A modified Stroop task incorporating a semantic manipulation was used to study processing of fear-related information in patients diagnosed with obsessive-compulsive disorder (OCD). Twenty-three OCs with washing rituals (washers), 10 OCs without washing rituals (nonwashers), and 14 normals were administered a modified Stroop task in which they were asked to color-name contamination words, general threat words, neutral words, and nonwords. Each word was preceded by one of four priming stimuli in a randomized order; XXXXX, danger, disturb, fruit. OC washers evidenced longer response latencies to contamination words than to neutral words. Also, their latencies to contamination words were longer than the latencies of OC nonwashers and of normals, but only the latter difference reached significance. OC nonwashers evidenced longer latencies to general threat words than to nonwords and normals were slower in color naming neutral words than either contamination or general threat words. The hypothesis that all subjects would show an effect of priming for the fruit words was supported by the data; longer color-naming latencies were found for fruit words when preceded by the prime fruit. The hypothesis that washers would show longer latencies for contamination words when they were primed by danger than when they were primed by the nonword XXXXX was not confirmed. The theoretical implication of the findings related to the priming manipulation, the specificity of the modified Stroop effects, and their clinical relevance are discussed