| |
Abstract:
Linguists and psycholinguists have traditionally argued that
meaning comes primarily or exclusively from words. On this view,
syntactic structures reflect the meanings of the words which
inhabit them, but lack semantic identities of their own. In this
poster, I present evidence that the subcategorization frames of
verbs carry their own semantics, independent of the words which
appear in them.
I presented subjects with several common frames whose content
words had been converted to nonsense and asked them to rate the
likelihood that the verb involved different event-related
properties identified by linguists (e.g., Jackendoff, 1983; Levin,
1993; Pinker, 1989) as relevant to verb semantics. For example,
after reading the sentence
The rom gorped the blick to the dax,
subjects had to rate how likely it was that
gorping
involved someone or something "changing possession", "moving",
"engaging in mental activity", "causing someone or something else
to do something", and so on. As predicted, subjects assigned
specific meanings to each verb as a function of its frame. To the
nonsense verb in the transitive frame, for instance, they ascribed
contact and causation, while to the nonsense verb in the sentence
complement frame they ascribed perception and mental activity (but
not the reverse).
To the extent that subcategorization frames have their own
semantics, the goodness of fit between a verb and a frame should
depend importantly on the semantic similarity between the two. To
test this hypothesis, I presented a series of real verbs to a new
set of subjects, who once again rated the likelihood that these
verbs involved different semantic properties. Another group of
subjects rated the naturalness of sentences which paired different
verbs and frames, with all content words except the main verb
turned to nonsense (e.g.,
The rom thought the blick to the dax).
With a couple of (important) exceptions, the semantic overlap
between a verb and frame predicted the naturalness of their pairing
fairly well. The power of overlap to predict naturalness varied
with the range of verbs which appear in that frame. For instance,
the correlation between overlap and naturalness was nearly zero for
the massively common transitive frame, but roughly 0.80 for the
less common sentence complement frame.
Together these results indicate that syntactic structures bear
meaning independent of words, and that comprehenders take into
account the overlap in meaning between verbs and their frames when
judging naturalness.
References
Jackendoff, R. (1983).
Semantics and Cognition.
Cambridge, MA: MIT Press.
Levin, B. (1993).
English Verb Classes and Alternations: A Preliminary
Investigation.
Chicago, IL: University of Chicago Press.
Pinker, S. (1989).
Learnability and Cognition.
Cambridge, MA: MIT Press.
|
