Click here to load reader

Perspectives on a Key Notion in Linguistics - CSLI LinGO lingo. · PDF filePerspectives on a Key Notion in Linguistics ... This work was supported in part by grant BCS ... structure

  • View
    213

  • Download
    0

Embed Size (px)

Text of Perspectives on a Key Notion in Linguistics - CSLI LinGO lingo. · PDF filePerspectives on a...

  • i

    i

    features 2009/3/18 17:19 page iii #1i

    i

    i

    i

    i

    i

    Features:

    Perspectives on a Key Notion in Linguistics

    Anna Kibort and Greville G. Corbett

    CLARENDON PRESS . OXFORD

    2009

  • i

    i

    features 2009/3/18 17:19 page iv #2i

    i

    i

    i

    i

    i

  • i

    i

    features 2009/3/18 17:19 page v #3i

    i

    i

    i

    i

    i

    Contents

    List of figures viList of tables vii

    1 Feature Geometry and Predictions of Locality 1

    Ivan A. Sag

    1.1 Introduction 11.2 Background 21.3 Locality of Construction 151.4 Sign-Based Construction Grammar 171.5 Predictions of Locality 221.6 Some Analytic Challenges 271.7 Conclusion 38

  • i

    i

    features 2009/3/18 17:19 page vi #4i

    i

    i

    i

    i

    i

    List of figures

    1.1 An Illicit Selection of an Embedded NP 41.2 Kajitas (1968) Clause Structure 61.3 Subcategorization by Subjunctive Clause (Kajita) 71.4 head Features in GPSG 81.5 Lexical ID Rules in GPSG 91.6 Head Feature Identity within a Headed Complement

    (GPSG) 101.7 Head and Foot Feature Identity within a VP Complement

    (GPSG) 111.8 Feature Geometry of Pollard and Sag 1994 161.9 A Tag-Question 171.10 The Sign in SBCG 181.11 A Clausal Construct 191.12 Type Declarations 191.13 A SBCG Type Hierarchy 201.14 Interaction of the Subcategorization and Head Feature

    Principles 241.15 A For-To Clause 281.16 A Filler-Gap Dependency in SBCG 301.17 Case Government in Martuthunira 321.18 xarg Analysis of Genitive-Embedding NP 351.19 xarg Analysis of your fancy 351.20 Literal and Idiomatic Strings Relations 38

  • i

    i

    features 2009/3/18 17:19 page vii #5i

    i

    i

    i

    i

    i

    List of tables

  • i

    i

    features 2009/3/18 17:19 page 1 #6i

    i

    i

    i

    i

    i

    1

    Feature Geometry and

    Predictions of LocalityIVAN A. SAG

    1.1 Introduction

    This chapter deals with a number of issues having to do with locality innatural language.1 Locality of selection is the problem of delimitingwhat syntactic and semantic information lexical items select. Relatedissues include the proper analysis of idiomatic expressions, control ofovert pronominals, and cross-linguistic variation in lexical sensitivity tofiller-gap dependencies. Closely related to selectional locality is the issueof locality of construction the problem of delimiting the syntacticand semantic information accessible to grammar rules. These issues haveconsiderable history in the field, though matters of locality are sometimesleft implicit in theoretical discussions.After providing some necessary background, I will propose a version ofgrammatical theory that embodies a particular hypothesis about locality.In the general theory I outline, the feature geometry serves to delimit thegrammatical information accessible for lexical selection or constructionalconstraints.

    1 For valuable discussions about locality, I would like to thank Emily Bender,Ann Copestake, Grev Corbett, Bill Croft, Bruno Estigarribia, Charles Fillmore,Dan Flickinger, Adele Goldberg, Andreas Kathol, Paul Kay, Bob Levine, DetmarMeurers, Laura Michaelis, Carl Pollard, Jan Strunk, and Tom Wasow. I amparticularly grateful to Detmar Meurers and Stefan Muller for detailed commentson an earlier draft of this chapter and to Grev Corbett and Anna Kibort forexcellent editorial advice. This work was supported in part by grant BCS-0094638from the National Science Foundation to Stanford University and in part bythe Research Collaboration between NTT Communication Science Laboratories,Nippon Telegraph and Telephone Corporation and CSLI, Stanford University.

  • i

    i

    features 2009/3/18 17:19 page 2 #7i

    i

    i

    i

    i

    i

    2 Ivan A. Sag

    1.2 Background

    The locality of selection is one of the theoretical issues that were hotlydebated during the 1960s. For example, Chomsky (1965, Ch. 2) proposedthat the lexical entries of verbs and other lexical formatives includestrict subcategorization restrictions like those shown in (1):2

    (1) a. prove, V, [+ NP]

    b. run, V, [+ dir]

    Context-sensitive lexical insertion transformations (which involved thesubstitution of a lexical formative for a dummy symbol ) were subjectto a matching condition that required the subcategorization restrictionsto match the local context in the deep structure phrase marker. Chomskyproposed that the matching condition obeyed a principle of strictlocality, which stipulated that strict subcategorization restrictions likethose illustrated in (1) could only make reference to (could only bematched against) elements that are dominated by the VP directlydominating the V in deep structure subtrees like (2a,b):

    (2) a. VP

    V

    NP

    ...

    b. VP

    V

    PP[dir]

    ...

    Strict locality imposed an upper bound on the domain ofsubcategorization, but not a lower bound. That is, an element referred toby a subcategorization restriction did not have to be a sister of the V; itcould be an element embedded within a sister of the V, as in (3):

    (3) a. believe, V, [+ that S]

    2 The field has fallen into an oddly mutated use of the verb subcategorize. Onefrequently finds in the literature expressions like This verb subcategorizes forX or This verb is subcategorized for X. Here and throughout, I will maintainwhat I believe is the original way of describing the dependencies in question, e.g.This verb is subcategorized by X. That is, the particular syntactic environmentX is used to classify the verb in question into the given subcategory.

  • i

    i

    features 2009/3/18 17:19 page 3 #8i

    i

    i

    i

    i

    i

    1.2 Background 3

    b. VP

    V

    S

    COMP

    that

    S

    ...

    But strict locality sharply distinguished subcategorization restrictionsfrom selectional restrictions, the similar device introduced by Chomsky toanalyze semantic cooccurrence restrictions. The selectional restrictions ofa verb, for example, were permitted to access properties of the subjectNP, but the strict subcategorization restrictions were not.This matter was taken up anew by Kajita (1968), who argued thatChomskys notion of strict locality was both too strong and too weak. Inparticular, Kajita (p. 96) argued, on the basis of contrasts like (4a,b),that subcategorizational domains should be extended to include a verbssubject:3

    (4) a. That Kim was right bothered me.

    b.*That Kim was right loved me.

    Although contrasts like this might be explained away as semantic(selectional) in nature, there are other minimal pairs that perhaps makeKajitas point more convincingly:

    (5) a. The question of whether Kim was right perplexed me.

    b. *?Whether Kim was right perplexed me.

    In any case, it is now well established that many languages have verbsthat select a subject with idiosyncratic case properties (e.g. Icelandicverbs requiring a quirky dative, accusative or genitive subject; seeThrainsson 1979). Hence, given that case information is (at least partly)

    3 This argument, of course, turns on the assumption that English has hier-archical clause structure, and not the flat structure assumed, for example, ina number of proposals for German, Japanese and other language with con-siderable word order freedom. Assuming the flat structure for clauses, thesubject is accessible to a verb without modifying Chomskys theory of strictsubcategorization.

  • i

    i

    features 2009/3/18 17:19 page 4 #9i

    i

    i

    i

    i

    i

    4 Ivan A. Sag

    VP

    V

    S

    NP VP

    V NP

    Fig. 1.1 An Illicit Selection of an Embedded NP

    syntactic in nature, permitting the syntactic selection of subjects, asKajita suggested, provides the most straightforward account of quirkysubject case and related phenomena.Chomskys strict locality proposal was too weak, Kajita argued, becauseit allowed subcategorization restrictions to access elements deeplyembedded within a verbs complement. For example, under Chomskysdefinition of strict locality, an object within a clause would be locallyaccessible to a verb that selected that clause as a complement, as inFigure 1.1. The objection runs as follows: although we commonly findverbs like prove, which require a direct object NP (and disallow a PPcomplement), there are no languages (as far as we know) where we find averb like prove that imposed the same requirement on thecomplementation pattern realized within its sentential complement. Thatis we would not expect to find a verb evorp whose selectional propertiesproduced contrasts like the following:

    (6) a. Lee evorped that someone bought the car .

    b.*Lee evorped that someone died .

    c.*Lee evorped that someone ran into the room .

    Kajita is to my knowledge the first to point out the theoreticalimportance of characterizing the lower bound on subcategorizationrestrictions.Kajita also considered examples like the following, arguing that the verbserve requires an infinitival VP complement (an S, in his system) thatcontains a direct object NP:

  • i

    i

    features 2009/3/18 17:19 page 5 #10i

    i

    i

    i

    i

    i

    1.2 Background 5

    (7) a. The ice served to chill the beer.

    a.*The ice served to melt

    To accommodate this contrast and other data he considers, Kajita (1968:105) proposed that the upper bound of a verbs subcategorization domainbe the minimal S node that dominates it and that the lower bound bedetermined by a constraint requiring that the path from the upper boundto the selected constituent contain at most one S node.4 Kajitas theorymust be underst

Search related