Brian McElree,1 Martin Pickering,2 Matthew Traxler3 & Steven Frisson4
1 New York University, 2 University of Edinburgh, 3 University of South Carolina, 4 University of
Antwerp
A standard view of comprehension holds that syntactic processes build structured expressions and that semantic processes assign an interpretation by combining lexical representations according to their syntactic position. Recent work in lexical semantics has challenged the latter view, arguing that composition is context-sensitive: For example, default interpretations of individual expressions can be modified by other expressions through operations like "type shifting" (Bach, 1986; Partee, 1992), and interactions between expressions can introduce semantic structure not explicitly represented in surface form (Jackendoff, 1997; Pustejovsky, 1995). We present self-paced reading and eye-tracking studies that provide behavioral evidence for these types of enriched composition, and that provide insights into which aspects of enriched composition cause processing difficulty.
McElree et al. (2001) found that people experienced difficulty with "The author began the book", in comparison to cases where "began" was replaced with "read" or "wrote". The verb "begin" semantically requires an event, but "the book" does not refer to an event. We interpreted this as evidence for a processing cost associated with an additional operation of semantically coercing "the book" into an event interpretation (e.g., "writing the book"). In support of this view, we present two recent experiments examining the processing of event nominals. In both self-paced reading and eye-tracking, we found no difficulty with "The boy started the fight" versus "The boy saw the fight", where the default interpretation of the nominals is as an event, though we did find difficulty with "The boy started the puzzle" versus "The boy saw the puzzle", where the default interpretation of the nominal is as an entity but the verb "started" requires an event interpretation.
Type shifting per se could be responsible for the difficulty in processing structures like "began/started the book". However, Frisson and Pickering (1999, submitted) found that people experienced no difficulty processing metonymies involving place-for-institution ("talked to the school"), place-for-event ("protested during Vietnam"), and producer-for-product ("read Dickens") type shifts, provided there is a type of institution or well-known event or product associated with the noun. These expressions do not, however, require the generation of additional semantic structure as expressions like "began the book" do, which are typically interpreted as either "began to read the book" or "began to write the book". We present an eye tracking study that contrasts expressions like "The gentleman spotted/read/started Dickens while waiting for a friend...", which indicates that difficulty arises only in the "started" case where the interpretation requires the generation of additional semantic structure.
References
Bach, E. (1986). The algebra of events. Linguistics and Philosophy, 9, 5-16.
Frisson, S., & Pickering, M. J. (1999). The processing of metonymy: Evidence from eye-movements. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25, 1366-1383.
Jackendoff, R (1997). The architecture of the language faculty. Cambridge, MA: MIT Press.
McElree, B., Traxler, M. J., Pickering, M. J., Jackendoff, R., & Seely, R. E. (2001). Coercion in on-line semantic processing. Cognition, 78, B17-B25.
Partee, B. (1992). Syntactic categories and semantic type. In M. Rosner & R. Johnson (Eds.), Computational Linguistics and Formal Semantics. Cambridge: Cambridge University Press.
Pickering, M. J. & Frisson, S. (submitted). Why reading Dickens is easy (and reading Needham can be hard): Contrasting familiarity and figurativeness in language comprehension.
Pustejovsky, J. (1995). The Generative Lexicon. Cambridge, MA: MIT Press.