Illusory correlation experiments indicate that people overestimate the association between random presentations of snake slides and shock, but do not overestimate the association between random presentations of slides of damaged and exposed electric outlets (DEEOs) and shock. To investigate whether reports of covariation biases might be attributable to expectancy biases, we had Ss rate the a priori probabilities with which they would expect slides of snakes (or DEEOs), flowers, and mushrooms to be paired with shock, a tone, or nothing. In Study 1, Ss reported a pattern of a priori slide/outcome probability estimates that is nearly identical to that reported by Ss who have just undergone an illusory correlation procedure involving phylogenetic fear-relevant stimuli (e.g. snakes). Therefore, postexperimental estimates of covariation involving such stimuli appear at least partly attributable to pre-experimental expectancy biases rather than solely attributable to on-line processing biases. Study 2 revealed that Ss also display inflated a priori probability estimates for DEEO slides and shock, unlike Ss who have just undergone an illusory correlation procedure involving such stimuli. Taken together, these studies suggest that random slide/outcome pairings easily abolish pre-experimental expectancy biases for ontogenetic, but not phylogenetic, fear-relevant stimuli.