Colin Phillips, of the University of Maryland, gave an ICESL-sponsored talk in Machmer E-37 at 3:30 on Friday March 2nd.
Electrophysiology and Language Architecture
In this talk I will discuss two recent lines of research in the cognitive neuroscience of language processing that started from independent puzzles, but that appear to be converging on a common solution. The first puzzle involves apparent discrepancies in the localization results obtained using fMRI and MEG measures of ‘semantic’ processing. The same tasks and materials yield conflicting localizations when measured using different tools. The second puzzle involves a recent series of studies that undermine received wisdom about the functional status of ERP components and, more interestingly, challenge the widespread view in linguistics and psycholinguistics that semantic composition is tightly coupled to the syntax of a sentence. The solution to both puzzles involves recognizing that the N400, a neural response component traditionally associated with compositional semantic interpretation, is more closely linked to lexical processes and to word-level expectations. This also provides an account of the split personality of the N400 – it is sometimes very ‘smart’, responding to fine details of the compositional semantic interpretation and pragmatic congruity of linguistic input, but at times it is surprisingly ‘dumb’, sensitive only to the lexical properties of a word and to associative relations among words. I show how it is possible to turn the dumb N400 into the smart N400. I show that rather than undermining widespread assumptions about language architecture, the electrophysiological evidence instead provides finer-grained evidence on how interpretations are computed on-line. The evidence is drawn from studies in English, Spanish, and Chinese.