|
Abstract:
Psycholinguistic research has established that sentences with
object-relative clauses (where the modified NP is the object in the
embedded clause) are more difficult to understand than those with
subject-relative clauses (where the modified NP is the subject in
the embedded clause),(see Gibson, 1998; MacWhinney & Pleh,
1988). Our first experiment examined the effect on this
subject-object difference of having an indexical pronoun (you) in
the embedded clause. It used self-paced reading time of single
sentences followed by a comprehension question, generally following
the method of King and Just (1991). Sentences included either
subject-relative clauses (first example below) or object-relative
clauses (second example below) and either a definite description of
a role term (e.g., the reporter) or the indexical pronoun "you" in
the embedded clause.
Description
Pronoun
The senator that attacked the reporter/you admitted the error. 549
497
The senator that the reporter/you attacked admitted the error. 602
479
The numbers to the right of the examples are the mean reading
time per word (ms) for the region after the relative pronoun
("that") and through the verb of the main clause ("admitted").
Reading times were significantly slower for the object-relative
clauses than for the subject-relative clauses when there was a
definite description but not when there was a pronoun. This pattern
is consistent with Gibson's (1998) analysis that embedded indexical
pronouns could cause a reduction in the subject-object difference.
It is not expected under other explanations such as MacWhinney and
Pleh's (1988) perspective maintenance account. However, the next
experiment (shown below) showed that the subject-object difference
is also eliminated when the embedded clause includes a name, a
pattern not expected under Gibson's theory because a name should
impose memory cost just like a definite description.
Description
Name
The senator that attacked the reporter/Ben admitted the error. 602
626
The senator that the reporter/Ben attacked admitted the error. 701
622
We hypothesize that understanding object-relative clauses is
hard because of difficulties in representing the order of the two
NPs when they are confusable; little problem is seen with pronouns
or names because they are not confused with definite descriptions.
The next two experiments attempt to determine the semantic basis of
confusability, increasing the difference between the NPs by having
an indefinite NP in the embedded clause and by having the two NPs
match or mismatch in grammatical number.
Definite
Indef.
The senator that attacked the/a reporter admitted the error. 601
541
The senator that the/a reporter attacked admitted the error. 726
629
Match
Mismatch
The senator/s that attacked the reporter/s admitted the error. 458
466
The senator/s that the reporter/s attacked admitted the error. 548
532
Neither of these manipulations caused a significant reduction in
the subject-object difference suggesting that neither
distinctiveness of the NPs on the structural semantic
characteristic of definite-indefinite nor the lexical semantic
characteristic of number is sufficient to reduce confusability.
Consideration is given to other possible semantic bases for the
non-confusability of pronouns and names with definite
descriptions.
Gibson, E. (1998). Linguistic complexity: Locality of syntactic
dependencies. Cognition, 68(1), 1-76.
King, J., & Just, M. A. (1991). Individual differences in
syntactic processing: The role of working memory. Journal of Memory
& Language, 30(5), 580-602.
MacWhinney, B., & Pleh, C. (1988). The processing of
restrictiverelative clauses in Hungarian. Cognition, 29(2), 95-141.
|
