Abstract This study investigated whether response variability and turn taking were controlled by the play partner's responding. Variability and turn taking was assessed in two conditions following an ABA experimental design. Participants were three preschool aged children with autism and eight nondisabled preschool aged children. During the A-condition, the partner emitted play responses that were topically related to the participant's preceding play response, and during the B-condition, the partner emitted play responses that were topically unrelated to the participant's preceding play response. Results showed marked reduction in play variability and turn taking during the B-condition for all participants, demonstrating that variability and turn taking in cooperative play may be affected by the partner's play during the interaction. Key Words: cooperative play, variability, turn taking, typical children and children with autism. Introduction The analysis of social interaction in terms of both structural and functional properties is critical as a basis for understanding the occurrence of enduring social interaction and for the development of effective programs to facilitate social interaction in people with limited skills in this area. Behaviors that are evoked and maintained by properties of one or several other persons behavior are usually considered social, and the contingencies in which these behaviors are evoked are usually labeled social interaction (Skinner, 1957). Reciprocity, as an overarching concept, is said to be the nature of such interaction (Strain, Shores, & Timm, 1977) indicating a dependent relation between the responses involved. Functional aspects of social interaction have been discussed by several researchers, and often within the context of cooperation (e.g., Guerin, 1994; Hake, & Olvera, 1978; Schmitt, 1984). A cooperative contingency is according to Hake and Olvera (1978) identified when the reinforcers for each person involved is at least partially dependent upon the other person's responses. Reaching such a conclusion however, may require a further analysis of structural features of response patterns that occurs between the partners involved. Especially since an interaction as such seldom involves arbitrary reinforcers (e.g., food or money) but idiosyncratic responses, supposedly containing both reinforcing and discriminative stimulus functions. Jahr, Eldevik and Eikeseth (2000) argued that response alternation and response variability constitute some of the structural properties of a cooperative contingency from which functional properties may be derived. In this study, a procedure to establish cooperative play in children with autism was investigated and later replicated with typical children as participants (Jahr & Eldevik, 2002). The training elements included verbal description and imitation of modeled play episodes, utilizing a multiple exemplar strategy. Following training, participants in both studies could initiate play episodes and sustain episodes initiated by their play partner, and they were able to take turns in play episodes that contained a larger number of play responses than were practiced during training. Moreover, a considerable increase in play variability was evident for each child. Hence, the children seldom repeated their own play responses or persevered on specific objects. Instead they utilized a variety of play materials and their play responses were usually an adequate supplement to their play partners' responses and thereby a thematic extension of the play episode. That the children and their partner were playing within the same topic was confirmed by independent raters during social validity measures. The observed increase in turn taking and the emergent ability to vary responding in accordance to both the play partners' response and own previously emitted responses was assumed to indicate a functional relationship between the responses involved (Jahr et al. …