Three experiments used an illusory correlation paradigm to assess the effects of fear on the perception of the covariation between fear-relevant stimuli and shock. In Experiment 1, high- and low-fear women were exposed to 72 trials during each of which a fear-relevant (snake or spider) or fear-irrelevant (mushroom and flower) slide was followed by a shock, a tone, or nothing. Although the relation between slide types and outcomes was random, high-fear subjects markedly overestimated the contingency between feared slides and shock. Experiment 2 showed that this bias was due to the aversive, rather than more generally salient, features of shock. Low-fear subjects demonstrated biases equivalent to those of high-fear subjects only when the base rate of shock was increased from 33% to 50% in Experiment 3. It is concluded that fear may be linked to biases that serve to confirm fear. The relevance of the present findings to preparedness theory is also discussed.