Alternating Quantifier Scope in CCG* Mark Steedman Division of Informatics, University of Edinburgh, 2 Buccleuch Place, Edinburgh EH8 9LW, UK steedman@cogsc i. ed. ac. uk Abstract equivalent computational operations of "quantify- The paper shows that movement or equivalent ing in" or "storage" at the level of logical form. computational structure-changing operations of any However, such accounts present a problem for kind at the level of logical form can be dispensed monostratal and monotonic theories of grammar with entirely in capturing quantifer scope ambi- like CCG that try to do away with movement or guity. It offers a new semantics whereby the ef- the equivalent in syntax. Having eliminated non- fects of quantifier scope alternation can be obtained monotonic operations from the syntax, to have to by an entirely monotonic derivation, without type- restore them at the level of logical form would be changing rules. The paper follows Fodor (1982), dismaying, given the strong assumptions of trans- Fodor and Sag (1982), and Park (1995, 1996) in parency between syntax and semantics from which viewing many apparent scope ambiguities as arising the monotonic theories begin. Given the assump- from referential categories rather than true general- tions of syntactic/semantic transparency and mono- ized quantitiers. tonicity that are usual in the Frege-Montague tra- dition, it is tempting to try to use nothing but the 1 Introduction derivational combinatorics of surface grammar to deliver all the readings for ambiguous sentences like It is standard to assume that the ambiguity of sen- (1). Two ways to restore monotonicity have been tences like (1) is to be accounted for by assigning proposed, namely: enriching the notion of deriva- two logical forms which differ in the scopes as- tion via type-changing operations; or enriching the signed to these quantifiers, as in (2a,b): 1 lexicon and the semantic ontology. (1) Every boy admires some saxophonist. It is standard in the Frege-Montague tradition to (2) a. Vx.boy' x -+ 3y.saxophonis/ y A admires' yx begin by translating expressions like "every boy" b. 3y.saxophonis/ y A Vx.bo/x -+ admires'yx and "some saxophonist" into "generalized quanti- tiers" in effect exchanging the roles of arguments The question then arises of how a grammar/parser like NPs and functors like verbs by a process of can assign all and only the correct interpretations to "type-raising" the former. In terms of the notation sentences with multiple quantifiers. and assumptions of Combinatory Categorial Gram- This process has on occasion been explained mar (CCG, Steedman 1996) the standard way to in- in terms of "quantifier movement" or essentially corporate generalized quantifiers into the semantics * Early versions of this paper were presented to audiences at of CG deterbainers is to transfer type-raising to the Brown U., NYU, and Karlov2~ U. Prague. Thanks to Jason lexicon, assig~g the following categories to deter- Baldridge, Gann Bierner, Tim Fernando, Kit Fine, Polly Ja- miners like every and some, making them functions cobson, Mark Johnson, Aravind Joshi, Richard Kayne,Shalom from nouns to "type-raised" noun-phrases, where Lappin, Alex Lascarides, Suresh Manandhar, Jaruslav Peregrin, the latter are simply the syntactic types correspond- Jong Park, Anna Szabolcsi, Bonnie Webber, AlistairWillis, and the referees for helpful comments. The work was supported in ing to a generalized quantifier: part by ESRC grant M423284002. (3) every := (T/(T\NP))/N : ~,p,~l.Vx.px -+ qx tThe notation uses juxtapositionfa to indicate application of a functor f to an argument a. Constants are distinguished every := (T\(T/NP))/N : kp.kq.Vx.px --+ qx from variables by a prime, and semantic functors like admires' (4) some := (T/(T\UP))/U:~,p.~l.3x.pxAqx are assumed to be "Curried". A convention of "left associativi- some := (T\(T/NP))/N:Lp.~l.3x.pxAqx ty" is assumed, so that admires'yx is equivalentto (admires'y)x. 301 (T is a variable over categories unique to each in- but narrow scope with respect to every girl that dividual occurrence of the raised categories (3) and is, where the boys all admire John Coltrane, but (4), abbreviating a finite number of different raised the girls all detest possibly different saxophonists. types. We will distinguish such distinct variables as There does not even seem to be a reading involving T, T', as necessary.) separate wide-scope saxophonists respectively tak- Because CCG adds rules of function composition ing scope over boys and girls--for example where to the rules of functional application that are stan- the boys all admire Coltrane and the girls all detest dard in pure Categorial Grammar, the further in- Lester Young. clusion of type-raised arguments engenders deriva- These observations are very hard to reconcile tions in which objects command subjects, as well as with semantic theories that invoke powerful mech- more traditional ones in which the reverse is true. anisms like abstraction or "Quantifying In" and its Given the categories in (3) and (4), these alterna- relatives, or "Quantifier Movement." For example, tive derivations will deliver the two distinct logi- if quantifiers are mapped from syntactic levels to cal forms shown in (2), entirely monotonically and canonical subject, object etc. position at predicate- without involving structure-changing operations. argument structure in both conjuncts in (5b), and However, linking derivation and scope as simply then migrate up the logical form to take either wide and directly as this makes the obviously false pre- or narrow scope, then it is not clear why some saxo- diction that in sentences where there is no ambi- phonist should have to take the same scope in both guity of CCG derivation there should be no scope conjuncts. The same applies if quantifiers are gener- ambiguity. In particular, object topicalization and ated in situ, then lowered to their surface position. 2 object right node raising are derivationally unam- Related observations led Partee and Rooth biguous in the relevant respects, and force the dis- (1983), and others to propose considerably more placed object to command the rest of the sentence general use of type-changing operations than are in derivational terms. So they should only have the required in CCG, engendering considerably more wide scope reading of the object quantifier. This is flexibility in derivation that seems to be required by not the case: the purely syntactic phenomena that have motivated CCG up till now. 3 (5) a. Some saxophonist, every boy admires. While the tactic of including such order- b. Every boy admires, and every girl detests, preserving type-changing operations in the gram- some saxophonist. mar remains a valid alternative for a monotonic Both sentences have a narrow scope reading in treatment of scope alternation in CCG and related which every individual has some attitude towards forms of categorial grammar, there is no doubt that some saxophonist, but not necessarily the same sax- it complicates the theory considerably. The type- ophonist. This observation appears to imply that changing operations necessarily engender infinite even the relatively free notion of derivation provided sets of categories for each word, requiring heuris- by CCG is still too restricted to explain all ambigu- tics based on (partial) orderings on the operations ities arising from multiple quantifiers. concerned, and raising questions about complete- Nevertheless, the idea that semantic quantifier ness and practical parsability. All of these ques- scope is limited by syntactic derivational scope has tions have been addressed by Hendriks and others, some very attractive features. For example, it imme- but the result has been to dramatically raise the ratio diately explains why scope alternation is both un- of mathematical proofs to sentences analyzed. bounded and sensitive to island constraints. There It seems worth exploring an alternative response is a further property of sentence (5b) which was to these observations concerning interactions of sur- first observed by Geach (1972), and which makes it seem as though scope phenomena are strongly re- 2Such observations have been countered by the invocation stricted by surface grammar. While the sentence has of a "parallelism condition" on coordinate sentences, a rule of one reading where all of the boys and girls have a very expressively powerful "transderivational" kind that one strong feelings toward the same saxophonist--say, would otherwise wish to avoid. John Coltrane--and another reading where their 3For example, in order to obtain the narrow scope object reading for sentence (5b), Hendriks (1993), subjects the cate- feelings are all directed at possibly different saxo- gory of the transitive verb to "argument lifting" to make it a phonists, it does not have a reading where the sax- function over a type-raised object type, and the coordination ophonist has wide scope with respect to every boy, rule must be correspondingly semantically generalized. 302 face structure and scope-taking. The present paper under any of the alternatives discussed so far. follows Fodor (1982), Fodor and Sag (1982), and Park (1995, 1996) in explaining scope ambiguities 2 Donkeys as Skolem Terms in terms of a distinction between true generalized One example of an indefinite that is probably better quantifiers and other purely referential categories. analyzed as an arbitrary object than as a quantified For example, in order to capture the narrow-scope NP occurs in the following famous sentence, first object reading for Geach's right node raised sen- brought to modern attention by Geach (1962): tence (5b), in whose CCG derivation the object must command everything else, the present paper fol- (6) Every farmer who owns a donkey/beats it/. lows Park in assuming that the narrow scope read- The pronoun looks as though it might be a variable ing arises from a non-quantificational interpretation bound by an existential quantifier associated with a of some scecophonist, one which gives rise to a read- donkey.