|
Abstract:
Coordinate structures entail a local ambiguity which provides
fertile testing ground for theoretical models of sentence
processing. In a sentence such as "Mary bought vegetables and..."
it is not apparent whether "and" must be attached low in the
phrase structure tree to conjoin two noun phrases, whether it
must be attached higher in the tree to conjoin two verb phrases,
or even higher to conjoin two clauses.
Gibson, Schutze, and Salomon (1996) and Cowart (1991)
independently noted that the word "both" often marks coordinate
structures. Of course, "both" serves many functions, including
pronoun (I want both.), (pre-)determiner (I want both desserts.),
and quantifier (I both want and need dessert.). Although in some
environments "both" need not mark the onset of coordinated NPs,
in environments such as (1b) it does signal an upcoming
coordinate structure.
We collected sentence completions and conducted a word by
word, "stop making sense," reading task to investigate how "both"
influences the attachment site of "and."
(1a) Mary sold vegetables and bought seeds at the fair.
(1b) Mary both sold vegetables and bought seeds at the
fair.
(1c) Mary sold vegetables and seeds at the fair.
(1d) Mary sold both vegetables and seeds at the fair.
The online results showed the expected garden path effect at
"bought" in (1a) compared to the same word position, "seeds," in
(1c). This result is compatible with most theories of sentence
processing. For example, Late Closure (Frazier, 1978) predicts
processing difficulty when the second conjunct is a VP since
"and" must be detached from the most recent constituent and
reattached higher up the tree. In other theories, processing
difficulty might be predicted because of an additional event in
the discourse model or an overall preference for coordinated
NPs.
Surprisingly, while prenominal "both" (1d) increased the
number of coordinated NPs in the completions, preverbal "both"
(1b) did not reliably increase the number of coordinated VPs. In
the online task, participants actually preferred 1A to 1B. Thus,
both the sentence completions and "stop making sense" data show
that preverbal "both" is not a lexical marker for coordinated VPs
in this subject pool. Rather the scope of preverbal "both" may be
the same as that of "just" or "only": Mary (only/just) sold
vegetables, not fruits.
Cowart, W. (1991). Mental representations of conjoined noun
phrases. Manuscript, Language Sciences Laboratory, University of
Southern Maine.
Frazier, L. (1978). On comprehending sentences: Syntactic
parsing strategies. Doctoral Dissertation, University of
Connecticut.
Gibson, E., Schutze, C., and Salomon, A. (1996). "The
relationship between the frequency and the processing complexity
of linguistic structure."
Journal of Psycholinguistic Research,
25:1, 59--92.
|
