In a study of the stability of social interaction, participants maintained a social interaction diary, the Rochester Interaction Record (RIR), for 4 1-week periods during their freshman year at college. Stability was operationalized in 3 ways: absolutely, in terms of the similarity across the 4 periods of amount of interaction and of reactions to interactions; relatively, in terms of correlations between interaction measures taken at different times; and in terms of the stability of social networks, defined as the consistency across time of participants' close friends. Social interaction was found to be more stable over shorter periods of time than over longer periods, and stability increased over time. In addition, opposite-sex social interaction was less stable than same-sex interaction, and this was particularly true for interactions with close friends. The present results are interpreted by considering social enviornments as social systems with an emphasis on the importance of social norms as mediators of social interaction.