Predictions derived from misattribution and negative-reinforcement theories concerning the mediators of attraction in aversive environments were examined. Male subjects, expecting to receive either high- or low-intensity shocks, participated with either a male or female confederate who was or was not present during the description of shock intensity. Subjects' arousal, attributions for their arousal, and attraction to the confederate were measured. Results were not consistent with the misattribution hypothesis. The negative-reinforcement hypothesis accurately predicted the arousal and attribution measures. The presence of the confederate did reduce the perceived aversiveness of the shock and made the subjects feel more calm. Additionally, subjects attributed their arousal to the threat and a calming influence to the confederate. Attraction to the confederate was most consistent with an incorporation of both negative-reinforcement and classical conditioning processes. Confederates who were absent during the high-shock threat were found to be most attractive and significantly more attractive than those who were present during the threat. This suggests that although both confederates reduced arousal, the one present during threat also took on some of the negative qualities of that threat, as would be expected to result from classical conditioning processes.