| |
Abstract:
Recently there has been debate about the nature of syntactic
representation in language production, and how syntax is
represented with respect to orthography and phonology (e.g.
Caramazza, 1997). Most lemma based production accounts assume a
modality neutral lemma (e.g., Levelt, Roelofs, & Meyer,
1999). Shelton and Caramazza (1999) proposed an account where
orthography and phonology are separately represented in language
production and where there is no lemma node. Hence, syntax must
be represented at the (modality-dependent) lexeme level. Within
this account, there are 2 possibilities for the representation of
syntax: that phonological and orthographic lexeme nodes are
associated with modality-specific syntactic nodes, or that the
lexeme nodes share one set of syntactic nodes. We report
syntactic priming studies that attempt to distinguish these
accounts by testing for priming both between and within
modalities.
Subjects were presented with sentence fragments on a computer
screen and were prompted to either read them aloud and complete
them, or write them down and complete them. The prime sentence
fragments were designed to either prompt a prepositional object
completion (e.g. The millionaire loans the painting to the
gallery) or a double object completion (e.g. The millionaire
loans the gallery the priceless painting). The prime was either a
written or a spoken sentence. A spoken target sentence then
followed (e.g. The surgeon hands) which could be completed with
either construction. We examined whether the syntactic form of
the target completion was affected by the syntactic form of the
prime completion and whether any priming effect was affected by
whether the prime was written or spoken. The prime either
contained the same verb as the target, or it contained a
different verb.
There was a significant priming effect: Subjects produced 16%
more target completions of the same structure as the prime than
of the alternative structure (14% with written primes and 18%
with spoken primes). Crucially, there was no significant
interaction between the magnitude of priming and whether the
prime was written or spoken.
The results demonstrate that syntactic priming occurs even
when the prime and target differ in modality, and that the
magnitude of priming is not significantly affected by this
difference. The results rule out the version of Caramazza's model
in which syntactic information is stored separately at modality
specific lexical nodes. They are difficult to reconcile with the
version of the model where independent lexical nodes share
syntactic nodes: such an account would most naturally predict
less priming between modalities compared with within a modality,
on the assumption, following Pickering and Branigan (1998), that
priming is at least partly based on residual activation of the
links between nodes as well as the nodes themselves. We argue
that the results are consistent with the view of a modality
neutral lemma representation (e.g. Pickering and Branigan, 1999),
which fits well with the finding of priming between comprehension
and production (Branigan et al., in press).
|
