, D. (1996), “The structure and acquisition of relative clauses in Serbo-Croatian.” Language Acquisition, 5, 285–314. Goodwin, Charles (1981), Conversational Organization. Interaction between Speakers and Hearers, New York: Academic Press. Green, Georgia M. (1976), “Main clause phenomena in subordinate clauses,” Language, 52, 382–97. Greenberg, Joseph H. (ed.) (1963), Universals of Language, Cambridge, MA: Massachusetts Institute of Technology Press. — (1963), “Some universals of grammar with particular reference to the order of meaningful elements,” in J. H. Greenberg (ed.) (1963), Universals of Language, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 73–113. Grewendorf, Günther (2002), “Left dislocation as movement,” in S. Manck and J. Mittelstät (eds), Georgetown University Working Papers in Theoretical Linguistics, pp. 31–81. Grewendorf, Günther and Joachim Sabel (1994), “Long scrambling and incorporation,” Linguistic Inquiry, 25, 263–308. — (1999), “Scrambling in German and Japanese,” Natural Language and Linguistic Theory, 17, 1–65. Grewendorf, Günther and Wolfang Sternefeld (1989), Scrambling and Barriers, Amsterdam: John Benjamins. Grimes, Joseph E. (1975), The Thread of Discourse, The Hague: Mouton & Co. Grimshaw, Jane (1990), Argument Structure, Cambridge, MA: Massachusetts Institute of Technology Press. — (1997), “Projections, heads, and optimality,” Linguistic Inquiry, 28, 373–422. Grimshaw, Jane and Sara T. Rosen (1990), “Knowedge and obedience: The developmental status of the binding theory,” Linguistic Inquiry, 21, 187–222.

498

9781441124609_Ref_Final_txt_print.indd 498

11/20/2012 5:54:07 PM

Bibliography Grodzinsky, Yosef and Gitit Kave (1993/1994), “Do children really know Condition A,”? Language Acquisition, 3, 41–54. Grolla, Elaine (2005), “Resumptive pronouns as last resort: Implications for language acquisition,” in S. Arunachalam, T. Scheffler, S. Sundaresan, and J. Tauberer (eds), Proceedings of the 28th Annual Penn Linguistics Colloquium, Philadelphia, PA: University of Pennsylvania Working Papers in Linguistics, 11, pp. 71–84. Grosu, Alexander (1988), “On the distribution of genitive phrases in Rumanian,” Linguistics, 26, 931–49. Guardiano, Cristina (2003), Struttura e storia del sintagma nominale nel greco antico: ipotesi parametriche. PhD dissertation, Università di Pisa. — (2011), “Genitives in the greek nominal domain: Parametric considerations,” in M. Janse, B. Joseph, and A. Ralli, Proceedings of the 3rd International Conference on Modern Greek Dialects and Linguistic Theory, Cyprus, June 14–16, 2007. Guardiano, Cristina and Giuseppe Longobardi (2005), “Parametric comparison and language taxonomy,” in M. Batllori, M. L. Hernanz, C. Picallo, and F. Roca (eds), Grammaticalization and Parametric Variation, Oxford: Oxford University Press, pp. 149–74. Guasti, Maria Teresa (1996), “Acquisition of Italian interrogatives,” in H. Clahsen (ed.), Generative Perspectives on Language Acquisition, Amsterdam and Philadelphia: John Benjamin, pp. 241–69. — (2000), “An excursion into interrogatives in early English and Italian,” in M. A. Friedeman and L. Rizzi (eds), The Acquisition of Syntax, Singapore: Pearson Education Limited, pp. 105–28. — (2002), Language Acquisition. The Growth of Grammar, Cambridge, MA: Massachusetts Institute of Technology Press. Guasti, Maria Teresa and Anna Cardinaletti (2003), “Relative clause formation in Romance child’s production,” Probus, 15, 47–88. Guasti, Maria Teresa and Gennaro Chierchia (1999/2000), “Reconstruction in child language,” Language Acquisition, 8, 129–70. Guasti, Maria Teresa and Luigi Rizzi (1996), “Null Aux and the acquisition of residual V2,” in A. Stringfellow, D. Cahana-Amitay, E. Hughes, and A. Zukowski (eds), Proceedings of the 20 Boston University Conference on Language Development, Sommerville, MA: Cascadilla Press, pp. 284–95. Guasti, Maria Teresa, Rosalind Thornton, and Ken Wexler (1995), “Negation in children’s questions: The case of English,” in D. MacLaughlin and S. McEwen (eds), Proceedings of the 19 Boston University Conference on Language Development, Sommerville, MA: Cascadilla Press, pp. 228–39. Günthner, Susanne (1996), “From subordination to coordination? Verb-second position in German causal and concessive constructions,” Pragmatics, 6, 323–370. — (1999), “Wenn-Sätze im Vor-Vorfeld: ihre Formen und Funktionen in der gesprochenen Sprache,” Deutsche Sprache, 3, 209–35. Günthner, Susanne and Paul J. Hopper (2010), “Zeitlichkeit und sprachliche Strukturen: Pseudoclefts im Englischen und im Deutschen,” Gesprächsforschung, Online-Zeitschrift zur verbalen Interaktion, 11, 1–18. Haegeman Liliane (1991), Introduction to Government and Binding Theory. Oxford: Blackwell. — (1994), Introduction to Government and Binding, Oxford: Blackwell. — (1995), “Root infinitives, tense and truncated structures in Dutch.” Language acquisition, 4, 205–55. — (1995), The Syntax of Negation, Cambridge: Cambridge University Press. — (2002), “West Flemish negation and the derivation of SOV order in West Germanic,” Nordic Journal of Linguistics, 25, 154–89. — (2006), “Argument fronting in English, Romance ClLD and the left periphery,” in R. Zanuttini, H. Campos, E. Herburger, and P. Portner (eds), Negation, Tense and Clausal Architecture: Cross-linguistic Investigations, Georgetown: Georgetown University Press, pp. 27–52.

499

9781441124609_Ref_Final_txt_print.indd 499

11/20/2012 5:54:08 PM

Bibliography Haegeman, Liliane and Raffaella Zanuttini (1991), “Negative heads and the Neg criterion,” The Linguistic Review, 8, 233–51. — (1996), “Negative concord in West Flemish,” in A. Belletti and L. Rizzi (eds), Parameters and Functional Heads. Essays in Comparative Syntax, Oxford: Oxford University Press, pp. 117–79. Haegeman, Liliane and Terje Lohndal (2010), “Negative concord and multiple agree: A case study of West Flemish,” Linguistic Inquiry, 41, 181–211. Haiman, John (1978), “Conditionals are topics,” Language, 54, 564–89. — (1985a), Iconicity in Syntax, Amsterdam: John Benjamins. — (1985b), Natural Syntax, Cambridge: Cambridge University Press. Haiman, John and Sandra A. Thompson (1984), “Subordination’ in universal grammar,” Berkeley Linguistics Society, 10, 510–23. Hajičová, Eva (2009), “From Prague structuralism to Treebank annotation,” TLT8 International Workshop, Milan, Catholic University of the Sacred Heart, December 4–5, 2009. Hale, Kenneth (1981), On the Position of Warlpiri in a Typology of the Base, Bloomington: Indiana University Linguistics Club. — (1983), “Warlpiri and the grammar of non-configurational languages,” Natural Language and Linguistics Theory, 1, 5–47. — (1994), “Core structures and Adjunctions in Walpiri syntax,” in N. Corver and H. van Riemsdijk (eds), Studies on Scrambling: Movement and Non-Movement Approaches to Free Word-Order Phenomena, The Hague: Mouton, pp. 185–219. Hale, Kenneth and Samuel Jay Keyser (1997), “On the complex nature of simple predicators,” in A. Alsina, J. Bresnan, and P. Sells (eds), Complex Predicates, Stanford, CA: CSLI Publications, pp. 29–65. Hale, Kenneth and Danilo Salamanca (2002), “Theoretical and universal implications of certain verbal entries in dictionaries of the Misumalpan languages,” in W. Frawley, K. C. Hill, and P. Munro (eds), Making Dictionaries: Preserving Indigenous Languages of the Americas, Berkeley: University of California Press, pp. 25–59. Halle, Morris and Jean-Roger Vergnaud (1987), An Essay on Stress. Cambridge, MA: MIT Press. Halliday, M. K. A. (1994), An Introduction to Functional Grammar, London: Edward Arnold. Halliday, M. A. K. and Ruqaiya Hasan (1976), Cohesion in English, London: Longman. Halpern, Aaron (1995), On the Placement and Morphology of Clitics, Stanford: CSLI Publications. Halpern, Aaron and Arnold M. Zwicky (eds) (1996), Approaching Second: Second Position Clitics and Related Phenomena, Stanford: CSLI Publications. Hamann, Cornelia (2000), “The acquisition of constituent questions and the requirements of interpretation,” in M. A. Friedemann and L. Rizzi (eds), The Acquisition of Syntax, London: Longman, pp. 170–201. — (2006), “Speculations about early syntax: The production of Wh-questions bynormally developing French children and French children with SLI,” Catalan Journal of Linguistics, 5, 143–89. Hamann, Cornelia, Odette Kowalski, and William Philip (1997), “The French delay of principle B effect,” in M. Hughes and A. Greenhill (eds), Proceedings of the 21st Annual Boston University Conference on Language Development, Somerville, MA: Cascadilla Press, pp. 205–19. Hamblin, Charles (1973), “Questions in Montague English,” Foundations of Language, 10, 41–53. Harbert, Wayne (2007), The Germanic Languages, Cambridge: Cambridge University Press. Harris, Alice C. (1990), “Alignment typology and diachronic change,” in W. P. Lehmann (ed.), Language Typology 1987.Systematic Balance in Language. Papers from the Linguistic Typology Symposium, Berkeley, 1–3 December 1987, Amsterdam: John Benjamins, pp. 67–90. — (2008), “Reconstruction in syntax: Reconstruction of patterns,” in G. Ferraresi and M. Goldbach (eds), Principles of Syntactic Reconstruction, Amsterdam: John Benjamins, pp. 73–95.

500

9781441124609_Ref_Final_txt_print.indd 500

11/20/2012 5:54:08 PM

Bibliography Harris, Alice C. and Lyle Campbell (1995), Historical Syntax in Cross-Linguistic Perspective, Cambridge: Cambridge University Press. Harrison, S. P. (2003), “On the limits of the Comparative Method,” in B. D. Joseph and R. D. Janda (eds), The Handbook of Historical Linguistics, Oxford: Blackwell, pp. 343–68. Hashimoto, Mantaro J. (1988), “The Chinese passive construction,” in M. Shibatani (ed.), Passive and Voice, Amsterdam: John Benjamins, pp. 329–55. Haspelmath, Martin (1990), “The grammaticization of passive morphology,” Studies in Language, 14(1), 25–72. — (1995), “The converb as a cross-linguistically valid category,” in M. Haspelmath and E. König (eds), Converbs in Cross-linguistic Perspective. Structure and Meaning of Adverbial Verb Forms—Adverbial Participles, Gerunds, Berlin: Mouton de Gruyter, pp. 1–56. — (2003), “The geometry of grammatical meaning: Semantic maps and cross-linguistic comparison,” in M. Tomasello (ed.), The New Psychology of Language, Vol. II: Cognitive and Functional Approaches to Language Structure, Mahwah, NJ: Erlbaum, pp. 211–42. — (2004a), “Coordinating constructions: An overview,” in M. Haspelmath (ed.), Coordinating Constructions, Amsterdam: John Benjamins, pp. 3–39. — (2004b), “Explaining the Ditransitive Person-Role Constraint: A usage-based account,” Constructions, 2(49) (free online journal, University of Düsseldorf). — (2005), “Nominal and verbal conjunction,” in M. Haspelmath, M. S. Dryer, D. Gil, and B. Comrie (eds), The World Atlas of Language Structures, Oxford: Oxford University Press, pp. 262–5. — (2007), “Coordination,” in T. Shopen (ed.), Language Typology and Linguistic Description (2nd edn), Cambridge: Cambridge University Press, pp. 1–51. Haspelmath, Martin, Matthew Dryer, David Gil, and Bernard Comrie (eds) (2008), The World Atlas of Language Structures Online, Munich: Max Planck Digital Library. (online at http://wals.info/feature/39; accessed July 7, 2010). Haspelmath, Martin and Thomas Müller-Bardey (2004), “Valency change,” in G. Booij, C. Lehmann, J. Mugdan, and S. Skopeteas (eds), Morphology: A handbook on Inflection and Word Formation, 2 vols, Berlin: de Gruyter, pp. 1130–45. Hawkins, John A. (1979), “Implicational universals as predictors of word order change,” Language, 55, 618–48. — (1983), Word Order Universals, San Diego: Academic Press. — (1984), “Modifier-head or function-argument relations in phrase structure: The evidence of some word order universals,” Lingua, 63, 107–38. — (2001), “Why are categories adjacent?” Linguistics, 37, 1–34. Hayes, Bruce (1995), Metrical Stress Theory. Chicago, IL: The University of Chicago Press. Heidinger, Steffen (forthcoming), “French anticausatives in a diachronic perspective (with reference to other Romance languages),” in S. Kittilä and L. Kulikov (eds), Diachronic Typology of Voice and Valency-changing Categories, Amsterdam: Benjamins. Heine, Bernd and Tania Kuteva (2002), World Lexicon of Grammaticalization, Cambridge: Cambridge University Press. Helasvuo, Marja-Liisa (2001), Syntax in the Making: The Emergence of Syntactic Units in Finnish Conversation, Amsterdam and Philadelphia: John Benjamins. — (2003), “What can intonation tell us about constituency?” in T. Nakayama, T. Ono, and H. Tao (eds), Recent Studies in Empirical Approaches to Language, Santa Barbara Papers in Linguistics 12. Santa Barbara, CA: University of California, Santa Barbara, pp. 22–35. — (2004), “Shared syntax: The grammar of co-construction,” Journal of Pragmatics, 36, 1315–36. — (this volume), “Constituency.” Helasvuo, Marja-Liisa, Minna Laakso, and Marja-Leena Sorjonen (2004), “Searching for words: Syntactic and sequential construction of word searches in conversations of Finnish speakers with aphasia,” Research on Language and Social Interaction, 37(1), 1–37. Henry, Alison (1995), Belfast English and Standard English, Oxford, UK: Oxford University Press.

501

9781441124609_Ref_Final_txt_print.indd 501

11/20/2012 5:54:08 PM

Bibliography Higginbotham, James (1983), “Logical form, binding and nominals,” Linguistic Inquiry, 14, 395–420. Hinterhölzl, Roland (2006), Scrambling, Remnant Movement, and Restructuring in West Germanic, Oxford: Oxford University Press. Hiraiwa, Ken (2001), “Multiple agree and the defective intervention constraint in Japanese,” MIT Working Papers in Linguistics, 40, 67–80. Hiramatsu, Kazuko and Diane Lillo-Martin (1998), “Children who judge ungrammatical what they produce,” in A. Greenhill, M. Hughes, H. Littlefield, and H. Walsh (eds), Proceedings of the 22nd Annual Boston University Conference on Language Development, Somerville, MA: Cascadilla Press, pp. 337–47. Hirsh–Pasek, Kathryn and Roberta M. Golinkoff (1996), The Origins of Grammar, Cambridge, MA: Massachusetts Institute of Technology Press. Hock, Hans Henrich 1996. “Who’s on first? Toward a prosodic account of P2 clitics,” in Halpern and Zwicky (eds), Approaching Second: Second Position Clitics and Related Phenomena, Stanford: CSLI Press, pp. 199–270. Hockett, Charles (1958), A Course in Modern Linguistics, New York: Macmillan. Hoffner, Harry A. and H. Craig Melchert (2008), A Grammar of the Hittite Language. Winona Lake: Eisenbrauns. Holmberg, Anders (1986), Word Order and Syntactic Features in the Scandinavian Languages and English. PhD dissertation, Stockholm: University of Stockholm. — (1991), “The distribution of Scandinavian weak pronouns,” in H. van Riemsdijk and L. Rizzi (eds), Clitics and Their Hosts, University of Geneva and Tilburg, pp. 155–74. — (1999), “Remarks on ‘Holmberg’s Generalization’,” Studia Linguistica, 53, 1–39. Holmberg, Anders and Christer Platzack (1995), The Role of Inflection in Scandinavian Syntax, Oxford: Oxford University Press. Hopper, Paul (1979), “Aspect and foregrounding in discourse,” in T. Givón (ed.), Syntax and Semantics 12, New York: Academic Press, pp. 213–41. — (2001), “Grammatical constructions and their discourse origins: Prototype or family resemblance?” in M. Pütz and S. Niemeier (eds), Applied Cognitive Linguistics: Theory, Acquisition, and Language Pedagogy, Berlin: Mouton/De Gruyter, pp. 109–30. Hopper, Paul and Elizabeth Traugott (2003), Grammaticalization, Cambridge: Cambridge University Press. Hopper, Paul and Sandra A. Thompson (1980), “Transitivity in grammar and discourse,” Language, 56(2), 251–99. — (2008), “Projectability and clause combining in interaction,” in Ritva Laury (ed.), Crosslinguistic Studies of Clause Combining: The Multifunctionality of Conjunctions, Amsterdam: John Benjamins (Typological Studies in Language Series), pp. 99–123. Horn, Stephen W. and Kerri L. Russell (2010), The Oxford Corpus of Japanese (online at http:// vsarpj.orinst.ox.ac.uk/files/HornRussell.OI.pdf). Hornstein, Norbert (2009), A Theory of Syntax, Cambridge: Cambridge University Press. Hornstein, Norbert and Jairo Nunes (2008), “Adjunction, labeling, and bare phrase structure,” Biolinguistics, 2, 57–86. Hornstein Norbert, Jairo Nunes, and Kleanthes K. Gromann (2005), Understanding Minimalism, Cambridge: Cambridge University Press. Horvath, Julia (1986), FOCUS in the Theory of Grammar and the Syntax of Hungarian, Dordrecht: Foris. Hout, Theo P.J. van den (2010), “Studies in the Phraseological Construction, II. Its Origin.” Hethitica 16, 191–204. Hróarsdóttir, Thorbjörg (2000), Word Order Change in Icelandic: From OV to VO, Amsterdam: John Benjamins. Huang, James (1982), Logical Relations in Chinese and the Theory of Grammar, Doctoral dissertation, Massachussets Institute of Technology. — (1995), “Logical Form,” in G. Webelhuth (ed.), Government and Binding Theory and the Minimalist Program, Oxford, UK: Blackwell.

502

9781441124609_Ref_Final_txt_print.indd 502

11/20/2012 5:54:08 PM

Bibliography Huddleston, Rodney and Geoffrey K. Pullum (2000), The Cambridge Grammar of the English Language, Cambridge: Cambridge University Press. Hulk, Aa e and Zuckermann, Shalom (2000), “The interaction between input and economy: Acquiring optionality in French Wh-questions,” in S. Howell, S. Catherine, S. A. Fish, and T. Keith-Lucas (eds), Proceedings of the 24th Annual Boston University Conference on Language Development, Somerville, MA: Cascadilla Press, pp. 438–49. Hunston, Susan (2003), “Frame, phrase or function: A comparison of frame semantics and local grammars,” in D. Archer, P. Rayson, A. Wilson, and T. McEnery (eds), Proceedings of the Corpus Linguistics 2003 Conference, Lancaster University: UCREL Technical Paper no. 16, pp. 342–58. — (2009), “A corpus-driven lexical grammar of English: Observation and theory,” Anglistik: International Journal of English Studies, 20, 125–38. Hunston, Susan and Gill Francis (2000), Pattern Grammar. A Corpus-driven Approach to the Lexical Grammar of English, Amsterdam and Philadelphia: Benjamins. Hyman, Larry, and Alessandro Duranti (1982), “On the object relation in Bantu,” in Paul Hopper and Sandra Thompson (eds), Syntax and Semantics 15, Studies in Transitivity, New York: Academic Press, pp. 217–39. Ide, Nancy and Laurent Romary (2003), “Encoding syntactic annotation,” in A. Abeillé (ed.), Treebanks: Building and Using Parsed Corpora, Dordrecht: Kluwer, pp. 281–96. Jackendoff, Ray (1977), X-Bar Syntax: A Study of Phrase Structure, Cambridge, MA: Massachusetts Institute of Technology Press. — (1983), Semantics and Cognition, Cambridge, MA: Massachusetts Institute of Technology Press. — (1997), The Architecture of the Language Faculty, Cambridge, MA: Massachusetts Institute of Technology Press. Jacobson, Pauline (2006), “Constituent structure,” in Encyclopedia of Language & Linguistics, 2nd edn, Oxford: Elsevier, pp. 58–71. Jaeggli, Osvaldo (1986), “Passive,” in Linguistic Inquiry, 17(4), 587–622. Jakubowicz, Celia (1984), “On markedness and binding principles,” Proceedings of the Northeastern Linguistic Society, 14, 154–82. Jankuhn, Harald (1969), Die passive Bedeutung medialer Formen untersucht an der Sprache Homers, Göttingen: Vandenhoeck & Ruprecht. Janse, Mark (2000), “Convergence and divergence in the development of the Greek and Latin clitic,” in R. Sornicola, E. Poppe, and, A. Shisha-Halevy (eds), Stability, Variation and Change of Word-Order Patterns over Time, Amsterdam: Benjamins, pp. 231–58. Jelinek, Eloise (1984), “Empty categories, case, and configurationality,” Natural Language and Linguistic Theory, 2, 39–76. — (1996), “Definiteness and second position clitics in Straits Salish,” in Halpern, Aaron L. and Arnold M. Zwicky (eds), Approaching Second: Second Position Clitics and Related Phenomena, Stanford: CSLI Publications, pp. 271–97. Jespersen, Otto (1909–49), A Modern English Grammar on Historical Principles, 7 vols, London and Copenhagen: Allen and Unwin. Johannessen, Janne B. (1998), Coordination, Oxford: Oxford University Press. Johnson, Mark (1980), Ergativity in Inuktitut (Eskimo) in Montague Grammar, and in Relational Grammar, Bloomington: Indiana University Linguistics Club. Johnson-Laird, P. N. and Ruth M. J. Byrne (2002), “Conditionals: A theory of meaning, pragmatics, and inference,” Psychological Review, 109, 646–78. Joseph, Brian (2002), “‘Word’ in Modern Greek,” in R. M. W. Dixon and A. Aikhenvald (eds) Word: A Cross-Linguistic Typology, Cambridge: Cambridge University Press, pp. 243–65. Joseph, Brian and Irene Philippaki-Warburton (1987), Modern Greek, London: Croom Helm. Kahnemuyipour, Arsalan (2004), The Syntax of Sentential Stress. PhD dissertation, University of Toronto. Kaplan, Ronald M. and Joan Bresnan (1982), “Lexical-Functional Grammar: A formal system for grammatical representation,” in J. Bresnan (ed.), The Mental Representation of

503

9781441124609_Ref_Final_txt_print.indd 503

11/20/2012 5:54:09 PM

Bibliography Grammatical Relations, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 173–281. Karimi, Simin (2003), Word Order and Scrambling, Oxford: Wiley-Blackwell. — (2005), A Minimalist Approach to Scrambling: Evidence from Persian, Berlin: Mouton de Gruyter. Kärkkäinen, Elise, Marja-Leena Sorjonen, and Marja-Liisa Helasvuo (2007), “Discourse structure,” in Timothy Shopen (ed.), Language Typology and Syntactic Description, 2 vols, Cambridge: Cambridge University Press, pp. 301–71. Kaufman, Diana K. (1988), Grammatical and Cognitive Interactions in the Study of Children’s Knowledge of Binding Theory and Reference Relations. Doctoral dissertation, Temple University of Philadelphia, PA. Kay, Paul (2002), “English subjectless tag sentences,” Language, 78, 453–81. — (2004), Null Complementation Constructions. Unpublished MS, University of California, Berkeley. Kay, Paul and Charles J. Fillmore (1999), “Grammatical constructions and linguistic generalizations: The What’s X doing Y? construction,” Language, 75(1), 1–33. Kay, Paul and Laura Michaelis (forthcoming), “Constructional meaning and compositionality,” in C. Maienborn, K. von Heusinger, and P. Portner (eds), Semantics: An International Handbook of Natural Language Meaning, Berlin: Mouton de Gruyter. Kay, Paul and I. Sag (forthcoming), “How hard a problem would this be to solve?” in Proceedings of the Ninth Conference on Head-Driven Phrase Structure Grammar. Kayne, Richard. S. (1984), Connectedness and Binary Branching, Dordrecht: Foris. — (1989), “Null subjects and clitic climbing,” in O. Jaeggli and K. Safir (eds), The Null Subject Parameter, Dordrecht: Kluwer Academic Publishers, pp. 239–63. — (1994), The Antisymmetry of Syntax, Cambridge, MA: Massachusetts Institute of Technology Press. — (2002), “On some prepositions that look DP-internal: English of and French de,” Catalan Journal of Linguistics, 1, 71–116. — (2005), “Some notes on comparative syntax, with special reference to English and French,” in G. Cinque and R. Kayne (eds), The Oxford Handbook of Comparative Syntax, New York: Oxford University Press, pp. 3–69. Kazanina, Nina and Colin Phillips (2001), “Coreference in child Russian: Distinguishing syntactic and discourse constraints,” in A. H.-J. Do, L. Domínguez, and A. Johansen (eds), Proceedings of the 25th annual Boston University Conference for Language Development, Somerville, MA: Cascadilla Press, pp. 413–24. Kazenin, Konstantin (2001a) “The passive voice,” in M. Haspelmath et al. (eds), Language Typology and Language Universals. An International Handbook, 2 vols, Berlin etc.: Walter de Gruyter, pp. 899–916. — (2001b), “Verbal reflexives and the middle voice,” in M. Haspelmath et al. (eds), Language Typology and Language Universals. An International Handbook, 2 vols, Berlin etc.: Walter de Gruyter, pp. 916–27. Keenan, Edward L. (1972), “Relative clause formation in Malagasy (and some related and some not so related languages),” in P. M. Peranteau, J. N. Levi, G. C. Phares (eds), The Chicago Which Hunt, University of Chicago, CLS. pp. 169–89. — (1976a), “Toward a universal definition of ‘subject’,” in C. Li (ed.), Subject and Topic, New York: Academic Press, pp. 303–33. — (1976b), “Remarkable subjects in Malagasy,” in C. Li (ed.), Subject and Topic, New York: Academic Press, pp. 247–301. — (1980), “Passive is phrasal: Not sentential or lexical,” in T. Hoekstra, H. van der Hulst, and M. Moortgat (eds), Lexical Grammar, Dordrecht: Foris, pp. 181–214. — (1985), “Passive in the world’s languages,” in Timothy Shopen (ed.), Language Typology and Syntactic description (1st edn), Cambridge: Cambridge University Press, pp. 243–82. — (1994), Creating Anaphors: An Historical Study of the English Reflexive Pronouns. MS, University of California at Los Angeles.

504

9781441124609_Ref_Final_txt_print.indd 504

11/20/2012 5:54:09 PM

Bibliography — (2002), “Explaining the creation of reflexive pronouns in English,” in D. Minkova and R. Stockwell (eds), Studies in the History of the English Language: A Millennial Perspective, Berlin and New York: Mouton de Gruyter, pp. 325–54. — (2008), “Voice and relativization without movement in Malagasy,” Natural Language and Linguistic Theory, 26, 467–97. — (2009), “Linguistic theory and the historical creation of English reflexives,” in P. Crisma and G. Longobardi (eds), Historical Syntax and Linguistic Theory, Oxford: Oxford University Press, pp. 17–40. Keenan, Edward L. and Alan Timberlake (1985a), “Predicate formation rules in universal grammar,” in J. Goldberg, S. MacKaye, and, M. T. Wescoat (eds), Proceedings of WCCFL, 4, Stanford: Stanford Linguistics Association, pp. 123–38. — (1985b), “Valency affecting rules in extended categorial grammar,” Language Research, 21(4), 415–34. Keenan, Edward L. and Bernard Comrie (1977), “Noun phrase accessibility and universal grammar,” Linguistic Inquiry, 8, 63–99. Keenan, Edward L. and Cécile Manorohanta (2001), “A quantitative study of voice in Malagasy,” Oceanic Linguistics, 40(1), 67–85. Keenan, Edward L. and Denis Paperno (eds) (2012), Handbook of Quantifiers in Natural Language, New York: Springer. Keenan, Edward L. and Maria Polinsky (1998), “Malagasy morphology,” in A. Spencer and A. M. Zwicky (eds), The Handbook of Morphology, London: Blackwell, pp. 563–624. Keenan, Edward L. and Matthew S. Dryer (2007), “Passive in the world’s languages,” in T. Shopen (ed.), Language Typology and Syntactic description (1st edn), Cambridge: Cambridge University Press, pp 321–61. Kemmer, Suzanne (1993), The Middle Voice, Typological Studies in Language 23, Amsterdam: Benjamins. — (1995), “Emphatic and reflexive-self: Expectations, viewpoint, and subjectivity,” in D. Stein and S. Wright (eds), Subjectivity and Subjectivisation: Linguistic Perspectives, Cambridge: Cambridge University Press, pp. 55–82. Keyser, Samuel J. and Thomas Roeper (1984), “On the middle and ergative constructions in English,” Linguistic Inquiry, 15(3), 381–416. Kiguchi, Hirohisa and Rosalind Thornton (2004), “Binding principles and ACD constructions in child grammars,” Syntax, 7, 234–71. Kikusawa, Ritsuko (2003), “The development of some Indonesian pronominal systems,” in B. J. Blake, K. Burridge, and J. Taylor (eds), Historical Linguistics 2001: Selected Papers from the 15th International Conference on Historical Linguistics, Melbourne, 13–17 August 2001, Amsterdam: John Benjamins, pp. 237–68. Kimenyi, Alexandre (1980), A Relational Grammar of Kinyarwanda, Berkeley, CA: University of California Press. — (1988), “Passives in Kinyarwanda,” in M. Shibatani (ed.), Passive and Voice, Amsterdam: John Benjamins, pp. 355–87. Kiparsky, Paul (1997), “The rise of positional licensing,” in A. van Kamenade and N. Vincent (eds), Parameters of Morphosyntactic Change, Cambridge, UK: Cambridge University Press, pp. 460–94. Kirsner, Robert (1976), “On the subjectless ‘pseudo-passive’ in standard Dutch,” in C. Li (ed.), Subject and Topic, New York: Academic Press, pp. 387–415. Kiss, Katalin É. (1987), Configurationality in Hungarian, Dordrecht: Reidel. — (1996), “Two subject positions in English,” Linguistic Review, 13, 119–42. — (2008), “Free word order (non)configurationality, and Phases,” Linguistic Inquiry, 441–75. Kittilä, Seppo and Leonid Kulikov (eds) (forthcoming), Diachronic Typology of Voice and Valency-Changing Categories, Amsterdam: Benjamins. Klaiman, M. H. (1991), Grammatical Voice, Cambridge: Cambridge University Press. Klavans, Judith (1985), “The independence of syntax and phonology in cliticization,” Language, 61, 95–120.

505

9781441124609_Ref_Final_txt_print.indd 505

11/20/2012 5:54:09 PM

Bibliography Knjazev, Ju. P. (2007), Grammatičeskaja semantika: Russkij jazyk v tipologičeskoj perspektive [Grammatical semantics: Russian language in a typological perspective], Moscow: Jazyki slavjanskix kul’tur (Studia philologica). Ko, Heejeong (2007), “Asymmetries in scrambling and cyclic linearization,” Linguistic Inquiry, 38, 49–84. Koller, Bernhard (forthcoming), “Hittite pai- ‘come’ and uwa- ‘go’ as Restructuring Verbs.” Journal of Historical Linguistics 3/1. König, Christa (2006), “Marked nominative in Africa,” Studies in Language, 30, 655–732. König, Ekkehard and Peter Siemund (1999), “Intensifiers and reflexives: A typological perspective,” in Z. Frajzyngier and T. S. Curl (eds), Reflexives: Form and Function, Amsterdam: Benjamins, pp. 41–74. König, Ekkehard and Volker Gast (2006), “Focused assertion of identity: A typology of intensifiers,” Linguistic Typology, 10(2), 223–76. — (eds) (2008), Reciprocals and Reflexives—Theoretical and Cross-Linguistic Explorations, Berlin: Mouton. Koopman, Hilda and Dominique Sportiche (1991), “The position of subjects,” Lingua, 85, 211–58. Kornfilt, Jaklin (1997), Turkish, London and New York: Routledge. Kortmann, Bernd (1997), Adverbial Subordination. A Typology and History of Adverbial Subordinators based on European Languages. Berlin and New York: Mouton de Gruyter. Kozinsky, Isaac S., Vladimir P. Nedjalkov, and Maria S. Polinskaja (1988), “Antipassive in Chukchee,” in M. Shibatani (ed.), Passive and Voice, Amsterdam: John Benjamins, pp. 651–706. Kramer, Ruth (2009), “VSO and SVO word order in Middle Egyptian,” in C. G. Häberl (ed.), Afroasiatic Studies in Memory of Robert Hetzron: Proceedings of the 35th Annual North American Conference on Afroasiatic Linguistics, Cambridge, MA: Cambridge Scholars Press, pp. 92–147 (proof downloaded online at www9.georgetown.edu/faculty/rtk8/Research. html; accessed August 28, 2010). Krishnamurthy, Ramesh (ed.) (2004), English Collocation Studies: The OSTI Report by John Sinclair, Susan Jones and Robert Daley, London: Continuum. Kroch, Anthony (1989), “Reflexes of Grammar in Patterns of Language Change,” Language Variation and Change, 1, 199–244. — (2001), “Syntactic change,” in M. Baltin and C. Collins (eds), Handbook of Contemporary Syntactic Theory, Oxford: Blackwell, pp. 629–739. Kroch, Anthony and Beatrice Santorini (2010), “The comparative evolution of word order in English and French: the evidence from parsed corpora,” Paper read at conference on Subjects in Diachrony December 2–4, 2010, Regensburg (online at http://rhssl1.uni-regensburg.de/SlavKo/conferences/gces/abstracts/Kroch-Regensburg.pdf). Kulikov, Leonid (2001), “Causatives,” in M. Haspelmath et al. (eds), Language Typology and Language Universals. An International Handbook, 2 vols, Berlin: de Gruyter, pp. 886–98. — (2006), “Passive and middle in Indo-European: Reconstructing the early Vedic passive paradigm,” in Werner Abraham and Larisa Leisiö (eds), Passivization and Typology: Form and Function, Amsterdam: Benjamins, pp. 62–81. — (2009), “Valency-changing categories in Indo-Aryan and Indo-European: A diachronic typological portrait of Vedic Sanskrit,” in A. Saxena and Å. Viberg (eds), Multilingualism. Proc. of the 23rd Scandinavian Conference of Linguistics, Uppsala University, 1–3 October 2008 (Studia Linguistica Upsaliensis; 8), Uppsala: Uppsala Universitet, pp. 75–92. — (2011), “Voice typology,” in J. J. Song (ed.), The Oxford Handbook of Linguistic Typology, Oxford: Oxford University Press, pp. 368–98. Laakso, Minna and Marja-Leena Sorjonen (2010), “Cut-off or particle: Devices for initiating self-repair in conversation,” Journal of Pragmatics, 42, 1151–72. Labelle, Marie (1990), “Predication, wh-movement, and the development of relative clauses,” Language Acquisition, 1, 95–119.

506

9781441124609_Ref_Final_txt_print.indd 506

11/20/2012 5:54:09 PM

Bibliography Ladusaw, William (1992), “Expressing negation,” in C. Barker and D. Dowty (eds), Proceedings of SALT II, Columbus: Ohio State University, pp. 237–59. Laka, Itziar (1990), Negation in Syntax. Doctoral dissertation, Massachusetts Institute of Technology. Lakoff, George (1984), “Performative subordinate clauses,” Berkeley Linguistics Society, 10, 472–80. Lakoff, Robin (1971), “If’s, and’s, and but’s about conjunction,” in C. J. Fillmore and D. T. Lagendoen (eds), Studies in Linguistic Semantics, New York: Holt, Rinehart and Winston Inc., pp. 115–10. Lambrecht, Knud (1994), Information Structure and Sentence Form, Cambridge: Cambridge University Press. — (2000), “When subjects behave like objects: an analysis of the merging of S and O in sentence focus constructions across languages,” Studies in Language, 24, 611–82. Lang, Ewald (1984), The Semantics of Coordination, Amsterdam: John Benjamins. — (2000), “Adversative connectors on distinct levels of discourse: A re-examination of Eve Sweetser’s three-level approach,” in E. Couper-Kuhlen and B. Kortmann (eds), Cause, Condition, Concession, Contrast, Berlin and New York: Mouton de Gruyter, pp. 235–56. Langacker, Ronald W. (1987/1991), Foundations of Cognitive Grammar, vols. I & II: Theoretical Prerequisites, Stanford: Stanford University Press. — (1994), “Conceptual grouping and constituency in cognitive grammar,” Linguistics in the Morning Calm, 3, 149–72. — (1997), “Constituency, dependency, and conceptual grouping,” Cognitive Linguistics, 8(1), 1–32. — (2001), “Discourse in cognitive grammar,” Cognitive Linguistics, 12(2), 143–88. Langdon, Margaret (1970), A Grammar of Diegueño. University of California Publications in Linguistics, 66, Berkeley: University of California Press. LaPolla, Randy (1990), Grammatical Relations in Chinese: Synchronic and Diachronic Considerations. PhD dissertation, Berkeley: University of California. Larson, Richard K. (1988), “On the double object construction,” Linguistic Inquiry, 19, 335–91. Lasnik, Howard and Tim Stowell (1991), “Weakest Cross-over,” Linguistic Inquiry, 22, 687–720. Lasnik, Howard and Juan Uriagereka (2005), A Course in Minimalist Syntax, London: Blackwell. Lazard, Gilbert (1995), “Préverbes et typologie,” in A. Rousseau (ed.), Les préverbes dans les langues d’Europe, Villeneuve d’Asq: Septention, pp. 23–30. Lazdina, Tereza B. (1966), Teach Yourself Latvian, London: English Universities Press. Lecarme, Jacqueline (1994), Construct State and Genitive Assignment in Somali. MS, Sophia Antipolis: CNRS. Leddon, Erin M. and Jeffrey L. Lidz (2006), “Reconstruction effects in child Language,” in D. Bamman, T. Magnitskaia, and C. Zaller (eds), Proceedings of the 30th Boston University Conference on Language Development, Sommerville, MA: Cascadilla Press, pp. 328–39. Lee, Felicia A. (2006), Remnant Raising and VSO Clausal Architecture, Dordrecht: Springer. Lee, J. and D. Haug (2010), “Porting an Ancient Greek and Latin Treebank,” LREC (online at www.lrec-conf.org/proceedings/lrec2010/pdf/631 Paper.pdf). Lee, Young-Suk and Beatrice Santorini (1994), “Towards resolving Webelhuth´s paradox: Evidence from German and Korean,” in N. Corver and H. van Riemsdijk (eds), Studies on Scrambling: Movement and Non-Movement Approaches to Free-Order phenomena, The Hague: Mouton, pp. 257–301. Legendre, Geraldine (2001), “Introduction to OT syntax,” in G. Legendre, S. Vikner, and J. Grimshaw (eds), OT Syntax, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 1–27. Lehmann, Christian (1974), “Prinzipien für Universal 14,” in H. Seiler (ed.), Linguistic Workshop II, München: Wilhem Fink, pp. 69–97.

507

9781441124609_Ref_Final_txt_print.indd 507

11/20/2012 5:54:09 PM

Bibliography — (1988), “Towards a typology of clause linkage,” in J. Haiman and S. Thompson (eds), Clause Combining in Grammar and Discourse, Amsterdam: John Benjamins, pp. 181–225. Lehmann, Winfred P. (1973), “A structural principle of language and its implications,” Language, 49, 47–66. — (1974), Proto-Indo-European Syntax, Austin: University of Texas Press. Lerner, Gene (1991), “On the syntax of sentences in progress,” Language in Society, 20(3), 441–58. — (1994), “Responsive list construction,” Journal of Language and Social Psychology, 13(1), 20–33. — (1996), “On the ‘semi-permeable’ character of grammatical units in conversation: Conditional entry into the turn space of another speaker,” in E. Ochs, E. A. Schegloff, and S. A. Thompson (eds), Interaction and Grammar, Cambridge: Cambridge University Press, pp. 238–76. Li, Charles N. (ed.) (1975), Word Order and Word Order Change. Austin: University of Texas Press. — (ed.) (1976), Subject and Topic, New York: Academic Press. Li, Charles, and Sandra Thompson (1976), “Subject and topic: A new typology of language,” in Charles Li (ed.), Subject and Topic, New York: Academic Press, pp. 457–90. Liddicoat, Anthony L. (2007), An Introduction to Conversation Analysis, London: Continuum. Lightfoot, David W. (1979), Principles of Diachronic Syntax, Cambridge: Cambridge University Press. — (1999), The Development of Language: Acquisition, Change and Evolution, Oxford: Blackwell. — (2002), “Myths and the prehistory of grammars,” Journal of Linguistics, 38, 113–36. — (2006), How New Languages Emerge, Cambridge: Cambridge University Press. Lillehaugen, Brook Danielle (2008), Word Order and Focus in Tlacolula de Matamoros Zapotec. MS, paper presented at the Workshop on American Indigenous Languages, University of California, Santa Barbara. Lindstad, Arne Martinus (2007), Analyses of Negation. Structure and Interpretation, Doctoral dissertation, University of Oslo. Linell, Per (2010), Formal Written-Language-Biased vs. Dialogical Linguistics on the Nature of Language, Paper read at the international conference on “Cognitive Dynamics in Linguistic Interactions: Theoretical and Applied Perspectives,” Irkutsk, Russia, June 2010. Longacre, Robert E. (1976), An Anatomy of Speech Notions, Lisse: Peter de Ridder. — (1983), The Grammar of Discourse, New York: Plenum. — (1985), “Sentences as combinations of clauses,” in T. Shopen (ed.), Complex Constructions, 2nd volume of Language typology and syntactic description, Cambridge: Cambridge University Press, pp. 235–86. Longobardi, Giuseppe (1994), “Reference and proper names,” Linguistic Inquiry, 25, 609–65. — (1996), The Syntax of N-raising: A Minimalist Theory, Research Institute for Language and Speech, University of Utrecht: OTS Working Papers. — (2001a), “How comparative is semantics? A unified parametric theory of bare nouns and proper names,” Natural Language Semantics, 9, 335–69. — (2001b), “Formal syntax, diachronic Minimalism, and etymology: The history of French chez,” Linguistic Inquiry, 32, 275–302. Luraghi, Silvia (1990a), Old Hittite Sentence Structure, London and New York: Routledge. — (1990b), “Osservazione sulla Legge di Wackernagel e la posizione del verbo nelle lingue indoeuropee,” in P. Ramat, M. E. Conte, and A. Giacalone (eds), Dimensioni della Linguistica, Milano: Franco Angeli, pp. 31–60. — (1993), “Verb serialization and word order. Evidence from Hittite,” in H. Aertsen and R. Jeffers (eds), Historical Linguistics 1989, Amsterdam: Benjamins, pp. 267–81.

508

9781441124609_Ref_Final_txt_print.indd 508

11/20/2012 5:54:10 PM

Bibliography — (1995), “The function of verb initial sentences in some ancient Indo-European languages,” in M. Noonan and P. Downing (ed.), Word Order in Discourse, Amsterdam: Benjamins, pp. 355–86. — (1996), “Processi di grammaticalizzazione in ittita,” Archivio Glottologico Italiano, 81(1), 45–75. — (1997), “Omission of the direct object in Classical Latin,” Indogermanische Forschungen, 102, 239–57. — (1998), “Omissione dell’oggetto diretto in frasi coordinate: dal latino all’italiano,” in P. Ramat, E. Roma (a cura di), Sintassi Storica. Atti del xxx Congresso SLI, Roma: Bulzoni, pp. 183–96. — (2000), Hittite, München and New Castle: Lincom Europa. — (2001), “The development of local particles and adverbs in Anatolian as a grammaticalization process,” Diachronica, 28(1), 31–58. — (2003), “Definite referential null objects in Ancient Greek,” Indogermanische Forschungen, 108, 169–96. — (2010a), “The rise (and possible downfall) of configurationality,” in S. Luraghi and V. Bubenik (eds), Continuum Companion to Historical Linguistics, London and New York: Continuum International Publishing Group, pp. 212–29. — (2010b), “Transitivity, intransitivity and diathesis in hittite,” in ИНДОЕВРОПЕЙСКОЕ ЯЗЫКОЗНАНИЕ И КЛАССИЧЕСКАЯ ФИЛОЛОГИЯ – XIV, 2 vols, St Petersburg: Nauka, pp. 133–54. — (2012), “Basic valency orientation and the middle voice in Hittite,” Studies in language 36(1), 1–32. Luraghi, Silvia and Claudia Parodi (2008), Key Terms in Syntax and Syntactic Theory, London: Continuum Publishing. Lyons, John (1968), Introduction to Theoretical Linguistics, Cambridge: Cambridge University Press. Mahajan, Anoop (1990), The A/A-Bar Distinction and Movement Theory. PhD dissertation, Cambridge, MA: Massachusetts Institute of Technology. — (1994), “Towards a unified theory of scrambling,” in N. Corver and H. van Riemsdijk (eds), Studies on Scrambling: Movement and Non-Movement Approaches to Free-Order phenomena, The Hague: Mouton, pp. 301–33. Mahootian, Shahrzad (1997), Persian, London: Routledge. Mair, Christian and Geoffrey Leech (2006), “Current change in English syntax,” in B. Aarts and A. MacMahon (eds), The Handbook of English Linguistics, Oxford: Blackwell, pp. 318–42. Malchukov, Andrej (2004), “Towards a semantic typology of adversative and contrast marking,” Journal of Semantics, 21, 177–98. Maling, Joan and Sigríður Sigurjónsdóttir (2002), “The ‘new impersonal’ construction in Icelandic,” Journal of Comparative Germanic Linguistics, 5, 97–142. Malouf, Robert (2003), “Cooperating constructions,” in E. Francis and L. Michaelis (eds), Mismatch: Form-Function Incongruity and the Architecture of Grammar, Stanford: CSLI Publications, pp. 403–24. Manley, Timothy M. (1972), Outline of the Sre Structure, Honolulu: University of Hawaii Press. Manney, Linda Joyce (2000), Middle Voice in Modern Greek: Meaning and Function of an Inflectional Category, Studies in Language Companion Series, 48, Amsterdam: Benjamins. Manning, Christopher (2003), “Probabilistic syntax,” in R. Bod, J. Hay, and S. Jannedy (eds), Probabilistic Linguistics, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 289–341. Marchese, Lynell (1987), “On the role of conditionals in Godie procedural discourse,” in R. Tomlin (ed.), Coherence and Grounding in Discourse, Amsterdam: John Benjamins, pp. 263–80. Marušič, Franc (2008), “Slovenian clitics have no unique syntactic position,” Formal Approaches to Slavic Linguistics, 16, 266–81.

509

9781441124609_Ref_Final_txt_print.indd 509

11/20/2012 5:54:10 PM

Bibliography Mason, Oliver and Susan Hunston (2004), “The automatic recognition of verb patterns: A feasibility study,” International Journal of Corpus Linguistics, 9(2), 253–70. Matthiessen, Christian and Sandra A. Thompson (1988), “The structure of discourse and subordination,” in J. Haiman and S. A. Thompson (eds), Clause Combining in Grammar and Discourse, Amsterdam: John Benjamins, pp. 275–329. Mauri, Caterina (2008a), Coordination Relations in the Languages of Europe and Beyond, Berlin, New York: Mouton de Gruyter. — (2008b), “The irreality of alternatives: Towards a typology of disjunction,” Studies in Language, 32(1), 22–55. — (2008c), “The parallelisms of clausal coordination,” Revue de Sémantique et Pragmatique, 24, 145–75. — (2010), “Semantic Maps or Coding Maps? A unified account of the coding degree, coding distance and coding complexity of coordination relations,” in Linguistic Discovery 8(1), 210–32. (special issue, M. Haspelmath, A. Malchukov, and M. Cysouw (eds)). May, Robert (1985), Logical Form, Cambridge, MA: Massachusetts Institute of Technology Press. Mazuka, Reiko (1996), “Can a grammatical parameter be set before the first word? Prosodic contributions to early setting of a grammatical parameter,” in J. Morgan and K. Demuth (eds), Signal to Syntax, Hillsdale, NJ: Lawrence Erlbaum, pp. 313–30. McCarthy, John J. (2002), A Thematic Guide to Optimality Theory, Cambridge: Cambridge University Press. McCawley, James D. (1970), “English as a VSO language,” Language, 46, 286–99. McDaniel, Dana, Cecile McKee, and Judy B. Bernstein (1998), “How children’s relatives solve a problem for minimalism,” Language, 74, 308–64. McEnery, Tony, Richard Xiao, and Yukio Tono (2006), Corpus-based Language Studies: An Advanced Resource Book, London: Routledge. McKee, Cecile (1992), “A comparison of pronouns and anaphors in Italian and English acquisition,” Language Acquisition, 2, 21–55. Mel’čuk, Igor A. (1993) “The inflectional category of voice: Towards a more rigorous definition,” in B. Comrie and M. Polinsky (eds), Causatives and transitivity, Amsterdam: Benjamins, pp. 1–46. — (1994), Cours de morphologie générale (théorique et descriptive). Vol. II(2): Significations morphologiques, Montréal: Les Presses de l’Université de Montréal. Merchant, Jason (2000), The Syntax of Silence, Oxford, UK: Oxford University Press. Merlan, Francesca (1985), “Split intransitivity: Functional oppositions in intransitive inflection,” in J. Nichols and A. Woodbury (eds), Grammar Inside and Outside the Clause, Cambridge: Cambridge University Press, pp. 24–362. Meurers, W. Detmar and Stefan Müller (2009), “Corpora and syntax,” in A. Lüdeling and M. Kytö (eds), Corpus Linguistics: An International Handbook, Volume 2, Berlin and New York: de Gruyter, pp. 920–32. Michaelis, Laura (2001), “Exclamative constructions,” in M. Haspelmath, E. König, W. Österreicher, and W. Raible (eds), Language Universals and Language Typology: An International Handbook, Berlin: Walter de Gruyter, pp. 1038–50. — (2009), “Sign-based construction grammar,” in B. Heine and H. Narrog (eds), The Oxford Handbook of Syntactic Analysis, Oxford: Oxford University Press, pp. 155–76. — (forthcoming), “Making the case for construction grammar,” in H. Boas and I. Sag (eds), Sign-Based Construction Grammar, Stanford, CA: CSLI Publications. Michaelis, Laura and Josef Ruppenhofer (2001), Beyond Alternations: A Constructional Account of the Applicative Pattern in German. Stanford, CA: CSLI Publications. Michaelis, Laura and Knud Lambrecht (1996), “Toward a construction-based model of language function: The case of nominal extraposition,” Language, 72, 215–47. Miestamo, Matti (2005), Standard Negation: The Negation of Declarative Verbal Main Clauses in a Typological Perspective, Berlin and New York: Mouton de Gruyter. Mikulová, Marie and Štěpánek, Jan (2009), “Annotation quality checking and its implications for design of Treebank (in Building the Prague Czech-English Dependency Treebank),” in TLT 8: Proceedings of eighth Treebanks and Linguistic Theories Workshop, pp. 137–48.

510

9781441124609_Ref_Final_txt_print.indd 510

11/20/2012 5:54:10 PM

Bibliography Milner, Jean-Claude (1982), Ordres et Raisons de Langue, Paris: Éditions du Seuil. Mithun, Marianne (1988), “The grammaticization of coordination,” in J. Haiman and S. Thompson (eds), Clause Combining in Grammar and Discourse, Amsterdam: John Benjamins, pp. 331–60. — (1991), “Active/agentive case marking and its motivations,” Language, 67, 51–546. — (1994), “The implications of ergativity for a Philippine voice system,” in B. Fox and P. Hopper (eds), Voice: Form and Function, Amsterdam: John Benjamins, pp. 247–77. Miyagawa, Shigeru (1989), “Structure and case marking in Japanese,” Syntax and Semantics, 22. — (1997), “Against optional scrambling,” Linguistic Inquiry, 28, 1–25. — (2005), “EPP and semantically vacuous scrambling,” in J. Sabel and M. Saito (eds), The Free Word Order Phenomenon: Its Syntactic Sources and Diversity, Berlin: Mouton de Gruyter, pp. 181–220. — (2006), “On the ‘undoing’ property of scrambling: A response to Bošković,” Linguistic Inquiry, 37, 607–24. Monachesi, Paola (1993), “Object clitics and clitic climbing in Italian HPSG grammar,” in S. Krauwer, M. Moortgat, and L. des Tombe (eds): Proceedings of the 6th Conference of the European Chapter of the Association for Computational Linguistics, April 21–3, 1993, Utrecht, The Netherlands. The Association for Computer Linguistics, 437–42. Moravcsik, Edith A. (1972), “Some crosslinguistic generalisations about intensifier constructions,” in P. Peranteau et al. (eds), Papers from the eighth regional meeting of the Chicago Linguistic Society. Chicago Linguistic Society, Chicago, pp. 271–77. Moscati, Vincenzo (2006), The Scope of Negation. Doctoral dissertation, University of Siena. Mous, Maarten (1993), A Grammar of Iraq, Hamburg: Buske. — (2000), “Middle voice in Iraqw,” Languages of Tanzania: Studies Dedicated to the Memory of Prof. Clement Maganga, 141–59. — (2004), “The middle in cushitic languages,” in A. Simpson (ed.), Proceedings of the 27th Annual Meeting of the Berkeley Linguistics Society, March 22–25, 2001: Special Session on Afroasiatic Languages, 75–86. Müller, Gereon (1996), “A constraint on remnant movement,” Natural Language and Linguistic Theory, 14, 355–407. — (2000), Elemente der optimalitätstheoretischen Syntax, Tübingen: Stauffenburg Verlag. Müller, Gereon and Wolfgang Sternefeld (1994), “Scrambling as A’-movement,” in N. Corver and H. van Riemsdijk (eds), Studies on Scrambling: Movement and NonMovement Approaches to Free Word-Order phenomena, The Hague: Mouton, pp. 331–87. Munn, Alan (1994) “The possessor that stayed close to home,” in V. Samiian and J. Schaeffer (eds), Proceedings of the Western Conference on Linguistics, WECOL 24, pp. 181–95. Munro, Pamela (1985), “A new “harmony” within Type 24,” MS. — (2006), “From parts of speech to the grammar,” Studies in Language, 30, 307–50. Munro, Pamela and Lynn Gordon (1982), “Syntactic relations in Western Muskogean: A typological perspective,” Language, 58, 81–115. Nedjalkov, Vladimir P. (ed.) (2007), Reciprocal Constructions, 4 vols, Typological studies in language 71, Amsterdam: Benjamins. Nedjalkov, Vladimir P. and Viktor P. Litvinov (1995), “The St Petersburg/Leningrad Typology Group,” in M. Shibatani and Th. Bynon (eds), Approaches to Language Typology, Oxford: Clarendon Press, pp. 215–71. Nedyalkov, Igor V. (1991), “Recessive-accessive polysemy of verbal affixes,” Languages of the World, 1, 4–31. Neeleman, Ad (1994), “Scrambling as D-Structure phenomena,” in N. Corvert and H. van Riemsdijk (eds), Studies on Scrambling: Movement and Non-Movement Approaches to Free Word-Order Phenomena, The Hague: Mouton, pp. 387–431. Nelson, Gerald, Sean Wallis, and Bas Aarts (2002), Exploring Natural Language: Working with the British Component of the International Corpus of English, Amsterdam and Philadelphia: Benjamins.

511

9781441124609_Ref_Final_txt_print.indd 511

11/20/2012 5:54:10 PM

Bibliography Nespor, Marina, Maria Teresa Guasti, and Anne Christophe (1996), “Selecting word order: The rhythmic activation principle,” in U. Kleinhenz (ed.), Interfaces in Phonology, Berlin: Academic Verlag, pp. 1–26. Nespor, Marina, Mohinish Schukla, Rubeb van der Vijver, Cinzia Avesani, Hanna Schraudolf, and Caterina Donati (2008), “Different phrasal prominence realization in VO and OV languages,” Lingue e Linguaggio, 7, 1–29. Nespor, Marina and Irene Vogel (1986), Prosodic Phonology, Dordrecht: Foris. — (2007), Prosodic Phonology, Berlin and New York: Mouton de Gruyter. Neu, Erich (1968a), Das hethitische Mediopassiv und seine indogermanischen Grundlagen, Wiesbaden: Harrassowitz. — (1968b), Interpretation der hethitischen mediopassiven Verbalformen, Wiesbaden: Harrassowitz. Nevis, Joel A., Brian Joseph, Dieter Wanner, and Arnold M. Zwicky (1994), Clitics: A Comprehensive Bibliography 1892–1991, Amsterdam: Benjamins. Nichols, Johanna (1992), Linguistic Diversity in Space and Time, Chicago: University of Chicago Press. Nishigauchi, Taisuke (1990), Quantification in Syntax, Dordrecht: Reidle. Nissenbaum, Jonathan (2000), Investigations of Covert Phrase Movement. Doctoral dissertation, Massachussets Institute of Technology. Nunes, Jairo and Ellen Thompson (1998) “Appendix,” in J. Uriagereka (ed.), Rhyme and Reason: An Introduction to Minimalist Syntax, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 497–521. Ochs, Elinor, Emanuel A. Schegloff, and Sandra A. Thompson (1996), “Introduction,” in E. Ochs, E. Schegloff, and S. A. Thompson (eds), Interaction and Grammar, Cambridge: Cambridge University Press, pp. 1–51. Ono, Tsuyoshi and Elizabeth Couper-Kuhlen (2007), “Turn increments in cross-linguistic perspective: Introductory remarks,” Pragmatics, 17(4), 505–12. Ono, Tsuyoshi and Sandra A. Thompson (1996), “Interaction and syntax in the structure of conversational discourse: Collaboration, overlap, and syntactic dissociation,” in E. H. Hovy and D. R. Scott (eds), Computational and Conversational Discourse: Burning Issues – An Interdisciplinary Account, Berlin: Springer. Ordóñez, Francisco (1998), “Postverbal asymmetries in Spanish,” Natural Language and Linguistic Theory, 313–45. — (2012), “Scrambling,” (this volume). Ortiz de Urbina, Jon (1987), Parameters in the Grammar of Basque, Dordrecht: Foris. Otten, Heinrich (1981), Die Apologie Hattusilis III, Wiesbaden: Harrassowitz. Özkaragöz, Inci (1988), “Monoclausal double passives in Turkish,” in D. Slobin and K. Zimmer (eds), Studies in Turkish Linguistics, Amsterdam: Benjamins. Palmer, Frank R. (1994), Grammatical Roles and Relations, Cambridge: Cambridge University Press. Paul, Ileana and Lucie Rabaovololona (1998), “Raising to object in Malagasy,” in I. Paul (ed.), UCLA Occasional Papers in Linguistics 20: The Structure of Malagasy, vol. II, Dept of Linguistics, UCLA, pp. 50–65. Payne, Doris (1980), “Switch-reference in Chickasaw,” in P. Munro (ed.), Studies of SwitchReference, UCLA Papers in Syntax, 8, pp. 89–118. — (1982), “Chickasaw agreement morphology: A functional explanation,” in P. Hopper and S. Thompson (eds), Syntax and Semantics 15: Studies in Transitivity, New York: Academic Press, pp. 351–78. — (1987), “Information structuring in Papago narrative discourse,” Language, 63, 783–804. — (1997), “Semantic role and argument structure in the Maasai ‘external possession’ construction,” Proceedings of the 1997 23rd Meetings of the Berkeley Linguistics Society, 98–115. — (2008), “Review of Semantic Role Universals and Argument Llinking: Theoretical, Typological, and Psycholinguistic Perspectives, ed. by Ina Bornkessel, Matthias Schlesewsky, Bernard Comrie and Angela Friederici. Berlin: Mouton de Gruyter.” Linguistic Typology, 12, pp. 310–22.

512

9781441124609_Ref_Final_txt_print.indd 512

11/20/2012 5:54:10 PM

Bibliography Payne, Doris and Thomas Payne (1990), “Yagua,” in D. Derbyshire and G. Pullum (eds), Handbook of Amazonian Languages, 2, Berlin: Mouton de Gruyter, pp. 249–474. Payne, John (1980), “The decline of ergativity in Pamir languages,” Lingua, 51, 147–186. — (1985), “Complex phrases and complex sentences,” in T. Shopen (ed.), Complex Constructions; Language Typology and Syntactic Description, vol. 2, Cambridge: Cambridge University Press, pp. 3–41. Payne, Thomas (1992), Participant Coding in Yagua Narrative, Berkeley: University of California Press. — (1997), Describing Morphosyntax, Cambridge: Cambridge University Press. — (1999), “A functional typology of inverse constructions,” in E. Rakhilina and Y. Testelets (eds), Typology and Linguistic Theory: From Description to Explanation, Moscow: Languages of Russian Culture, pp. 245–554. Penka, Doris (2007), Negative Indefinites. Doctoral dissertation, University of Tübingen. Pérez-Leroux A. T. (1995) “Resumptives in the acquisition of relative clauses.” Language acquisition, 4, 105–38. Perlmutter, David M. (1983), Studies in Relational Grammar I, Chicago: University Of Chicago Press. Perlmutter, David M. and Paul M. Postal (1983a), “Some proposed laws of basic clause structure,” in D. Perlmutter (ed.), Studies in Relational Grammar 1, Chicago: University of Chicago Press, pp. 81–128. — (1983b), “Toward a universal characterization of passivization,” in D. M. Perlmutter (ed.), Studies in Relational Grammar 1, Chicago: University of Chicago Press, pp. 3–30 (Based on BLS III 394–417). — (1984), “The 1-Advancement Exclusiveness Law,” in D. Perlmutter and C. Rosen (eds), Studies in Relational Grammar 2, Chicago: University of Chicago Press, pp. 81–125. Perlmutter, David and Carol Rosen (eds) (1984), Studies in Relational Grammar 2, Chicago: University of Chicago Press. Pesetsky, David (1987), “WH-in-situ: Movement and unselective binding,” in E. Reuland and A. G. B. Ter Meulen (eds) The Linguistic Representation of (In)definiteness, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 98–129. — (1997), “Optimality Theory and syntax: Movement and pronunciation,” in D. Archangeli and D. Terence Langendoen, Optimality Theory, an Overview, Oxford: Blackwell, pp. 134–70. — (2000), Phrasal Movement and its Kin, Cambridge, MA: Massachusetts Institute of Technology Press. Pesetsky, David and Esther Torrego (2001), “T-to-C movement: Causes and consequences,” in M. Kenstowicz (ed.), Ken Hale: A Life in Language, Cambridge, MA: Massachusetts Institute of Technology Press, pp. 355–426. Philip, William and Peter Coopmans (1996), “The double Dutch delay of Principle B effect,” in A. Stringfellow, D. Cahana-Amitay, E. Hughes, and A. Zukowski (eds), Proceedings of the 20th annual Boston University Conference on Language Development, Somerville, MA: Cascadilla Press, pp. 576–87. Pike, Kenneth (1982), Linguistic Concepts: An Introduction to Tagmemics, Lincoln: University of Nebraska Press. Piñar, Larruba Pilar (1996), Negative Polarity Licensing and Negative Concord in the Romance Languages. Doctoral dissertation, University of Arizona. Pinault, George (1995), “Le problème du préverbe en indo-européen,” in A. Rousseau (ed.), Les Préverbes Dans les Langues d’Europe, Villeneuve d’Asq: Septention, pp. 35–59. Pintzuk, Susan (1991), Phrase Structure in Competition: Variation and Change in Old English Word Order. PhD dissertation, University of Pennsylvania. Pires, Acrisio and Sarah G. Thomason (2008), “How much syntactic reconstruction is possible?” in G. Ferraresi and M. Goldbach (eds), Principles of Syntactic Reconstruction, Amsterdam: John Benjamins, pp. 27–72. Plank, Frans (ed.) (1979), Ergativity: Towards a Theory of GRs. New York: Academic Press.

513

9781441124609_Ref_Final_txt_print.indd 513

11/20/2012 5:54:10 PM

Bibliography — (2003), Noun Phrase Structure in the Languages of Europe (Empirical Approaches to Language Typology, EUROTYP, 20–7), Berlin, Mouton de Gruyter. Plann, Susan (1982), “Indirect questions in Spanish,” Linguistic Inquiry, 13, 297–312. Plunkett, Bernadette (2004), “Early peripheries in the absence of C,” in D. Adger, C. De Cat, and G. Tsoulas (eds), Peripheries. Syntactic Edges and Their Effect, Dordrecht: Kluwer, pp. 383–405. Poletto, Cecilia (2000), The Higher Functional Field. Evidence from Northern Italian Dialects, New York: Oxford University Press. — (2002), “On V2 and V3 sequences in Rhaetoromance,” in S. Barbiers, L. Cornips, and S. van der Kleij (eds), Syntactic Microvariation (online at www.meertens.nl/books/synmic/ pdf/poletto.pdf). — (2012), “On V2 types,” (this volume). Pollard, Carl and Ivan Sag (1987), Information-based Syntax and Semantics, Volume 1: Fundamentals, Stanford, CA: CSLI Publications. — (1994), Head-driven Phrase Structure Grammar. Chicago: University of Chicago Press. Pollock, Jean-Yves (1989), “Verb movement, Universal Grammar, and the structure of IP,” Linguistic Inquiry, 20, 365–424. Pooth, Roland (in preparation), Diathese und Verbsemantik im Vedischen und im Urindogermanischen. PhD diss. Leiden. Postal, Paul M. (1986), Studies of Passive Clauses, New York: State University of New York Press. Prince, Alan (1983), “Relating to the Grid,” Linguistic Inquiry, 14, 19–100. Prince, Alan and Paul Smolensky (1993), “Optimality theory: constraint interaction in generative grammar,” RuCCS Technical Report, 2, Rutgers University Center for Cognitive Science. — (2004), Optimality Theory: Constraint Interaction in Generative Grammar, Oxford: Blackwell. Prince, Ellen (1978), “A comparison of WH-clefts and it-clefts in discourse,” Language, 54, 883–906. Pullum, Geoffrey K. (1982), “Free Word Order and Phrase Structure Rules”. In James Pustejovsky and Peter Sells (eds), Proceedings of the Twelfth Annual Meeting of the North Eastern Linguistic Society, 209-20. Graduate Linguistics Student Association, University of Massachusetts, Amherst, Massachusetts. Puskas, Genoveva (2000), Word Order in Hungarian, Linguistics Today, Amsterdam and Philadelphia: John Benjamins Publishing Company. Quicoli, Carlos (1982), The Structure of Complementation, Ghent: Story-Scientia. Quicoli, Carlos (2008), “Anaphora by Phase,” Syntax, 11, 299–329. Quirk Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik (1985), A Comprehensive Grammar of the English Language, London: Longman. Radanović Kocić, Vesna (1996), “The placement of Serbo-Croatian clitics: a prosodic approach,” in A. L. Halpern and A. M. Zwicky (eds) (1996), Approaching Second: Second Position Clitics and Related Phenomena, Stanford: CSLI Publications, in Halpern and Zwicky, pp. 429–46. Radford, Andrew (2004), English Syntax: An Introduction, Cambridge: Cambridge University Press. — (2010), On the Doubly Filledo COMP Filter and Complementizer Spellout in English. MS, University of Essex. Randriamasimanana, Charles (1986), The Causatives of Malagasy, Honolulu: University of Hawaii Press. Rappaport Hovav, Malka and Beth Levin (1998), “Building verb meanings,” in M. Butt and W. Geuder (eds), The Projection of Arguments, Stanford, CA: CSLI Publications, pp. 97–134. Rasolofo, Andoveloniaina (2006), Malagasy Transitive Clause Types and their Functions. PhD dissertation, University of Oregon. Real Academia Española (2009), Nueva Gramática de la Lengua Española, Madrid: Espasa Calpe.

514

9781441124609_Ref_Final_txt_print.indd 514

11/20/2012 5:54:11 PM

Bibliography Rebuschi, Georges (2005), “Generalizing the antisymmetric analysis of coordination to nominal modification,” Lingua, 115, 445–59. Reinhart, Tanya (1976), The Syntactic Domain of Anaphora. PhD dissertation, Massachusetts Institute of Technology. — (1997), “Quantifier scope: How labor is divided between QR and choice functions,” Linguistics and Philosophy, 20, 335–97. — (1981), “Definite NP-anaphora and c-command domains,” Linguistic Inquiry, 12, 605–35. — (1983), Anaphora and Semantic Interpretation, London: Croom Helm. Renzi, Lorenzo and Laura Vanelli (1983), “I pronomi soggetto in alcune varietà romanze,” in AA. VV (a cura di), Scritti in onore di G.B. Pellegrini, Pisa: Pacini, pp. 120–45. Richards, Marc (2004), Object Shift and Scrambling in North and West Germanic: A Case Study in Symmetrical Syntax. PhD dissertation, Robinson College, Cambridge, England. Richards, Norvin (2001), Movement in Language. Oxford, UK: Oxford University Press. — (2010), Uttering Trees, Cambridge, MA: Massachusetts Institute of Technology Press. Ritter, Elizabeth (1988), “A Head-movement Approach to Construct-State Noun Phrases,” Linguistics, 26, 909–929. — (1991) “Two functional categories in noun phrases: evidence from modern Hebrew,” in S. Rothstein (ed.), Syntax and Semantics 26, San Diego: Academic Press, pp. 37–62. Rizzi, Luigi (1982), Issues in Italian Syntax, Dordrecht: Foris Publications. — (1990), Relativized Minimality, Cambridge, MA: Massachusetts Institute of Technology Press. — (1996), “Residual verb second and the Wh Criterion,” in A. Belletti and L. Rizzi (eds), Parameters and Functional Heads, Oxford: Oxford University Press, pp. 63–90. — (1997), “The fine structure of the Left Periphery,” in Liliane Haegeman (ed.), Elements of Grammar, Dordrecht: Kluwer, pp. 281–337. — (2000), Comparative Syntax and Language Acquisition, London: Routledge. — (2001), “On the position INT(ERROGATIVE) in the Left Periphery of the clause,” in G. Cinque and G. Salvi (eds), Current Studies in Italian Syntax, Amsterdam: Elsevier, pp. 287–96. — (2004), “Locality and Left Periphery,” in A. Belletti (ed.), Structures and Beyond. The Cartography of Syntactic Structures, vol. 3, New York: Oxford University Press, pp. 104–31. — (2006), “On some properties of subjects and topics,” in L. Brugè, G. Giusti, N. Munaro, W. Schweikert, and G. Turano (eds), Contributions to the XXX Incontro di Grammatica Generativa, Venezia: Cafoscarina, pp. 203–24. — (2010), “Some consequences of criterial freezing,” in P. Svenonius (ed.), Festschrift for Tarald Taraldsen, Oxford: Oxford University Press. Rizzi, Luigi (ed.) (2004), The Structure of CP and IP—The Cartography of Syntactic Structures vol 3, Oxford: Oxford University Press. Roberts, Craige (2010), Topic. MS, Ohio State University. Roberts, Ian (2004), “The C-system in Brythonic Celtic languages, V2 and the EPP,” in Rizzi, Luigi (ed.), The Structure of CP and IP – The Cartography of Syntactic Structures vol 3, Oxford: Oxford University Press, pp. 297–328. — (2007), Diachronic Syntax, Oxford: Oxford University Press. Roberts, Ian and Kyle Johnson (eds) (1999), Beyond Principles and Parameters: Essays in Memory of Oswaldo Jaeggli, Dordrecht: Kluwer Academic. Roberts, John R. (1987), Amele, London: Croom Helm. Roeper, Thomas (1987), “Implicit arguments and head-complement relation,” Linguistic Inquiry, 18, 267–310. Rögnvaldsson, Eiríkur (1994/1995), “Breytileg orðaröð í sagnlið [Word order variation in the Verb Phrase],” Íslenskt mál, 16–17, 27–66. — (1995), “Old Icelandic: A non-configurational language?” North-Western European Language Evolution, 26, 3–26. — (1996), “Word order variation in the VP in old Icelandic,” Working Papers in Scandinavian Syntax, 58, 55–86.

515

9781441124609_Ref_Final_txt_print.indd 515

11/20/2012 5:54:11 PM

Bibliography Rögnvaldsson, Eiríkur and Höskuldur Thráinsson (1990), “On Icelandic word order once more,” in J. Maling and A. Zaenen (eds), Modern Icelandic syntax, volume 24 of Syntax and semantics, San Diego, CA: Academic Press, pp. 3–40. Rooth, Mats (1996), “Focus,” in S. Lappin (ed.), The Handbook of Contemporary Semantic Theory, Oxford: Blackwell, pp. 271–97. Rosch, Eleanor (1978), “Principles of categorization,” in E. Rosch and B. Lloyd (eds), Cognition and Categorization, Hillsdale, NJ: Lawrence Erlbaum, pp. 27–48. Rosen, Carol (1984), “The interface between semantic roles and initial grammatical relations,” in David Perlmutter and Carol Rosen (eds), Studies in Relational Grammar 2, University of Chicago Press, pp. 38–77. Rosenbaum, Peter S. (1967), The Grammar of English Predicate Complement Constructions, Cambridge, MA: Massachusetts Institute of Technology Press. Ross, John R. (1967), Constraints on Variables in Syntax. PhD dissertation, Massachusetts Institute of Technology. Rousseau, André (ed.) (1995), Les préverbes dans les langues d’Europe, Villeneuve d’Asq: Septention. Rouveret, Alain (1994), Syntaxe du gallois: le groupe nominal. MS, Université de Paris VIII. Rowlands, Evan C. (1969), Teach Yourself Yoruba, London: The English Universities Press. Rudin, Catherine (1988), “On multiple questions and multiple wh fronting,” Natural Language and Linguistic Theory, 6, 445–501. Rudolph, Elisabeth (1996), Contrast. Adversative and Concessive Expressions on Sentence and Text Level, Berlin, New York: Walter de Gruyter. Ruppenhofer, Josef (2004), The Interaction of Valence and Information Structure. Unpublished doctoral dissertation, Department of Linguistics, University of California, Berkeley. Russi, Cinzia (2008), Italian Clitics. An empirical study, Berlin and New York: Mouton de Gruyter. Rutkowski, Paweł (2007), “Some remarks on the syntax of genitival phrases in Lithuanian,” in N. Carter, L. Hadic-Zabala, A. Rimrott, and D. R. Storoshenko (eds), Simon Fraser University Working Papers in Linguistics, vol. I: Proceedings of the 22nd NorthWest Linguistics Conference (NWLC22), Burnaby, SFU Linguistics Graduate Students” Association, Simon Fraser University, pp. 221–31. Rutkowski, Paweł and Ljiljana Progovac (2005), “Classification Projection in Polish and Serbian: The position and shape of classifying adjectives,” in S. Franks, F. Y. Gladney, and M. Rasseva-Kurktchieva (eds), Formal Approaches to Slavic Linguistics 13, The South Carolina Meeting, Ann Arbor Michigan: Michigan Slavic Publications, pp. 289–99. Ryan, Robert (2004), The “Reflexive” as an Interpretation. PhD thesis, Växjö University. Sabel, Joachim and Mamoru Saito (2005), The Free Word Order Phenomenon: Its Syntactic Sources and Diversity, The Hague: Mouton de Gruyter. Sacks, Harvey (1992 [1967–68]), Lectures on Converstion, edited by Gail Jefferson, Oxford: Blackwell. Sadler, Louisa (1998), Welsh Syntax, London: Croom Helm. Sag, Ivan (2007), “Remarks on locality,” in S. Müller (ed.), Proceedings of the HPSG07 Conference, Stanford University, Stanford, CA: CSLI Publications, pp. 394–414. — (2010), “English filler-gap constructions,” Language, 86, 486–545. — (2012), “Sign-based construction grammar: An informal synopsis,” in H. Boas and I. Sag (eds), Sign-Based Construction Grammar, Stanford, CA: CSLI Publications. Sag, Ivan, Thomas Wasow, and Emily M. Bender (1999), Syntactic Theory: A Formal Introduction, 2nd edn, Stanford: CSLI. Saito, Mamoru (1985), Some Asymmetries in Japanese and their Theoretical Implications, Cambridge, MA: Ph.D Thesis, MIT. — (1992), “Long-distance scrambling in Japanese,” Journal of East Asian Linguistics, 1, 69–118. — (2003), “A derivational approach to the interpretation of scrambling chains,” Lingua, 113, 481–518. — (2010), Sentence Types and the Japanese Right Periphery. MS, Nanzan University, Nagoya.

516

9781441124609_Ref_Final_txt_print.indd 516

11/20/2012 5:54:11 PM

Bibliography Saltarelli, Mario (1988), Basque, London: Croom Helm. Salvi, Giampaolo (2004), La formazione della struttura di frase romanza: ordine delle parole e clitici dal latino alle lingue romanze antiche, Tubingen: Niemeyer. Samek-Lodovici, Vieri (1996), Constraints on Subjects. An Optimality Theoretic Analysis, Ph.D. Dissertation, Rutgers University, New Brunswick, NJ. — (2005), “Prosody-Syntax interaction in the expression of focus,” Natural Language and Linguistic Theory, 23, 687–755. — (2006), “When right dislocation meets the left-periphery. A unified analysis of Italian non-final focus,” Lingua, 116, 836–73. — (2009), “Topic, focus, and background in Italian clauses,” in A. Dufter and D. Jacob (eds), Focus and Background in Romance Languages, Studies in Language Companion Series 112, Amsterdam: John Benjamins, pp. 333–57. Samper-Padilla, José, A. (1990), On the Definition of Binding Domains in Spanish. Dordrecht: Kluwer. Santelmann, Lynn (1998), “The acquisition of verb movement and spec-head relationships in child Swedish,” in D. Adger, S. Pintzuk, B. Plunkett, and G. Tsoulas (eds), Specifiers: Minimalist Approaches, Oxford: Oxford University Press, pp. 271–98. Santorini, Beatrice (1989), The Generalization of the Verb-Second Constraint in the History of Yiddish. PhD thesis, University of Pennsylvania. Sapir, Edward (1917), “Review of C.C. Uhlenbeck ‘Het passieve karakter van het verbum transitivum of van het verbum actionis in talen van Noord-America’.” International Journal of American Linguistics, 1, 82–86. — (1921), Language, an Introduction to the Study of Speech, San Diego: Harcourt, Brace Jovanovich. Schachter, Paul (1977a), “Constraints on coordination,” Language, 53, 86–103. — (1977b), “Reference-related and role-related properties of subjects,” in P. Cole and J. Sadock (eds), Syntax and Semantics 8: Grammatical Relations, New York: Academic Press, pp. 279–306. — (1984), “Semantic-role-based syntax in Toba Batak,” in P. Schachter (ed.), Studies in the Structure of Toba Batak, UCLA Occasional Papers in Linguistics 5, 122–49. Schäufele, Steven (1996), “Now that we’re all here, where do we sit? Phonological ordering in the Vedic clause-initial string,” in A. L. Halpern and A. M. Zwicky (eds), Approaching Second: Second Position Clitics and Related Phenomena, Stanford: CSLI Publications, pp. 447–75. Schegloff, Emanuel A. (1979), “The relevance of repair to syntax-for-conversation,” in T. Givòn, (ed.), Syntax and Semantics 12: Discourse and Syntax, New York: Academic Press, pp. 261–86. — (2007), Sequence Organization in Interaction: A Primer in Conversation Analysis, Cambridge: Cambridge University Press. Schegloff, Emanuel A., Gail Jefferson, and Harvey Sacks (1977), “The preference in selfcorrection in the organization of repair in conversation,” Language, 53, 361–82. Scheibman, Joanne (2002), Point of View and Grammar: Structural Patterns of Subjectivity in American English Conversation, Amsterdam and Philadelphia: John Benjamins. Schmidt, Karl Horst (1965), “Indogermanisches Medium und Sataviso im Georgischen,” Bedi Kartlisa, 19–20, 129–35. Schmidtke-Bode, Karsten (2009), A Typology of Purpose Clauses, Amsterdam: John Benjamins. Schwartz, Arthur (1972), “The VP constituent of SVO languages,” in J. Kimball (ed.), Syntax and Semantics, 1, New York: Seminar Press, pp. 213–35. Schwartz, Bonnie and Sten Vikner (1996), “The verb always leaves IP in V2 Clauses,” in A. Belletti and L. Rizzi (eds), Parameters and Functional Heads, New York: Oxford University Press, pp. 11–62. Seefranz-Montag, Ariane von (1983), Syntaktische Funktionen und Wortstellungsveränderung: Die Entwicklung “subjektloser” Konstruktionen in einigen Sprachen. Munich: University of Munich dissertation. [Published as Studien zur theoretischen Linguistik 3, Munich: Fink, 1983.]

517

9781441124609_Ref_Final_txt_print.indd 517

11/20/2012 5:54:11 PM

Bibliography Seiler, Walter (1985), Imonda, a Papuan Language, Canberra: Australian National University. Selkirk, Elisabeth (1995), “Sentence prosody: Intonation, stress, and phrasing,” in J. A. Goldsmith (ed.), The Handbook of Phonological Theory, Cambridge: Blackwell, pp. 550–69. Shibatani, Masayoshi (1988), “Voice in Philippine languages,” in M. Shibatani (ed.), Passive and Voice, Amsterdam: John Benjamins, pp. 85–143. — (2004), “Voice,” in G. Booij et al. (eds), Morphology: A handbook on inflection and Word Formation, Vol. 2, Berlin: de Gruyter, pp. 1145–65. — (2006), “On the conceptual framework for voice phenomena,” Linguistics, 44(2), 217–69. Shibatani, Masayoshi (ed.) (1988), Passive and Voice, Amsterdam: John Benjamins. — (2001), The Grammar of Causation and Interpersonal Manipulation, Amsterdam: John Benjamins. Shlonsky, Ur (2000), “Remarks on the complementizer layer of standard Arabic,” in J. Lecarme, J. Lowenstamm, and U. Shlonsky (eds), Research in Afro-Asiatic Grammar, Amsterdam: Benjamins, pp. 48–59. Shopen, Timothy (ed.) (1985), Language Typology and Syntactic Description, Cambridge: Cambridge University Press. — (2007), Language Typology and Syntactic Description, Volume 1: Clause Structure (2nd edn) (first edn 1985), Cambridge: Cambridge University Press. Sidhakarya, Wayan (1995), Syntactic Roles and Constituent Order in Balinese. MA Thesis, University of Oregon. Sidnell, Jack (ed.) (2009), Conversation Analysis: Comparative Perspectives, Cambridge: Cambridge University Press. Siewierska, Anna (1984), The Passive: A Comparative Linguistic Analysis, London: Croom Helm. Sigurjónsdóttir, Sigríður and Nina Hyams (1992), “Reflexivization and logophoricity: Evidence from the acquisition of Icelandic,” Language Acquisition, 2, 359–413. Siloni, Tal (1990), Hebrew Noun Phrases: Generalized Noun Raising. MS, Université de Genève. — (1994), Noun Phrases and Nominalizations. Doctoral dissertation, Université de Genève. Silverstein, Michael (1976), “Hierarchy of features and ergativity,” in R. M. W. Dixon (ed.), Grammatical Categories of Australian Languages, Canberra: Australian Institute of Aboriginal Studies, pp. 112–71. Silvestri, Giuseppina (2011) “Prepositionless Genitive: Cases of Juxtaposition Genitive and N-to-D Raising in Certain Central and Southern Italian Dialects,” (paper presented at Italian Dialect in Diachrony (IDD) Conference, University of Leiden, May 19–21, 2011). Simpson, Andrew (2000), Wh-movement and the Theory of Feature-Checking, Amsterdam and Philadelphia: John Benjamins. Sinclair, John (1991), Corpus, Concordance, Collocation. Oxford: Oxford University Press. — (1996), “The search for units of meaning,” Textus, 9(1), 75–106 (reprinted in Sinclair, 2004a). — (1998), “The lexical item,” in E. Weigand (ed.), Contrastive Lexical Semantics, Amsterdam and Philadelphia: Benjamins, pp. 1– 24 (reprinted in Sinclair, 2004a). — (2003), Reading Concordances, London: Longman. — (2004a), Trust the Text. Language, Corpus and Discourse, London: Routledge. — (2004b), “New evidence, new priorities, new attitudes,” in J. Sinclair (ed.), How to Use Corpora in Language Teaching, Amsterdam and Philadelphia: Benjamins, pp. 271–99. — (forthcoming 2012), Essential Corpus Linguistics, London: Routledge. Sinclair, John and Anna Mauranen (2006), Linear Unit Grammar, Amsterdam and Philadelphia: Benjamins. Smith, Marcus (2007), “What is the word order of Pima?”. MS. Smyth, Herbert Weir (1920, 1984), Greek Grammar, Cambridge, MA: Harvard University Press. Snyder, William (2007). Child Language: The Parametric Approach, New York: Oxford University Press.

518

9781441124609_Ref_Final_txt_print.indd 518

11/20/2012 5:54:12 PM

Bibliography Sobin, Nicholas J. (1985), “Case assignment in Ukranian morphological passive constructions,” Linguistic Inquiry, 16, 649–62. Sohn, Ho-Min (1994), Korean, London and NewYork: Routledge. Song, Jae Jung (1990), “On the rise of causative affixes: A universal-typological perspective,” Lingua, 82, 151–200. — (1996), Causatives and Causation: A Universal-Typological Perspective, London: Addison Wesley Longman. — (2001), Linguistic Typology: Morphology and Syntax, Harlow: Pearson Longman. — (2005a), “Periphrastic causative constructions,” in M. Haspelmath, M. Dryer, D. Gil, and B. Comrie (eds), The World Atlas of Language Structures, Oxford: Oxford University Press, pp. 446–9. — (2005b), “Nonperiphrastic causative constructions,” in M. Haspelmath, M. Dryer, D. Gil, and B. Comrie (eds), The World Atlas of Language Structures, Oxford: Oxford University Press, pp. 450–3. Song, Jae Jung (ed.) (2011), The Oxford Handbook of Linguistic Typology, Oxford: Oxford University Press. Sornicola, Rosanna (1981), Sul parlato, Bologna: Il Mulino. Speer, Shari R. and Kiwako Ito (2008), “Prominence and phrasing in spoken discourse processing,” Proceedings of the Fifth Interantional Natural Language Generation Conference, Salt Fork, Ohio. The Association for Computational Linguistics (online at www.aclweb.org/ anthology-new/W/W08/W08–1101.pdf; accessed August 23, 2010). Spevak, Olga (2010), Constituent Order in Classical Latin Prose, Amsterdam and Philadelphia: John Benjamins Publishing Company. Sportiche, Dominique (2005), “Division of labor between merge and move: Strict locality of selection and apparent reconstruction paradoxes,” in Proceedings of the La Bretesche Borchard Workshop on the Division of Linguistic Labor. http://ling.auf.net/lingbuzz/000163. Stassen, Leon (1985), Comparison in universal grammar, Oxford: Basil Blackwell. — (2001), “Noun phrase coordination,” in M. Haspelmath, E. König, W. Österreicher, and W. Raible (eds), Language Typology and Language Universals. An International Handbook, Berlin, New York: Walter de Gruyter, pp. 1105–11. Stavrou Melita and Ianthi M. Tsimpli (2011), Definite Agreement in Complex NPs. MS, Aristotle University of Thessaloniki. Steele, Susan (1977), “Clisis and diachrony,” in C. N. Li (ed.), Mechanisms of Syntactic Change, Austin: Texas UP, pp. 539–79. Stein, Gabrielle (1979), Studies in the Function of the Passive, Tuebingen: Gunter Narr Verlag. Stowell, Timothy (1977), The Role of Clausal Embedding in a Grammar of Okinawan. MA thesis, University of Toronto. — (1981), Origins of Phrase Structure. PhD dissertation, Cambridge, MA: Massachusetts Institute of Technology. — (1983), “Subjects across categories,” The Linguistic Review, 2, 285–312. — (1989), “Raising in Irish and the Projection Principle,” Natural Language and Linguistic Theory, 7, 317–59. Stromswold, Karin (1990), Learnability and the Acquisition of Auxiliaries. Doctoral dissertation, Cambridge, MA: Massachusetts Institute of Technology. Suñer, Margarita (1994), “V-movement and the licensing of argumental Wh phrases in Spanish,” Natural Language and Linguistic Theory, 12, 335–72. Svartvik, Jan (1966), On Voice in the English Verb, The Hague: Mouton de Gruyter. Sweetser, Eve (1990), From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Semantic Structure, Cambridge: Cambridge University Press. Szabó, Zoltán (2007), “Compositionality,” in E. Zalta (ed.), The Stanford Encyclopedia of Philosophy, Spring 2007 edn (online at http://plato.stanford.edu/archives/spr2007/entries/ compositionality/). Szabolcsi, Anna (1981), “The possessive construction in Hungarian: A configurational category in a non-configurational language,” Acta Linguistica Academiae Scientiarum Hungaricae, 31, 261–89.

519

9781441124609_Ref_Final_txt_print.indd 519

11/20/2012 5:54:12 PM

Bibliography Szendröi, Kriszta (2001), Focus and the Syntax-Phonology Interface. PhD dissertation, London: University College London. Tada, Hiroaki (1993), A/A’ Partition in Derivation. PhD, Cambridge, MA: Massachusetts Institute of Technology. Takano, Yuji (2010), “Scrambling and control,” Linguistic Inquiry, 41, 83–110. Tanaka, Hiroko (1999), Turn-Taking in Japanese Conversation, Pragmatics & Beyond New Series 56, Amsterdam and Philadelphia: John Benjamins. Tao, Hongyin (1996), Units in Mandarin Conversation: Prosody, Discourse, and Grammar, Studies in Discourse and Grammar 5, Amsterdam and Philadelphia: John Benjamins. Taraldsen, Knut Tarald (1990), “D-projections and N-projections in Norwegian,” in M. Nespor and J. Mascaró (eds), Grammar in Progress, Dordrecht: Foris, pp. 419–31. Taylor, Ann (1990), Clitics and Configurationalilty in Ancient Greek. PhD dissertation, Phildelphia, PA: University of Pennsylvania. — (2003), “The distribution of object clitics in Koine Greek,” in M. Southern (ed.), IndoEuropean Perspectives. Journal of Indo-European Studies Monograph Series 41,Washington, DC: Institute for the Study of Man; an earlier version appeared in Penn Working Papers in Linguistics, 2, 121–39. Taylor, Ann, Mitchell Marcus, and Beatrice Santorini (2003), “The Penn Treebank: An overview,” in A. Abeillé (ed.), Treebanks: Building and Using Parsed Corpora, Dordrecht: Kluwer, pp. 5–22. Taylor, Charles (1985), Nkore-Kiga, London: Croom Helm. Tegey, Habibullah (1977), The Grammar of Clitics: Evidence from Pashto and Other Languages. PhD thesis, University of Illinois. Thiersch, Craig (1978), Topics in German Syntax. PhD dissertation, Cambridge, MA: Massachusetts Institute of Technology. Thomason, Richmond H. (1976), “Some extensions of Montague Grammar,” in B. Partee (ed.), Montague Grammar, London: Academic Press, pp. 77–119. Thompson, Sandra A. (1985), “Grammar and written discourse. Initial and final purpose clauses in English,” in T. Givón (ed.), Quantified Studies in Discourse, Special issue of Text, 5, pp. 55–84. — (1987), “Subordination’ and narrative event structure,” in R. Tomlin (ed.), Coherence and Grounding in Discourse, Amsterdam: John Benjamins, pp. 435–54. — (2002), “Object Complements and conversation: Towards a realistic account,” Studies in Language, 26(1), 125–64. Thompson, Sandra A., Robert E. Longacre, and Shin Ja J. Hwang (2007), “Adverbial clauses,” in T. Shopen (ed.), Language Typology and Syntactic Description, Vol II: Complex Constructions, Cambridge: Cambridge University Press, pp. 237–300. Timberlake, Alan (1982), “The impersonal passive in Lithuanian,” in M. Macaulay and O. D. Gensler (eds), BLS, 8, pp. 508–35. Tomaselli, Alessandra (1990), La sintassi del verbo finito nelle lingue germaniche, Padova: Unipress. Tomić, Olga Mišeska (1996), “The balkan slavic nominal clitics,” in Halpern and Zwicky (eds), Approaching Second: Second Position Clitics and Related Phenomena, Stanford: CSLI Press, pp. 511–35. Tomlin, Russell (1995), “Focal attention, voice, and word order: An experimental, cross-linguistic study,” in P. Downing and M. Noonan (eds), Word Order in Discourse, Amsterdam: John Benjamins, pp. 517–54. Trask, Robert L. (2003), “The noun phrase,” in J. I. Hualde and J. Ortiz de Urbina (eds), A Grammar of Basque, Berlin and New York: Mouton de Gruyter, pp. 113–70. Truckenbrodt, Hubert (1995), Phonological Phrases: Their Relation to Syntax, Focus, and Prominence. PhD dissertation, Cambridge, MA: Massachusetts Institute of Technology. — (1999), “On the relation between syntactic phrases and phonological phrases,” Linguistic Inquiry, 30, 219–55. Trueswell, John C., Irina A. Sekerina, Nicole M. Hill, and Marian L. Logrip (1998), “The kindergarten-path effect: Studying on-line sentence processing in young children,” Cognition, 73, 89–134.

520

9781441124609_Ref_Final_txt_print.indd 520

11/20/2012 5:54:12 PM

Bibliography Tsunoda, Tasaku (1988), “Antipassives in Warrungu and other Australian languages,” in M. Shibatani, Passive and Voice, Amsterdam: John Benjamins, pp. 595–651. Uhmann, Susanne (2001), “Some arguments for the relevance of syntax to same-sentence self-repair in everyday German conversation,” in M. Selting and E. Couper-Kuhlen (eds), Studies in Interactional Linguistics, Amsterdam and Philadelphia: John Benjamins, pp. 373–404. Utzeri, Irene (2007), Produzione e acquisizione delle frasi relative sul soggetto e sull’oggetto in italiano: un contributo sperimentale allo studio comparativo (Production and comprehension of subject and object relative clauses in Italian: an experimental contribution to comparative studies), MA thesis, University of Siena. Vallduvi, Enric (1992), The Informational Component. Outstanding Dissertations in Linguistics Series, New York: Garland. Valois, Daniel (1991), The Internal Syntax of DP. Doctoral dissertation, University of California, Los Angeles. Van den Berg, René (1989), A Grammar of the Muna Language, Dordrecht: Foris. Van der Auwera, Johan (1998), “Conclusion,” in J. van der Auwera and D. P. Ó Baoill (eds), Adverbial Constructions in the Languages of Europe, Empirical Approaches to Language Typology – EUROTYP 20–3, Berlin: Mouton de Gruyter, pp. 813–36. Van Riemsdijk, Henk (1999), Clitics in the Languages of Europe, Berlin and New York: Mouton de Gruyter. Van Valin, Robert D. Jr, and Randy LaPolla (1997), Syntax, Structure, Meaning, and Function, Cambridge: Cambridge University Press. Varlokosta, Spyridoula (2000), “Lack of clitic-pronoun distinctions in the acquisition of Principle B in child Greek,” in S. C. Howell, S. Fish, and T. Keith-Lucas (eds), Proceedings of the 24th annual Boston University Conference on Language Development, Somerville, MA: Cascadilla Press, pp. 738–48. Vennemann, Theo (1974), “Theoretical word order studies: Results and problems,” Papiere zur Linguistik, 7, 5–25. Vergnaud, Jean.-Roger (1974), French Relative Clauses. Doctoral dissertation, Cambridge, MA: Massachusetts Institute of Technology. — (1977), Unpublished Letter to Noam Chomsky and Howard Lasnik, dated April 17, 1977. Verstraete, Jean-Christophe (2004), “Initial and final position of adverbial clauses in English: the constructional. Basis of the discoursive and syntactic differences,” Linguistics, 42, 819–853. — (2007), Rethinking the Coordinate-Subordinate Dichotomy. Interpersonal Grammar and the Analysis of Adverbial Clauses in English, Berlin: Mouton de Gruyter. Vikner, Sten (1994), “Scandinavian object shift and West German,” in N. Corver and H. van Riemsdijk (eds), Studies on Scrambling: Movement and Non-Movement Approaches to Free Word-Order phenomena, The Hague: Mouton, pp. 487–517. — (1995), Verb Movement and Expletive Subjects in the Germanic Languages, New York: Oxford University Press. Vilkuna, Maria (1989), Free Word Order in Finnish: Its Syntax and Discourse Functions, Helsinki: Finnish Literature Society. Villalba, Xavier (2008), “Exclamatives: A thematic guide with many questions and few answers,” Catalan Journal of Linguistics, 7, 9–40. Vitale, Anthony J. (1981), Swahili Syntax, Dordrecht: Foris. Wackernagel, Jacob (1892), “Über ein Gesetz der indogermanischen Wortstellung,” Indogermanische Forschungen, 1, 333–436. Wälchli, Bernhard (2005), Co-Compounds and Natural Coordination, Oxford: Oxford University Press. Wallis, Sean (2008), “Searching treebanks and other structured corpora,” in A. Lüdeling and M. Kytö (eds), Corpus Linguistics: An International Handbook. Volume 1, Berlin and New York: de Gruyter, pp. 738–58. Wanner, Dieter (1987), The Development of Romance Clitic Pronouns from Latin to Old Romance, Berlin and New York: Mouton de Gruyter.

521

9781441124609_Ref_Final_txt_print.indd 521

11/20/2012 5:54:12 PM

Bibliography Wasow, Thomas (1977), “Transformations and the lexicon,” in P. W. Culicover, T. Wasow, and A. Akmajian (eds), Formal Syntax, New York: Academic Press, pp. 327–60. Watanabe, Akira (2002a), “Loss of wh-movement in Old Japanese,” in D. Lightfoot (ed.), Syntactic Effects of Morphological Change, Oxford: Oxford University Press, pp. 179–95. — (2002b), “Wh in situ, subjacency and chain formation,” Occasional Papers in Linguistics 2, Department of linguistics and philosophy, Massachusetts Institute of Technology. Watkins, Calvert (1976) “Toward Proto-Indo-European syntax: Problems and pseudoproblems,” in S. B. Steever, C. A. Walker, and S. S. Mufwene (eds), Proceedings of the Chicago Linguistics Society: Papers from the parasession on diachronic syntax, Chicago: Chicago Linguistics Society, pp. 305–26. Webber, David John (1989), A Grammar of Huallaga (Huánuco) Quechua, Berkeley: University of California Press. Webelhuth, Gert (1992), Principles and Parameters of Syntactic Saturation, Oxford: Oxford University Press. Williams, E. (1980), “Predication,” Linguistic Inquiry, 11, 203–38. Wood, Mary McGee (1993), Categorial Grammars, London: Routledge. Wordick, Frank J. F. (1982), The Yindjibarndi Language (Pacific Linguistics, C-71), Canberra: ANU. Wouk, Fay (2008), “The syntax of intonation units in Sasak,” Studies in Language, 32(1), 137–62. Yip, Moira, Joan Maling, and Ray Jackendoff (1987), “Case in tiers,” Language, 63, 217–50. Yuasa, Etsuyo and Jerry Sadock (2002), “Pseudo-subordination: A mismatch between syntax and semantics,” Journal of Linguistics, 38, 87–111. Zagona, Karen (2002), The Syntax of Spanish. Cambridge: Cambridge University Press. Zanuttini, Raffaella (1997), Negation and Clausal Structure, Oxford: Oxford University Press. — (2001), “Sentential negation,” in M. Baltin and C. Collins (eds), The Handbook of Contemporary Syntactic Theory, Malden: Blackwell, pp. 511–35. Zavala, Roberto (2000), Inversion and Other Topics in the Grammar of Olutec (Mixean). PhD dissertation, University of Oregon. Zeijlstra, Hedde (2004), Sentential Negation and Negative Concord. Doctoral dissertation, University of Amsterdam. Zigmond, Maurice L., Curtis G. Booth, and Pamela Munro (1990), Kawaiisu: Grammar and Dictionary, with Texts. University of California Publications in Linguistics, 119, Berkeley: University of California Press. Zipf, George K. (1949), Human Behavior and the Principle of Least Effort, Cambridge, MA: Addison-Wesley. Zubizarreta, Maria Luisa (1980), “Pour une restructuration thématique,” Recherches Linguistiques, 9, Université de Paris VIII, 141–87. — (1994), “On some prosodically governed operations,” in G. Cinque, J. Koster, J. Y. Pollock, L. Rizzi, and R. Zanuttini (eds), Paths Toward Universal Grammar. Studies in Honor of Richard S. Kayne, Washington, DC: Georgetown University Press, pp. 473–85. — (1998), Prosody, Focus, and Word Order, Cambridge, MA: Massachusetts Institute of Technology Press. Zúñiga, Fernando (2006), Deixis and Alignment. Inverse Systems in Indigenous Languages of the Americas, Amsterdam: John Benjamins. Zwicky, Arnold (1970), “The free-ride principle and two rules of complete assimilation in English,” in M. A. Campbell et al. (eds), Papers from the Sixth Regional Meeting of the Chicago Linguistic Society, Chicago: Chicago Linguistic Society, pp. 579–88. — (1977), On Clitics, Bloomington: Indiana University Club. — (1985), “Clitics and particles,” Language, 61, 283–305. — (1994), “Dealing out meaning: Fundamentals of grammatical constructions,” Berkeley Linguistics Society Proceedings, 20, 611–25.

522

9781441124609_Ref_Final_txt_print.indd 522

11/20/2012 5:54:12 PM

Author Index Aarts, Bas 51, 56 Abeillé, Anne 53 Abney, Steven 88, 196, 214, 317 Aboh, Enoch O. 446, 448 Ackema, Peter 455 Acquaviva, Paolo 364 Adams, James N. 178, 193 Adger, David 14, 16, 327 Agbayani, Brian 193 Aikhenvald, Alexandra Y. 240, 252, 270 Aissen, Judith 219 Alexiadou, Artemis 88 Allan, Rutger 271, 273–4 Alsina, Alex 249 Alves, Flávia 234 Andersen, Paul Kent 261, 280 Anderson, Mona 91 Anderson, Stephen R. 135, 141–2, 168, 171, 172, 192, 211 Andrews, Avery 220, 368 Antinucci, Francesco 456 Applegate, Richard Brian 122, 141 Arregi, Karlos 316 Artiagoitia, Xabier 101 Auer, Peter 417–19 Austin, Peter 67 Avrutin, Sergey 400 Baauw, Sergio 399, 401 Badecker, William 401 Bailyn, John F. 152 Baker, Mark 1–2, 19, 22–31, 89, 92, 139, 142, 144, 146, 153, 226, 240, 251, 435 Barðdal, Jóhanna 381–8 Barnbrook, Geoff 54 Barnes, Michael 368 Bartonek, Antonín 184 Bayer, Josef 144, 150 Behaghel, O o 365 Belle i, Adriana 365 Belloro, Valeria A. 192 Beltrami, Giulia 53–4 Benediktsdó ir, Ásbjörg 377 Benincà, Paola 160–1, 164, 189, 192, 448, 450 Bentzen, Kristine 141, 159, 163 Berinstein, Ava 260

Bernstein, Judy 88 Berre a, Monica 192 Berruto, Gaetano 395 Bhaskararao, P. 202 Bha , Rakesh M. 163, 438 Bianchi, Valentina 315, 448, 451 Biber, Douglas 55 Biberauer, Theresa 96, 379 Bidese, Ermenegildo 163 Biktimir, Tuvana 254 Bion, Ricardo A. H. 391 Blanche-Benveniste, Claire 395 Bloom, Paul 399 Bloomfield, Leonard 218 Bocci, Giuliano 440, 445 Boeckx, Cedric 87 Bögel, Tina 170 Boeder, Winfried 270 Böhmová, Alena 56 Boley, Jacqueline 182 Booth, Curtis G. 132 Borer, Hagit 114, 318, 435 Borsley, Robert D. 292–3 Bošković, Željko 145, 150, 152, 322, 328 Bosque, Ignacio 340 Bossong, Georg 179, 236 Brandi, Luciana 192 Bresnan, Joan 67, 204, 219–20, 228, 236, 240, 243, 292, 449 Broadwell, George A. 233 Brody, M. 316, 448 Broekhuis, Hans 452, 462 Brown, Jerrold C. 142 Brown, Roger 392–3 Brown, Sue 357–8 Browning, Marguerite A. 450 Brune i, Lisa 462 Büring, Daniel 458 Bybee, Joan 15, 51, 74–5, 171, 173, 428 Calabrese, Andrea 456 Camacho, José 292 Campbell, Lyle 258, 260, 365, 367, 373 Cardinale i, Anna 395–6, 401, 460 Carlson, Robert 239, 343 Castroviejo, Elena 314 Cecche o, C. 402, 451

523

9781441124609_Aut_Index_Final_txt_print.indd 523

11/20/2012 9:26:07 AM

Author Index Cennamo, Michela 276–7 Chafe, Wallace L. 39, 47, 68, 343, 347 Chametzky, Robert A. 87 Cheng, Lisa 317 Chien, Yuchin 398–400 Chierchia, Gennaro 401 Chomsky, Noam 5, 7, 22–8, 31, 73, 76–7, 79, 81–7, 93, 98–9, 117, 139–40, 156, 195, 204, 206–9, 212, 214, 219, 223, 240, 315, 317–18, 322, 324, 327, 338–9, 366, 423–4, 439, 445, 450–5, 457, 462 Christophe, Anne 391 Chumbow, Beban S. 119 Chung, Sandra 142 Cinque, Guglielmo 90, 129, 136, 142, 189, 192, 436, 443, 445, 447, 450, 456 Citko, Barbara 87 Clahsen, Harald 391, 393 Cole, Peter 130, 287 Collins, Chris 87, 240 Comrie, Bernard 230, 232, 247, 272, 284–5, 287, 398 Conrad, Susan 55 Conroy, Anastasia 399–401 Coopmans, Peter 400 Cooreman, Ann 258 Cordin, Patrizia 192 Cormack, Annabel 292–3 Cornilescu, Alexandra 103 Corver, Norbert 144, 317 Couper-Kuhlen, Elizabeth 406, 417 Couro, Ted 131 Craig, Cole e 259 Crain, Stephen 401 Creissels, Denis 236, 239 Crisma, Paola 88, 95, 99, 379 Cristofaro, Sonia 294, 300, 307–8, 351 Cro , William 1, 11–21, 30–1, 68, 219–20, 229–30, 236, 276, 310 Cruschina, Silvio 444, 450 Cruse, D. Alan 17 Dahlstrom, Amy 235, 237 Davies, John 288 Davies, William D. 255 Dayal, Veneeta 319 De Haan, Gerard 348 De Hoop, Helen 455 Deane, Paul 294 Dehé, Nicole 458 Dekkers, Joost 452, 455 DeLancey, Sco  228, 234, 236, 238–9 Delbrück, Berthold 192, 365 Delfi o, Denis 95, 100 Delsing, Lars-Olof 379–80 Demonte, Violeta 340

Den Besten, Hans 155, 159 Déprez, Viviane 147 Devine, Andrew M. 188 Di Tullio, Ángela 340 Diesing, Molly 150, 169 Diessel, Holger 5, 341–53 Dik, Simon 228, 230, 293, 295 Dirven, René 279 Dixon, Robert M. W. 221, 225, 230, 232, 234–5, 240, 259, 270 Donohue, Mark 234–5 Dover, Kenneth J. 184 Dowty, David 228 Dryer, Ma hew S. 102, 123–4, 129–30, 141, 236, 240, 244, 249, 251, 260 Dubinsky, Stanley 255 DuBois, John 238 Duffield, Nigel 95 Duranti, Alessandro 232 Egerland, Verner 179 Etxepare, Ricardo 101 Eythórsson, Thórhallur 5–6, 202–3, 365–88 Faarlund, Jan Terje 139, 141–2 Fanselow, Gisbert 145, 323 Farmer, Ann K. 209–10 Fillmore, Charles 57, 62, 201, 220, 230, 424, 426, 429 Fiorentino, Giuliana 395 Fischer, Olga 242 Flobert, Pierre 273, 275 Fontana, Josep 160, 190, 192 Ford, Cecilia E. 343, 347, 350, 352, 409, 417 Forrest, Linda 231 Fortescue, Michael 259 Fox, Barbara A. 70–2, 240, 413 Fox, Danny 153 Fraenkel, Eduard 184 Francis, Gill 51, 58, 60 Frascarelli, Mara 439, 448, 451 Freidin, Robert 327 Friedemann, M. A. 502, 504 Friedmann, Na’ama 398 Fuss, Eric 162, 164 Garside, Roger 56 Gast, Volker 269, 279–80 Gazdar, Gerald 429 Gelderen, Elly Van 355 Geniušienė, Emma 265, 268, 278, 280 George, Leland 254 Giacalone Ramat, Anna 309 Giannakidou, Anastasia 357, 364 Gianollo, Chiara 94, 96 Gildea, Spike 224, 234, 237, 239

524

9781441124609_Aut_Index_Final_txt_print.indd 524

11/20/2012 9:26:07 AM

Author Index Gildersleeve, Basil L. 240 Gili Gaya, Samuel 340 Ginzburg, Jonathan 317 Giorgi, Alessandra 89–91, 96, 105, 107, 117 Givón, Talmy 65, 68, 219–20, 228, 230, 238, 240, 251, 306, 342–4, 405 Gleason, Henry Allan Jr. 218 Goldberg, Adele 2, 227–8, 430–1, 434 Goldstein, David M. 184–8, 191–3 Golinkoff, Roberta M. 390 Golston, Chris 193 Gonda, Jan 273, 275 Goodluck, H. 395 Goodwin, Charles 70 Gordon, Lynn 235 Green, Georgia M. 346, 348 Greenberg, Joseph H. 3, 28, 97, 124, 126, 128–30, 135, 139–41, 210, 245 Grewendorf, Günther 144, 153, 448 Grimes, Joseph E. 405 Grimshaw, Jane 228, 399, 457, 462 Grodzinsky, Yosef 399 Grolla, Elaine 400 Grosu, Alexander 103 Guardiano, Cristina 94, 97, 99, 116, 381 Guasti, Maria Teresa 5–6, 163, 389–402 Günthner, Susanne 348, 413–14, 416 Haegeman, Liliane 5, 327, 354–64, 393–4, 438 Haiman, John 287, 293, 303, 307, 342, 350 Hajičová, Eva 57 Hale, Kenneth 67, 138–9, 141, 144, 170, 199, 365 Halliday, M. A. K. 296, 342 Halpern, Aaron 169 Hamann, Cornelia 392, 401 Hamblin, Charles 318–19 Harbert, Wayne 384 Harris, Alice C. 220, 238, 365, 367, 373, 381 Harrison, S. P. 381 Hasan, Ruqaiya 296 Hashimoto, Mantaro J. 243 Haspelmath, Martin 129, 170, 220, 240, 243, 280, 288, 293, 295–7, 300–1, 342 Hawkins, John A. 102, 125, 129, 256 Hayes, Bruce 459 Heidinger, Steffen 277 Heine, Bernd 276–7 Helasvuo, Marja-Liisa 2–3, 6, 65–75, 328, 417 Henry, Alison, 314 He erle, Katja 352–3 Higginbotham, James 89 Hinterhölzl, Roland 144, 149, 153, 448, 451

Hiraiwa, Ken 357, 359 Hiramatsu, Kazuko 393 Hirsh–Pasek, Kathryn 390 Hock, Hans Henrich 170, 182, 184, 192–3 Hocke , Charles 218 Hoffner, Harry A. 182 Holmberg, Anders 147, 153 Hopper, Paul 6, 51, 72, 171, 240, 265, 273, 350, 405–20 Hornstein, Norbert 87, 208, 327 Horvath, Julia 316 Hout, Theo P.J. van den 182 Hróarsdó ir, Thorbjörg 379 Huang, James 319–21 Huddleston, Rodney 62 Hulk, Aa e 392 Hunston, Susan 51, 58, 60–1 Hyams, Nina 399 Hyman, Larry 232 Ide, Nancy 62 Ito, Kiwako 67 Jackendoff, Ray 87, 139–40, 228, 370, 425 Jacobson, Pauline 66 Jaeggli, Osvaldo 91, 240 Jakubowicz, Celia 399 Jankuhn, Harald 273 Janse, Mark 188 Jasperson, Robert 70–2 Jelinek, Eloise 144, 150 Jespersen, O o 365, 368 Johannessen, Janne B. 292, 300 Johnson, Kyle 92, 240 Johnson, Mark 260 Johnson-Laird, P. N. 350 Joseph, Brian 166, 188, 289 Kahnemuyipour, Arsalan 457, 459–61 Kaplan, Ronald M. 204 Karimi, Simin 144 Kärkkäinen, Elise 70 Kaufman, Diana K. 400 Kave, Gitit 399 Kay, Paul 62, 422, 424–6, 432–3 Kayne, Richard. S. 29, 79, 87, 102, 139–40, 210, 315, 355, 394, 437 Kazanina, Nina 401 Kazenin, Konstantin 265 Keenan, Edward L. 4, 88, 218–19, 221, 224–5, 240–60, 264, 267, 280, 284–5, 367, 378, 398 Kemmer, Suzanne 265–6, 275–7, 279 Keyser, Samuel Jay 280, 435 Kiguchi, Hirohisa 399

525

9781441124609_Aut_Index_Final_txt_print.indd 525

11/20/2012 9:26:07 AM

Author Index Kikusawa, Ritsuko 381 Kimenyi, Alexandre 244, 248, 250, 255, 257 Kiparsky, Paul 137 Kirsner, Robert 251 Kiss, Katalin É. 146, 217, 448 Klaiman, M. H. 265 Klavans, Judith 167–8, 170, 182 Knjazev, Ju. P. 273 Ko, Heejeong 153 Koller, Bernhard 192 König, Christa 233–4 König, Ekkehard 269, 279–80 Koopman, Hilda 216 Kornfilt, Jaklin 144, 150, 254, 308 Kortmann, Bernd 303 Kozinsky, Isaac S. 258 Kramer, Ruth 120, 139 Krishnamurthy, Ramesh 58 Kroch, Anthony 366, 464 Kulikov, Leonid 4, 261–80 Kuteva, Tania 276–7 Laakso, Minna 70 Labelle, Marie 395–6 Ladusaw, William 364 Laka, Itziar 357 Lakoff, George 346, 348 Lakoff, Robin 296 Lambrecht, Knud 236, 343–5, 434–5 Lang, Ewald 295–7 Langacker, Ronald W. 15, 73–4, 219, 227, 230, 292–3 Langdon, Margaret 131, 141 LaPolla, Randy 219–21, 225–7, 229–30, 238 Larson, Richard K. 79, 217 Lasnik, Howard 240, 444 Lazard, Gilbert 182 Lazdina, Tereza B. 251 Lecarme, Jacqueline 97 Leddon, Erin M. 401 Lee, Felicia A. 129 Lee, J. 466 Lee, Young-Suk 148 Leech, Geoffrey 51 Legendre, Geraldine 455 Lehmann, Christian 301, 342, 365 Lehmann, Winfred P. 123 Lerner, Gene 71 Levin, Beth 430 Li, Charles N. 238, 405 Liddicoat, Anthony L. 418 Lidz, Jeffrey L. 401 Lightfoot, David W. 366, 368, 377, 379, 381 Lillehaugen, Brook Danielle 140–1 Lillo-Martin, Diane 393

Lindstad, Arne Martinus 355 Linell, Per 407 Litvinov, Viktor P. 280 Lodge, Gonzalez 240 Lohndal, Terje 354–64 Longacre, Robert E. 220, 296, 301, 405 Longobardi, Giuseppe 3, 88–117, 195–6, 379, 381 Luraghi, Silvia 4, 8, 116, 134, 139, 141–2, 165–93, 274–5, 327–8, 463–8 Lyons, John 87 Mahajan, Anoop 148–51 Mahootian, Shahrzad 342 Mair, Christian 51 Malchukov, Andrej 300 Maling, Joan 370, 373, 375, 378 Malouf, Robert 423 Manley, Timothy M. 245 Manney, Linda Joyce 277 Manning, Christopher 51, 53 Manorohanta, Cécile 251 Marchese, Lynell 343 Marušič, Franc 169, 192 Mason, Oliver 61 Ma hiessen, Christian 342 Mauranen, Anna 60 Mauri, Caterina 5, 292–310, 342 May, Robert 321 Mazuka, Reiko 391 McCarthy, John J. 455 McCawley, James D. 139 McDaniel, Dana 396 McEnery, Tony 52, 56 Mchombo, Sam 249 McKee, Cecile 399, 401 Mel’čuk, Igor A. 280 Merchant, Jason 313 Merlan, Francesca 235 Meurers, W. Detmar 53–4 Michaelis, Laura 6, 421–35 Miestamo, Ma i 119 Mikulová, Marie 467 Milner, Jean-Claude 90 Mithun, Marianne 2, 32–50, 232, 235, 296, 299 Miyagawa, Shigeru 150, 152–3, 248 Monachesi, Paola 192 Moravcsik, Edith A. 276 Moscati, Vincenzo 355 Moshi, Lioba 219, 236 Mous, Maarten 275 Müller, Gereon 53–4, 149–50, 153, 280, 455 Müller, Stefan 53–4 Müller-Bardey, Thomas 280

526

9781441124609_Aut_Index_Final_txt_print.indd 526

11/20/2012 9:26:07 AM

Author Index Munn, Alan 117 Munro, Pamela 3, 118–42, 144, 152, 154–5, 198, 200, 235

Progovac, Ljiljana 117 Pullum, Geoffrey K. 62, 142 Puskas, Genoveva 448

Nedjalkov, Vladimir P. 269, 280 Neeleman, Ad 144, 147, 149, 151 Nelson, Gerald 56 Nespor, Marina 173, 191, 391, 459 Neu, Erich 273–5 Nevis, Joel A. 165 Nichols, Johanna 102–3 Nishigauchi, Taisuke 321 Nissenbaum, Jonathan 321 Nunes, Jairo 2, 66, 76–87, 327

Quicoli, Carlos 5, 80, 312, 325–40, 351 Quirk, Randolph 415

Ochs, Elinor 406–7 Ono, Tsuyoshi 71, 417 Ordóñez, Francisco 3, 133, 143–53, 155, 164 Ortiz de Urbina, Jon 316 O en, Heinrich 180 Özkaragöz, Inci 251, 254 Palmer, Frank R. 219–20 Paperno, Denis 88 Paradisi, Paola 95, 100 Parodi, Claudia 5, 80, 312, 325–40, 351 Paul, Ileana 257 Payne, Doris 4, 138, 197, 204, 218–39, 238, 284, 287 Payne, John 234, 295, 297–8 Payne, Thomas 220, 225, 234, 236–7 Pérez-Leroux A. T. 395 Penka, Doris 358 Perlmu er, David M. 204, 219–20, 226, 240 Pesetsky, David 5, 153, 219, 317, 321–2, 324 Philip, William 400 Philippaki-Warburton, Irene 289 Phillips, Colin 401 Pike, Kenneth 218 Pinault, Georges-Jean 182 Pintzuk, Susan 380 Piñar, Larruba Pilar 364 Pires, Acrisio 381 Plank, Frans 88, 232 Plann, Susan 438 Plunke , Bernade e 392 Pole o, Cecilia 3, 134, 136, 154–64, 192, 210–11, 448, 450 Polinsky, Maria 254 Pollard, Carl 219, 424 Pollock, Jean-Yves 208, 217, 355 Pooth, Roland 275 Postal, Paul M. 196, 204, 219, 226, 240, 251 Prince, Alan 7, 452, 455, 459 Prince, Ellen 415

Radanović Kocić, Vesna 170 Radford, Andrew 109, 438 Randriamasimanana, Charles 249 Rappaport Hovav, Malka 430 Rasolofo, Andoveloniaina 224, 232 Rebuschi, Georges 292 Reinhart, Tanya 116, 322 Renzi, Lorenzo 192 Richards, Marc 144, 321 Richards, Norvin 322 Ri er, Elizabeth 100, 113 Rizzi, Luigi 7, 153, 159, 161, 196, 199, 217, 313, 317, 393–4, 396, 436–50, 457, 460–2 Roberts, Craig 441 Roberts, Ian 366, 373, 379, 381, 438 Roberts, John R. 288 Roeper, Thomas 92, 280 Rögnvaldsson, Eiríkur 139, 159, 379 Romary, Laurent 62 Rooth, Mats 444 Rosch, Eleanor 219 Rosen, Carol 220, 226, 238, 240 Rosen, Sara T. 399 Ross, John R. 136–7, 143, 294, 313, 450 Rouveret, Alain 95 Rowlands, Evan C. 298 Rudin, Catherine 322 Rudolph, Elisabeth 296 Ruppenhofer, Josef 427, 431 Russi, Cinzia 169, 191 Rutkowski, Pawel 117 Ryan, Robert 268 Sabel, Joachim 144, 153 Sacks, Harvey 71 Sadler, Louisa 243 Sadock, Jerry 292 Sag, Ivan 219, 240, 317, 422–5, 428 Saito, Mamoru 144, 147–9, 151–3, 439 Saltarelli, Mario 289 Salvi, Giampaolo 176, 178–9 Samek-Lodovici, Vieri 7, 439, 452–62 Samper-Padilla, José A. 399 Santelmann, Lynn 392–3 Santorini, Beatrice 148, 159, 464 Sapir, Edward 220, 232, 235 Schachter, Paul 224–5, 293–4 Schäufele, Steven 182, 192

527

9781441124609_Aut_Index_Final_txt_print.indd 527

11/20/2012 9:26:07 AM

Author Index Schegloff, Emanuel A. 70, 407 Scheibman, Joanne 15, 74 Schmidt, Karl Horst 270 Schmidtke-Bode, Karsten 351 Schwartz, Arthur 142 Schwartz, Bonnie 159 Seefranz-Montag, Ariane von 386 Seiler, Walter 351–2 Selkirk, Elisabeth 459 Selting, Margret 406, 409 Shibatani, Masayoshi 240, 246, 261, 265, 280–1, 287 Shlonsky, Ur 95, 456 Shopen, Timothy 220 Shyldkrot, Hava Bat-Zeev 276 Sidhakarya, Wayan 224 Sidnell, Jack 407 Siewierska, Anna 240, 280 Sigurjónsdó ir, Sigríður 373, 375, 378, 399 Siloni, Tal 113, 117 Silverstein, Michael 219, 231 Silvestri, Giuseppina 3, 88–117, 195 Simpson, Andrew 317 Sinclair, John 51, 54, 58–61 Smith, Marcus 138 Smith, Neil 292–3 Smolensky, Paul 7, 452, 455 Smyth, Herbert Weir 240 Snyder, William 28 Sobin, Nicholas J. 253 Sohn, Ho-Min 298 Song, Jae Jung 4–5, 236, 281–91 Sorjonen, Marja-Leena 70 Sornicola, Rosanna 395 Speer, Shari R. 67 Spevak, Olga 142, 177 Sportiche, Dominique 216–17 Starke, Michael 401 Stassen, Leon 300–1 Stavrou, Melita 94 Steele, Susan 170 Stein, Gabrielle 251 Stephens, Laurence D. 188 Sternefeld, Wolfang 144, 149–50, 153 Stowell, Timothy 4, 107, 194–215, 220, 334, 444 Straub, Kathleen 401 Stromswold, Karin 393 Subbarao, K. V. 202 Suñer, Margarita 438 Svartvik, Jan 251 Sweetser, Eve 345–6 Szabó, Zoltán 429–30 Szabolcsi, Anna 88 Szendröi, Kriszta 458

Tada, Hiroaki 148–9 Takahashi, Daiko 145, 150, 152 Takano, Yuji 152 Tamanji, Pius N. 119 Tanaka, Hiroko 71 Tao, Hongyin 68–9 Taraldsen, Knut Tarald 117 Taylor, Ann 55, 62, 193, 463 Taylor, Charles 301 Tegey, Habibullah 170 Thiersch, Craig 211 Thomason, Richmond H. 256 Thomason, Sarah G. 381 Thompson, Ellen 87 Thompson, Sandra A. 48, 71, 238, 265, 273, 342–5, 349, 351, 409–11, 413, 418 Thornton, Rosalind 399 Timberlake, Alan 253–4 Tomaselli, Alessandra 156 Tomlin, Russell 219, 231 Tomić, Olga Mišeska 168 Torrego, Esther 5, 317, 324 Trask, Robert L. 101–2 Truckenbrodt, Hubert 457–9 Tsimpli, Ianthi M. 94 Tsunoda, Tasaku 259 Uhmann, Susanne 70–1 Uriagereka, Juan 240 Utzeri, Irene 397 Vallduvi, Enric 439 Valois, Daniel 88 Van den Berg, Helma 230 Van den Berg, René 124 Van der Auwera, Johan 164 Van der Wurff, Wim 242 Van Riemsdijk, Henk 144, 191 Van Valin, Robert D. Jr. 219–21, 225–6, 229–30, 301 Vanelli, Laura 192 Varlokosta, Spyridoula 395, 401 Vennemann, Theo 123 Vergnaud, Jean-Roger 203, 213, 215, 394, 459 Verstraete, Jean-Christophe 345, 347–8 Vikner, Sten 147, 156–9, 164 Vilkuna, Maria 67 Villalba, Xavier 314 Vitale, Anthony J. 289 Vogel, Irene 173, 191, 459 Wackernagel, Jacob 3–4, 134–5, 168, 170, 172, 178–9, 181, 184, 187, 192, 365 Wälchli, Bernhard 296

528

9781441124609_Aut_Index_Final_txt_print.indd 528

11/20/2012 9:26:08 AM

Author Index Wallis, Sean 56 Wanner, Dieter 167, 176, 190, 192 Wasow, Thomas 243 Watanabe, Akira 130, 321 Watkins, Calvert 365, 380 Webber, David John 243–4, 248 Webelhuth, Gert 147, 149 Westergaard, Marit 163 Wexler, Kenneth 398–400 Wichmann, Søren 234–5 Williams, E. 212 Wood, Mary McGee 219 Wordick, Frank J. F. 249 Wouk, Fay 68–70

Yip, Moira 370, 372 Yoon, James 438 Yuasa, Etsuyo 292 Zagona, Karen 340 Zanu ini, Raffaella 363 Zavala, Roberto 232 Zeijlstra, Hedde 356–61, 363–4 Zigmond, Maurice L. 132 Zipf, George K. 303 Zubizarreta, Maria Luisa 90, 455–7, 459, 461 Zuckermann, Shalom 392 Zúñiga, Fernando 237 Zwicky, Arnold M. 13, 166, 171–3, 423

529

9781441124609_Aut_Index_Final_txt_print.indd 529

11/20/2012 9:26:08 AM

Subject Index

absolute transitive 267 accentuation 172, 173 accidental combination 296 accusative genitive 387 acquisition of syntax  pronominal reference  nonreflexive pronouns 399 principle B errors  399, 400, 401 wh-movement 391–8 inversion of the verb/auxiliary 393 preference for wh-in-situ 392 relative clauses 394–7 resumptive pronouns 395–6 word order 390–1 additional word order correlations 129–30 adjective phrases (AP) 328 adjunct clause (islands) 313 adjunction 85–6 containment and dominance 85, 86 preserves bar-level status 86 adjuncts 326–7 VP (verb phrase) 327 X-bar theory 327 adverbial clauses 341–2 as coordinate sentences 346 final 345, 346–8 initial 343–5, 346–8 preposed 345 adverbial subordination 341–53 clause order and pragmatic function 342–6 semantic link 349–53 causal clauses 350, 352 conceptual frame or mental model 350 conditional clauses 349, 350 purposive clauses (final/second) 349, 350–1 temporal clauses 350 syntactic structure 346–9 complex sentences 346–7, 349 initial/final adverbial clauses 346–8 “main clause phenomena” 348 adverbializer 42 adversative coordinators 304 affixes 170–4 agglutinating languages 283

Agree 318 agreement 16 alignment 232–8 common patterns 233 ergative-absolutive 233–4 inverse–direct and hierarchical 237 languages and constructions without GRs 238 neutral 233–4 nominative–accusative 233–4 object 236 split-S 235–6 tripartite 233–4 see also grammatical relations (GRs) alternative, coordination relation 295, 297 choice-aimed alternative 297 simple alternative 297 AN (adjective-noun), 124 anaphor 112 Ancient Greek clitics  demise of Wackernagel’s Law 187–8 P2 clitics outside P2 184–5 prosodic boundaries inside sentences 185–7 And-But-Or languages 309 anticausative 272–3 antipassives (ANTIP) 258–60 antisymmetry 139 antitopic construction 41–2 A-positions 106 and evidence for empty categories 106–9 argument alignment 201 articulatory–perceptual system 100 asymmetric V2 161 asyndetic coordination 299 atemporal combination 296 attitudinal features or semantic prosodies 60 attraction, strength of 60 “auto-benefactive” 270 bare phrase structure 81 adjunction 85–6 containment and dominance 85, 86 preserves bar-level status 86 functional determination of bar-levels 81–3

530

9781441124609_Sub_Index_Final_txt_print.indd 530

11/27/2012 11:06:49 AM

Subject Index lexical items 81–4 redundancy 81 specifiers and complements 83 terminal nodes 81, 82 vacuous projections 81–3 operation merge  creation of vacuous projections 84 specifiers, restricted to one 84 two syntactic objects 84 bar-levels 78, 81–3 basic word order 118–24, 137–8 definition 119–21 in intransitive sentences 123–4 not morpheme order 121–2 types of 122–3 because-clause 352 before-clause 345 behavioral phenomena 221 Berkeley Construction Grammar (BCG) 424 binary Agree 357 binding theory 199, 204, 399 bisyndetic (two coordinators) 299 bottom-up (inductive) approaches 26 bound pronominal 16 “bridge verbs” 156 building block model 6, 18–19, 406 cartography of syntactic structures 436– 50 see also left periphery, cartography Case Hierarchy (CH) theory 284–5 case marking  of arguments, changes in 371–2 with arguments in Insular Scandinavian 368–9 changes in subject 369–70 case theory 115, 213, 339 causal clauses 350, 352 causative constructions 283–7 areal typology 290–1 nonsyntactic causatives 291 purposive and sequential subtypes 290–1 causing/caused event 281–2 continuum and types 286–7 directive or indirect causation 287 lexical/morphological/syntactic form 286 manipulative causation 287 grammatical relation of Causee NP 283–5 Case Hierarchy (CH) theory 284–5 isolating/agglutinating/fusional languages 283 lexical causative type/morphological type 283

semantic typology 285–6 direct vs. indirect causation 285–6 manipulative vs. directive causation 285–6 semantics and case marking 287 grammatical relations hierarchy 287 causatives 281–91 constructions 283–7 nonsyntactic 291 syntactic 288–90 see also individual entries C-command 112, 116, 318, 322, 359 choice-aimed alternative 297 Chomskian generativism 220 “Chomskian” linguistics 22 Chomsky’s I-language 366 Chomsky’s theory 214 clausal asymmetric coordinations 301 clausal bipartite system 93 clause bound scrambling 147–51 clause coordination 300 different verb forms (non-parallel construction) 300 same verb form (parallel construction) 300 CLAWS (“Constituent Likelihood Automatic Word-tagging System”) 62 clitic and postpositives in P2 179–88 Ancient Greek clitics  demise of Wackernagel’s Law 187–8 P2 clitics outside P2 184–5 prosodic boundaries inside sentences 185–7 clitic chains in Hittite  slots 180–2 trends 182–3 clitic climbing 175–6 clitic doubling 174–5 Clitic Left Dislocation 437, 439–40, 450 clitics  and affixes: Romance clitic pronouns 174–9 clitic climbing 175–6 clitic doubling 174–5 from Latin Weak Pronouns to Romance Clitics 176–9 enclisis and proclisis 189–90 order of clitics in clusters 169–70 constituent-based and P2 clitics 170 pronominal clitics 169 possible hosts for P2 clitics  Wackernagel’s Law 168 and postpositives in P2 179–88 clitic chains in Hittite 

531

9781441124609_Sub_Index_Final_txt_print.indd 531

11/27/2012 11:06:50 AM

Subject Index simple vs. special clitics 166–7 two options for clitic placement  to specific constituents/position 167–8 versions of GB, Minimalism, optimality 165 words, affixes, simple clitics, special clitics 170–4 accentuation and deaccentuation 173, 174 prosody 173 coding of coordination relations 302–5 combination–alternative coding complexity implication 303–5 combination–alternative coding implication 302–3 combination–contrast coding complexity hierarchy 303 combination–contrast coding implication 302 coding phenomena 221 coherent infinitives 151 colligation 58 collocation 2, 74, 173, 191 combination, coordination relation 295–6 of entities  accidental 296 natural 296 of SoAs  atemporal 296 sequential 296 simultaneous 296 common noun phrase (CNP) 165, 195 Comparative Method 381 complement clauses 313 complementation 325–40 adjuncts 326–7 complements 325–6 complex complement structures 336–40 phrasal complements 328–30 sentential complements 330–6 complementizer phrases (CPs) 331 complements 325–6 direct object complement 325, 326 “ditransitive” of “bitransitive” verbs 326 indirect object complement 326 intransitive verb 326 phrasal see phrasal complements prepositional phrase (PP) 326 sentential see sentential complements complex complement structures 336–40 Cyclic Principle 337, 339 NP movement 336–7, 339 Principles of Case Theory 339–40 wh-movement 337–9 complex sentences 346–7, 349

cross-linguistic comparisons 46–7 functions of subordination 43–6 subordinate marking 44 subordinated, definition 42–3 Yup’ik subordinatives 45, 46 computing parsimony 15 conceptual frame or “mental model” 350 conceptual grouping 73 conditional clauses 349, 350 configurational languages 67 Configurationality 3, 146, 188 Conjunction Reduction 376 Conjunctive coordinators 304 constituency 37–9 and intonation  conversational discourse 68 and grammatical structure in conversational Mandarin 69 “major constituent boundaries” 67 nominal units 69 prosodic phrasing 67–8 units and grammatical structure, relationship 68 in service of conversational activities 70–2 expansions and completions 71–2 Finnish conversation 72 German data, studies with 71 Japanese data, studies with 70–1 recycling 70–1 relationship between syntax and repair practices 70 constituency-based paradigms 55 constituent  constituency and frequency of occurrence  “Linear Fusion Hypothesis” 74–5 constituency and intonation  conversational discourse 68 intonation and grammatical structure 68–9 major constituent boundaries 67 nominal intonation units 69 prosodic phrasing 67–8 constituency in conversational activities 70–2 expansions and completions 71–2 Finnish conversation 72 German data, studies with 71 Japanese data, studies with 70–1 recycling 70–1 syntax and repair practices, relationship 70 definition 65–6 simple clause, subject and predicate 65–6

532

9781441124609_Sub_Index_Final_txt_print.indd 532

11/27/2012 11:06:50 AM

Subject Index emergence of constituency 72–4 kinds of conceptual and phonological groupings 73 Langacker’s view 73 and phrase structure 66–7 bracketed structure of the clause 66 structural similarities between clauses 67 constituent order 39–40, 197–200 constituent structure 65 constituent-based and P2 clitics 170 constraint conflict 454, 459 see also optimality theory for minimalist program constraint ranking 8, 454, 458, 472 construct 428 state (or Status Constructus) construction 94, 95, 99, 100, 103, 112–13 see also construction grammar and syntax-semantics interface; genitive construction  combinatory 425, 426, 427 derivational 432, 434 inflectional 434 lexical-class 426–7 construction grammar and syntax-semantics interface  facts 429–35 composition requires constructions 429–30 license complements see constructions license complements formalism 424–9 “doctrine of syntactically transparent composition” 425 Head-Driven Phrase-Structure Grammar (HPSG) 424 locality 425–8 variable granularity 428–9 foundations  Government and Binding (GB) theory 423 topicalization and Filler-Gap construction 423–4 constructions license complements 430–5 valence variation 430–4 RHL model 431 weird sisterhood 434–5 nominal extraposition 434 containment 85 contrast, coordination relation 295, 296–7 corrective 296 oppositive 296 control infinitive 386

controller 107–8, 152 conversation 33 coordinate clauses 342 coordinating constructions  NP/VP or clause conjunction 298 relative and main clauses 298 semantic domain 300 types of coordination relations 298 coordination 292–310 across languages 301–9 And-But-Or languages 309 coding of coordination relations  302–5 Hierarchical Coding Map 306–7 implication of coordination parallelism 307–8 constructions, syntactic domain of 298 NP/VP or clause conjunction 298 relative and main clauses 298 types of coordination relations 298 definition 293 morphosyntactic properties  (A)symmetry of the construction 300–1 coordinating markers 299–300 explicit coding of relation 299–300 relation, notion of 292–5 as adjunct phrases 292 conceptual parallelism 293 as conjunction phrases 292 coordinate construction 293 coordinated entities/properties, or SoAs 294 level of pragmatics 294 semantic parallelism 293 relation, types of 295–301 alternative 295, 297 combination 295–6 contrast 295, 296–7 coordination parallelism 307–8 cross-linguistic tendency 308 syntactic (non-)parallelism 307–8 principle of iconicity of independence 307–8 principle of syntagmatic economy 307–8 coordination relation see coordination corpus annotation 52 corpus methods  annotation 52 attitudinal features or semantic prosodies 60 colligation 58 grammatical to lexical 60 linear unit grammar 60 MSUs—Sinclair’s last legacy 60

533

9781441124609_Sub_Index_Final_txt_print.indd 533

11/27/2012 11:06:50 AM

Subject Index parsed corpora and research in syntax 54–8 pattern grammar 60–1 random set of concordances for node block 59 syntactic annotation 52–4 corrective contrast 296 co-subordination 301 counterexpectative contrast 297 CP expletives see V2 Types creak 42 cross-clause phenomena 225 cross-linguistic comparisons 46–7 cross-linguistic extension of language-specific analyses 14 cross-linguistic strong form-function isomorphism 17–18 cross-linguistic variation 355 Cyclic Principle 337, 339 data 33 “Dative Sickness” 368, 369–70 deaccentuation 173 deagentivized (intransitivized) syntactic patterns 265–6 dedicated (monofunctional) coordinators 299 definite suffixes 103 degrammaticalization (of middle) 277 “deobjective” 267 dependency grammar 55 dependency tree 465, 466 dependent marking 175, 179 “deponent” 275 derivation, valency-decreasing 272–3 descriptive adequacy 27, 28 desemanticization 172 Diachronic Corpus of Present-Day Spoken English (DCPSE) 56 diachronic perspective 276–8 evolution of middle 276–8 origins 276 direct causation 286–7 direct object and passive construction 18 direct object complement 325, 326 direct object (DO) relation 284 direct-indirect object alignment 236 directionality parameter 139 directive causation 285–6 discontinuous constituent 57, 67, 136, 138–9, 142, 169 discourse-given constituents 456 distributional analysis 11–21 basis for forming linguistic hypotheses 12

mapping 13 roles of the constructions 12 set of constructions 12 example  analysis of certain argument phrases in English 11 English active voice construction 11 English passive voice construction 12 pattern of relations among formal structures 13 unwarranted assumptions  cross-linguistic extension of languagespecific analyses 14 cross-linguistic strong form-function isomorphism 17–18 form-function isomorphism 17 Free Ride Principle 13 global extension of construction 13–14 strong form-function isomorphism 17–18 weak form-function isomorphism 17 distributional method 12–13, 20 “ditransitive” of “bitransitive” verbs 326 documentation of highly endangered languages 34 dominance 168 “Do-support” 318 DP (determiner phrase) 196 elicitation, traditional linguistic 25 “emergent constituentization” 72 emphatic reflexive 279–80 empty category 107, 109–10, 115 empty pronominals, synopsis of  null subjects 111 obligatoriness of PRO 114–15 optional PRO 112–14 enclisis and proclisis, differences between 189–90 endocentricity 80 English active voice construction 11 English passive voice construction 12 ergative-absolutive alignment 233–4 Ergative–Absolutive alignment 4, 232, 234, 239 EVENT (component of causation) 289 Exceptional Case Marking Construction (ECM) 334 exclamative clauses 314 explanatory adequacy 27, 424 Extended Projection Principle (EPP) 462 see also optimality theory for minimalist program Extended Standard Theory (EST), 77

534

9781441124609_Sub_Index_Final_txt_print.indd 534

11/27/2012 11:06:50 AM

Subject Index “external argument” 228 “external thematic roles” (Williams’s theory) 215 field methods in syntactic research  complex sentences  cross-linguistic comparisons 46–7 functions of subordination 43–6 subordinate marking on independent sentences 44 subordinated, definition 42–3 Yup’ik subordinatives 45, 46 data 33 dynamicity of syntactic structure  Tuscarora adverbial clause 49 Tuscarora, an Iroquoian language 48– 9 methodology 33–4 conversation 33 documentation of highly endangered languages 34 speakers, variation in 34 Mohawk demonstrative construction 38 simple sentences  constituency 37–9 constituent order 39–40 lexical categories 35–7 Mohawk topicalization 40 nuclear clause boundaries 40–2 fieldwork-style interaction 25 finite complement 330–3 see also complementation Fin(iteness) 438, 439 focus-movement 316 form-function isomorphism 17–18, 303 FrameNet 57 free genitive 95 consistency 96 uniformity 96–7 Free Ride Principle 13 Free word order see scrambling functional genitive 95–6 form of 99–100 GenS and GenO 100 licensing 97–9 fusional languages 283 Fuzzy system 2, 51, 56, 170, 227, 342 general (multifunctional) coordinators 299 generality 16 generalized (“macro”-semantic) roles 228 generalized semantic roles 228 generative enterprise, goals of 76 generative grammar  positions and structures 205–13

grammatical case assignment 213 hierarchical structure 205–9 languages with canonical SOV or VSO order 209–12 sentence 197–204 argument alignment 201 constituent order 197–200 DP identification 196 Government-Binding theory 204 grammatical case 202–3 subject-verb agreement 203–4 thematic hierarchy 201 thematic roles 200–1 theme argument 200 subject, definition 194–5 subject position 214–17 syntactic constituents as subjects 195–6 generative or “Chomskian” linguistics 22 generative syntactic theory 4 genitive  free genitive 96–7 consistency 96 uniformity 96–7 functional genitive 97–100 form of 99–100 GenS and GenO 100 licensing 97–9 genitive 93 and definite suffixes 103 nominative and accusative 93 head marking 102–3 postpositional genitives 101–2 theory of genitive 100–1 types 93–6 genitive suffixes 103 genitive-noun (GN) 124–5 Germanic type 155 Germanic V2 155–8 Global Extension of Construction 13–14 GOAL (component of causation) 289 gold standard testing 30 Government-Binding (GB) theory 199, 204, 423 grammatical relations (GRs)  conceptual side: Signifié  “pragmatics” and cognition, 230–1 Proto-Agent and Proto- Patient, 228 salience, 231–2 semantic roles and related concepts, 228–30 syntactic functions, 230 valence, 227 conventionalization: l’ arbitraire du signe, 238–9

535

9781441124609_Sub_Index_Final_txt_print.indd 535

11/27/2012 11:06:50 AM

Subject Index formal side: signifiant, 221–3 “coding” phenomena, 221 “split-ergative” languages, 234 behavioral and pivot properties, 225–6 case, verb agreement/bound pronominals and others, 221–3 constituency and constituent order, 223–5 inter-clausal phenomena, 221 primitives, 219 terminology, 220 typology: alignment, 232–8 common patterns, 233 ergative-absolutive, 233–4 inverse–direct and hierarchical alignment, 237 languages and constructions without GRs, 238 neutral, 233–4 nominative–accusative, 233–4 object alignment, 236 split-S, 235–6 tripartite, 233–4 universality, 220 Greenbergian sense 28 Greenberg’s findings 124–5 Hamblin-style set of propositions 319 harmony 130 head marking 102–3 head-final 125, 127, 130, 209–11 head-initial 125, 127, 139, 209–11 “heavy reflexives” 279 Hierarchical Coding Map 306–7 Hittite, clitic chains in  slots in clitic chains 180–2 trends in clitic clustering 182–3 Holmberg’s generalization 147 host 180, 182–3 hypothesis testing in Minimalist Program 22–31 “bottom-up” (inductive) approaches 26 construction grammars 23 fieldwork-style interaction 24 generative linguists 24 generative or “Chomskian” linguistics 22 generative practice 26 gold standard testing 30 Greenbergian sense 28 implications 23–4 macro/micro-comparative approach 29 “parametric clusters” 28 psycholinguistic techniques 25 test of explanatory adequacy 27

test of “mere” descriptive adequacy 27, 28 theoretical hypothesis and observable data, relation 24 “top-down” (deductive) fashion 26 traditional linguistic elicitation 25 usage-based approaches 23 ICE-GB component of ICE project 56 “idiom principle” 58 increments 417 indirect causation 285–6 indirect object complement 326 indirect object (IO) relation 284 Inertia Hypothesis 378–80 infinitive complement 410 inflected infinitive 336 innateness 28, 111, 204, 210 Insular Scandinavian, case marking 368–9 interactional linguistics 406 interface phenomenon 378 International Corpus of English (ICE) 56 International Workshop on Treebanks and Linguistic Theories 54 interpretable feature [ineg] 359 interrogative clauses  matrix 312–13 adjunct clause (islands) 313 complement clauses 313 subject clause 312 subordinate 313–14 intonational units 68 “intransitives” 326, 328 islands (adjunct clause) 313 isolating languages 283 labelled bracketing (Treebank) 55 Lancaster Parsed Corpus 56 Langacker’s view 73 Last Resort condition 84 Latin weak pronouns to Romance clitics 176–9 left periphery, cartography  of clause 437–9 clitic left dislocation construction 437 interface with semantics/pragmatics, old/new information  contrastive focus 443 “understood partitive” context 442 “scope-discourse” semantics 436 topic and focus  compatibility with clitic 445 contrastive focus movement 439–40 differences between 444–5 scope-discourse properties 446 uniqueness of focus 447

536

9781441124609_Sub_Index_Final_txt_print.indd 536

11/27/2012 11:06:50 AM

Subject Index topic and mod 449–50 preposed adverbial 449–50 referential adverbials 449 lexeme 432–3 lexical categories 35–7 lexical causative 283, 287 Lexical Functional Grammar (LFG) 219 lexical (thematic and idiosyncratic) case 370–1 lexicalization 60–1, 231, 277 Linear Fusion Hypothesis 74–5 “linear restriction” 155 “linear unit grammar” 60 linguistic hypotheses  characteristics 19–21 fundamental properties 21 mapping 13 roles of the constructions 12 set of constructions 12 listeme see lexeme local particles, -kan 180 locality 425–8 long distance scrambling 151–2 Longman Grammar of Spoken and Written English 55 macro-comparative approach 29 “main clause phenomena” 345, 348, 352 “major constituent boundaries” 67 manipulative causation 286–7 meaning shift units (MSUs) 60 merge 453 methodological opportunism 18 cross-linguistic comparison 19 subjectivity 19 micro-comparative approach 29 middle and reflexive  cluster of deagentivized (intransitivized) patterns 265–6 diachronic perspective 276–8 evolution of middle 276–8 origins 276 middle types/patterns see middle types/ patterns middle voice in Indo-European 273–5 preliminaries 261–5 definitions and terminology 262–4 historical notes 261 syntactic pattern and voice 264–5 reflexive sensu stricto vs. “reflexive” as a cover term 278–80 “reflexive” as a synonym of “middle” 278 “reflexive” and “(emphatic) reflexive” 279–80

middle types/patterns 266–73 derived patterns/voices sensu stricto 266–8 object-demoting diatheses: antipassive and deobjective 267–8 subject-demoting patterns: canonical/ agentless 267 syntactic changes preserving semantic roles inventory 268–70 auto-benefactive 270–1 reciprocal 269–70 reflexive 268–9 valency-decreasing derivation: anticausative 272–3 middle voice in Indo-European 273–5 Minimalist program 357 Mohawk demonstrative construction 38 Mohawk topicalization 40 monosyndetic (single coordinator) 299 morpheme order 121–2 morphological causative 284, 288 morphosyntactic leveling 366, 368 morphosyntactic non-parallelism 301 Multiple Agree analysis 357, 359–64 application of 359–60 problems for 360–4 n-constituents 361–3 multiple wh-movement 322–3 Narrative Inversion 384 natural conjunction 296 n-constituents 361–3 simple 363 NEG feature 359 negation 354–64 and agree 357–8 binary Agree 357 Minimalist program 357 “Multiple Agree,” 357 is not multiple agree 358–64 interpretable/valued or uninterpretable/unvalued features 358 multiple agree analysis 359–64 [uNEG] feature/NEG feature 359 sentential, variation in expression of 354–7 cross-linguistic variation 355 “Negation Phrase” (NegP) 355 negative arguments 356 “negative concord” 356 negative expressions (“n-words,” “n-constituents”) 357 in West Flemish 358 “Negation Phrase” (NegP) 355

537

9781441124609_Sub_Index_Final_txt_print.indd 537

11/27/2012 11:06:50 AM

Subject Index “negative concord” 356 neg-preposing 316 neutral alignment 233–4 New Passive 373 emergence of 375–7 in Icelandic 367 vs. Canonical Passive 373–5 NG (noun-genitive) 124–5, 127, 128 No Tempering Condition (NTC) 453 node 81 nominative case 197–204 nominative genitive 93 Nominative Substitution 366, 369, 372 Nominative–Absolutive alignment 234 Nominative–Accusative alignment 233–4 non-configurationality 138–9 non-configurational languages 67 non-harmonic word order pattern 130–2 non-parallel clausal coordinations 300–1 non-parallel construction (different verb forms) 300 nonredundancy in representation 15 nonstructural arguments 431 non-subordinate clauses 342 nonsyntactic causatives 291 noun phrase coordination 301 “and-languages” 301 “with languages” 301 Noun Phrases (NP) 327–8 NP Movement 336–7, 339 NPs, properties of thematic structure  hierarchies of arguments 89–91 P(ossessor).…S(ubject)…O(bject) 89–91 UTAH 89 nominative/ergative 92–3 supposed “passivization” properties 91–2 NPs, structure of  poverty of stimulus 110–11 properties of thematic structure 89–91 synopsis of empty pronominals 111–15 syntactic realization of arguments 104–10 varieties of genitive 93–101 nuclear clause boundaries 40–2 null complementation 432 null subjects 111 “n-words” 357 object shift 147 “object suppressive” 267 oblique subject 371–2, 385–6 old Germanic languages 382 old Romance V2 160–2 omitting subject pronoun 160 Tobler-Mussafia law 160–1 oppositive contrast 296

optimality theory for minimalist program  OT as choice for pursuing MP goals 455–61 conflict-free analyses 459–61 conflicts with interface constraints 456–9 relation between the MP and OT  conflict-free on closer inspection 453 interface constraints 453 optimality theory 454–5 sensory-motor interface 453 Optimality Theory (OT) 7 OSV 119, 122–3 OV (object-verb) order 39, 123, 138, 209, 343, 379–80 OVS 122–3, 198 parallel construction (same verb form) 300 “parametric clusters” 28 paratactic sentences 342 parse tree 56 parsed corpora and research in syntax  54–8 analysis of tagged representative corpora 55 constituency-based paradigms 55 dependency grammar 55 FrameNet lexical database 57 ICE-GB component of ICE project 56 Lancaster Parsed Corpus 56 Longman Grammar of Spoken and Written English 55 Penn Treebank 55–6 Prague Dependency Treebank (PDT) 56–7 “replace” vs. “substitute” 57 “Treebank” (labelled bracketing) 55 parsing see parsed corpora and research in syntax partial wh-movement 323 heterogeneous phenomenon 323 matrix Spec,C, 323 participant roles 228 part-of-speech (POS) tagging 52 passives 240–57 agent phrases 251–2 with control verbs 254–5 impersonal 251–2 iterated 254 periphrastic 242–4 predicate-level treatment 241 of raising to object 255–7 strict morphological 244–7 asymmetric 244–5 symmetric 245–7

538

9781441124609_Sub_Index_Final_txt_print.indd 538

11/27/2012 11:06:51 AM

Subject Index of three argument verbs 248–51 applicatives 248–51 causatives 248–51 ditransitives 248–51 two passives from the same source 253 “passivization” of NPs 91 pattern grammar 60–1 Pear Stories 68 Penn Treebank 55–6 “personalization,” 366, 368 phonological bootstrapping hypothesis 391 phonological grouping 73 phrasal complements 328–30 with APs 328 complements and adjuncts 329–30 with NPs 329 subcategorization 329 transitive and intransitive adjectives 328 with VPs 328 “pivot” 228 PN (preposition-noun) 124–5 poliysynthetic languages or headmarking languages 144 possessive-reflexive 271 possessives 103 possessor 89 postpositional genitives 101–2 poverty of stimulus 110–11 Prague Dependency Treebank (PDT) 56–7 Pred 19 prepositional phrase (PP) 326 primary linguistic data (PLD) 378 primary-secondary object alignment 236 principle-C effect 149 Principles of Case Theory 339–40 PRO 213 obligatoriness of 114–15 optional 112–14 Probe 318 proclitic 169, 174, 176, 178 pronominal clitics 169 prosodic phrasing 67–8 prosody 173 pseudocleft or wh-cleft 415 psycholinguistic techniques 25 purposive causative 290 purposive clauses (final/second) 349, 350–1 radical reconstruction effects 151 raising  to n 109–10 subject-to-object 384–5 subject-to-subject 385–6 reanalysis 366–7 reciprocal 269–70

Reconstructing Case Frames 382–3 recursion 337 reflexive sensu stricto vs. “reflexive” as a cover term 278–80 “reflexive” and “(emphatic) reflexive” 279–80 “reflexive” as a synonym of “middle” 278 “reflexive voice” 268 rejection 297 relative clauses 314 “replace” vs. “substitute” 57 restricted Romance construct state construction 112 RESULT (component of causation) 289 right dislocated constituents 162 Romance clitic pronouns 174–9 clitic climbing 175–6 clitic doubling 174–5 from Latin Weak Pronouns to Romance Clitics 176–9 Santa Barbara Corpus of Spoken American English 407 scope-discourse semantics 436 scrambling 132–3 base generation approaches  configurationality parameter 144 major problems 145 mixed position 146 theta roles 145 definition 143–4 movement approaches  clause bound scrambling 147–51 long distance scrambling 151–2 object shift 147 second-position clitic 134 second-position phenomena 133–4 “self-beneficent” 270 “semantic negation” 359 semantic roles (SR) 228 sentence-initial position 197–204 argument alignment 201 constituent order 197–200 DP identification 196 Government-Binding theory 204 grammatical case 202–3 subject-verb agreement 203–4 thematic hierarchy 201 thematic roles 200–1 theme argument 200 sentential complements 330–6 finite 330–3 with APs 331 complementizer phrases (CPs) 331

539

9781441124609_Sub_Index_Final_txt_print.indd 539

11/27/2012 11:06:51 AM

Subject Index N-bar (N’) node. 333 with NPs 331–2 relative clauses 332 with VPs 330–1 infinitival 333–6 abstract element PRO 333–5 with AP 334–5 within CPs 334 ECM construction 334 with NP 335–6 within VP 333 sequencing 407 sequential causative see syntactic causatives sequential conjunction 290 sequentiality 75 shallow parsing 54 Sign-Based Construction Grammar (SBCG) 421 simple alternative 297 simple clitics 170–4 simple disjunction 295 simple sentences  constituency 37–9 constituent order 39–40 lexical categories 35–7 Mohawk topicalization 40 nuclear clause boundaries 40–2 simple vs. special clitics 166–7 simultaneous combination 296 small clause constructions 215 sound laws 381 SOV 209–12 speakers, variation in 34 special clitics 4, 170–4 specifiers 78, 83 SpecNegP 359 split system 233 split-ergative languages 234 split-S system 233, 235, 236, 238 storage parsimony 15 strong form-function isomorphism 17–18 subcategorization 329 “subject” 225, 231 Subject Auxiliary Inversion (SAI) 393 subject case marking 369–70 “Dative Sickness” 369–70 Nominative Substitution (personalization) 369 subject clause 312 subject clitic inversion 162 subject, definition 194–5 subject inversion in Cimbrian (German dialect) 162–3 subjective version 270

subject-to-object raising 384–5 subject-to-subject raising 385–6 subject-verb agreement 203–4 subordinate clause 341–2 subordination in usage-based syntax 410–20 biclausal, constructions 413–16 complementation 410–11 incrementation 416–18 projection 418–20 relativization 411–13 superiority 322–3 superordinate clause 341–2 supposed “passivization” properties 91–2 SV 39, 73, 128, 131 SVO  English: another mixed 127–8 Italian and Norwegian 126–7 symbolization 73 symmetric V2 160 symmetrical object systems 236 symmetry 14 syndetic coordination 299 single coordinator (monosyndetic) 299 two coordinators (bisyndetic) 299 syntactic annotation 52–4 ambiguity tags 53 annotated corpora 53 annotation errors 53 “false positives” 53 “flat vs. full” annotation 54 lexical matching on a stochastic basis 52 parsing 52 part-of-speech (POS) tagging 52 syntactic frames 53 syntactic parsing 52 syntactic argumentation 315 syntactic causatives 288–90 sequential and purposive subtypes 288–9 theoretical implications 289–90 causative affixes 290 EVENT/RESULT/GOAL component 289 syntactic change 365–88 directionality of morphosyntactic change  arguments in Insular Scandinavian 368–9 changes in case marking of arguments 371–2 changes in subject case marking  369–70 structural case and (thematic and idiosyncratic) lexical case 370–1 Inertia Hypothesis 378–80 loss of OV Order in the VP in Icelandic 379–80

540

9781441124609_Sub_Index_Final_txt_print.indd 540

11/27/2012 11:06:51 AM

Subject Index reanalysis of ambiguous structures 373–8 emergence of New Passive 375–7 nature of change 377–8 New Passive vs. Canonical Passive 373–5 syntactic reconstruction 380–8 predictive power 387 pros and cons 381 reconstructing case frames 382–3 reconstructing grammatical relations 383–6 syntactic constituents as subjects 195–6 syntactic domain of coordinating constructions 298 NP/VP or clause conjunction 298 relative and main clauses 298 types of coordination relations 298 syntactic gradience 171 syntactic negation 359 syntactic pattern and voice 264–5 syntactic positions and structures 205–13 generative theories of grammatical case assignment 213 hierarchical structure 205–9 languages with canonical SOV or VSO order 209–12 syntactic realization of arguments  A-positions and evidence for empty categories 106–9 order and number of arguments 104–6 raising to n 109–10 syntactic reconstruction 380–8 predictive power of comparative method 387 pros and cons 381 reconstructing case frames 382–3 reconstructing grammatical relations 383–6 syntactic structure, dynamicity of  Tuscarora adverbial clause 49 Tuscarora, an Iroquoian language 48–9 Syntactic Tree 55 syntagmatic nonredundancy 15–16 taxonomic epiphenomena 423 temporal clauses 349, 350 ternary to binary branching 79 test of explanatory adequacy 27 test or criterion 11–12 thematic grid 212 thematic hierarchy 201 thematic roles 200–1 theme 197, 231, 248, 249, 258, 371, 415 argument 200–1, 427 role 201

theory of external arguments (Williams) 213, 215 theory of genitive 100–1 theory of Mohawk syntax 28 theory of phraseology 61 theory of predication (Williams) 214–15 theory of Relational Grammar 204 Tlacolula Valley Zapotec 125 Tobler Mussafia’s Law 178 “top-down” (deductive) fashion 26 topic see left periphery, cartography trace 311 traditional linguistic elicitation 25 transitive adjectives 328 “Treebank” (labelled bracketing) 55 tree-diagram 55, 66 tripartite alignment 233–4 T-to-C movement 317–18 turn constructional unit (TCU) 406 Tuscarora adverbial clause 49 Tuscarora, an Iroquoian language 48 type hierarchy 433 see also construction grammar and syntax-semantics interface UAH (Universal Alignment Hypothesis) 226, 239 underlying vs. surface word order 138 uNEG feature (uninterpretable feature) 359 Universal Theta Assignment Hypothesis 226 Universality of syntactic strutures 220 universals 129 usage and syntax  example of TCUs 408–9 on-line grammar 407–8 syntax and speaking practices 409–20 “subordination” in usage-based syntax 410–20 TCUs and clauses 409–10 usage and the study of syntax 405–7 usage-based approaches 23 UTAH (Uniformity of Theta Assignment Hypothesis) 89 V2 types  Old Romance V2 160–2 possibility of omitting the subject pronoun 160 Tobler-Mussafia law 160–1 standard view on Germanic V2 155–8 V2 correlates 155 subject inversion in Cimbrian (German dialect) 162–3

541

9781441124609_Sub_Index_Final_txt_print.indd 541

11/27/2012 11:06:51 AM

Subject Index V2 in Scandinavian languages 158–60 V2 phenomenon 154 vacuous projection 84 valence 227 verb first 156, 207, 210–11, 413 verb form 300 verb phrases (VP) 139, 328 verb second 134, 210–11 verb-final 123, 136 verb-initial 123, 222, 237 verb-medial 123, 130 verb-second constituent order 210 voice 264–5 VO (verb-object) order 379 VOS 123–4, 133, 198 VP-internal subject hypothesis 195 VS (verb subject) 123–4, 128 VSO language, prototypical 125 Wackernagel’s law 168, 179, 365 Wackernagel’s position 134 Weak Crossover (WCO) 147 weak form-function isomorphism 17 wh-clause 415 wh-clefts 413 wh-in-situ 319–22 Attract theory 322 Chinese-style 320 clauses with multiple wh-phrases 320, 321–2 echo question 320 superiority effect 322 wh-movement 311–24 constructions 312–16 exclamative clauses 314 matrix interrogative clauses 312–13 relative clauses 314 subordinate interrogative clauses 313–14 multiple 322–3 partial 323 wh-in-situ 319–22 wh-phrase and wh-clause 316–19 as determiner (Abney’s hypothesis) 317 “Do-support” 318 semantics of the complementizer 319 T-to-C movement 317–18 word order  basic word order 118–24 definition 119–21 in intransitive sentences 123–4

is not morpheme order 121–2 types of basic 122–3 change 140–1 typology see word order typology variation 132–7 order of constituent elements 136–7 scrambling 132–3 second-position phenomena 133–4 verb second 134–6 word order and case marking 137 word order typology 124–32 additional word order correlations  129–30 English: another mixed SVO pattern 127–8 Greenberg’s findings 124–5 Japanese: prototypical SOV language 126 non-harmonic word order pattern 130–2 SVO order: Italian and Norwegian 126–7 and theoretical syntactic analysis  137–40 “basic word order” 137–8 non-configurationality 138–9 underlying vs. surface word order 138 Tlacolula Valley Zapotec: prototypical VSO language 125 word order and the category “adjective” 128–9 World Atlas of Language Structures 220, 288 X-bar theory 139, 207–9, 211, 213, 214 XP (X-phrase) 78, 139 X’-structure and Minimalism  bare phrase structure  adjunction 85–6 functional determination of bar-levels 81–3 operation merge 83–5 generative enterprise, goals of 76 properties  bar-levels 78 binary branching 79 endocentricity 80 uniqueness of mother nodes 79–80 proposal reliance on hypothesis 76–8 remarks and current issues 86–7 structurally comparable structures 77 X’-theory 77, 81, 156 Yup’ik subordinatives 45

542

9781441124609_Sub_Index_Final_txt_print.indd 542

11/27/2012 11:06:51 AM

Language Index

Abkhaz 283 Afro-Asiatic 297 Algonquian 237 Amharic 267 Arabic 94, 98, 103, 211, 244, 463, 467–8 Armenian 465 Bafut 119 Balinese 224 Bantu 232, 236, 244, 249–51, 258 Basque 101–2, 233, 243, 289, 316 Bulgarian 93, 103, 169–70, 322–3, 463 Old 169 Cariban 224, 234, 239 Catalan 97, 309, 463 Celtic 93–6, 156 Chichewa 19 Chickasaw 121–3, 133, 137–8, 140–1, 221–2, 225, 229, 233, 235–6 Chinese, 39, 52, 199, 203, 238, 243, 319–21, 463, 467 Choctaw 233, 235 Chukchee 258, 267 Cimbrian 155, 162–3 Croatian 463 Czech 56, 169, 309, 323, 359, 463–4 Danish 147, 158, 309, 380, 463 Diegueño 131–2, 138 Dutch 147, 151, 154–5, 158, 160, 163, 251, 265, 309, 390–1, 393, 399–400, 463 Dyirbal 234, 259 Edo 19 Egyptian, Middle 120, 122–3, 140 English passim  American 53–4, 407 British 53–4, 56 Early Middle 386 Early Modern 379, 464 Middle 372, 379, 380, 464 Modern 382, 383 Modern British 464 Old 95, 242, 372, 382–3, 387, 464 Eskimo 232, 259 Estonian 463

Faroese 159–60, 366, 368–72, 382 Finnish 2, 67–8, 70, 72, 463, 467 French 14, 17, 90, 92–3, 136, 142, 164, 179, 190–2, 212, 236, 249, 264, 267, 269, 278–9, 295, 299–300, 302–3, 309, 354–6, 358, 391–3, 395–6, 399, 401, 463–4 Old 95, 100, 160, 369, 372 Garifuna 123, 130, 137, 140–1 Georgian 235, 270, 283 German 70, 71, 89, 92, 94–6, 100, 104–6, 109, 120, 122, 134–8, 141, 143, 147–51, 153–8, 160–3, 202, 210, 225, 227, 242, 252, 269, 279, 283, 295, 297, 299–300, 304, 309, 323, 348–9, 365, 368, 372, 382, 383, 390–1, 393, 413–4, 447, 463, 467 Old Germanic 160–2, 382–4, 386–7 Old High 164, 382 Germanic 3, 29, 93–5, 99, 135, 137, 155, 158, 160–3, 179, 264, 276–7, 309, 367–8, 372–3, 382–7, 448 Proto-Germanic 382–3, 385–7 East-Germanic 385 West-Germanic 385–7 North-Germanic 386–7 Gothic 382, 384–7, 465 Greek 95–6, 139, 184–5, 188, 193, 240, 261, 264, 267, 273–4, 277, 280, 309, 384–5, 401, 465 Ancient 99, 139, 165–7, 182, 184–5, 187, 188, 261–3, 271, 274, 463–4 Classical 94, 97, 166, 169, 174, 184–5, 188, 191 Homeric 179 Modern 139, 166, 170, 187–8, 192, 289, 395, 463 Mycenaean 189 Grisons Swiss German 137 Gungbe 446, 448 Hausa 304 Hebrew 94, 99–100, 113, 241, 244, 299, 399, 463 Hindi 98, 101–2, 150, 244, 252, 463 Hittite 165–6, 168–70, 179–80, 182–3, 185, 188–90, 192–3, 273–5 Old 180, 182–3 Hungarian 94, 96, 98, 103, 116, 143, 146, 297, 316, 322, 448, 463

543

9781441124609_Lang_Index_Final_txt_print.indd 543

11/20/2012 5:56:37 PM

Language Index Icelandic 6, 95, 117, 137, 147, 159–60, 163–4, 202, 309, 365–80, 388, 399, 400, 463 Modern 382–3 Old 379, 382–3, 385–7 Ineseño Chumash 122 Iranian 234, 236 Irish 211, 309 Modern 211 Italian 7–8, 91–2, 104, 111–12, 126–8, 160–2, 164, 169, 174–6, 179, 182, 189, 191–2, 196, 199, 264, 268, 299–300, 302, 304, 309, 353–7, 391–7, 399, 401, 437–39, 441–44, 447–48, 450, 452, 456–8, 460, 462–3 Old 161, 164, 179, 189, 192 Sianese 440 Jacaltec 259 Japanese 2, 69–71, 125–6, 145, 150, 202–3, 209–11, 248, 283, 319, 343, 439, 463, 467 Old 130 Kashmiri 163 Kawaiisu 132, 137–8, 141 Kunuz 131 Kutenai 251 Latin 94, 96, 99, 117, 136–9, 141–3, 167, 175–9, 181, 187–9, 192, 199, 240, 243–5, 251–3, 264, 267, 273, 275, 277–8, 297, 463–5, 466 Late 192–3 Classical 199 Latvian 251, 265, 267 Lepcha 283 Luxembourgish 309 Maa 222–3, 229, 233–4 Northern Samburu 223 Southern Kenyan 223 Malagasy 169, 224–6, 232, 234, 241, 246, 248–9, 251, 254–5, 257 Mangarayi 304 Mohawk 2, 28, 35–42, 139, 142 Muna 124 Muskogean 232, 235 Nakh-Dagestanian 235 Nakh-Daghestan 267 Navajo 42–4, 46–7 Ngiyambaa 169 Norwegian 103, 126–9, 135–6, 141, 147, 159, 309, 463

Okinawan 210 Old Church Slavonic 273, 465 Norse, Old 139, 386 Olutec 232 Persian 342, 463 Pilagá 235 Polish 253, 265, 304, 323, 463 Portuguese 86, 309, 335–6, 463–4 Quechua 243, 467 Bolivian 287 Huallaga 243–4, 248 Imbabura 130 Romance 3–4, 29, 89–90, 93–6, 99–100, 104–7, 109–15, 117, 136, 142, 154, 161–4, 165–7, 171, 173–8, 182, 188–9, 191–2, 264, 276–7, 309, 355, 367, 439, 441, 447–8, 455 Modern 160, 175, 264, 278 Old 155, 160, 161–3, 178, 192 Rosani 234 Rumanian 103, 309 Russian 143, 202, 245, 268–70, 272–3, 279, 304, 357, 463 North 253 Salishan 19 Samoan 17–18 Sanskrit 170, 172, 182, 184, 192, 199, 262–4, 267, 280 Vedic 179, 271, 273–6, 279 Sardinian 309 Sasak 68–9 Scandinavian 6, 94, 101, 109, 142, 147, 154–5, 157–9, 164, 372, 379–80 Insular 366, 368–9, 372 Mainland 366, 379–80 Semitic 93–5, 113–14, 244 Serbo-Croatian 168–70, 192, 395 Siouan 232 Slavic 94–6, 98, 117, 142, 268, 273, 276–8, 322 Slovenian 169, 192 Somali 97–8, 297, 300 Spanish 121–2, 128, 148, 150, 153, 160, 175, 192, 196, 199, 222, 236, 282, 309, 332, 335, 340, 395, 399, 401, 438, 463, 467 Old 190, 192 Swedish 147, 275, 309, 391–3, 463, 467 Swiss German 137 Tagalog 246 Thai 234, 243, 463

544

9781441124609_Lang_Index_Final_txt_print.indd 544

11/20/2012 5:56:37 PM

Language Index Tibetan 4, 236, 238–9 Toba Batak 224 Tohono O’odham 138–9 Tolkapaya Yavapai 122, 138 Turkic 267 Turkish 150, 210, 240, 244, 251, 254–5, 284–5, 308, 390–1, 463 Tuscarora 48–9 Tuvaluan 304 Tzeltal 243 Uralic 98 Urdu 463 Uto-Aztecan 132–3, 138–9

Vietnamese 243, 463 Walpiri 144 Warlpiri 139, 169, 199, 365 Welsh 243, 252, 438 West Flemish 5, 358–64 Yiddish 159–60 Yup’ik 44–7 Zapotec Tlacolula Valley 125, 134, 137 Isthumus 344

545

9781441124609_Lang_Index_Final_txt_print.indd 545

11/20/2012 5:56:38 PM