| |
Abstract:
We investigated the time course and possible interactions of
semantic and syntactic processes in language comprehension using
event-related brain potentials. Previous studies had frequently
shown an early left anterior negativity (ELAN) and a late
parietal positivity (P600) in correlation with phrase structure
violations whereas semantic integration processes are correlated
with a N400. The aim of the present study was to find out whether
or how these different processes might interact.
Four types of sentences were presented auditorily. The
sentences ended with a target word that was either (a) correct,
(b) semantically incorrect, i.e. violating the selectional
restriction of the verb, (c) syntactically incorrect, i.e.
violating the phrase structure, or (d) semantically and
syntactically incorrect, i.e. violating both the selectional
restriction and the phrase structure. In the first experiment
subjects were asked to judge the overall correctness of each
sentence. The ERP- results with regard to conditions (b) and (c)
replicated earlier results: we observed a N400 for semantic
violations and an ELAN component followed by a P600 for phrase
structure violations. The combined violation (d), however,
elicited the same result pattern as the pure syntactic condition,
i.e. ELAN and P600. Interestingly, there was no N400 despite the
clear selectional restriction error. This indicates that the
semantic integration of a word in a sentence may only be
initiated in case of a successful prior integration of the
element into the phrase structure of the sentence.
In a second experiment we presented the same sentences but
changed the instruction to the subjects. In this experiment they
were asked to judge exclusively the semantic coherence of each
sentence thereby ignoring syntactic aspects of the sentence. That
is, participants had to classify semantically correct sentences
containing only a phrase structure error as being correct. The
ERP-results were clear-cut: even when focussing on semantic
aspects of the sentence an ELAN component was elicited in the
pure syntactic violation condition (c) as well as in the combined
semantic-syntactic violation condition (d). This illustrates the
highly autonomous character of first-pass parsing processes.
Interestingly, under the semantic coherence instruction there was
no P600 for phrase structure violations which suggests that the
processes reflected in this component are of a fairly controlled
nature. A further important result of this experiment was that,
in contrast to Experiment 1, a N400 component was elicited also
in the combined violation condition (d). The fact that the N400
was dependent on the instruction suggests that the processes
underlying this component are also under the participants'
strategic control.
|