| |
Abstract:
There are at least two sorts of mechanisms which can account
for context effects in language processing (viz. the finding that
related semantic context yields faster reaction times, and smaller
N400 components in the event- related potential). Lexical context
mechanisms rely on pre-stored associative relationships
(coffee/tea),and are thought to be applied quickly and
automatically. In contrast, a distinct and slower mechanisms has
been proposed to account for sentential context effects, which
cannot be stored (because there are an infinite number of
sentences), and must be computed. ERPs were recorded to the final
words of four sentence conditions, in which lexical and sentential
context were orthogonally varied ("The waitress was good about
bringing refills of COFFEE and TEA/TALKING" and "After dinner, they
spent a long time over COFFEE just TALKING/TEA."). Sentence-level
congruity influenced N400 amplitude, while lexical context did not.
Subsequent to the N400, the largest late positivity was observed
when sentential and lexical context conflicted (i.e., for words
which were sententially anomalous, but lexically related). These
results suggest that readers focus on relationships at the highest
level supported by the input, and that lexical context does not act
automatically. Subjects with lower performance on behavioral tests
of verbal ability showed smaller sentence congruity effects, but
also evidenced no sign of a pure lexical context effect.
|
