This study compared the effects of exposure to masked and unmasked phobic stimuli on phobic behavior. Participants were identified as spider-phobic with a questionnaire and a Behavioral Avoidance Test (BAT) with a live tarantula (N = 101). One week later, they were administered one of three types of exposure: very brief (25 ms) or clearly visible (120 ms) images of spiders, or very brief images of flowers. They reported ratings of subjective distress just before and after these exposures, and engaged in the BAT again thereafter. Two weeks later, 57 participants returned for a follow-up BAT. The results indicated a double dissociation between the effects of very brief and clearly visible exposure: the former reduced avoidance of the tarantula and did not affect distress, whereas the latter increased distress but did not affect avoidance. The behavioral effect lasted for two weeks. These findings suggest that avoidance of a feared object can be reduced without full conscious awareness. The theoretical implications are discussed.