| |
Abstract:
This research focuses on the activation and integration of
prosodic, semantic, and syntactic information during the
comprehension of spoken sentences.
Event-related brain potentials (ERPs) were recorded while
subjects listened to sentences in Dutch that were temporarily
syntactically ambiguous between two grammatical structures: (a) a
conjoint NP, or (b) a sentence conjunction. The conjoint and
conjunction structures are associated with distinct prosodic
profiles in spoken Dutch. In the experiment, subjects listened to
sentences with prosodic profiles that either matched or mismatched
with a conjoint or conjunction analysis. In addition, the sentences
were either semantically neutral with respect to the syntactic
assignment, or contained a semantic bias towards one analysis. For
example:
No semantic bias:
(1) The tourist photographs the waiter and the guide films the
church.
Semantic bias:
(2) John sandpapered the plank and the carpenter fixed the
door.
Before the second verb (i.e., films or fixed) has been heard,
both sentences can, in terms of a purely syntactic analysis,
receive either a conjoint NP or a sentence conjunction analysis
(i.e., either [the waiter and the guide] are taken together in a
conjoint NP, or [the guide] is the head of a new sentence). In the
semantic bias condition, however, the selectional restrictions of
the first verb (i.e., sandpapered) are not compatible with the (by
default preferred) conjoint analysis. In the experimental design,
we crossed the bias condition with the prosodic condition, thereby
enabling an assessment of the separate and possibly interactive
effects of prosodic, semantic, and syntactic information on on-line
spoken language comprehension.
The waveforms show that the prosodic manipulation has a clear
and early impact on syntactic analysis. Sentences spoken with a
prosodic contour that is compatible with a conjunction analysis,
show a positive-polarity shift on the noun of the first NP (i.e.,
waiter or plank). This effect is not elicited when the same
sentences are pronounced with a prosodic contour that is compatible
with a conjoint-NP. In addition, the results indicate that semantic
information can have an early impact on syntactic processing. In
the Semantic Bias condition, the ERPs reveal that listeners
experience syntactic processing difficulty already at the second NP
(i.e., carpenter) when listening to sentences with a conjoint-NP
prosodic contour. This effect is not present when the same
sentences are spoken with a conjunction prosodic contour. Taken
together, the results point towards an interactive effect of
prosody and semantics on on-line syntactic analysis.
|
