| |
Abstract:
According to some sentence processing models, ungrammatical
and dispreferred continuations of a sentence are initially treated
similarly; they only differ in whether a correct reading can be
computed. These models predict that ungrammatical and dispreferred
continuations show similar early ERP components (100-500 ms), but
may differ with respect to the P600, which is said to reflect
reanalysis. ERPs were recorded from 18 subjects while they read,
and judged for acceptability, 40 Preferred, Dispreferred, and
Ungrammatical sentences ('I cut the [pizza next to the cakes /
pizzas next to the cake/ pizza next to the cake] that WERE brought
by Jill.'). In contrast to prediction, an anterior negativity
between 300-500 ms was seen at WERE for the Ungrammatical condition
only. Dispreferred and Ungrammatical conditions elicited a
posterior P600, which started earlier and lasted longer for the
Ungrammatical condition, suggesting (a) that error repair was
initiated earlier, and (b) that the attempts to repair the
structure were longer lasting. To test whether the P600 onset
indeed reflects the ease of determining the error, unambiguous
conditions were tested in which only one noun phrase was relevant
to determine the ungrammaticality. The P600 had an earlier onset
compared to the previous ambiguous conditions. In addition, the
P600 was broadly distributed, and longer lasting, suggesting that
ambiguity affects the processing of ungrammatical
continuations.
|
