|
Abstract:
Syntactic priming is the tendency to reuse syntactic
structure during sentence production (grammatical encoding) or
sentence perception (grammatical decoding). Processing a particular
syntactic construction in encoding or decoding modality tends to
facilitate processing this structure in a subsequent sentence, in
the same or the other modality. During an initial priming trial,
participants process a sentence embodying a specific syntactic
structure (e.g. a double object construction); this is followed by
a target trial involving a sentence that embodies one of two
alternative, semantically equivalent structures (e.g. double object
or prepositional object). The priming effect is typically measured
in terms of response bias (increased proportion of compatible
responses embodying the prime construction).
We explored online syntactic decoding-to-encoding priming in
Dutch, measuring response facilitation (reduced processing time for
compatible responses), instead of response bias. Dutch features a
semantically neutral contrast between Subject-Verb order in
subordinate clauses (introduced by subordinating conjunctions; e.g.
terwijl vogels zingen, while-birds-sing) and Verb-Subject order in
main clauses (if introduced by adverbs; e.g. hier zingen vogels,
here-sing-birds). In every trial, participants first silently read
a prime clause embodying either an SV or VS construction. Five
seconds later, a picture was shown depicting an action performed by
plural agents (e.g. babies crying). Participants described it in
either SV or VS format, as prompted by a lead-in, i.e. a
subordinating conjunction or an adverb. We measured Utterance Onset
Latencies for the targets (without lead-in). The lead-in appeared
either simultaneously with the picture (SOA=0) or 2 seconds earlier
(SOA=-2000). In 1/3 of the trials, the prime sentence contained the
same words as the to-be-produced target sentence (identical
trials); in the remaining trials, the description of another
picture served as prime (non-identical). There were as many
compatible (same word order in prime and target) as incompatible
trials.
The unexpected result was a clear negative priming effect in the
non-identical trials, for both SV and VS targets, and for both
SOAs: longer UOLs for compatible prime and target trials. This
contrasts with the facilitatory effects hitherto obtained in
syntactic priming studies. The identical trials did yield a
positive priming effect: repeating noun and verb of the prime in
the same order was somewhat faster than producing these words in
reversed order.
In the human performance literature, systematic contrasts
between positive and negative priming effects have been reported
for analogous non-linguistic tasks. According to the recent code
occupation theory, the direction of the priming effect depends on
whether the internal representation (code) of the prime stimulus
has been released (discarded), or is still being maintained
(occupied) in order to be integrated with the response under
construction. Positive priming is expected when the prime
representation has been released (but still has a relatively high
level of activation). Negative priming is predicted when parts of
the prime representation (but not the representation as a whole) is
integrated in the target response.
This theory and our results shed new light on the phenomenon of
syntactic priming, in particular on the crosstalk between
grammatical decoding and encoding.
|
