Abstract for Matthew Wagers' talk

Representing the control of agreement in real-time
Matthew Wagers (University of Maryland)
October 30, 2007 (Tuesday)
6:30 PM - 8:00 PM; Room 7102, The CUNY Graduate Center

In English, a finite verb typically agrees with the noun phrase in subject position (1a). But a nearby noun phrase can sometimes wrest away control of agreement, in a phenomenon known as 'agreement attraction' (1b) (Jespersen 1913; Bock & Miller 1991).

(1) (a) The path is/*are littered with bottles.
(b) The path to the monuments is/??are littered with bottles.

This phenomenon is most robustly reflected in production, but it also impacts judgments and processing difficulty in comprehension. In this talk, I'll discuss the alternative accounts of this phenomenon found in the syntactic and psycholinguistic literature. I'll first consider one of the most prominent claims, that the constituents of complex subjects are combined in such a way, as to sometimes allow the subject projection to effectively be valued with the number of a noun other than its head (e.g., Eberhard, Cutting & Bock, 2005; den Dikken 2001). Based on a series of six comprehension studies (self-paced reading and speeded grammaticality) which carefully compare individuals' responses to grammatical and ungrammatical sentences, I'll argue that subjects do not erroneously value the subject projection in comprehension, and that attractor nouns only impact reanalysis. In addition, these studies also examine a less well-studied configuration, in which subject-verb agreement within a relative clause seems to be sensitive to the number of the relative clause head (Kimball & Aissen, 1971):

(2) (a) The runner that the driver sees/*see on her commute always waves.
(b) The runners that the driver sees/??see on her commute always waves.

In the second part of the talk, I'll argue that agreement attraction in comprehension reflects the interaction between predictive structure building and cue-based retrieval processes, and discuss how it can be modeled in an associative memory. Close parallels can be drawn between this explanation and a recent proposal for production by Solomon & Pearlmutter (2004), that emphasizes the simultaneity of constituents in processing. Finally I'll discuss implications of this research for understanding under what circumstances online processes are grammatically faithful.

Note 1: The work reported in this talk was undertaken in collaboration with Ellen Lau (University of Maryland).

Note 2: A manuscript reporting parts of the research can be found at
http://ling.umd.edu/~matt/papers/wagerslau2007draft.pdf