| |
Abstract:
Most models of lexical access in speaking hold that the
retrieval of semantic-syntactic information (i.e., lemma
information) and the retrieval of phonological information (i.e.,
wordform information) ought to be distinguished. In particular, the
retrieval of a noun's grammatical gender is assumed to be mediated
by access to the noun's lemma, but to be independent of the
retrieval of the noun's phonological form. In a series of
experiments we tested this claim by investigating the production of
gender-marked pronouns in German, using variants of the
picture-word interference paradigm (cf., Schriefers, Meyer, &
Levelt, 1990, JML). Participants named or described pictures of
simple objects by using the appropriate pronoun while ignoring
auditory distractor words. If grammatical gender needed for
determining the form of the pronoun is accessed from the noun's
lemma, we expected inhibition from distractors semantically related
to the object name as compared to unrelated controls. If the noun's
wordform is also activated, we additionally expected to find
inhibition from distractors phonologically related to the object
name.
In our experiments, we consistently observed semantic inhibition
in both pronoun production and noun production. This inhibition
effect was similar in size. By contrast, distractors phonologically
related to the noun facilitated noun production but did not
influence pronoun production. This pattern was replicated in a
number of experiments. It proved to be independent of whether the
pronoun was used to establish anaphoric reference to one out of two
previously mentioned candidate antecedents, whether the pronoun was
produced in isolation, or whether it was uttered in a cataphoric
context.
In one further experiment with distractors related to the
pronoun's own phonology, we demonstrated that pronoun production
was in principle sensitive to the phonological content of the
distractors. In a final experiment, only one possible candidate
antecedent was introduced. Now the semantic interference effect
disappeared, suggesting that under certain circumstances speakers
may retrieve grammatical gender information in pronoun production
from an episodic memory trace rather than from the mental lexicon
(see also Meyer & Bock, 1999, JML).Together, our data suggest
that in generating pronouns, speakers can reaccess the lemma of the
corresponding noun, while its wordform is not substantially
activated. This interpretation stands in contrast to the one
recently advanced by Schmitt, Meyer, & Levelt (1999,
Cognition). Possible reasons for the discrepant findings will be
discussed.
|
