|
Abstract:
A review of the literature on reduced relatives (Townsend and
Bever, in press) extends prior observations (e.g., MacDonald, 1994)
that verb type and preposition affect whether and where reduced
relatives lead to increased reading time. We constructed a study
designed to confirm the findings from our post hoc analysis of the
literature. Using a two-word at a time, self-paced reading study we
contrasted:
a) The past participle form of verbs that require objects
(transitive only), as in (i), vs. the past participle form of verbs
that do not (potentially intransitive) as in (ii):
i) The horse caught by the wolf fell.
ii) The horse raced by the jockey fell.
b) Initial NPs that are good agents for the verb vs. good patients
c) The preposition /by/ with an agent phrase vs. some other
prepositional phrase
We found a distinct pattern of the reduction effect for
sentences with transitive only past participles compared to
sentences with potentially intransitive past participles. For
potentially intransitive sentences, the effect was greatest at the
main verb, while for the transitive only sentences, the effect
occurred earlier, at the prepositional phrase. In addition, the
reduction effect was much greater for sentences with a non-/by/
phrase (about 350 ms) than for those with a /by/ phrase (about 70
ms), but only for sentences with potentially intransitive
verbs.
There was no overall reduction effect of the good agent/good
patient manipulation. However, good agents elicited consistently
longer reading times at the noun phrase following the /by/ or
preposition, whether or not the relative clause was reduced.
All results are consistent with an early application of a
canonical sentence schema regardless of whether the initial noun is
a good agent or good patient for the embedded verb. Potentially
intransitive verbs can form a complete sentence at the point of the
embedded verb; a preposition other than /by/ enhances the complete,
intransitive reading of the verb. The parsing mistake goes
undiscovered until the main verb. In contrast, transitive only
verbs are not complete at the point of the verb. The reduction
effect during the prepositional phrase for transitive only
sentences follows from the fact that the canonical schema is
blocked right at reading the verb+preposition. More processing is
then required to find a different analysis. The main effect of good
agent/good patient follows from the fact that the good agent will
induce a misleading tendency towards the canonical sentence schema,
regardless of whether the relative clause is reduced or not.
MacDonald, M. C. (1994). Probabilistic constraints and syntactic
ambiguity resolution. Language and Cognitive Processes, 9,
157-201.
Townsend, D., and Bever, T. G. (in press). The case for the
sentence. Cambridge, MA: MIT Press.
|
