|
Abstract:
One account of sentence processing places the burden on
lexical properties (MacDonald et al., 1994). With this in mind,
we investigated gap filling in VP-ellipsis constructions
containing verbs that were either obligatorily reflexive or
required an inalienably possessed body part object:
1. The optometrist who had signed the release [1] form
asserted herself [2], and the pilot who needed to pass [3] the
training exam did [4] too.
2. The gambler who won ten hands in [1] a row winked his [2]
eyes, and the pit boss who was in on [3] the elaborate scheme did
[4] too.
In (1), the verb "assert" allows only the reflexive reading;
only the 'sloppy' interpretation is allowed in the second clause
(the "pilot" can only assert herself, and not the "optometrist").
Similarly, in (2) the verb "wink" requires an object (a body
part) that is possessed by the subject NP (the "pit boss" cannot
typically wink the "gambler's" eyes). Though other continuations
exist ("assert control," "wink" as intransitive), we found that
off-line judgments from an independent subject group presented
with sentences like (1) and (2) revealed that only the strict
interpretation was available.
Sentences were presented over headphones with lexical decision
probes presented visually. The probes were either related to the
subject NP of the first clause (strict reading) or to the subject
NP of the second clause (sloppy reading), or were unrelated
controls; these were presented in one of four positions, as shown
in (1) and (2). If lexical and contextual properties influence
gap-filling, we should observe only the sloppy reading at the
gap.
We found that at position [2], re-access of the subject NP was
observed (no activation of the subject NP was found at position
[1]). We also found priming at the gap ([4]) but not pre-gap
([3]) position for the subject NPs from the first and second
clauses. Thus, both sloppy and strict interpretations were
available at the gap. One possibility is that we have observed
end-of-clause 'wrap-up' effects where all NPs could be active.
However, only a syntactically relevant filler is (re)activated at
a gap, even when the gap is at a clausal boundary (e.g., Love and
Swinney, 1996). Furthermore, Shapiro and Hestvik (1995) found
that subordinated ellipsis constructions do not elicit re-access
of the subject NP at the gap/clausal boundary, only later. We
interpret our findings in light of constraint-satisfaction models
and alternative accounts that suggest that gap filling is driven
by syntactic considerations.
|
