The variability of the information on which a belief was initially bas
ed should moderate the degree to which the belief is revised when new
disconfirming information is received. Reasoning from everyday intuiti
ons, psychological theories of concept representation, and a rational
mathematical analysis, we inferred that as variability increased, resi
stance to disconfirming information would also increase. Participants'
judgments of the central tendencies of distributions of numerical inf
ormation confirmed this prediction.