| |
Abstract:
We used a lexical priming technique (Sereno, 1995) to
investigatelexical representation of argument structure and its
role in on-line syntactic processing. We expanded on previous
results in which briefly displayed prime verbs influenced
resolution of the direct object/sentential complement ambiguity
(Trueswell and Kim, 1998). In the current studies, we found that
these priming effects arise in other syntactic ambiguities and
across major category boundaries (i.e., the argument structure of a
noun can prime verb argument structure).
In one experiment, we investigated the processing of sentences
containing a local transitive/intransitive ambiguity such as 1.
1) While the camera crew filmed the rioters stormed into the
parliament building.
Target verbs (e.g., "filmed") were ambiguous but biased toward
the direct-object interpretation, suggesting that they would cause
erroneous direct object interpretations. Ambiguity was manipulated
by comma insertion or omission ("filmed, the").
Stimuli were presented via self-paced reading in combination
with the brief display of a prime word. When the participant
pressed a button to reveal the target verb, a prime verb was
displayed in the target position. The prime was displayed for 39ms
and then replaced by the target, which was displayed until the next
button press. Prime verbs were structurally biased toward either
the direct object analysis (DO-Primes; e.g., "seized") or the
intransitive analysis (I-Primes; e.g., "prayed"), and were matched
for overall frequency. In post-experimental interviews participants
reported only rarely identifying primes and typically perceived
priming events as flickers.
Reading times were longer in the disambiguating region for
Ambiguous sentences than for Unambiguous sentences, indicating
erroneous direct object commitments. However, the Ambiguity effect
was significantly reduced by I-Primes, resulting in a reliable
interaction between Ambiguity and Prime at the first disambiguating
word "stormed" (F(1,44) = 7.40, p < 0.01). A related experiment
in which post-verbal noun phrases were highly implausible as direct
objects also showed large ambiguity effects, but reliable
interactions with Prime were delayed to the third word of the
disambiguating region (F(1,53) = 4.11, p < 0.05).
In another experiment, we found that argument structure of noun
primes can have an effect much like that of verb primes.
Participants read locally ambiguous sentential-complement
constructions like 2.
2) The ice skater doubted the judges would keep her from
competing.
At the position of the ambiguous verb ("doubted"), a noun prime
was presented as described above. Primes were of three types:
S-Comp, which are nouns that frequently take a sentential
complement (e.g., "opinion"), Concrete (e.g., "machine"), and
Abstract (e.g., "freedom"). Reading times showed effects of Prime
at the ambiguous verb (F(2,70) = 5.11, p < 0.01) and in the
disambiguating region "would be" (F(2,70) = 7.00, p < 0.005),
with reading times being faster for S-Comp primes than for the
other two types. Abstract primes did not pattern with S-Comp
primes, ruling out an explanation in which the mere abstractness of
S-Comp primes facilitated the propositional meaning of the target
verb.
We have shown that the fast priming of verb argument structure
occurs in a range of syntactic ambiguities, and can help illuminate
the role of lexical representations in syntactic processing. We
suggest that the representations of nouns, like verbs, encode
detailed combinatory information. The representation of such
combinatory information is accomplished by mechanisms that are
common to the entire lexicon. Nouns are able to prime verbs by
engaging these common mechanisms, and the resulting processing
affects sentence-level syntactic processing.
|
