Combinatory Linguistics


Bozşahin H. C.

Mouton de Gruyter , Berlin, 2012

  • Publication Type: Book / Research Book
  • Publication Date: 2012
  • Publisher: Mouton de Gruyter
  • City: Berlin

Abstract

This book is about the place and role of combinators in linguistics, and through it, in cognitive science, computational linguistics and philosophy. It traces the history of Combinatory Categorial Grammar (CCG) and presents its linguistic implications. It aims to show how combinatory theories and models can be built, evaluated and situated in the realm of the four fields. The introductory remarks in the beginnings of early chapters can hopefully be excused because of the wide target readership.

The book examines to what extent knowledge of words can be construed as the knowledge of language, and what that knowledge might look like, at least on paper. It studies the semantic mechanism that engenders directly in- terpretable constituents, the combinators, and their limits in a grammar. More specifically, it investigates the mediating relation between constituents and their semantics insofar as it arises from combinatory knowledge of words and syntacticized combinators. It is not about forms or meanings per se.

Its key aspect is to promote the following question as a basic scientific inquiry of language: why do we see limited dependency and constituency in natural language syntax? We owe the question to David Hume by a series of links, some of which are covered in the book. The reader might be puzzled by this claim, knowing that Hume had said very little about language. I believe it was for a good reason, but the question goes back to him nevertheless, as I try to argue in the book.

It seems that thinking syntax is syntax and semantics is semantics in their own structure isn’t going to take us too far from the knowledge we have ac- cumulated on grammars, about what they can and cannot do regarding code- terminism in forms and meanings, and about the coconspiracy of forms and meanings. The same goes, I am sure, to thinking discourse is discourse, mor- phology is morphology etc. The book focuses on the relationship between syntax and semantics.

Many explanans about syntactic processes become explananda when we readjust our semantic radar, a term which I use as a metaphor for looking at semantic objects with a syntactic eye. As all metaphors are, it is somewhat misleading in the beginning, which I hope becomes less of a metaphor as we proceed. If we open the radar too wide, we are forced to do syntax with semantic types, and run the risk of missing the intricate and complex syntactic dependencies, which in turn might miss an opportunity to limit “possible human languages”. If it is too narrow, we must do semantics with syntac- tic types, and that might take us to the point of having syntaxes rather than syntax. Both extremes need auxiliary assumptions to provide a constrained theory of language.

Many syntactic dependencies turn out to be semantic in nature, and these dependencies seem to arise from a single resource. This resource is conjec- tured to be adjacency. The conjecture of semantics arising from order goes back about a century in mathematics, to Schönfinkel, and almost half a cen- tury in philosophy, linguistics and cognitive science, to Geach, Ades and Steedman. The natural meeting point of the two historically independently motivated theorizing about adjacency, the semantic and the syntactic one about combinators, is the main story of the book.

In this regard, the book was bound to be a historical account from the beginning. However, it came to provide, in some detail, ways of theory and model construction for linguistics and cognitive science in which there is no degree of freedom from adjacency. This pertinacious course seems to set up the crucial link between forms and meanings with as little auxiliary as- sumptions as its progenitors can think of. I believe it sets up creative links in theorizing about the computational, linguistic, cognitive and philosophical aspects of grammar. I exemplify these connections one by one.

When we look at combinators as functions they are too powerful, equiv- alent to the power of a Turing machine. As such they cannot do linguistic work because natural language constituency narrows down the expressible semantic dependencies manifested by functions. The linguistic theorizing be- gins when we syntacticize the combinators and establish some criteria about which combinator must be in the grammar and which one can materialize in the lexicon. An explanatory force can be reached if the process reveals predic- tions about possible constituents, possible grammars and possible lexicons, without the need for variables and within a limited computational power. Structure-dependence of natural language strings can be predicted too, rather than assumed.

Every intermediate constituent will be immediately interpretable, and non- constituents will be uninterpretable by this process. In other words, being a constituent, being derivable and being immediately interpretable are three manifestations of the same property: natural grammars are combinatory type- dependent. These are the narrow claims of Combinatory Categorial Grammar.

The notion of grammar is trivialized if there is no semantics in it. Some, like Montague, went as far as claiming that syntax is only a preliminary for semantics. On the other hand, language would not be worth theorizing about if the semantics we have in mind for grammars is the semantics out there. All species do this all the time without language, to mean things as they see fit.

1

Words would be very unnecessary, as one songwriter put it in the early 90’s.

2

Although I have tried to keep it to a minimum to compare the present the- ory with others, for the sake of brevity and focus, the historical perspective in the book makes unavoidable points of contact with different ways of theo- rizing about grammars. Some examples are worth noting from the beginning. (a) Steedman’s and Jacobson’s use of combinators for syntax differs when it comes to reference and quantifier scope. (b) Kayne claims that structure determines order, with directionally-constrained syntactic movement being the key element in explanations. Order determines structure in the combina- tory theory, and no-movement is the key to explanations. (c) HPSG is another type-dependent theory of syntax like the one presented in the book. HPSG’s types are related to each other by subtyping, whose semantics do not nec- essarily arise from order. (d) Type-logical grammar in particular and Mon- tague’s use of type-theoretic language in general use semantic types to give rise to meaningful expressions, that is, to syntax. Order is not necessary or sufficient for a set-based type’s construal, therefore it need not be the basis for meaningful expressions. (e) Obviously not all categorial grammars are combinatory categorial grammars. The telltale signs of the latter kind, which is the main topic of the book, are no use of phonologically null types, no use of surface wrap, some use of type combination that goes above function ap- plication, and the insistence on an order-induced syntax-semantics for every rule and lexical category, as opposed to for example order and structural unifi- cation. (f) Dependency grammars take dependency as an asymmetric relation of words in a string, i.e. as a semantic relation between syntactic objects, but leave open why there are limited kinds of dependencies, and why these de- pendencies relate to surface constituency and interpretability in predictable ways. (g) Chomsky’s program can be seen as a concerted effort to squeeze as much compositional semantics into syntax as possible. The A-over-A principle, the X-bar model, subjacency, cyclicity, filters, functional categories, main thematic condition, chains, crash and the process of derivation-by-phase do to syntactic trees what they cannot do by themselves: constrain the possible semantic interpretations of the syntactic objects in them hypergrammatically.

The apparent similarities of these theories must be put in context. As Pol- lard points out frequently, most theories subscribe to some form of syntac- tocentrism because they conceive the relation between forms and meanings as indirect. It must be mediated by syntax. The theory covered in the book is syntactocentric in Pollard’s sense. The syntactocentrism that will be argued against here is the one that sees semantics as an appendix to syntax. The theory presented here is neither the first nor the only remaining one on this stance.

We need only look thirtysomething years before the rise of that kind syn- tactocentrism to find an alternative foundation for bringing semantics into syntax. Two historically independent programs, radical lexicalization and codeterminacy of syntax and semantics, culminate in a theory where adja- cency is the only fundamental assumption. Two aspects will figure promi- nently: dependency and constituency. Both will get their interpretation from a single source, the semantics of order.

For the reader the book is organized in such a way that the technical ma- terial that gets in the way of linguistics have been moved to appendices. This leaves some aspects of combinators, grammars and computing to the appen- dices (mostly definitions and basic techniques). Linguistic theorizing about the combinators is in the main text. There is no reference to the appendices from the main matter, or from appendices to the chapters in the main body. The back matter is organized such that the latter ones might refer to earlier ones. Reading all the appendices in the given order might help readers who are unfamiliar with some of the terminology.

Now to pay some debts old and new academic and personal. This book started as my I-see-what-they-mean project, although I am not sure about the end result. It was an attempt at a collective understanding of Moses Schön- finkel, Mark Steedman, Anna Szabolcsi, Pauline Jacobson, Noam Chom- sky, Richard Montague, Haskell Curry, Emmon Bach and John Robert Ross, among others. I hope I’ve done them justice.

At a more personal level, my first contact with Mark and his theory was in the years 1992–1994, and since then it has become a major part of my aca- demic life. I have asked so many questions to Mark (not to mention numerous requests for judgments on difficult English examples) that I am slightly embarrassed I am getting away with an acknowledgment. Before then I was for- tunate to be taught by great teachers, whom I’m honored to list in somewhat chronological order: Türkân Barkın, Metin Ünver, ̇Ibrahim Nis ̧ancı, late Esen Özkarahan, Nicholas Findler and Leonard ‘Aryeh’ Faltz. Some friends and family taught me more on academic affairs than I was able to acknowledge so far. There is a bit of them in the book but I cannot exactly point where. Thank you Canus ̧, née Cihan Bozs ̧ahin, Nezih Aytaçlar, Zafer Aracagök, Ug ̆ur Atak, Ragıp Gürkan, Justin Coven, Uttam Sengupta, Halit Og ̆uztüzün, Samet Bag ̆çe, Sevil Kıvan, Aynur Demirdirek, Mark Ellison and Stuart Al- lardyce.

Mark Steedman, Ash Asudeh and Frederick Hoyt provided comments on much earlier drafts. Umut Özge was less fortunate to have gone through sev- eral drafts. I owe some sections to discussions with him, and with Ceyhan Temürcü, Mark Steedman and Aravind Joshi. Elif Gök, Yag ̆mur Sag ̆, Süley- man Tas ̧çı, Deniz Zeyrek and Alan Libert suggested corrections and clarifi- cations for which I am grateful. The Mouton team, Uri Tadmor, Birgit Sievert and Julie Miess, have been encouraging and supportive from the very start. Livia Kortvelyessy of Versita helped me get the project going.

I am solely responsible for not heeding good advice of so many good people.