| |
Abstract:
Previous ERP studies found an early left anterior negativity
followed by a late positivity in response to syntactic violations.
Are these components natural language specific, or can they also be
found with a well-trained artificial grammar? In a 10-hour intense
training procedure subjects learned the syntactic rules of a simple
artificial language by speaking and listening to word sequences.
The words, being non-words in known languages, referred to objects
and actions in a computer game that was intended to motivate the
acquisition of the language. When the subjects' mastery approached
perfection they entered the ERP experiment.
Both correct and syntactically incorrect sentences were presented
auditorily. To assure attentive listening the subjects had to
perform a probe detection or grammaticality judgment task (randomly
mixed), without knowing the type of task while hearing the phrase.
ERPs time-locked to the onset of the violation show highly
significant differences: a negative component from 100 to 200 ms
with a fronto-central distribution, a partially overlapping
negativity peaking at 300 ms at parietal and occipital sites, and a
broad positivity starting at 400 ms. The study focused on the early
negativity which is thought to reflect automatic first pass parsing
processes (specifically, the detection of word category errors).
The results suggest that this component not only occurs in native
natural language contexts, but more generally reflects activity of
neural systems involved in processing language related structured
sequences. However, in contrast to similar natural language
studies, the effects are not left lateralized, possibly indicating
the coactivation of other cortical areas.
|
