Chapter : Phrase Structure Grammar*

James P. BlevinsUniversity of Cambridge

Ivan A. SagStanford University

To appear in M. den Dikken (ed.)The Cambridge Handbook of Generative Syntax

. Origins of phrase structure analysisTo understand the properties of modern phrase structure grammars, it is useful toplace their development in a wider formal and historical context. Phrase structuregrammars and associated notions of phrase structure analysis have their proximateorigins in models of Immediate Constituent (IC) analysis. Although inspired by theprogrammatic syntactic remarks in Bloomfield (), these models were princi-pally developed by Bloomfield’s successors, most actively in the decade between thepublication of Wells () and the advent of transformational analyses in Harris() and Chomsky (). The central intuition underlying models of IC analysiswas that the structure of an expression could be exhibited by dividing the expressioninto parts (its immediate constituents), further subdividing these parts, and contin-uing until syntactically indivisible units were obtained. This style of analysis wasmotivated in part by a belief in the of syntactic relations, in particular theview that the most important relations held between immediate constituents.

The process of analyzing syntax is largely one of finding successive lay-ers of ICs and of immediate constructions, the description of relation-ships which exist between ICs and the description of those relationshipswhich are not eciently described in terms of ICs. The last is generallyof subsidiary importance; most of the relationships of any great signif-icance are between ICs. (Gleason : )

Within the Bloomfieldian tradition, there was a fair degree of consensus regard-ing the application of syntacticmethods aswell as about the analyses associatedwithdierent classes of constructions. Some of the general features of IC analyses find anobvious reflex in subsequent models of analysis. Foremost among these is the ideathat structure involves a part-whole relation between elements and a larger superor-dinate unit, rather than an asymmetrical dependency relation between elements at

*We thank Stefan Müller and an anonymous reviewer for comments on an earlier version of thischapter.

the same level. The Bloomfieldians’ preference for binary branching analyses like-wise reemerges in later models of phrase structure, and their practice of extendingsyntactic analysis below the word level, to include stems and inflectional forma-tives, survives largely intact in the transformational tradition. Some other featuresof IC analyses are less faithfully preserved. These include general properties suchas the recognition of discontinuous and overlapping constituents or the representa-tion of intonation. More specific proposals, such as the classification of elements(notably coordinating conjunctions) as ‘markers’ (Hockett : ) were not re-habilitated until nearly years later (Gazdar et al. ,Sag et al. ,Pollard andSag : chapter ). The encoding of dependency relations within a part-wholeanalysis (Nida ) was also suppressed until the development of feature-basedmodels such as LFG (Kaplan and Bresnan and chapter of this volume) andHPSG (Pollard and Sag and section .. below) that could explicitly expressvalence dependencies within syntactic representations.

.. Procedures of IC analysisThe development of constituent structure analysis within the Bloomfieldian tradi-tion was held back by, among other things, the lack of a perspicuous format forrepresenting syntactic analyses. The formats explored by the Bloomfieldians werecumbersome, ranging from annotated circuit diagrams in Nida () through thechart representation in Table . or the ‘Chinese box’ arrangements in Table ..

John is here

Table .: Chart-based IC Analysis (Hockett : chapter )

John P can → go

Table .: Chinese box-based IC Analysis (Gleason : )

The shortcomings of these representational formats were particularly evident inthe analysis of the discontinuous and overlapping constituents recognized by the

A partial list of constructions that were analyzed as discontinuous by the Bloomfieldians wouldinclude parentheticals (Bloomfield : , Nida : ), coordination (Bloch : ), ditransi-tives (Pike : ), complex predicates (Nida : ), phrasal verbs (Wells : ), polar ques-tions (Pike : , Hockett : , Gleason : , Nida : ), non-subject relatives (Nida: ), non-subject constituent questions (Nida : , Gleason : ).

The analysis in Table . also represents functional and even dependency information, as it illustratesthe convention that “the arrow points towards the head” in a modifier-head construction and that “the Palways faces the predicate” in a subject-predicate construction (Gleason : ).

Bloomfieldians. While generally preferring continuous (and binary) analyses, theyalso admitted a range of constructions that violated these preferences.

Most linguists operate on the principle that cuts will be made binarywhenever possible, but that cuts giving three or more ICs will not beexcluded a priori. In the same way, they will make cuts giving contin-uous ICs wherever possible, but discontinuous ICs are not excluded onprinciple. (Gleason : )

The descriptive challenges that arose in extending these formats to the descrip-tion of discontinuous dependencies are illustrated by the representation of phrasalverb constructions, whichwere taken to be discontinuous from at leastWells ().

Verb phrases of the type verb+prepositional adverb (up, away, through,etc.) may seem to deserve being treated as constituents even when theyare discontinuous: wake up your friend and wake your friend up are al-most synonymous. (Wells : )

Expressions such as wake up your friend presented no new diculties. However,the ‘shifted’ order in which the object intervened between the verb and particle re-quired a means of indicating that ICs formed units at non-adjacent levels One ofthe representational extensions explored by Hockett () is shown in the chart inTable .. Box diagrams provided a somewhat more flexible format, as illustrated inTable ..

wake your friend up

↕

Table .: Chart-based Analysis of ‘shifted’ Phrasal Verb

wake your friendup

Table .: Box-based Analysis of ‘shifted’ Phrasal Verb

.. Phrase structure analysisAs suggested by the contrast between the analyses in Table . and Table ., graphtheory provided the natural formalization of the intuitions underlying models ofIC analysis, though this idea was not developed until McCawley (). Instead, IC

analyses were interpreted as representing the successive segmentation of an expres-sion into sub-expressions, each of which was annotated with a word class label and,usually, other types of information. It was not until the early transformational ac-counts that IC analyses were incorporated into explicit grammar formalisms ratherthan treated as procedures of classification, and the fact that these procedures werefirst formalized by the Bloomfieldians’ successors had the eect of simplifying them,much as the Bloomfieldians had themselves simplified Bloomfield’s more intricateconstructional perspective (Manaster-Ramer and Kac ). In Chomsky (),phrase structure grammars are proposed as “the form of grammar [that] correspondsto [the] conception of linguistic structure” expressed by IC analysis (). Chom-sky’s insight consisted in recognizing how informal procedures for segmenting andclassifying expressions could be expressed by means of rules of the form A → ωthat would ‘rewrite’ a single word class label A by a stringω (which could consistof labels along with words and formatives). Thus a rule such as S→ NP VP wouldrewrite a sentence S by a subject NP and a VP predicate, and the rule V→ tookwouldclassify took as a verb.

By startingwith the sentence label ‘S’ and applying a sequence of phrase structurerules, one could define a ‘derivation’ that terminated in the expression that wouldbe the starting point for procedures of IC analysis. The syntactic analysis assignedto an expression by a phrase structure grammar was conventionally represented bya phrase structure tree, though in Chomsky’s initial formulations, analyses are rep-resented by stringsets that he termed . These sets contain stringsfrom equivalence classes of derivations diering from one another solely in thatthey apply the same rules in a dierent order (e.g., a derivation where the subjectNP is rewritten before rewriting the VP and a second derivation where the VP isrewritten first).

. Extended phrase structure systems

As clarified particularly in Scholz and Pullum () and Pullum (), phrasestructure (and transformational) grammars represent linguistic applications of thegeneral string rewriting systems developed in Post (, ). Despite the evidentsuccess attained by grammatical models based on rewriting systems, it was soonapparent that standard systems were not always ideally suited to the description ofnatural languages.

.. ‘The dicult question of discontinuity’In particular, initial formulations of phrase structure grammars were incapable ofrepresenting the classes of discontinuous constituents recognized by the Bloomfiel-dians, a point that Chomsky () was initially freely prepared to concede.

Chomsky appears to maintain the representational assumptions in Chomsky (: chapter VII)when he later suggests that “We take these objects [i.e. levels of linguistic representation, JPB/IAS] tobe phrase markers in the familiar sense, (represented conventionally by trees or labelled bracketings)”(Chomsky : ).

This [the treatment of ‘long components’ in the sense of Harris ]is an important question, deserving a much fuller treatment, but it willquickly lead into areas where the present formal apparatus may be in-adequate. The dicult question of discontinuity is one such problem.Discontinuities are handled in the present treatment by construction ofpermutationalmappings fromP [the level of phrase structure, JPB/IAS]toW [the level of word structure, JPB/IAS/], but it may turn out thatthey must ultimately be incorporated somehow into P itself. (Chomsky: )

The transformational tradition never did reconsider whether discontinuities couldbe handled better within a phrase structure analysis and no general approach to thisissue was explored within constituency-based grammars until the development ofHead Grammars (Pollard ) and linearization-based models of HPSG (Reape; Müller , ; Kathol ). These models rehabilitated many of thesame intuitions about syntactic and semantic units that had been explored in ‘wrap-ping’ analyses in the Montague grammar tradition, particularly in the accounts ofBach () and Dowty (). However, Chomsky sought to reinforce the casefor ‘permutational mappings’ (i.e., transformations) by disputing the feasibility ofapplying procedures of IC analysis to ‘derived’ constructions such as polar and in-formation questions.

The case for indirect representation, not based on the relation of mem-bership, becomes even stronger when we consider such sentences as“did they see John” or “whom did they see”. T - –, i.e. no one would ask how theycan be subdivided into two or three parts, each of which has severalconstituents, going on to use this subdivision as the basis for analysisof other sentences, and so on. Yet there is nothing in the formulationof principles of procedure for IC analysis that justifies excluding thesesentences, or treating them somehow in terms of sentences already an-alyzed. (Chomsky : f.; emphasis added JPB/IAS)

In the emphasized passage, as elsewhere inChomsky’swritings about theBloom-fieldians, a position possibly consistent with the practice of Zellig Harris is incor-rectly attributed to the Bloomfieldians as a group. Virtually all leading Americanlinguists of the time, including Hockett, Gleason, Nida, Pike andWells, among oth-ers, not only considered applying – but in fact apply – procedures of IC analysisto questions in English. In particular, the analysis of polar questionswas regarded asa solved problem and presented as such in the introductory textbooks of the day. Inthe passage below, Gleason gives what he takes to be an uncontroversial IC analysisof polar questions to exemplify the notion of discontinuous constituents.

In English, discontinuous constituents occur. One common instanceoccurs in many questions: Did the man come? This is clearly to be cutdid … come | the man. (Gleason : )

This discrepancy between procedures of IC analysis and phrase structure gram-mars is of more than purely historical interest. One of the criticisms levelled byChomsky against phrase structure grammars turned on their inability to representdiscontinuous dependencies, particularly within auxiliary verb phrases.

To put the same thing dierently, in the auxiliary verb phrase we re-ally have discontinuous elements… But discontinuities cannot be han-dled within [Σ, F] grammars [i.e. phrase structure grammars, JPB/IAS].(Chomsky : )

.. Generalized Phrase Structure Grammar (GPSG)For the most part, modern phrase structure systems preserve Chomsky’s preferencefor describing discontinuous dependencies indirectly, usually in terms of relationsbetween dierent parts of a single structure or correspondences between dierenttypes of structures. However other restrictions on phrase structure systems havebeen more comprehensively revised. The most severe of these was the assump-tion that the ‘nonterminal’ vocabulary of a phrase structure grammar should consistsolely of atomic labels such as ‘S’, ‘NP’ ‘V’, etc. The case for relaxing this restrictionis made initially by Harman (), who objects that “it is irrational to restrict theamount of information expressed by the grammar to statements about grammaticalcategory” (). The response in Chomsky (: f) dismisses Harman’s pro-posal out of hand as a “terminological equivocation” and appears to construe anyrefinement of phrase structure grammars as a case of a patent infringement ratherthan as a genuine attempt to understand the scope and limits of constituent struc-ture grammars. Partly as a consequence, Harman’s ‘defense of phrase structure’ hadlittle direct influence on the field at the time. Hence, the descriptive potential of fea-ture ‘decompositions’ of atomic symbols was not fully realized until the later workon unbounded dependencies and coordination (Gazdar ).

By this time, a limited amount of feature decomposition had been incorporatedinto transformational models that adopted some version of the X-bar conventions.However, features were assigned a tightly circumscribed role in Chomsky (),and these restrictions were preserved in subsequent accounts. Two constraints wereparticularly decisive. The first of these restricted propagation through an endocen-tric X-bar projection to the word class features ±N and ±V (Chomsky : f.),excluding other types of lexical and inflectional properties. The second constraintlimited feature ‘percolation’, as it came to be known, more generally by “tak[ing] fea-ture complexes to be associated only with lexical categories, and permit[ting] com-plex symbols to dominate a sequence of elements only within the word” (Chomsky: ). These restrictions precluded the use of constituent structure links asa conduit for the propagation of complex feature bundles. Likewise, although the

In the continuation of this passage, Chomsky notes that the second constraint has merely been car-ried over from Chomsky (: f.) and appears willing to countenance the idea “that certain featuresshould also be associated with nonlexical phrase categories”. Yet the accompanying footnote immediatelycharacterizes the arguments supporting previous proposals as “very weak”, and Chomsky does not in factpropose a general relaxation of the constraints on complex feature bundles that would allow the inflec-tional features associated with a lexical category to be propagated to or shared with a phrasal projection.

‘nondistinctness’ condition on complex symbols in Chomsky (: ) anticipatedthe unification operations of later constraint-based formalisms, this condition couldplay no role in regulating the distribution of features within a projection.

... Nonlocal dependencies

As with the representational constraints that barred discontinuities, restrictions onthe ‘flow’ of feature information prevented feature-basedmechanisms fromencroach-ing on the role reserved for structure-changing rules and derivational operations intransformational models. By relaxing these restrictions, extended phrase structuremodels could exploit the descriptive value of feature information for describing localand nonlocal grammatical dependencies. Unbounded dependencies had long beentaken to require the power of a transformational grammar, or at any rate to defyanalysis in terms of phrase structure grammars, as suggested in the quotation fromChomsky () on p. above. Hence the rehabilitation of phrase structure analysisbegan, somewhat counterintuitively perhaps, with an analysis of unbounded depen-dencies that was developed in the late s but first published in Gazdar ().The simple intuition developed in this work was that the constituent structure linksof a phrase structure tree provided a suitable conduit for the flow of informationabout displaced elements. The components of the analysis were equally straightfor-ward: feature attributes that could take categories as values, the insight that infor-mation about ‘missing’ elements could be treated in terms of a feature (Bear ),and feature ‘passing’ conditions that could match features between the ‘mother’ and‘daughter’ nodes in a phrase structure tree. By passing the value of a category-valuedattribute along a chain of local mother-daughter nodes, a phrase structure analysiscould match the properties of a ‘missing’ element at an ‘extraction site’ with those ofthe ‘dislocated’ element that typically occurred at the periphery of a construction.

The components of what came to be known as the ‘slash category’ analysis ofunbounded dependencies are exhibited in the analysis of the English indirect ques-tion in Figure . below. The lowest occurrence of the category-valued featureencodes the properties of the missing object NP that is governed by the transitiveverb saw. These properties are passed up successively to the superordinate VP andS nodes until they can be matched against the properties of the ‘filler’ what.

To a large degree, the early phrase structure analyses carried over prevailing as-sumptions about the structure of unbounded dependency constructions from trans-formational accounts. In contrast to the IC analyses adumbrated in the descriptivisttradition, the structure in Figure . assumes that the dislocated element what ishigher in the tree as well as to the left of the extraction site. This assumption isretained in most subsequent analyses of unbounded dependencies. In addition, thestructure in Figure . preserves the assumption that the extraction site is occupiedby an empty placeholder ‘gap’. Since this assumption had no internal motivationwithin phrase structure models, the analysis developed in Sag and Fodor ()and Bouma et al. () dispensed with null terminals. These analyses neverthe-less retain the strategy of using dedicated attributes to represent information aboutextracted elements. In this respect unlike categorial analyses, such as Steedman(), which use the slash notation both to indicate the argument of a functor and

..S.

NP

.

what

.

S[ ]

.

NP

.

Max

.

VP[ ]

.

V

.

saw

.

NP[ ]

.

e

Figure .: ‘Slash category’ analysis of indirect question

to encode information about extracted elements. In the categorial analysis in Fig-ure ., the category ‘(S\NP)/NP’ marks the transitive verb saw as a functor thatlooks rightward for an NP to form a functor that in turn looks leftward for an NP toform an S. The overloading of this notation is reflected in the fact that the category‘S\NP’ encodes the ‘missing’ object NP in the expressionMax saw.

what Max saw

S/(S\NP) (S\NP)/((S\NP)/NP) (S\NP)/NPS\NP

S

Figure .: Gap-free categorial analysis (cf. Steedman )

As recognized by thoseworking to extend the empirical coverage of phrase struc-ture models, category-valued features oered a novel perspective on a range of phe-nomena that interacted with unbounded dependencies. In particular, the assump-tion that information about missing constituents formed part of the syntactic in-formation associated with a node interacted with the independent assumption thatcoordination was restricted to syntactically like elements. One immediate conse-quence was an account of the parallelism that Ross () had termed ‘across-the-board’ extraction. The central observation was that in a coordinate structure, if oneconjunct contained an extraction site, then all of the conjuncts must. In transfor-mationalmodels, this condition had been attributed to dedicated devices, such as theCoordinate Structure Constraint (Ross ) or theAcross-the-Board convention ofWilliams (), which, as Gazdar et al. (b) noted, incorporated a construction-specific and somewhat imprecise extension to the notion of phrase marker. In con-

See Steedman and Baldridge () for recent synopsis of combinatory categorial approaches.Though subsequent work has called into question whether this parallelism is restricted to coordi-

nate structures (Postal ) and whether the constraint ultimately reflects more general semantic ordiscourse factors (Goldsmith ; Lako ; Kehler ).

trast, the parallelism requirement on extraction from coordinate structures followedon a phrase structure analysis. Two conjuncts of category X[ ] were syntac-tically alike, whereas a conjunct of category X[ ] and one of category X werenot. In the analysis in Figure ., the two conjuncts of category S[ ] aresyntactically alike and can be conjoined, but neither could be conjoined with a full Sto yield unacceptable examples such as *what Felix heard and Max saw the intruderor *what Felix heard the intruder and Max saw.

..S[ ].

S[ ]

.

NP

.

Felix

.

VP[ ]

.

V

.

heard

.

NP[ ]

.

e

.

and

.

S[ ]

.

NP

.

Max

.

VP[ ]

.

V

.

saw

.

NP[ ]

.

e

Figure .: ‘Across-the-board’ extraction from coordinate indirect question

Gazdar () also clarified how constraints on extraction, which had typicallybeen described in terms of conditions on rule application, could be recast in terms ofrestrictions on the ‘paths’ of category-valued features that connected extraction sitesto dislocated fillers. In classical transformational accounts, there had been no rea-son why information about missing constituents should trace a path along the con-stituent structure links of a tree. But once extraction was characterized in terms ofthe sharing of category-valued features along a sequence of mother-daughter links,it became clear that any restrictions on the extraction of elements out of specified‘island’ domains (Ross ) would correspond to paths in which those domainsoccurred somewhere along the path between extraction sites and fillers.

... Local dependencies

The demonstration that complex-valued features could provide an analysis of un-bounded dependencies inspired surface-based analyses of more local syntactic phe-nomena within the nascent community that had begun to explore the potential ofmonostratal models. The English auxiliary system had long been an obvious can-didate for reanalysis. The system consisted of a finite inventory of modal and auxil-iary elements, which were subject to ordering constraints that determined a (short)

The analysis in Figure . introduces the coordinating conjunction as a sister of the conjuncts, ratherthan associating it with the second conjunct, though nothing here hinges on this dierence.

Indeed, there was a considerable delay before the tradition even addressed the challenge of assigninga derived constituent structure to transforms, an issue that had been raised as early as Stockwell ().

See Ladusaw () for discussion of the contrast between linguistic ‘levels’ and ‘strata’.

maximum expansion. The expansions were indeed almost as restricted as pronom-inal clitic sequences in Romance languages, and, like these sequences, exhibitedsome of the ordering rigidity characteristic of morphological formations. Even theselectional dependencies tended to relate pairs of adjacent elements. So there wasnothing that presented any intrinsic diculties for a phrase structure analysis.

The ‘ax hopping’ analysis of Chomsky () had long been held as one of thecrowning achievements of transformational approaches. However, Gazdar et al.(a) showed that the strategy of ‘hopping’ axes from one point in a terminalstring to another was a solution to a self-inflicted problem and hence dispensable inamodel with complex-valued features. If one auxiliary element could select the verbform of the head of a phrasal complement, there was no need to assemble inflectedforms in the course of a syntactic derivation. Instead, the admissible expansionscould be determined by the subcategorization demands of individual elements. Thefirst component of this analysis is a feature classification of verbal elements thatdistinguishes tense, aspect and voice properties, along with form variants, such asparticiples, infinitives, etc. The second is a generalization of the X-bar feature con-ventions that allows these ‘head’ features to be shared between a mother and headdaughter node. The final ingredient is, again, category-valued features that permita verbal element to select a complement headed by a particular form variant.

These components are set out in detail in Gazdar et al. (a) and in muchsubsequent work within Generalized Phrase Structuremodels. One type of analysisthat they define is illustrated in Figure . below. The advantages of this analysisare summarized in Gazdar et al. (a: .), though one immediate benefit wasthe avoidance of the formal problems that had plagued the ‘ax-hopping’ analysissince its initial formulation (see, e.g., Akmajian andWasow ; Sampson ).

..V.

V[+]

.

must

.

V[+,+]

.

V[+,+]

.

have

.

V[+,+,+]

.

V[+,+,+]

.

been

.

V[+,–,+]

.

V[+,–,+]

.

being

.

V[+]

.

persecuted

Figure .: Passive auxiliary expansion (cf. Gazdar et al. a: )

The analyses inGazdar et al. (a) thus established that the samebasic feature-

passing strategy used in the treatment of unbounded dependencies could provide anaccount of local dependencies. Patterns of subject-auxiliary inversionwere amenableto a similar analysis using grammar rules systematically related to the basic rulesviametarules, a device whose utility in the grammar of programming languages hadpreviously been established. Figure . exhibits the analysis of the polar questioncited by Gleason () on p. above. The invertibility of modals and auxiliaries isencoded here via compatability with the [+] specification that is required of theverbal head in a phrase structure rule licensing the ‘inverted’ structure. Independentmotivation for this feature comes from lexical restrictions on the distribution andinterpretation of auxiliary elements. Some elements, such as sg aren’t, are obli-gatorily inverted, while others, such as better, are obligatorily uninverted, and yetothers, such as may, have a dierent range of meanings depending on whether ornot they are inverted.

..V[+]

.

V[+]

.

did

.

V[+]

.

NP

.

the man

.

V[+]

.

V[+]

.

come

Figure .: Subject-auxiliary ‘inversion’

.. Node admissibility and constraint satisfactionMore generally, it turned out that all of the alternations and dependencies that hadbeen described by transformational models had simple – and, in at least some cases,arguably superior – phrase structure analyses. One might have expected that thisresult would have produced a general rapprochement between transformational andphrase structure approaches and an attempt to arrive at broadly accepted criteria forevaluating the dierent strategies for describing these patterns. In fact, just the

The analyses of unbounded dependencies and auxiliary selection/inversion outlined above were fol-lowed by phrase structure treatments of, among others, an expanded range of extraction constructions(Pollard and Sag : chapters –, Levine and Hukari ), passives (Pollard and Sag : ,Manning and Sag ), control constructions (Sag and Pollard ), anaphoric binding (Pollard andSag ). Contemporary work in Lexical Functional Grammar (Dalrymple et al. ) and Tree Adjoin-ingGrammar (Joshi and Schabes ) explored a similar range of empirical extensions. See also Johnsonand Lappin () for a comprehensive comparison of constraint-based and derivation perspectives.

opposite occurred. Transformational models abandoned their flirtation with a ‘rep-resentational’ interpretation, a perspective that had been developed particularly inKoster (, ), and adopted a more resolutely derivational orientation.

While transformational accounts were following the developmental pathwaythat led to current Minimalist models (see chapter ), extended phrase structuremodels began to incorporate insights and perspectives from other monostratal ap-proaches. Following McCawley (), models of Generalized Phrase StructureGrammar (Gazdar et al. ) had already adopted – and, indeed, refined – a ‘nodeadmissibility’ interpretation of phrase structure rules. On this interpretation, a rulesuch as S→NPVP is interpreted as directly ‘licensing’ a local subtree in which S im-mediately and exhaustively dominates NP and VP daughters, and the NP daughterimmediately precedes the VP daughter. A node admissibility interpretation imme-diately eliminated the need for string-rewrite derivations and string-based repre-sentations of phrase structure (‘phrase markers’). Instead, rules could be regardedas partial of the subtrees that they sanctioned and the admissibility ofa tree could be defined in terms of the admissibility of the subtrees that it contained.

In large part, this reinterpretation of phrase structure productions supplied graph-theoretic modelling assumptions that were a better fit for the classes of IC analy-ses initially proposed by the Bloomfieldians. The schematization adopted withinmodels of X-bar Theory similarly deprecated phrase structure rules within transfor-mational models, though without substantially revising the string-based model ofphrase structure represented by phrase markers (as discussed in footnote ).

Furthermore, a node admissibility interpretation clarified the fact that conven-tional phrase structure rules bundle information about structure (mother-daughterlinks) together with information about order (linear arrangement of daughters).GPSG accounts showed how these two types of information could be expressed sep-arately, by means of a set of Immediate Dominance (ID) rules that just constrainedmother-daughter relations and a set of Linear Precedence statements that applied tosisters in a local tree. For example, the information represented by the phrase struc-ture rule S→ NP VP would be expressed by an ID rule S→ NP, VP and the generalLP statement NP ≺ VP. The absence of an applicable LP rule would not sanctionunordered trees, but rather trees in which the NP and VP occurred in either order.

An overriding consideration in the development of GPSG was the goal of keep-ing analyses as explicit as possible and the underlying grammatical formalism asformally restrictive as possible. The central role of context-free phrase structuregrammars largely reflected the fact that their properties were well-understood andprovided a formal basis for transparent analyses. In some cases, analyses were con-strained so that they did not take GPSGmodels outside the class of phrase structuregrammars. For example, requiring that sets of ID rules and LP statements must op-erate over the same local domains, ensured that they could in principle be ‘reconsti-tuted’ as phrase structure grammars. LP statements were thus restricted to apply tosister nodes. As a consequence, LP statements could allow free or partial orderingof VP-internal elements, but they could not impose any ordering of subjects andVP-internal elements other than those that followed from the ordering of a subjectand full VP expansion. Yet there was no direct empirical support for this restriction.

Hence the tight association between the domains of ID rules and LP statements

undermined the fundamental separation of structure and order in the ID/LP formatsince certain types of ordering variation dictated a flat structure. This was perhapsacceptable as long as there was some independent motivation for remaining withinthe class of context-free phrase structure grammars. But by , the demonstrationof non-context-free patterns in Swiss German subordinate clauses (Shieber )and Bambara compounds (Culy ) had weakened the empirical grounds for thisrestriction and the non-transformational community shifted their focus to identify-ing restricted classes of weakly context-sensitive grammars that were descriptivelyadequate. This was a natural development within the family of phrase structure ap-proaches, given that the interest in context-free grammars had been driven by aninterest in explicit formalisms with clearly-defined and well-understood properties.Hence the move from the limited word order freedom defined by the ID/LP for-mat in GPSG to ‘domain union’ in HPSG (Reape ) extended the dissociationof structure and order in ways that allow for the interleaving of non-sisters in anexplicit but non-context-free formalism.

. Model-theoretic grammarIn the subsequent development of phrase structure grammars, the interpretation ofrules as partial descriptions of trees provided the model for a more comprehensiveconstraint-based ormodel-theoretic perspective. As inmodels of Lexical-FunctionalGrammar (Kaplan and Bresnan ; chapter of this volume), rules and grammat-ical principles came to be construed as that described or were satisfiedby corresponding types of . This move to a uniform model-theoreticorientation permitted much of the complexity that had been associated with rep-resentations to be confined to the constraint language that described structures.In addition, a general model of constraint satisfaction provided a conception un-der which the diverse feature distribution principles of GPSG could be subsumed.The gradual accretion of constraints and conditions in GPSG had led over time toa theory in which the components that regulated feature information included fea-ture co-occurrence restrictions and feature specification defaults, in addition to theimmediate dominance rules, linear precedence statements and other devices, suchas metarules. As detailed in Gazdar et al. (: chapter ), the constraints in thesecomponents exhibited fairly intricate interactions. On a description-theoretic inter-pretation, these constraints and interactions could be modelled in a more uniformand transparent way.

The emergence of a constraint-based perspective was accompanied by the adop-tion of richer sets of structures and more expressive constraint languages. Thesedevelopments provided clearer conceptions of the lexicon, valence and valence al-ternations than had been possible in GPSG. The phrase structure systems proposedin Chomsky () had oered only very rudimentary treatments of the lexicon andvalence demands, and incorporated no notion of a lexical valence-changing process.The closest counterparts of ‘lexical entries’ in these simple systems were rules of theform V → sleep, which rewrote a nonterminal symbol as a terminal element. The

Reape () was widely circulated in draft form, as of

valence of a predicate was likewise represented implicitly by the other elements thatwere introduced in the same rule expansions. GPSG enriched this spartan concep-tion by locating terminal elements within lexical entries that specified distinctivegrammatical features of an element other than word class. Corresponding to thepreterminal rules of a simple phrase structure grammar was a class of ‘lexical IDrules’ which introduced lexical heads indexed by a subcategorization index. Thisindex (technically the value of a feature) was then cross-referenced with aclass of lexical entries. For example, the rule VP → H[] would license a local VPsubtree that dominated a unary tree whose mother was V[] and whose daughterwas an intransitive verb, such as sleep, whose entry contained the index .

In eect, the use of subcategorization indices achieved a limited type of contextsensitivity within a context-free formalism. Yet, as Jacobson (: .) pointedout, the fact that lexical items did not directly represent valence information cre-ated numerous complications in GPSG. Themost acute arose in connection with thetreatment of valence alternations. There was no way to formulate a passive rule thatmapped the transitive entry for devour onto a (syntactically) detransitivized entrydevoured, because entries themselves contained no direct representation of transi-tivity. This led to an analysis of passivization in terms of metarules that mapped a‘transitive expansion’ such as VP → W , NP to a ‘detransitivized expansion’ such asVP[]→W (whereW is any string). However, it then became necessary to con-strain metarules so that they only applied to lexical ID rules. But lexical ID ruleswere serving as proxies for underinformative entries, so the obvious solution lay inassociating valence information directly with lexical items and introducing a classof lexical rules to map between entries, as suggested by Pollard ().

.. Head-Driven Phrase Structure Grammar (HPSG)The models of Head-driven Phrase Structure Grammar outlined in Pollard and Sag(, ) develop a number of these revisions in the context of a broad constraint-based conception of grammar. A central component of these models is the set of as-sumptions that have come to be known as the ‘formal foundations’ of HPSG. As inLFG, grammatical constraints and lexical entries are interpreted as partial descrip-tions of structures, though the representational conventions and model theories ofthe two theories dier significantly. One representational dierence concerns theinterpretation of attribute-value matrices (AVMs). Whereas in LFG, AVMs of thesort illustrated in Figure .a are used to represent f(unctional)-structures, in HPSGthey represent descriptions, i.e. sets of constraints. The structures (orwhich ) a set of constraints are represented as graphs like the one in Fig-ure .b.

A distinctive aspect of the HPSG model theory is the role assigned to the system. The core idea is that each kind of structure is associated with certain sortsof attributes, and that each attribute is associated with a type of value. For example,a referential index (object of type ref in Figure .) is associated with the attributes(), () and (). Each attribute takes values from a partitioned

By virtue of the Head Feature Principle (aka the Head Feature Convention), the metavariable ‘H[]’would license a preterminal V[] that shared the word class features of the VP mother.

a.

ref rd sing masc

b.

..→ ·ref.

·masc

.

. ·sing.

.

·rd

.

Figure .: Descriptions and structures in HPSG

value space, which in the present case just represents traditional person, numberand gender contrasts. The empirical eects of this type system derive from twoadditional assumptions. The first is that structures must be -(Carpenter , chapter ) in the sense that they must be assigned a value foreach appropriate attribute. This constraint precludes, for example, the assignmentof a number-neutral structure as the analysis of English sheep, given that number isdistinctive for English nouns (each occurence of sheep is unambuously singular orplural). A separate requirement that structures must be - (Pollard andSag : ) permits only ‘fully specific’ feature values and thus bars disjunctivecase values from occurring in a wellformed structure. Hence sheep could not betreated as neutral by assigning the attribute a maximally general value such asnumber, which subsumes the resolved values sing and plur. Given that entries areinterpreted as descriptions of lexical structures, the English lexicon can still containa single underspecified for sheep, one that specifies either no attribute ora attribute with a non-sort-resolved value. But the lexical structures describedby the entry must be totally well-typed and sort-resolved.

These general assumptions have the eect of ensuring that structures are max-imally specific and that all underspecification is confined to descriptions. A neutraldescription is not satisfied by a correspondingly underspecified structure but by a setof structures, each of which supplies dierent, fully resolved values for underspeci-fied attributes. This technical point has a number of consequences. On the positiveside, the assumption that structures must be totally well-typed and sort-resolveddoes some of the work of the completeness and coherence conditions in LFG, andfacilitates type-based inferencing within HPSG. However, these assumptions alsolead to apparent diculties in accounting for the types of patterns described in Ingria(), in which the neutrality of an item seems to permit it to satisfy incompatibledemands simultaneously, most prominently in coordinate structures.

Note further that in a model theory that only contains fully specified structures,it is somewhat anachronistic to describe the processes that determine feature com-patibility in terms of feature structure , as had been the practice inGPSGand PATR-based formalisms (Shieber ). A more accurate characterization of a

See Blevins () for a recent review and discussion of these types of cases.

model-theoretic linguistic framework would be as -, a term thathas garnered a certain acceptance in the non-transformational community. WithinHPSG, configurations in which a single object occurs as the value of multiple at-tributes are described in terms of -, a term that refers to reen-trance in the graph-theoretic models typically assumed in HPSG.

... Valence, raising and control

Raising constructions illustrate how structure sharing interactswith complex-valuedattributes to provide an insightful analysis of grammatical dependencies. The term‘raising’ derives from transformational analyses in which the subject of a comple-ment is taken to be ‘raised’ to become an argument of the raising verb. However,complex-valued features permit an analysis in which raising involves the identifi-cation of arguments within the argument structure of a raising predicate. Patternsinvolving the sharing of purely morphological properties oer the clearest supportfor this analysis. As discussed by Andrews (), among others, modern Icelandiccontains verbs that may govern ‘quirky’ non-nominative subjects. One such verb isvanta ‘to want’, which occurs with the accusative subject hana ‘her’ in (.a). Thesequirky case demands are preserved by raising verbs such as virðist ‘seems’. As ex-ample (.b) shows, virðist is, in eect, ‘transparent’ to the accusative case demandsof vanta, which are imposed on its own syntactic subject.

(.) Quirky case in Icelandic raising constructions (Andrews )a. Hana

her.vantarlack.

peninga.money.

‘She lacks money.’b. Hana

her.virðistseem.

vantalack

peninga.money.

‘She seems to lack money.’

The constraints in Figure . first identify hana as a sg feminine accusativeNP, and indicate that the verb vanta selects an accusative subject and complement(though category is suppressed here). In place of the integer-valued featureof GPSG, HPSG represents the core valence demands of a verb by means of list-valued and features. The value of the attribute can either be anempty list or a singleton list, whereas the value may contain as many depen-dents as a verb can select. The boxed integers in the indicated constraints for vantarepresent the fact that the subject term corresponds to the first element of the lexi-cal argument structure (-) of vanta and the complement term corresponds tothe second term. This correspondence is not established by individual entries, butinstead reflects a general relationship between - and and lists. Bytreating the correspondence as canonical rather than as invariant, HPSG accommo-dates divergences between argument structure and grammatical relations (Manningand Sag ).

The analysis in Figure . then illustrates how these valence features regulatebasic valence requirements. Adapting the idea of ‘argument cancellation’ from cate-

hana:

3rd sg fem

acc

vanta:

⟨1

[ acc

]⟩

⟨2

[ acc

]⟩-

⟨1 , 2

⟩

Figure .: Constraints on accusative NP and ‘quirky’ accusative-governing verb

gorial approaches, elements are ‘popped o’ valence lists as arguments are encoun-tered. Hence the term in the list of the verb vantar is structure shared withthe syntactic object peninga in Figure ., producing a VP with an empty list.The subject term is in turn identified with the syntactic subject hana, yielding a ‘sat-urated’ clause, with empty and lists. The terms in the - list of theverb vanta are also structure-shared with the syntactic arguments. However, in ac-cordance with the locality constraints of HPSG, - values are only associatedat the lexical level, so that elements that combine syntactically with the clause inFigure . cannot access information about the dependents it contains.

..[ ⟨ ⟩ ⟨ ⟩

].

1

3rd sg fem

acc

.

hana

.

⟨1

⟩ ⟨ ⟩

.

⟨1

⟩

⟨2

⟩-

⟨1 , 2

⟩

.

vantar

.

2

3rd sg masc

acc

.

peninga

Figure .: Regulation of valence demands in HPSG

Given this general treatment of valence, the transparency of virðist can be repre-sented by the entry in Figure .. The cross-referencing of the two values (viathe boxed integer ‘ 1 ’) indicates that the attribute of virðist literally shares itsvalue with the value of its complement. Identifying the values of the two attributes ensures that any constraints that apply to the of the complement ofvirðist will apply to its own syntactic . Hence when vanta occurs as the headof the complement, as in Figure ., its accusative demands will be identifiedwith the demands of virðist. Only an accusative subject such as hana can satisfy

these demands. So this analysis forges a direct association between hana and thecomplement vanta peninga, but the association is established by means of structuresharing, rather than through constituent structure displacements.

virðist:

⟨1

⟩

⟨2

[

⟨1

⟩]⟩-

⟨1 2

⟩

Figure .: Lexical entry of subject raising verb

..[ ⟨ ⟩ ⟨ ⟩

].

1

3rd sg fem

acc

.

hana

.

⟨1

⟩ ⟨ ⟩

.

⟨1

⟩

⟨2

⟩-

⟨1 , 2

⟩

.

virðist

.

2

⟨1

⟩ ⟨ ⟩

.

vanta peninga

Figure .: Subject ‘raising’ as structure sharing

This analysis shows how the complex-valued features that provide an accountof basic valence demands in Figure . interact with structure-sharing to allow thesubject demands of a raising verb to be identified with those of its complement.Furthermore, precisely the same elements oer an analysis of ‘control’ construc-tions, in which the higher controller merely identifies the reference of the subject ofthe complement. The properties of control constructions are discussed in detail inSag and Pollard () but they can be broadly subsumed under the generalizationthat control verbs are not transparent to the syntactic demands of the head of theircomplement. The contrast with raising verbs is reflected in the fact that the sub-ject of the control verb ‘hope’ in (.b) follows the default nominative patternand does not inherit the accusative case governed by its complement in (.a)(repeated from (.a)).

A similar analysis is proposed within LFG in terms of ‘functional control’ (Bresnan ).

(.) Icelandic subject control constructions (cf. Andrews : )a. Hana

her.vantarlack.

peninga.money.

‘She lacks money.’b. Hún/*hana

she./*her.vonasthope.

til aðtoward

vantanot

ekkilack

peninga.money.

‘She hopes not to lack money.’

The intuition that the subject of a control verb merely identifies the referenceof its complement’s subject is expressed by the entry in Figure ., in which theindex values of the two values are identified (i.e. structure-shared). The fact

vanast:

⟨1

[ 3

]⟩

⟨2

[

⟨[ 3

]⟩]�-

⟨1 , 2

⟩

Figure .: Lexical constraints on subject control verb

that index but not case values are shared in this entry allows the subject of vonastto select a nominative subject and control a complement that selects an accusativesubject in Figure .. Exactly the same formal components determine the anal-yses in Figures . and .; there is no analogue to distinct ‘raising’ and ‘equi’transformations or to distinct PRO and ‘trace’ elements in the subordinate subjectpositions. Instead it is solely the locus of structure sharing that distinguishes thesesubconstructions.

... Lexical rules

The treatment of argument structure in terms of a list-valued - feature alsoprovides the formal basis for a lexical analysis of valence alternations in HPSG. Lex-ical rules can apply to an entry andmodify the - list in variousways, by adding,deleting, permuting or reassociating elements. The new entries that are defined bythese types of rules will then have dierent combinatory and interpretive proper-ties, due to the cross-referencing of - elements with valence features and withsemantic representations. For example, dierent versions of passive lexical rulesare proposed in Pollard and Sag (: ) and Manning and Sag (), and anumber of other valence-changing lexical rules are proposed in the HPSG literature(see, e.g., Wechsler and Noh () and Müller ()). However, the study ofvalence alternations has been less a primary focus of research within HPSG than in,say, LFG (see the discussion of Lexical Mapping Theory in chapter ).

There is more to say about Icelandic raising constructions and the mechanisms that allow quirky‘lexical’ case to take priority over default ‘structural’ case. See Andrews (, ) for some discussion.

..[ ⟨ ⟩ ⟨ ⟩

].

1

3

3rd sg fem

nom

.

hún

.

⟨1

⟩ ⟨ ⟩

.

⟨1

⟩

⟨2

⟩-

⟨1 , 2

⟩

.

vonast

.

2

⟨[

3

acc

]⟩ ⟨ ⟩

.

til að vanta ekki peninga

Figure .: Subject ‘control’ as index feature sharing

... The architecture of signs

Figures . and . illustrate the tree-based diagrams that are often used to exhibitHPSG analyses. These representations show the usefulness of tree-based diagramsfor isolating particular aspects of an analysis, in the present case the role of valencefeatures and structure sharing. However, the familiarity of this representational for-mat comes at a cost, as it slightly misrepresents the - nature of HPSG.In GPSG, feature structures are labels that annotate the nodes of a phrase struc-ture tree. But HPSG inverts this conception and incorporates constituent structurelinks within general data structures termed . Within the version of HPSG ex-pounded in Pollard and Sag (), a head-complement sign has the general struc-ture in Figure .. There are two innovative aspects of this analysis. The first is thatsyntactic and semantic features are consolidated into a single type of data structure,termed a synsem. The second is that constituent structure is represented by (‘daughters’) attributes that take signs or lists of signs as values. Hence the VP fromFigure . above is represented, albeit somewhat less perspicuously, by the sign inFigure . below.

Even the fairly rich analysis in Figure . suppresses syntactic detail (not tomention all of the semantic properties incorporated within synsem objects). Al-though the highly explicit nature of the HPSG formalism may seem somewhat im-posing, the formal character of the formalism is designed with the dual goals ofbroad-coverage theoretical description and large-scale practical implementation inmind. For students (and general linguists) who mainly want to understand the

See also Orgun () for a sign-based model of morphology compatible with HPSG assumptions.Repositories of information, publications and materials related to current HPSG implementations

can be found at http://lingo.stanford.edu/, http://hpsg.fu-berlin.de/Projects/core.html, and

phonological representation syntactic and semantic features

[- single sign- <list of signs>

]

Figure .: Structure of head-complement signs

<vantar peninga>

⟨1

⟩ ⟨⟩

-⟨ <vantar>

⟨1

[ acc

]⟩

⟨2

⟩-

⟨1 2

⟩

⟩

-⟨

2

<peninga>

3rd plu masc

acc

⟩

Figure .: Partial analysis of vantar peninga

basic intuitions and desiderata that underlie HPSGmodels, a more streamlined ver-sion of the formalism is presented in Sag et al. ().

.. Sign-Based Construction Grammar (SBCG)In much the same way that initial models of HPSG drew on ideas from categorialgrammar and adapted techniques fromAI and theoretical computer science, currentmodels of Sign-Based Construction Grammar integrate key empirical insights fromthe Berkeley Construction Grammar tradition (Goldberg ; Kay and Filmore). The conceptual unification of these traditions rests on the insight that therich construction inventories investigated in Construction Grammar can be mod-elled by organizing individual constructions into inheritance networks. The formalarchitecture required by this analysis is already fully present in standard models ofHPSG, in the form of the type hierarchies that cross-classify individual signs repre-senting words, phrases and clauses. The main prerequisite for a construction-basedextension of HPSG is then a type of feature structure that represents constructions.

http://hpsg.fu-berlin.de/∼stefan/Babel/. See also Müller (: f.).

As noted in Sag (b, ), feature structure counterparts of the local treesfrom GPSG provide suitable candidates. Individual constructions can be repre-sented by feature structures exhibiting the organization in Figure ., where represents themother sign and a list of daughter signs. Many of the construction-specific properties investigated in themodernConstructionGrammar literature (typ-ified by Kay and Filmore ()) can be integrated into these unified data structures.

construct sign0 <sign1,…,signn>

Figure .: General structure of a construct (Sag b: )

As inHPSG, the inheritance of properties within a construction inventory can bemodelled by type hierarchies. The partial hierarchy in Figure . represents naturalclasses of constructions relevant to the analysis of extraction in English.

..construct.

phrasal-cxt

.

…

.

headed-cxt

.

…

.

subject-head-cxt

.

head-filler-cxt

.

aux-initial-cxt

.

lexical-cxt

.

derivational-cxt

.

inflectional-cxt

.

…

Figure .: Partial construction type hierarchy (Sag b: )

The detailed treatment of English relative and filler-gap clauses in Sag (,a) presents a sustained argument for extending HPSGmodels to include a no-tion of construction. At the same time, these studies make a case for reconceptu-alizing grammatical constructions in the context of a constraint-based architecture,rather than in the exemplar-based terms assumed in traditional grammars.

These studies also illustrate the ways that phrase structure models continue toevolve, driven in part by the logic of their basic organizing principles, and in partby their ability to incorporate and extend insights from other traditions. From theirorigins in the string rewriting systems in Chomsky (), extended phrase struc-turemodels have assumed theirmodern form by successively integrating traditionalperspectives on grammatical features and units with more formal notions such asinheritance hierarchies and constraint satisfaction. In addition to providing analy-ses of a wide range of syntactic constructions, these models have clarified how ex-plicit mechanisms for regulating the distribution of grammatical informationwithin

a single syntactic representation can achieve any of the benefits that had, beginningwith the work of Harris (), been claimed to accrue to derivational analyses.

ReferencesAkmajian, Adrian andWasow, Thomas (). The constituent structure of VP and and the position of the verb . Linguistic Analysis, , –.

Andrews, Avery D. (). The representation of case in modern Icelandic. In TheMental Representation of Grammatical Relations (ed. J. Bresnan), pp. –.MIT Press, Cambridge.

Andrews, Avery D. (). Case structures and control in modern Icelandic. InSyntax and semantics : Modern Icelandic syntax (ed. J. Maling and A. Zaenen),pp. –. Academic Press, San Diego.

Bach, Emmon (). Control in Montague Grammar. Linguistic Inquiry, ,–.

Bear, John (). Gaps as syntactic features. Technical note, Indiana UniversityLinguistics Club.

Blevins, James P (). Feature-based grammar. See Borsley and Börjars ().

Bloch, Bernard (). Studies in colloquial Japanese II: Syntax. Language, ,–. Reprinted in Joos (: –).

Bloomfield, Leonard (). Language. University of Chicago Press, Chicago.

Borsley, Robert D. and Börjars, Kersti (ed.) (). Non-Transformational Syntax.Wiley-Blackwell, Oxford.

Bouma, Gosse, Malouf, Rob, and Sag, Ivan A (). Satisfying constraints onextraction and adjunction. Natural Language and Linguistic Theory, , –.

Bresnan, Joan (). Control and complementation. Linguistic Inquiry, –.

Carpenter, Bob (). The Logic of Typed Feature Structures. Cambridge UniversityPress, New York.

Chomsky, Noam (). Three models for the description of language. Institute ofRadio Engineers Transactions on Information Theory, II-, –. Reprinted inLuce et al. (), –.

Chomsky, Noam (). Syntactic Structures. Mouton, The Hague.

Chomsky, Noam (). A transformational approach to syntax. In Third TexasConference on Problems of Linguistic Analysis in English (ed. A. A. Hill), pp.–. University of Texas Press, Austin.

Chomsky, Noam (). Aspects of the Theory of Syntax. MIT Press, Cambridge,MA.

Chomsky, Noam (). Remarks on nominalization. In Readings in English Trans-formational Grammar (ed. R. A. Jacobs and P. S. Rosenbaum), pp. –. Ginnand Company, Waltham. Reprinted in Chomsky (), –.

Chomsky, Noam (). The Logical Structure of Linguistic Theory. University ofChicago Press, Chicago.

Chomsky, Noam (). Studies on Semantics in Generative Grammar. Mouton,The Hague.

Chomsky, Noam (). The Minimalist Program. MIT Press, Cambridge.

Culy, Christopher (). The complexity of the vocabulary of Bambara. Linguisticsand Philosophy, , –.

Dalrymple, Mary, Kaplan, Ronald M., Maxwell, III, John T., and Zaenen, Annie(ed.) (). Formal Issues in Lexical-Functional Grammar. CSLI, Stanford.

Dowty, David (). Grammatical relations and Montague Grammar. In The Na-ture of Syntactic Representation (ed. G. K. Pullum and P. Jacobson), pp. –.Reidel, Dordrecht.

Gazdar, Gerald (). Unbounded dependencies and coordinate structure. Lin-guistic Inquiry, , –.

Gazdar, Gerald, Klein, Ewan, Pullum, Georey K., and Sag, Ivan A (). Gener-alized Phrase Structure Grammar. Harvard University Press, Cambridge.

Gazdar, Gerald, Pullum, Georey K., and Sag, Ivan A (a). Auxiliaries andrelated phenomena in a restrictive theory of grammar. Language, , –.

Gazdar, Gerald, Pullum, Georey K., Sag, Ivan A., and Wasow, Thomas (b).Coordination and transformational grammar. Linguistic Inquiry, , –.

Gleason, Henry Allan (). An Introduction to Descriptive Linguistics. Holt, Rine-hart and Winston, New York.

Gleason, Henry Allan (). Linguistics and English Grammar. Holt, Rinehart andWinston, New York.

Goldberg, Adele (). A Construction Grammar Approach to Argument Structure.University of Chicago Press, Chicago.

Goldsmith, John A (). A principled exception to the coordinate structure con-straint. In Chicago Linguistic Society, Volume , pp. –.

Harman, Gilbert H (). Generative grammars without transformational rules.Language, , –.

Harris, Zellig S (). Co-occurrence and transformation in linguistic structure.Language, , –. Reprinted in Harris (), –.

Harris, Zellig S (). Papers in Syntax. Reidel, Dordrecht.

Hockett, Charles F (). A Course in Modern Linguistics. MacMillan, New York.

Ingria, Robert J (). The limits of unification. In Proceedings of the th An-nual Meeting of the Association for Computational Linguistics, Morristown, NJ,pp. –.

Jacobson, Pauline (). Review of Generalized Phrase Structure Grammar. Lin-guistics and Philosophy, , –.

Johnson, David and Lappin, Shalom (). Local Constraints vs Economy. CSLI,Stanford.

Joos, Martin (ed.) (). Readings in Linguistics I. University of Chicago Press,Chicago.

Joshi, Avarind and Schabes, Yves (). Tree Adjoining Grammars. In Hand-book of Formal Languages, Vol. (ed. G. Rosenberg and A. Salomaa), pp. –.Springer Verlag.

Kaplan, RonaldM. and Bresnan, Joan (). Lexical-functional grammar: A formalsystem for grammatical representation. In The Mental Representation of Gram-matical Relations (ed. J. Bresnan), pp. –. MIT Press, Cambridge.

Kathol, Andreas (). Linear Syntax. Oxford University Press, Oxford.

Kay, Paul and Filmore, Charles J. (). Grammatical constructions and linguisticgeneralizations: TheWhat’s X doing Y? construction. Language, (), –.

Kehler, Andrew (). Coherence, Reference, and the Theory of Grammar. CSLIPublications, Stanford.

Koster, Jan (). Locality Principles in Syntax. Foris, Dordrecht.

Koster, Jan (). Domains and Dynasties: The Radical Autonomy of Syntax. Foris,Dordrecht.

Ladusaw, William (). A proposed distinction between levels and strata. InLinguistics in the Morning Calm (ed. S.-D. Kim), pp. –. Hanshin, Seoul.

Lako, George (). Frame semantic control of the coordinate structure con-straint. In Chicago Linguistic Society, Volume , pp. –.

Levine, RobertD. andHukari, Thomas (). TheUnity ofUnboundedDependencyConstructions. CSLI Lecture Notes, No. . CSLI Publications, Stanford.

Luce, R., Bush, R., and Galanter, E (ed.) (). Readings in Mathematical Psychol-ogy . Wiley and Sons, New York.

Manaster-Ramer, Alexis and Kac, Michael B. (). The concept of phrase struc-ture. Linguistics and Philosophy, , –.

Manning, Christopher D. and Sag, Ivan A. (). Dissociations between argumentstructure and grammatical relations. In Lexical and Constructional Aspects of Lin-guistic Explanation (ed. G. Webelhuth, J.-P. Koenig, and A. Kathol), pp. –.CSLI, Stanford.

McCawley, James D. (). Concerning the base component of a tranformationalgrammar. Foundations of Language, , –.

McCawley, James D (). Parentheticals and discontinuous constituent structure.Linguistic Inquiry, , –.

Müller, Stefan (). Deutsche Syntax deklarativ. Head-Driven Phrase StructureGrammar für das Deutsche. Volume Linguistische Arbeiten. Niemeyer.

Müller, Stefan (). Complex Predicates: Verbal Complexes, Resultative Construc-tions, and Particle Verbs in German. Studies in Constrant-Based Lexicalism .CSLI Publications, Stanford.

Müller, Stefan (). Continuous or discontinuous constituents? a comparisonbetween syntactic analyses for constituent order and their processing systems.Research on Language and Computation, –.

Müller, Stefan (). Grammatiktheorie. Number in Stauenburg Einführun-gen. Stauenburg Verlag, Tübingen.

Nida, Eugene A (). A Synopsis of English Grammar. Mouton, The Hague.

Orgun, C. Orhan (). Sign-Based Morphology and Phonology: with special at-tention to Optimality Theory. Ph.D. thesis, University of California, Berkeley.

Pike, Kenneth L (). Taxemes and immediate constituents. Language, , –.

Pollard, Carl (). Generalized Phrase Structure Grammars, Head Grammars andNatural Language. Ph.D. thesis, Stanford.

Pollard, Carl and Sag, Ivan A (). Information-Based Syntax and Semantics.CSLI, Stanford.

Pollard, Carl and Sag, Ivan A. (). Anaphors in English and the scope of thebinding theory. Linguistic Inquiry, , –.

Pollard, Carl and Sag, Ivan A (). Head-driven Phrase Structure Grammar. Uni-versity of Chicago Press, Stanford.

Post, Emil L (). Formal reductions of the general combinatorial decision prob-lem. American Journal of Mathematics, , –.

Post, Emil L (). Recursive unsolvability of a problem of Thue. Journal of Sym-bolic Logic, , –.

Postal, Paul M (). Three Investigations of Extraction. MIT Press, Cambridge,MA.

Pullum, Georey K. (). On the mathematics of Syntactic Structures. Journal ofLogic, Language and Information, , –.

Reape, Mike (). Getting things in order. In Discontinuous Constituency (ed.H. Bunt and A. van Horck), pp. –. Mouton de Gruyter, Berlin.

Ross, John R (). Constraints on Variables in Syntax. Ph.D. thesis, MIT.

Sag, Ivan A. (). English relative clause constructions. Journal of Linguistics, ,–.

Sag, Ivan A. (a). English filler gap constructions. Language, , –.

Sag, Ivan A. (b). Feature geometry and predictions of locality. In Features:Perspectives on a Key Notion in Linguistics (ed. A. Kibort and G. G. Corbett), pp.–. Oxford University Press, Oxford.

Sag, Ivan A. (). Sign-based Construction Grammar: An informal synopsis. InSign-Based Construction Grammar (ed. I. A. Sag and H. Boas). CSLI Publications,Stanford.

Sag, Ivan A. and Fodor, Janet D. (). Extraction without traces. In Proceedingsof the Thirteenth West Coast Conference on Formal Linguistics (ed. R. Aranovich,W. Byrne, S. Preuss, and M. Senturia), Stanford University. CSLI.

Sag, Ivan A., Gazdar, Gerald, Wasow, Thomas, and Weisler, Steven (). Co-ordination and how to distinguish categories. Natural Language and LinguisticTheory, , –.

Sag, Ivan A. and Pollard, Carl (). An integrated theory of complement control.Language, , –.

Sag, Ivan A., Wasow, Thomas, and Bender, Emily (). Syntactic Theory: AFormal Introduction (nd edn). CSLI Publications, Stanford.

Sampson, Georey R. (). What was transformational grammar? Lingua, ,–.

Scholz, Barbara C. and Pullum, Georey K (). Tracking the origins of trans-formational generative grammar. Journal of Linguistics, , –.

Shieber, StuartM (). Evidence against the context-freeness of natural language.Linguistics and Philosophy, , –.

Shieber, Stuart M. (). An Introduction to Unification-based Approaches toGrammar. CSLI, Stanford.

Steedman, Mark (). The Syntactic Process. MIT Press, Cambridge, MA.

Steedman, Mark and Baldridge, Jason (). Combinatory categorial grammar.See Borsley and Börjars ().

Stockwell, Robert P. (). Discussion of ‘A transformational approach to syntax’.In Third Texas Conference on Problems of Linguistic Analysis in English (ed. A. A.Hill), Austin, pp. –. University of Texas.

Wechsler, Stephen M. and Noh, Bokyung (). On resultative predicates andclauses: Parallels between Korean and English. Language Sciences, , –.

Wells, Rulon S (). Immediate constituents. Language, , –.

Williams, Edwin (). Across-the-board rule application. Linguistic Inquiry, ,–.