NP interference in sentence processing

Peter C. Gordon & Randall Hendrick
University of North Carolina at Chapel Hill

pcg@email.unc.edu

Sentence processing involves integrating the meanings conveyed by various constituents according to syntactically specified relations; most centrally this involves using noun phrases (NPs) to determine the arguments of predicates conveyed by verbs.  Performing this integration successfully requires that the representations derived from the NPs be kept distinct and so that they can be integrated in the appropriate way.  A great deal of research on human memory has shown that the distinctiveness of representations is reduced as their similarity increases, and that this lack of distinctiveness is a source of interference in processing (Crowder, 1976).  Linguistic and psycholinguistic evidence has further indicated that similarity of NPs can yield interference that impairs the comprehension of syntactically complex sentences (Bever, 1974; Gordon, Hendrick & Johnson, 2001).

The current work uses eye tracking during reading to examine whether this type of interference occurs for a simple sentence made up of a single clause.  We varied whether the subject NP and first object NP were descriptions or names.

After the meeting Mark/the lawyer gave Brad/the doctor the large notebook.

The combinations of these NP types yielded two types of matched sentences (name-name and description-description) and two types of non-matched sentences (name-description and description-name).  Forty-eight sentences were created with different double-object verbs (culled from Levin, 1993) along with different common first names and role terms.  The sentences were divided among the combinations of NP types and assignment of sentences to conditions was counterbalanced across participants.  Eye movements were recorded as subjects read the experimental sentences (as well as an equal number of filler sentences); a comprehension question followed each sentence.

First-pass gaze duration and total gaze time are shown in the table for the reading of the object NP. 

On both reading time measures there was a significant interaction between type of subject NP and type of object NP [(first-pass gaze time: F1(1,11) = 5.11, p < .05; F2(1,47) = 2.27; p<.15), (total gaze time: F1(1,11) = 7.59, p<.025; F2(1,47) = 6.84; p<.025)].  These interactions showed that reading time for the NP was elevated when the preceding, subject NP was of a matched as compared to non-matched type.  This result shows that similarity-based interference can occur even in simple sentences with a single clause, indicating that processes analogous to those that occur in linking clauses together also occur in the determination of the arguments of a verb. 

References

Bever, T. G. (1974).  The ascent of the specious, or there's a lot we don't know about mirrors.  In D. Cohen (Ed.), Explaining Linguistic Phenomena (pp. 173-200).  Washington: Hemisphere.

Crowder, R. G. (1976).  Principles of Learning and Memory.  Hillsdale, NJ: Erlbaum.

Gordon, P. C., Hendrick, R., & Johnson, M. (2001).  Memory interference during language processing.  Journal of Experimental Psychology: Learning, Memory and Cognition, 27, 1411-1423.

Levin, B. (1993).  English Verb Classes and Alternations: A Preliminary Investigation.  Chicago: The University of Chicago Press.