Emotional response categorization - the grouping of stimuli on the basis of the emotional response they elicit - is an imporant complement to traditional accounts of categorization that rely on stimulus-based similarity. However, previous demonstrations of emotional response categorization have left open the possibility that stimuli were associated on the basis of their association with similar emotions in the past, or on the basis of semantic emotion knowledge, rather than on the basis of common emotions elicited at the time of judgment. In contrast, in the current study, mere exposure was used to 'attach' positive affect to neutral individuals. Participants judged which of two comparison individuals, one of whom had been previously seen in the experiment, was most likely to be friends with (or, for different participants, to have the same job as) a target individual, who himself displayed either a happy or a neutral expression. Participants were more likely to judge previously-seen individuals (who elicited positive affect through mere exposure) as friends with smiling targets (who elicited positve affect through emotion contagion), but were indifferent to emotional response when judging jobs. Because these individuals had no pre-existing emotional associations prior to the study, selective choice of positive comparision individuals represents the first demonstration of true momentary emotional response-based categorization. Several explanations of the domain specificicty of emotional response categorization are considered, including the honest predictiveness of emotional response for judging friendship, the relative importance of emotional response compared to other stimulus dimensions, and the selective engagement of an associative reasoning strategy.