| |
Abstract:
On one *autonomous* account of syntactic priming is an
entirely automatic process: After processing one utterance with a
particular syntactic structure, a speaker is more likely to produce
another utterance with the same structure. Such an account fits
well with traditional proposals (Bock, 1986; Bock & Loebell,
1990; Pickering & Branigan, 1998). Branigan, Pickering, and
Cleland (in press) found syntactic priming from comprehension to
production in dialogue. If this account of priming is correct, it
should not be affected by whether the speaker interacts with the
listener or not.
Many aspects of production, however, are affected by the
speaker*s relation with the listener (e.g., Brennan & Clark,
1996; Garrod & Anderson, 1987; Schober & Clark, 1989;
Wilkes-Gibbs & Clark, 1992). For example, speakers tend to
employ referring expressions that are compatible with listener
knowledge, and are more responsive to interlocutors (people they
are conversing with) than with overhearers. If syntactic processing
is similar, then priming should be enhanced when the speaker
addresses the person who just addressed the speaker.
In order to address this question, we employed the *confederate
priming* technique introduced by Branigan et al. (in press), but
modified the relationship between the dialogue participants. One
group of subjects replicated Branigan et al.: A subject and a
confederate participated in a dialogue game in which they both had
to place a set of cards in the same order as each other. The key
cards could be described using a prepositional-object form (e.g.,
the cook giving a flower to the policeman) or a double-object form
(e.g., the cook giving the policeman a flower). Unbeknownst to the
subject, the other participant was a stooge who produced scripted
responses in one or other of these forms. Subjects showed a highly
significant tendency to produce the same form of sentence as the
stooge, even though none of the content words stayed the same
between prime and target: 66% of responses shared the same form
between prime and target (vs. 50% chance).
The second group of subjects played a game involving three
participants, two confederates (A, a speaker; and B, a listener)
and one subject (C). First A spoke to B, producing one or other
form (on the critical trials), and B moved the appropriate card.
Then C spoke to B, and B moved the appropriate card; at which
point, A spoke again; and so on. Under these conditions, priming
was significant, but significantly reduced: 57% of responses shared
the same form.
This suggests that priming is stronger in a true dialogue than
in an overhearing situation (even a licensed overhearer, as in this
case); though, interestingly, it still occurs. However, an
alternative explanation for the reduction in priming is that the
subject paid less attention to the other speaker than in the true
dialogue. The third condition tested this by repeating the second
condition, except that the speaking confederate A made mistakes in
describing the cards, and the subject was instructed to correct A*s
mistakes. Although such corrections were invariably successful,
thereby showing that a high level of attention was retained in this
version of the task, priming was still reduced: Again, 57% of
responses shared the same form between prime and target. We
conclude by discussing the implications of these findings for the
autonomy of syntax in language production (e.g., Bock & Levelt,
1994;Garrett,1980).
|
