In contemporary dance, cognitive events are not necessarily restricted "to the skin or skull of an individual" (Hutchins in Int Encycl Soc Behav Sci 2068-2072, 2001) but distributed across dancers during collaborative improvisation. There is some experimental evidence of greater output when people perform problem-solving tasks alone. However, when a task is challenging and paired participants are familiar with each other, pairwise and emergent outcomes are more plentiful than solo outcomes. We investigate these factors in the context of dance with the broad hypothesis that innovation is enhanced when dancers improvise together compared with when they improvise alone. Dancers (N = 10) in a professional company improvised for 2 min alone and then with another dancer. Dancer familiarity (familiar, unfamiliar) and task (expressive, non-expressive) were crossed (within-subjects). The improvisations were video-recorded over 2 h in the dancers' usual improvisation space. After each improvisation, the dancers: stated the number of movement ideas expressed and rated task ease, satisfaction, interest, novelty, originality and clarity. In both tasks, there was a tendency for self-report of a greater number of movement ideas expressed in familiar and unfamiliar pairs than alone. Ratings of task ease, satisfaction, interest, clarity, etc. were slightly higher in the unfamiliar pair condition. In the non-expressive task, ratings of the task were higher in pairs (M = 3.02, SD 0.82) than in the solo (M = 2.67, SD 0.96) condition. Distributed creativity, relational cognition and social facilitation are used to interpret the results.