| |
Abstract:
A central issue in sentence processing research is whether
the parser can immediately access and use information on meaning
content (semantics) as well as form (syntax). Much effort has been
expended in trying to find evidence for temporal primacy of
syntactic over semantic processing, but the problem has proven
remarkably refractory to experimentation (Fodor, Ni, Crain, &
Shankweiler, 1996). Records of eye movements during reading and
results from cross-modal studies reveal that while both syntactic
and semantic information is available to the parser simultaneously,
this information leads to divergent behavior subsequently:
Syntactic anomalies caused abrupt disturbances that were quickly
resolved whereas semantic anomalies resulted in prolonged decision
responses.
Additional support for the hypothesis of modular organization of
the language processing apparatus comes from neurobiological
sources. Studies of Event-related brain potentials (ERP) have
yielded distinct electrophysiological signatures for semantic
anomaly detection (N400) and syntactic processes (P600/Syntactic
Positive Shift), but localizing information afforded by this
technique is very limited. Hemodynamic responses assessed by
functional Magnetic Resonance Imaging (fMRI) offer an unprecedented
opportunity to pinpoint site(s) of origin of task relevant brain
activation.
We conducted two coordinated experiments on sentence processing
using fMRI to investigate whether distinguishable cognitive demands
are posed by syntactic and semantic operations. While in the
scanner, 14 young, unimpaired adults listened to simple sentences
that were either nonanomalous or contained a syntactic anomaly
(e.g.,
*Trees can grew.),
or a semantic anomaly (e.g.,
*Trees can cry.).
In Experiment 1, subjects judged whether each sentence was
acceptable. A same/different tone pitch judgment task provided a
baseline for distinguishing brain activity associated with
linguistic processing from activity generated by other components
of the task. Sites selectively activated by sentence processing
were found in inferior, posterior frontal, middle and superior
frontal, posterior superior temporal and temporo-parietal regions.
There was nearly complete overlap between regions activated by each
type of anomaly. In Experiment 2, which preceded Experiment 1,
subjects heard isolated occurrences of anomalous sentences of one
type or the other interspersed among nonanomalous sentences.
Judgments of animacy were made, which did not require anomaly
detection. Experiment 2 yielded nonoverlapping areas of activation:
Syntactic anomaly triggered significantly increased activity in and
around Broca's area, whereas semantic anomaly activated several
other sites, among them Wernicke's area.
The results indicate that fMRI signals can distinguish brain
activities associated with syntactic and semantic operations during
automatic sentence processing. Hence, this is new neurobiological
evidence for the modularity hypothesis.
Reference
Fodor, J. D., Ni, W., Crain, S., & Shankweiler, D. (1996).
Tasks and timing in the perception of linguistic anomaly.
Journal of Psycholinguistic Research,
25, 26-57.
|
