|
Abstract:
Chang and colleagues have presented a computational model
which can simulate the behavior observed in psycholinguistic
experiments of syntactic priming (Chang, Griffin, Dell, & Bock,
1997). The type of learning in Chang et al.'s model is called
"implicit learning" because the computational model is able to
perform the simulation without being explicitly programmed to
recognize specific characteristics of the input. That is, the model
employs a computational rather than a cognitive definition of
implicit learning. The goal of the current series of studies is to
pursue a potential relationship between syntactic priming and
implicit learning from a cognitive perspective.
One standard measure of implicit knowledge is conceptual
fluency, which predicts that participants will rate previously
encountered stimuli more favorably during a subsequent test phase
than will participants in control conditions who were not
pre-exposed to the critical items (Bornstein, 1989; Gordon &
Holyoak, 1983; Seger, 1994). The following three experiments use
grammatical acceptability as a measure of conceptual fluency.
In Experiment 1 (24 participants, 20 critical sentences plus
fillers) we test the hypothesis that sentences read for
comprehension earlier in the experiment will be judged as more
grammatically acceptable during the subsequent rating task. The
results show a priming effect for previously encountered sentences.
(Note: all significance values are calculated for materials and for
subjects using one-tailed t-test comparisons of the ratings of
previously-encountered versus novel sentence structures with
materials counterbalanced across subjects.) We integrate this
finding with the literature on implicit memory and we interpret it
as support for the hypothesis that there is implicit memory for
recently encountered sentences. Experiment 2 (24 participants, 18
critical sentences plus fillers) examines whether priming and test
sentences which are syntactically similar (in terms of phrase
structures) but which do not share all of the same content words
will also show syntactic priming. We do find a priming effect. This
result supports the hypothesis that in addition to implicit memory
there is also implicit learning for recently encountered syntactic
structures, independent of lexical content.
An alternative explanation for the results of Experiment 2 is
that priming could be due to indirect lexical associations or
thematic similarities. The results of Experiment 3 (24 participants
rating 40 critical sentences plus fillers) eliminate this
alternative hypothesis. The significant effect observed in
Experiment 2 cannot be attributed to lexical or thematic influences
and should be attributed to shared syntactic structures defined
across the training and test materials. One implication of these
studies is the influence of the ease of processing on judgments of
grammatical acceptability. Our findings allow us to incorporate
purely psychological models of conceptual fluency (Jacoby, Kelley,
& Dywan, 1989) into the development of a psycholinguistic model
metalinguistic evaluation.
Our results also support the integration of models of implicit
cognition with models of syntactic processing. By extending the
research to include implicit knowledge, we are able to investigate
models of representation and processing which the syntactic priming
research alone has been unable to address.
|
