|
Abstract:
Which-N' phrases are d(iscourse)-linked: (1b) is only
felicitous if "mistakes" is already present or implied in the
discourse (Pesetsky 1987):
(1) a. What did Ted realize ___
b. Which mistake did Ted realize ___
(object: after he finished the paper, embedded subject: was
causing the problem)
I provide evidence that English which-N' phrases
preferentially occur in the subject position, and argue that this
follows because subject is the default topic in English, and
subject which-N' phrases are interpreted as "topical."
Under the Garden-Path model, object interpretation of the
wh-phrase in (1) should be easier: it allows for a more minimal
structure and earlier assignment of the filler. This was found
for (1a) both in an off-line completion study and in self-paced
reading, where the sentence-fragments and the appropriate
continuations were presented word-by-word. However, the subject
interpretation was preferred in both experiments for the which-N'
phrase (1b). This is as expected if the d-linked wh-phrase
preferentially occurs in the default topic position.
Further data show the topicality of which-N' in subject
position. "Who" as antecedent for reflexives is "dispreferred" in
questionnaire studies in ambiguous examples, and takes longer to
comprehend in self-paced reading in unambiguous examples like (2)
(Rado 1994).
(2) a. Who/Which boy did the neighbors say talked to Frances
about himself?
b. Who/Which boy did the neighbors say Frances talked to about
himself?
In English, which-N' doesn't behave like who. When it appears
as object, as in (2b), it is a bad antecedent, just like who. As
a subject, however, which-N' was found to be the preferred
antecedent of the reflexive. The results follow assuming that
topics and already given entities are prototypical antecedents
for reflexives. This conclusion is further confirmed by Hungarian
data testing the counterparts to (2) in questionnaire and pilot
on-line studies. Hungarian has syntactic topic and focus
positions: wh-constituents must appear in focus, not in topic. By
hypothesis the d-linked feature alone is not sufficient to make a
wh-phrase a preferred antecedent, thus which-N', just like who,
was predicted to be dispreferred both as subject and as object.
Indeed, who and which-N' were found to be equally dispreferred
antecedents for reflexives in Hungarian.
Finally, since the subject is not a default topic in
Hungarian, which-N' is expected to behave just like who in
gap-filling, both preferring object interpretation in
counterparts to (1). Pilot results confirm this prediction.
|
