This analysis is amplified by Rosch (1978; Rosch and Mervis 1975), who shows experimentally that categorization
judgments may contain a clineof“typicality,”ranging fro mtypical exe mplarsof the category (e.g. a robin is a“typical”
bird) to atypical exemplars (e.g. a penguin is an“atypical”bird). There are various sources for typicality effects; but
a mong the mis a set of conditions that for ma cluster concept.^183 Examples that satisfy fewer of the conditions are
generally regarded as less typical than examples that maximally satisfy the conditions.
The effects of cluster concepts can be observed in concepts involving as few as two conditions. A casefirst discussed
by Fillmore (1982) is the verbclimb. Consider the following examples:
(10) a.Bill climbed (up) the mountain.
b. Bill climbed down the mountain.
c. The snake climbed (up) the tree.
d. ?*The snake climbed down the tree.Climbinginvolves twoindependent conceptual conditions: (a)an individual is travelingupward, and (b)theindividual is
moving with characteristic effortful grasping motions (clambering). On the most likely interpretation of (10a), both
conditions are met. (10b) violates thefirst condition and, since snakes cannot clamber, (10c) violates the second; yet
both examples are acceptable instances of climbing. However, if both conditions are violated, as in (10d), the action
cannot be characterized as climbing. Thus neither of thetwoconditions is necessary, but either is sufficient. Moreover,
the default interpretation of (10a), in which both conditions are satisfied, is judged to be more prototypical climbing;
(10b) and (10c) are judged somewhat more marginal but still perfectly legitimate instances.
One possible account of this would be to say thatclimbis ambiguous or polysemous between the readings‘rise’and
‘clamber.’But this does violence to intuition: (10a) is not ambiguous between these two senses. Rather, other things
beingequal, it satisfies both of them. Another possibleaccountwould be to say that the meaning ofclimbis the logical
disjunction of the two senses:‘riseorclamber.’But this is too crude: a disjunction isn't“more prototypically satisfied”
LEXICAL SEMANTICS 353
(^183) Other sorts of typicality judgments occur, for instance, in color concepts, where a focal red is judged more typical than a red tinged with orange. Armstrong et al. (1983)
show that typicality judgments can be obtained even for sharply defined concepts; e.g. 38 is judged a less typical even number than 4, and a nun is judged a less typical
woman than a housewife.Thus typicalityin and of itself is only a symptom for a number of underlying phenomena in categorization. But the existenceof multiple sources
for typicality does not undermine the existence of cluster concepts, as Armstrong et al. claim it does. See Jackendoff (1983 : ch. 7, n. 6) and Lakoff (1987 : ch. 9).