Dancing is a universal human activity that involves exertive rhythmic movement to music. It is often conducted in a social environment and often involves synchronization. It has been found to cause dancers to bond socially. Like conversation, it has been suggested that dancing may be an inexpensive form of social bonding, in that both activities facilitate efficient group bonding by allowing multiple individuals to bond simultaneously. However, no previous study has systematically observed the size of naturally occurring dance groups. During unobtrusive observation of natural dance and conversation behavior, we found that the cumulative number of dance partners (cumulative group) was greater than the number of partners at any one moment (instantaneous group), whereas no such difference was found for conversation. Additionally, the length of uninterrupted engagement (bout) was negatively predicted by group size in conversation but not dance groups, and it was significantly longer in conversation groups than dance groups. Finally, instantaneous group size was significantly larger in dance than conversation groups, and also positively related to time spent synchronizing in dance groups. Together, these results suggest that dance may allow a larger number of individuals to simultaneously engage with each other than conversation does because (i) more rapid partner switching increases cumulative broadcast group; and (ii) synchrony facilitates simultaneous interaction with multiple individuals, allowing for larger instantaneous groups. We conclude that the capacity for information transfer provided by language comes at a cost in terms of social bonding, and that dance may have played an important role in bonding large hominin social groups.