Metaphors of the form X is a Y may be conceived as statements of prope
rty attribution, in which properties of the vehicle Y are attributed t
o the topic X. The properties attributed from the vehicle to the topic
are those that are epitomized by the vehicle, and may characterize a
dimension of within-category variation in the topic. We report two exp
eriments that tested the implications of this ''property attribution''
view of metaphor comprehension. One implication of this view is that
metaphors are not implicitly transformed into comparisons (X is like a
Y), but rather are understood as what they appear to be: class-inclus
ion assertions. If metaphors are understood as class-inclusion asserti
ons, they should either change meaning or become nonsensical when reve
rsed. Experiment 1 provides evidence for this claim. A second implicat
ion of the property attribution view is that topics and vehicles provi
de different kinds of information to guide and constrain the comprehen
sion process. Experiment 2 provides evidence for this claim. These fin
dings cannot he accounted for by models which assume that metaphors ar
e understood as implicit comparisons. (C) 1997 Academic Press.