Questions are raised about the possible ecological invalidity of the traditional “who said what?” (WSW) procedure. Three studies are reported which use a modified, quasi-naturalistic version of the WSW technique, in which rather than being passive recipients of presented information, participants instead play an active role in the production of such information in an interactive context. Data are presented that demonstrate that the social categorization effect typically obtained through the traditional WSW procedure generalizes beyond the circumstances in which it was initially demonstrated. Thus, in Study 1 it was established that social categories (gender, in this case) are cognitively used in real interactions between participants in order to organize information. Moreover, Study 2 demonstrated that the use of categories is accentuated when, as is typical in real world situations, recollections are not made immediately after the perception of the relevant stimuli but following a period of intervening cognitive activity. The findings of Study 3, however, suggest that the WSW technique may not inevitably lead to social categorization effects: typical effects did not appear when participants knew each other well, and therefore saw one another as unique, idiosyncratic individuals, rather than as category members.