Three experiments used process measures of concept activation to provide ev
idence-consistent with previous research, which had used only outcome and s
elf-report measures-that people use category exemplars to assess their soci
al category attitudes. The studies were based on the well-established princ
iple that individuals selectively activate different aspects of their knowl
edge about a topic, depending on which aspect is relevant for a specific ju
dgment. Exemplar activation was measured through ambiguous anagrams (Experi
ment 1), word fragment completions (Experiment 2), and response latencies (
Experiment 3). In each case, exemplar activation was greater when participa
nts had recently judged their attitudes toward a social category than when
they had recently judged a definition of the category. implications for the
ories of attitude change and attitude-behavior consistency are discussed.