|
Abstract:
The observation that pronominal and reflexive noun phrases in
English have a nearly complementary distribution has played a
central role in syntactic theory since the 1960's when researchers
began to formulate the structural conditions that now form the
basis of theories of "binding".
We present two experiments demonstrating that basic binding
phenomena with picture noun phrases are reproduced in an on-line
task in which subjects followed instructions to manipulate objects
while we monitored their eye movements. Three male dolls (Ken, Joe,
and Harry) were seated on a table. Pictures of each doll were
mounted on a board behind each doll in a column. Subjects were told
that the pictures directly behind each doll belonged to that
doll.
Experiment 1 used instructions such as:
(1) Look at Ken. Have Joe touch a picture of him/himself.
Binding theory predicts that with a pronoun, the relevant doll
(e.g., Joe) should touch a picture of the doll mentioned in the
LEAD-IN sentence (e.g., Ken); and indeed 96.3% of the responses
were to a picture of that doll. With a reflexive, the doll should
touch a picture of himself, and 96.3% of the responses were of that
type. The mean latency of the look to the picture that immediately
preceded the touch was 1574 ms from the offset of the pronoun and
1329 ms from the offset of the reflexive.
Experiment 2 used instructions such as:
(2) Look at Harry/Joe. Have Joe touch Ken's picture of
him/himself.
The lead-in sentence either referred to the same doll as the one
performing the action (SAME lead-in) or a different doll (DIFFERENT
lead-in). Binding theory predicts that there are two possible
referents for the pronoun, the NP in the lead-in sentence and the
subject NP in the target sentence. In the absence of strong
discourse/pragmatic constraints, this ambiguity should result in
conflict between the two potential antecedents, delaying looks to
the target picture (the one eventually chosen by the subject) in
the different lead-in condition. With the reflexive, however, the
ambiguity should disappear because only the NP in the picture noun
phrase (the possessor) should be a possible antecedent for the
reflexive.
These predictions were confirmed. For the pronouns, subjects had
the doll touch a picture of the lead-in on 78% of the trials, and a
picture of the subject NP on 20% of the trials. Eye movement
latencies were 669 ms slower in the different lead-in condition
compared to the same lead-in condition. There was still a 418 ms
difference for those trials in which a picture of the lead-in doll
was touched. In contrast, the reflexive pronouns showed no effect
of lead-in. Subjects touched a picture of the possessor on 88% of
the trials. Response latencies were 1750 ms and 1737 ms in the same
and different lead-in conditions, respectively.
The results establish the feasibility of using the eye-tracker
methodology for exploring binding. Some of the data patterns that
motivate binding theory are cleanly reproduced without explicit
meta-linguistic judgments, using a paradigm in which it will be
possible to manipulate non-structural constraints that have been
argued to influence binding. We have also shown that the presence
of a potential competitor can delay binding to a pronoun regardless
of the final choice. Use of competitor effects should be a powerful
tool for determining which potential antecedents are potentially
available for binding.
|
