Search

A novel logic-based framework for representing the syntax–semantics interface of natural language, applicable to a range of phenomena. In this book, Yusuke Kubota and Robert Levine propose a type-logical version of categorial grammar as a viable alternative model of natural language syntax and semantics. They show that this novel logic-based framework is applicable to a range of phenomena—especially in the domains of coordination and ellipsis—that have proven problematic for traditional approaches. The type-logical syntax the authors propose takes derivations of natural language sentences to be proofs in a particular kind of logic governing the way words and phrases are combined. This logic builds on and unifies two deductive systems from the tradition of categorial grammar; the resulting system, Hybrid Type-Logical Categorial Grammar (Hybrid TLCG) enables comprehensive approaches to coordination (gapping, dependent cluster coordination, and right-node raising) and ellipsis (VP ellipsis, pseudogapping, and extraction/ellipsis interaction). It captures a number of intricate patterns of interaction between scopal operators and seemingly incomplete constituents that are frequently found in these two empirical domains. Kubota and Levine show that the hybrid calculus underlying their framework incorporates key analytic ideas from competing approaches in the generative syntax literature to offer a unified and systematic treatment of data that have posed considerable difficulties for previous accounts. Their account demonstrates that logic is a powerful tool for analyzing the deeper principles underlying the syntax and semantics of natural language. The open access edition of this book was made possible by generous funding from Arcadia – a charitable fund of Lisbet Rausing and Peter Baldwin.

In this paper, we propose an analysis of pseudogapping in Hybrid Type-Logical Categorial Grammar (Hybrid TLCG; Kubota 2010,Kubota and Levine 2012). Pseudogapping poses a particularly challenging problem for previous analyses in both the transformational and nontransformational literature. The flexible notion of constituency countenanced in categorial grammar in general (including Hybrid TLCG) enables a simple analysis of pseudogapping that overcomes the problems for previous approaches. In addition, we show that Hybrid TLCG offers a useful platform for comparing different types of CG (and analytic ideas embodied in them) within a single framework with a linguist-friendly notation. In the context of an analysis of pseudogapping, this enables us to compare the adequacies of two types of treatments of discontinuous pseudogapping. We believe that this type of cross-framework comparison of different analytic ideas is useful for the purpose of testing the predictions of different types of analyses that can respectively be formulated in only one or the other of the more conservative types of CG.

This paper proposes an analysis of modal auxiliaries in English in Type-Logical Grammar. The proposed analysis captures the scopal interactions between different types of modal auxiliaries and negation by incorporating the key analytic idea of Iatridou and Zeijlstra [6], who classify English modal auxiliaries into PPI and NPI types. In order to technically implement this analysis, we build on Kubota and Levine’s [8, 10] treatment of modal auxiliaries as higher-order operators that take scope at the clausal level. The proposed extension of the Kubota/Levine analysis is shown to have several interesting consequences, including a formal derivability relation from the higher-order entry for auxiliaries to a lower-order VP/VP entry traditionally recognized in categorial grammar (CG) research. The systematic analysis of the scopal properties of auxiliaries and the somewhat more abstract meta-comparison between ‘transformational’ and ‘non-transformational’ analytic ideas that become possible in a type-logical setup highlight the value of taking a logical perspective on the syntax of natural language embodied in Type-Logical Grammar research.

National Institute for Japanese Language and Linguistics, Ochanomizu University, Ohio State University, The scope parallelism in the Geach sentence (Every boy loves, and every girl detests, some saxophonist) and the related parallel interpretation requirement in pronominal binding is a pervasive phenomenon found across different types of coordination and ellipsis phenomena. Previous accounts all resort to additional constraints of some sort that restrict the otherwise flexible syntax-semanticsinterface to avoid overgeneration. In this paper, we propose a novel approach to this long-standing problem. We show that, by taking a proof-theoretic perspective on natural language semantics and by viewing the ambiguity resolution for pronouns and indefinites as underspecification resolution thatinvokes extra proof search, a conceptually natural solution emerges for the problem of interpretive parallelism. The analysis is cast in Dependent Type Semantics, with Hybrid Type-Logical Categorial Grammar as the syntax-semantics interface backbone. For empirical illustration, we show how the proposed approach correctly accounts for the classical Geach paradigm and its pronominal variant.

Introduction: The Problem, and a Possible Solution Natural Language Data Linguists appear to be in an enviable position among scientific disciplines. The lifeblood of science is data, and unlike, say, glaciologists, who can only collect primary material for their research in remote and generally rather inhospitable parts of the planet, or particle physicists, who require access to massive, extremely expensive, and sometimes erratic machines – with demand for access to such machines far exceeding supply – linguists are literally surrounded by the kind of data that make up the target of their investigations. It's true that field linguists need informants at a considerable distance from where they themselves live, and experimental linguists often need laboratories with elaborate and sophisticated equipment. But for syntacticians – linguists who investigate the structure of sentences, a large fraction of whom (possibly a majority) study sentences in their own respective languages – matters are as convenient as they could possibly be. Syntacticians have intuitive access to all of the sentences made available by their own knowledge of their language, as well as the speech (and reactions) of their fellows in constant use around them, and ready-made corpora in the form of written materials and electronic records, many of which are available for searches based on word sequences (Google, for example, is a valuable source of data for both syntacticians and morphologists). Learning how to take advantage of this vast pool of readily available data is a major component of syntacticians’ training. In a sense, of course, the true data of syntax are not strings of words themselves, but judgments about the status of those strings of words. The syntactician's primary responsibility is to give an account of how it is that certain strings of words have the status of sentences, while other do not, and still others have a kind of shadowy intermediate status – not bad enough to be outright rubbish, but not good enough to pass completely unnoticed in conversation as utterly and tediously normal. For example, consider the status of the three word strings in (1): (1) a. I asked Robin to leave the room. b. I requested Robin to leave the room. c. I inquired (of) Robin to leave the room.

Categorial grammar is well-known for its elegant analysis of coordination enabled by the flexible notion of constituency it entertains. However, to date, no systematic study exists that examines whether this analysis has any obvious empirical advantage over alternative analyses of nonconstituent coordination available in phrase structure-based theories of syntax. This paper attempts precisely such a comparison. We compare the direct constituent coordination analysis of non-canonical coordinations (right-node raising, dependent cluster coordination and Gapping) in categorial grammar with an ellipsis-based analysis of the same phenomena in the recent HPSG literature. We provide a set of empirical evidence, consisting of cases in which non-canonical coordinations interact with scopal operators of various sorts, which systematically falsifies the predictions of the latter, ‘linearization-based’ ellipsis approach to coordination. We propose an alternative analysis in a variant of categorial grammar called Hybrid Type-Logical Categorial Grammar. The proposed framework builds on both the Lambek-inspired variants of categorial grammar and a more recent line of work modelling word order via a lambda calculus for the prosodic component. The flexible syntax–semantics interface of this framework straightforwardly captures the interactions between non-canonical coordinations and scopal expressions, demonstrating the broader empirical payoff of the direct constituent coordination analysis of non-canonical coordinations pioneered by Steedman (Language 61(3):523–568, 1985; Linguist Philos 13(2):207–263, 1990) and Dowty (Categorial grammars and natural language structures, 1988) hitherto not explicitly recognized in the literature.

This paper proposes a unified analysis of the ‘respective’ readings of plural and conjoined expressions, the internal readings of symmetrical predicates such as same and different, and the summative readings of expressions such as a total of $10000. These expressions pose significant challenges to compositional semantics, and have been studied extensively in the literature. However, almost all previous studies focus exclusively on one of these phenomena, and the close parallels and interactions that they exhibit have been mostly overlooked to date. We point out two key properties common to these phenomena: (i) they target all types of coordination, including nonconstituent coordination such as Right-Node Raising and Dependent Cluster Coordination; (ii) the three phenomena all exhibit multiple dependency, both by themselves and with respect to each other. These two parallels suggest that one and the same mechanism is at the core of their semantics. Building on this intuition, we propose a unified analysis of these phenomena, in which the meanings of expressions involving coordination are formally modelled as multisets, that is, sets that allow for duplicate occurrences of identical elements. The analysis is couched in Hybrid Type-Logical Categorial Grammar. The flexible syntax-semantics interface of this framework enables an analysis of ‘respective’ readings and related phenomena which, for the first time in the literature, yields a simple and principled solution for both the interactions with nonconstituent coordination and the multiple dependency noted above.

The scope anomaly observed in sentences like Mrs. J can’t live in Boston and Mr. J in LA (¬◊>∧) and No dog eats Whiskas or cat Alpo (¬∃>∨) is known to pose difficult challenges to many analyses of Gapping. We provide new arguments, based on both the basic syntactic patterns of Gapping and standard constituency tests, that the so-called ‘low VP coordination analysis’—the only extant analysis of Gapping in contemporary syntactic theories which accounts for this scope anomaly—is empirically untenable. We propose an explicit alternative analysis of Gapping in Hybrid Type-Logical Categorial Grammar, a variant of categorial grammar which builds on both the Lambek-inspired tradition and a more recent line of work modelling word order via a lambda calculus for the prosodic component. The flexible syntax-semantics interface of this framework enables us to characterize Gapping as an instance of like-category coordination, via a crucial use of the notion of hypothetical reasoning. This analysis of the basic syntax of Gapping is shown to interact with independently motivated analyses of scopal operators to immediately yield their apparently anomalous scopal properties in Gapping, offering, for the first time in the literature, a conceptually simple and empirically adequate solution for the notorious scope anomaly in Gapping.

The scope parallelism in the so-called Geach sentences in right-node raising (Every boy admires, and every girl detests, some saxophonist) poses a difficult challenge to many analyses of right-node raising, including ones formulated in the type-logical variants of categorial grammar (e.g. Kubota and Levine (2015)). In this paper, we first discuss Steedman’s (2012) solution to this problem in Combinatory Categorial Grammar, and point out some empirical problems for it. We then propose a novel analysis of the Geach problem within Hybrid Type-Logical Categorial Grammar (Kubota and Levine 2015), by incorporating Dependent Type Semantics (Bekki 2014) as the semantic component of the theory. The key solution for the puzzle consists in linking quantifiers to the argument positions that they correspond to via an anaphoric process. Independently motivated mechanisms for anaphora resolution in DTS then automatically predicts the scope parallelism in Geach sentences as a consequence of binding parallelism independently observed in right-node raising sentences.

In a 1982 paper in the journal Glossa, Pullum outlined a set of arguments for treating English infinitival to as a defective auxiliary verb. Twenty years later, in his entry on infinitival constructions in The Cambridge Grammar of the English Language (CGEL, 2002), Huddleston argues that several distributional facts about auxiliaries fit poorly with this hypothesis. He proposes, on the basis of significant structural parallels, that to is a subordinator (complementizer). I show that Huddleston's arguments constitute a flawed analysis in CGEL's otherwise superb coverage of English descriptive grammar, and that the facts run strongly counter to his claims, often falling out independently from generalizations about auxiliaries that Huddleston overlooks. Several of these points were anticipated in Pullum's paper, but recent research on an idiosyncratic auxiliary-specific pattern of English nonrestrictive relative clause formation provides a powerful new argument in support of the auxiliary claim. In this respect, as in all others, the assignment of to to the class of auxiliaries provides the simplest and broadest account of its syntactic behavior.

Collins et al. 2008 offers a principles-and-parameters-based analysis of an AAVE construction first described in Spears 1998, in which nominal phrases such as John's ass appear to have exactly the same denotation, and behavior with respect to familiar conditions on anaphora, as the possessor [ John , and similarly for pronominal possessors. Agreement, however, reflects not the properties of the possessor, but of the possessed nominal ass , which belongs to a small, closed class of lexical items that behave in parallel fashion and which the authors call 'mask' nominals. Collins and colleagues convincingly argue that the class of NPs consisting of possessors attached to mask nominals have the same syntactic structure as ordinary NPs displaying (pro)nominal possessors. In order to account for the split between anaphora and agreement, however, they are apparently forced to invoke a very complex derivational mechanism that includes a lowering rule, along with a number of other highly stipulative components, in order to encompass certain related constructions. I offer a far simpler and empirically more comprehensive alternative treatment in which mask nominals are nothing more than semantically parasitic heads, based on Kathol's (1999) dichotomy between AGR(eement) and INDEX specifications within head-driven phrase structure grammar representations. Collins and colleagues adduce what they take to be empirical arguments against such an approach, but these arguments are, as I show, all predicated on a basic technical misinterpretation of the nature of indices in the HPSG syntax/semantics interface, and thus have no force. Comparison of the two approaches is interesting not only in the context of the phenomenon described by Spears, but also in terms of broader, cross-framework issues-in particular, the question of whether or not movement and feature matching are merely two alternative, interconvertible ways of expressing linkages between structurally distant categories.

Extraction is treated in Head-driven Phrase Structure Grammar (HPSG) as a series of linked identities, enforced by a small number of interacting constraints, in the specification of a syntactic feature, slash. HPSG's monostratal formalism ensures that this strictly local determination of slash identities connects the relevant morphosyntactic properties of the filler to those imposed by conditions at the gap site. This connectivity mechanism operates in precisely the same manner whether the filler is linked to a single gap site or several, thus yield strictly parallel treatments for single and multiple gap phenomena, respectively. The inherently local nature of HPSG's connectivity mechanism is supported by data from languages in which all clausal levels in filler/gap pathways are flagged by grammatical effects restricted to such pathways. Interruption in connectivity, familiar in syntax as ‘island’ effects, are regarded as outside the formal combinatorics of the syntax itself, as argued in an increasingly broad and deep literature documenting the origins of these effects in real-time processing, pragmatic and other functional aspects of linguistic cognition. And this simple characterization of extraction connectivity is compatible with a review of arguments that purport to argue for a heterogeneous characterization of extraction constructions, but which prove to lack empirical motivation. Nonetheless, despite a homogeneous connectivity mechanism, extraction constructions differ from each other idiosyncratically in many respects, and these idiosyncrasies are the subject of much ongoing research in HPSG.

We argue that the phenomenon described and discussed in the literature as locative or stylistic inversion in English is actually a conflation of twoquite different constructions: on the one hand, light inversion (LI), inwhich the postverbal NP element can be phonologically and structurallyextremely simple, possibly consisting of a single name, and on the otherhand heavy inversion (HI), where the postverbal element is heavy in thesense of Heavy NP Shift.1 We present evidence that the preverbal PP in LI patterns with subjects but the PP in HI is a syntactic topic, usinga variety of tests which distinguish A-positions from Ā-positions.Other significant differences between HI and LI, such as the classesof verbs which support these two constructions respectively, and thedifferential behavior of HI and LI with respect to adverbial placement,provide support for interpreting HI as a case of Heavy NP Shiftapplying to subject constituents.

Once again we would like to thank our critics for replying to our comments and continuing the discussion of the issues that we raised in our original Topic-Comment piece. We feel that this debate is important as it touches on fundamental questions concerning the scientific status of linguistic theories and the way in which they are accepted. We will take up their main points in turn and then offer some concluding remarks.

Dans un article precedent, les As. ont critique les fondements theoriques et methodologiques du programme minimaliste en s'insurgeant contre le fait que les postulats et les principes de ce paradigme theorique aient ete aussi facilement adoptes par la communaute scientifique. Cet article a suscite des reactions de la part de A. Holmberg, E. Reuland, I. Roberts, M. Piattelli-Palmarini et J. Uriagereka, qui ont fait l'objet de publications dans le present numero. Les As. font ici le bilan des critiques emises a propos de leur article et montrent dans quelle mesure elles revelent des contradictions conditionnees par une tendance a accepter, sans discussion critique, des arguments provenant d'une autorite

We propose a unified analysis of 'respective' readings of plural and conjoined expressions and the internal readings of symmetrical predicates such as same and different. The two problems have both been recognized as significant challenges in the literature of syntax and semantics, but so far there is no analysis which captures their close parallel via some uniform mechanism. In fact, the representative compositional analyses of the two phenomena in the current literature Gawron and Kehler 2004 G&K on 'respective' readings and Barker 2007 on symmetrical predicates look superficially quite different from each other, where one Barker employs a movement-like nonlocal mechanism for mediating the dependency between the relevant terms whereas the other G&K achieves a similar effect via a chain of local composition operations. In this paper, we first point out the parallels and interactions between the two phenomena that motivate a unified analysis. We then briefly review G&K's and Barker's analyses and show that the G&K-style analysis can be modelled by the Barker-style analysis once we formulate the relevant rules within an explicit syntax-semantics interface couched in a variant of Type-Logical Categorial Grammar called Hybrid TLCG. After clarifying the hitherto unnoticed formal relations between the Barker-style nonlocal modelling and the G&K-style local modelling by focusing on the analysis of 'respective' readings, we present our unified analysis of 'respective' readings and symmetrical predicates and show how their parallel behaviors and interactions can be systematically accounted for.

In this paper, we propose an analysis of pseudogapping in Hybrid Type-Logical Categorial Grammar (Hybrid TLCG; Kubota 2010,Kubota and Levine 2012). Pseudogapping poses a particularly challenging problem for previous analyses in both the transformational and nontransformational literature. The flexible notion of constituency countenanced in categorial grammar in general (including Hybrid TLCG) enables a simple analysis of pseudogapping that overcomes the problems for previous approaches. In addition, we show that Hybrid TLCG offers a useful platform for comparing different types of CG (and analytic ideas embodied in them) within a single framework with a linguist-friendly notation. In the context of an analysis of pseudogapping, this enables us to compare the adequacies of two types of treatments of discontinuous pseudogapping. We believe that this type of cross-framework comparison of different analytic ideas is useful for the purpose of testing the predictions of different types of analyses that can respectively be formulated in only one or the other of the more conservative types of CG.

This chapter examines the syntactic structure of Stylistic Inversion. It argues that the phenomenon discussed in the literature as Locative or Stylistic Inversion in English is actually a conflation of two quite different constructions: firstly, light inversion (LI), in which the postverbal NP element can be phonologically and structurally extremely simple, possibly consisting of a single name; and, secondly, heavy inversion (HI), where the postverbal element is heavy in the sense of heavy NP shift. It presents evidence that the preverbal PP in LI patterns with subjects but the PP in HI is a syntactic topic.

Head-driven Phrase Structure Grammar represents the fruits of a decade of research in feature-based syntactic theory. It belongs to a select company of linguistic monographs which are both monumental and intensely provocative, including among their number Chomsky (I965, I98I), Ross (1967) and Gazdar et al. (I985). Like other works in this elite group, Pollard & Sag's book addresses simultaneously a wide variety of grammatical phenomena and a set of fundamental questions about the nature of syntactic representations which have shaped the direction of theoretical work by investigators throughout the field, leading to hundreds of papers and new hypotheses, including, inevitably, challenges and counterproposals to the analyses proposed by the authors. There is, in addition, a second domain in which Head-driven Phrase Structure Grammar has had major impact in linguistics the field of computational implementations, where the framework proposed by the authors that gives their book its name has rapidly become the theory of choice, a fact particularly evident in the European computational scene. The book's organization can be understood in terms of the syntax/ semantics dichotomy, and, within the syntactic discussion that takes up the first half of the book, the application of the technology outlined in the first chapter to progressively less local grammatical dependencies, insofar as locality was understood in the early generative era. Thus, after the first chapter, in which the feature-logic formalism of the theory is lightly sketched and the major constraints which the theory comprises are stated, or at least alluded to, we are given in chapter 2 a discussion of agreement. Chapter 3 is explicitly devoted to issues of subcategorization and complementation.

We argue that an approach to discontinuous constituency via prosodic lambda binding initiated by Oehrle (1994) and adopted by some subsequent authors (de Groote, 2001; Muskens, 2003; Pollard, 2011) needs to recognize higher-order prosodic variables to provide a fully systematic treatment of two recalcitrant empirical phenomena exhibiting discontinuity, namely, split gapping involving determiners and comparative subdeletion. Once we admit such higher-order prosodic variables, straightforward analyses of these phenomena immediately emerge. We take this result to provide strong support for recognizing such higher-order prosodic variables in this type of approach. We also touch on the more general issue of alternative approaches to discontinuity in categorial grammar, and suggest that an approach that recognizes (possibly higher-order) prosodic functors like the one we propose here leads to a more principled treatment of certain interactions between phenomena exhibiting complex types of discontinuity than competing approaches.

In current linguistic theory, the theoretical status of adjunct extractions, as in for example How often do you think Robin sees Kim? is, somewhat surprisingly, an unresolved issue, with some investigators arguing that only arguments extract syntactically, entailing analyses of adverbial gaps via fundamentally different mechanisms from those posited for argument extraction. We adduce extensive evidence against such positions from a number of languages which exhibit morphological or syntactic phenomena which are sensitive to binding (extraction) domains and where this morphosyntactic flagging is present in instances of adjunct extraction as well as argument extraction. We also present language-internal arguments for the syntactic nature of adjunct extraction in English, including the coextensiveness of adjunct and argument extraction and their parallelism with respect to strong/weak crossover effects. Finally, we discuss the challenge which binding domain effects pose for accounts of adjunct extraction in various frameworks.

We propose a version of Type-Logical Categorial Grammar (TLCG) which combines the insights of standard TLCG (Morrill 1994, Moortgat 1997) in which directionality is handled in terms of forward and backward slashes, and more recent approaches in the CG literature which separate directionality-related reasoning from syntactic combinatorics by means of Ł-binding in the phonological component (Oehrle 1994, de Groote 2001, Muskens 2003). The proposed calculus recognizes both the directionality-sensitive modes of implication (/ and \) of the former and the directionality-insensitive mode of implication tied to phonological Ł-binding in the latter (which we notate here as |). Empirical support for the proposed system comes from the fact that it enables a straightforward treatment of Gapping, a phenomenon that has turned out to be extremely problematic in the syntactic literature including CG-based approaches.

Since Noam Chomsky's 1977 paper ‘On wh-movement’, syntactic theorists almost universally have assumed that tough/too/enough constructions are to be treated by the same formal mechanism as wh-extraction constructions, in spite of well-known syntactic divergences between the two types. We argue that these divergences reflect a real dichotomy between unbounded dependencies with fillers in A(rgument) positions, e.g. tough constructions, and those whose fillers occupy non-argument (A-bar) positions, e.g. topicalization. We first show that strong external evidence exists supporting the GPSG-internal prediction of full syntactic connectivity between Aposition fillers and their gap sites, but that this result seems contradicted by the existence of case conflict between the filler and gap. To overcome this contradiction, we introduce a new gap-licensing feature GAP, and show how a number of other divergences between A- and A-bar filler/gap constructions follow as a consequence.

Head-driven phrase structure grammar is a monostratal theory of natural language grammar, based on richly specified lexical descriptions which combine according to a small set of abstract combinatory principles stated as formulae in a constraint logic regulating, for the most part, the satisfaction of valence and other properties of syntactic heads. These constraints, applying locally, determine the flow of information, encoded as feature specifications, through arbitrarily complex syntactic representations, and capture all syntactic dependencies – both local and non-local – in elegant and compact form requiring no derivational apparatus. Keywords: syntactic form; propostional structure; hierarchy; syntactic dependancies; anaphora

This paper compares transformation-based and constraint-based treatments of unbounded filler-gap dependencies, the latter specifically as articulated in terms of HPSG, and argues, contrary to the commonly made allegations of notational variance , that there is purely empirical evidence that is consistent with only the constraint-based account. Recent proposals to deal with parasitic gaps in terms of null pronominals and empty operators are unable to account for the phenomenon of symbiotic gaps, the apparent case mismatches found in parasitic gap constructions, or (in general) for the well-known across-the-board effects within coordinate structures.

This book explores a wide variety of theoretically central issues in the framework of Head-Driven Phrase Structure Grammar (HPSG), a major theory of syntactic representation, particularly in the domain of natural language computation. HPSG is a strongly lexicon-driven theory, like several others on the scene, but unlike the others it also relies heavily on an explicit assignment of linguistic objects to membership in a hierarchically organised network of types, where constraints associated with any given type are inherited by all of its subtypes. This theoretical architecture allows HPSG considerable flexibility within the confines of a highly restrictive, mathematically explicit formalism, requiring no derivational machinery and invoking only a single level of syntactic representation. The separate chapters consider a variety of problematic phenomena in German, Japanese and English and suggest important extensions of, and revisions to, the picture of HPSG.

I propose to summarize a number of findings from recent fieldwork and analysis which suggest that the evidence offered in support of the "classical" Na-Dene hypothesis (i.e., as set up by Sapir in his 1915 statement) is spurious, and that there is currently no empirical basis for including Haida in the Na-Dene grouping.' Writers on the Na-Dene hypothesis have in the past relied heavily on the significance of supposedly shared grammatical features; hence much of the debate, so far as Haida is concerned, hinges on the accuracy of Swanton's 1911 grammar for the Handbook of American Indian Languages.2 This is unfortunate, because much of Swanton's analysis contains major flaws, for example, inaccurate segmentation of individual morphemes and confusion of homophonous suffixes. The principal short

Farkas (1986) argues that earlier work by Horvath on Hungarian clause structure appealing to constituent-lowering movement is incorrect, and that no such ‘downgrading’ movement exists in natural languages. Farkas then claims that while this lack of downgrading movement must be stipulated in the GB framework, it follows as a theorem from GPSG formalism. I argue here that the formal structure of GPSG is completely compatible with the existence of down-grading constructions. A rule system implementing such constructions is presented together with a compositional semantics which yields the appropriate semantic representations for such constructions. It follows that the possibility of lower filler/higher gap linkages is irrelevant to theory comparison between GB and GPSG, or indeed, between any pair of theories both of which include a phrase structure component.

0. Kwakwala, like all other members of the Wakashan family, is characterized by a morphology which is unusually rich and complex, even by Amerindian standards.' Several elements of this morphology have syntactic effects which correspond to passive constructions in English and other Indo-European languages. In attempting to construct an insightful description of these Kwakwala forms, we encounter certain types of difficulties which classically "polysynthetic" languages present to re