Seeing clearly: Frame Semantic, Psycholinguistic, and Cross ...

CHAPTER 4. PSYCHOLINGUISTIC EXPERIMENTS 161
imagined a dictionary would have and others did not, but simply tried to construct their
own categories in their own way, this might produce variation across subjects in the type of
categorization. It seems unlikely that enough subjects carried out one such aninterpretation
of the instructions consistently enough to substantially a ect our results.
As noted earlier in this chapter, we expect to nd individual variation in internal-
ized semantic structures for see across speakers; these patterns would be expected to follow
the usual lines of geographical and social dialects, with educational level playing a major
role, but we were not able to determine anything about such distribution in this study. To
do so would require many more subjects from much more diverse backgrounds. Nor do these
experiments provide any evidence as to the relative importance (or temporal precedence)
of semantic vs. syntactic factors. As noted in Chapter 2, many senses have quite speci c
restrictions (syntactic and/or semantic) on their arguments, such asaccompany or deter-
mine. The subjects may be learning to distinguish at least some of senses from relatively
straightforward cues which are primarily syntactic (although the assumptions that syntactic
cues (a) can be easily separated from semantic cues and (b) are more straightforward than
semantic cues may becharacteristic of some linguists rather than most other speakers).
Even given these caveats, there are still a number of signi cant conclusions to be
drawn; let us review our predictions in the light of our results:
1. Since see is highly polysemous, we predicted that our subjects would produce more
senses on the sorting task than the subjects in Jorgensen 1990.
In Experiment 1, the mean number of categories per subject was 11.6, substantially
higher than the mean of 5.6 found by Jorgensen (1990), as we had predicted. For
Experiments 2 and 3, the median numbers of categories produced were 6 and 10,
which approximate the number of senses intended by the experimenters, i.e. 7 and
14 respectively. The di erence between the two medians is signi cant, meaning that
subjects recognized that more senses were present in Experiment 3 on the basis of the
stimuli alone.
2. We predicted broad agreement among our subjects as to what the central sense(s) of
see is.
Most of the subjects chose as a central sense either a category corresponding to
eye+faculty, oreye alone if they separated faculty from it.