心理语言学.doc

MEMORY AND KNOWLEDGE IN DISCOURSE
COMPREHENSION
Early models of discourse comprehension focused on the process of estaon in LTMis activated through a passive resonance process, with
level of activation for a given item determined by overlap in semantic features and
strength (but not type) of association between activated item and cue.
This model is able to account for a number of findings in the literature where
lengthened reading times are observed for sentences that are locally coherent but
that conflict with information presented in a distant sentence earlier in the text.
For example, Albrecht & O’Brien (1993) asked participants to read long passages
in which the main character Mary was introduced as a strict vegetarian. After 6
filler sentences, the reader encountered the target sentence in which Mary ordered
a cheeseburger and fries. This sentence was locally coherent with the preceding
sentence, but reading times for this sentence were longer than in a neutral version
of the paragraph that gave no preliminary information about Mary’s eating habits.
A memory-based model accounts for this result by claiming that the concept of
vegetarian was re-activated through a resonance process initiated by the word
“cheeseburger.” This allowed the reader to detect the inconsistency in the text.
Evidence that this resonance process has the dumb characteristic of earlier