Complementation by construction

Laura Michaelis, University of Colorado, Boulder

Where does a verb's frame come from? An emerging consensus is that it comes from 'top down' scene construal rather than 'bottom up' lexical projection (Goldberg 1995, Rappaport Hovav & Levin 1998, Partee & Borschev 2004). For example, as shown in (1), monovalent activity verbs can be accommodated by the 'locative inversion' pattern, resulting in an 'overlay' of the locative-theme presentational frame (Bresnan 1994):

(1) In Maria’s sticky hand melted a chocolate-chip ice-cream cone.

(Birner and Ward 1998:193)

Some linguists have argued that patterns like that in (1) are not merely construals but constructions, in particular argument-structure constructions. Because such constructions denote situation types, it makes sense that a verb's meaning and combinatory potential might change to fit the meaning of a given construction (Goldberg 1995, 2002, Michaelis & Ruppenhofer 2001). The problem is that phrasal patterns are not supposed to denote anything: they combine symbols rather than being symbols. Therefore, it would seem preferable to view valence augmentation and other construal-based semantic effects on verbs as the products of lexical derivations, as per Rappaport Hovav & Levin 1998 [RHL]. I will argue, however, that there are a number of compelling reasons to retain argument licensing by constructions. First, the full range of verb-valence variability, including null complementation, cannot feasibly be described by augmentative operations on event structure of the type described by RHL. Second, many verb frames specify syntactic sisterhood relations that would not be predicted from the general-purpose rule that combines heads and complements. Third, quantificational behaviors of argument NPs, including constraints on quantifier scope in certain argument structures and nominal type coercion in argument positions, require constructions: constructions have semantic frames, including quantifier frames, while Aktionsart templates do not. Fourth, certain constraints on argument interpretation and realization make sense only if viewed as the effects of one argument-structure construction overriding another (Zwicky 1994). In the course of discussing the four points above, I will outline a formal model of constructions that is currently being developed by Fillmore, Kay, Sag & Michaelis (in prep). According to this model, the basic object of grammatical description is the sign, a pairing of form and meaning that may represent a word, lexeme or phrase.