|
Abstract:
In two studies using event-related brain potentials we
investigated the processing of different aspects of verb
information, namely information about the number of arguments as
well as semantic information (selectional restrictions). We
therefore presented sentences with either (a) a violated argument
structure (two NPs but only one-place verb) or (b) violated
selectional restrictions (animate instead of inanimate object
NP). In the first experiment, both types of violation were
realized on the verb which followed its NP arguments, e.g.:
(1a) argument structure violation Er weiss, dass der Sekretaer
den Kanzler [lacht]. He knows that the secretary the chancellor
[laughs].
(1b) selectional restriction violation Er weiss, dass der
Sekretaer den Kanzler [kehrt]. He knows that the secretary the
chancellor [sweeps].
EEG activity of 16 subjects was recorded during sentence
reading. Subjects performed a grammaticality judgement. Compared
to correct sentences, both types of violation elicited
negativities between 300 and 500 ms with identical topography,
so-called N400 components, usually following semantic anomalies.
In addition, the sentences with an incorrect argument structure
elicited a late positivity between 700 and 1000 ms (P600
component), normally taken to indicate a phrase structure
revision process. In order to determine whether these results
were dependent on the position and/or the syntactic category of
the target, violations were realized not on the verb but on its
direct object argument (NP presented as a whole). We therefore
created sentences in which the verb preceded its arguments by
adding a sentence adverbial, e.g.:
(2a) argument structure violation Heute verstarb der Schueler
[den Sportler] im Krankenhaus. Today died the pupil [the
sportsman] in the hospital.
(2b) selectional restriction violation Heute strickte der
Schueler [den Sportler] im Unterricht. Today knitted the pupil
[the sportsman] during the lessons.
Again, 16 subjects read the sentences and performed a
grammaticality judgement while their EEG activity was recorded.
As in the first experiment, the ERP patterns in both violation
conditions showed topographically identical N400 components and
only the sentences with a violated argument structure
additionally exhibited a P600. These results suggest that in an
N400 time window both semantic and argument structure information
involve similar processes, whereas a mismatch between the number
of arguments licensed by the verb and the number of NP arguments
in the sentence additionally leads to a repair of phrase
structure. This is not dependent on whether the processing of
verb information precedes the processing of argument information
or vice versa.
|
