Sense and structure: Meaning as a determinant of verb subcategorization preferences

Readers are sensitive to the fact that verbs may allow multiple subcategorization frames that differ in their probability of occurrence. Although a verb s overall subcategorization preferences can be described probabilistically, underlying non-random factors may determine those probabilities. One potential factor is verb semantics: Many verbs show sense differences, and a verb s subcategorization profile can vary by sense. Thus, although find can occur with a direct object (DO) or a sentential complement (SC), when it is used to mean locate it occurs only with a DO, whereas in its realize sense it is SC-biased, but can take either frame. We used corpus analyses to identify verbs that occur with both frames, and found that their subcategorization probabilities differ by sense. Off-line sentence completions demonstrated that contexts can promote a specific sense of a verb, which subsequently influenced subcategorization probability. Finally, in a self-paced reading time experiment, verbs occurred in target sentences containing either a structurally unambiguous or ambiguous SC, following a context favoring the verb s DOor SC-biased sense. Sensebiasing context influenced reading times at that, and interacted with ambiguity in the disambiguating region. Thus, readers use sense-contingent subcategorization preferences during on-line language comprehension. 2002 Elsevier Science (USA). All rights reserved. The nature and use of information regarding the syntactic arguments that a verb can take is crucial to language comprehension and production. A critical component of knowing how to use a verb involves knowing its combinatory properties with other words, including both the range of possible syntactic complements it permits (subcategorization profile; MacDonald, Pearlmutter, & Seidenberg, 1994; Trueswell, Tanenhaus, & Kello, 1993), and the semantic constraints on its possible arguments (selectional restrictions, or more precisely, thematic fit, McRae, Ferretti, & Amyote, 1997). In addition, verb subcategorization information has been of strategic importance in testing psycholinguistic theories because theories of sentence processing differ regarding the time-course of the availability and use of this information during on-line comprehension. In particular, the constraint-based or expectation-driven framework argues that a broad range of information is available and used by listeners or readers in determining even initial structural analyses (Altmann, 1998, 1999; MacDonald et al., 1994; MacWhinney & Bates, 1989; Spivey & Tanenhaus, 1998). In this paper, in concert with Roland (2002), Roland and Jurafsky (1998, 2002), and Roland et al. (2000), we argue that knowledge of lexically specific structural constraints may be quite finegrained, and specifically that important aspects of subcategorization information are encoded not in relation to a verb, but to the verb s specific senses. One way to determine the types of information computed during comprehension, and to identify Journal of Memory and Language 48 (2003) 281–303 Journal of Memory and Language www.elsevier.com/locate/jml Corresponding author. Fax: +419-372-6013. E-mail address: [email protected] (M. Hare). 0749-596X/02/$ see front matter 2002 Elsevier Science (USA). All rights reserved. PII: S0749-596X(02 )00516-8 precisely when that information influences processing, is to study situations in which potentially available information sources are in conflict or underdetermine the interpretation of the input. Such situations arise whenever a word or a syntactic structure is temporarily ambiguous. For these reasons, ambiguity resolution has been a fruitful domain for testing theories of information use in sentence processing. One such structural ambiguity, which is the tool used in the current study, is the so-called direct object/sentential complement (DO/SC) ambiguity. This ambiguity arises at the post-verbal noun phrase (NP) in sentences such as The woman heard the dog had barked all night. In sentences such as these, the dog (at the point where it occurs) could be either the DO of heard (as in The woman heard the dog.) or the subject noun of an SC, as it ends up being in the first example sentence. This ambiguity arises because the complementizer that may optionally be omitted in such constructions. The true structure is thus not revealed until the verb in the SC (at had barked in the sentence above). This is referred to as the disambiguation region. Verb subcategorization biases and ambiguity resolution Although some verbs occur in both SC and DO constructions with equal likelihood, many exhibit a bias. The notion that subcategorization preferences such as these might play a role in sentence processing has been considered by a number of researchers (Clark & Clark, 1977; Connine, Ferreira, Jones, Clifton, & Frazier, 1984; Ferreira & McClure, 1997; Fodor, 1978; Ford, Bresnan, & Kaplan, 1982; Fodor & Garret, 1967). Some researchers, such as Ford et al., have suggested that comprehenders use their knowledge of the relative probability with which a verb occurs with different subcategorizations to guide syntactic analysis, though others, such as Frazier (1987a, 1987b) and Ferreira and Henderson (1990) have reported patterns of results consistent with the notion that lexically specific knowledge is used only in the revision process. In more recent years, the role of verb bias in comprehension has been subject to careful scrutiny in a number of studies involving the DO/SC ambiguity. Although some studies report late or no effects of verb bias (Ferreira & Henderson, 1990; Mitchell, 1987), more recent work has shown early effects (though see Kennison, 2001). Trueswell et al. (1993) contrasted sentence pairs that were structurally ambiguous at the post-verbal NP, and differed only in the bias of the main verb. These sentences were then contrasted with structurally unambiguous versions (i.e., containing the complementizer that), and the additional time taken to read the ambiguous version was referred to as the ambiguity effect. In a self-paced reading time experiment, Trueswell et al. found a large ambiguity effect for sentences containing DO-biased verbs at the point following the disambiguation toward an SC. In contrast, with SC-biased verbs, reading times for ambiguous sentences were similar to unambiguous controls. The bias by ambiguity interaction was replicated in an eyetracking experiment, although in this case it was found at the disambiguation itself, rather than in the post-disambiguation region. Finally, although there was a 34ms ambiguity effect for SC-biased verbs in first-pass reading times, this was nonsignificant. The increased reading times were attributed to a that-preference effect because they were correlated with lexically specific that-preference. Trueswell et al. (1993) used the same sentence frame for both a DOand an SC-biased verb. The post-verbal NP was always a plausible DO for the DO-biased verbs (e.g., The waiter confirmed the reservation was made yesterday. . .) but rarely or never a plausible DO for the SC-biased verbs (The waiter insisted the reservation was made yesterday. . .). As a result, bias was confounded with DO plausibility, and the results may have been influenced by the degree of commitment to a semantically plausible or implausible parse (cf. Pickering & Traxler, 1998). Garnsey et al. (1997) addressed this issue in an eyetracking study crossing bias with DO plausibility. Verbs of each bias type (including a third, equibiased condition) appeared in sentences in which the post-verbal NP was either plausible or implausible as a DO. Thus for the implausible conditions, the NP was syntactically ambiguous, but semantically anomalous if interpreted as a DO. An effect of verb bias was found nonetheless: At the disambiguation, reading times were longer in the ambiguous than the unambiguous conditions for DO-biased, but not SC-biased verbs. In addition, with DO-biased verbs, sentences with plausible DOs yielded a significant ambiguity effect whereas the effect for those with implausible DOs was not statistically reliable. Overall, then, recent results suggest that verb bias is one relevant source of information in guiding sentence interpretation.