Schott et al. (2007) have shown that the 'tragedy of the commons' can be overcome when individuals share their output equally in groups of optimal size and there is no communication. In this paper we investigate the impact of introducing communication groups that may or may not be linked to output sharing groups. Communication reduces shirking, increases aggregate effort and reduces aggregate rents, but only when communication groups and output-sharing groups are linked. The effect is stronger for fixed groups (partners treatment) than for randomly reassigned groups (strangers treatment). Performance is not distinguishable from the no-communication treatments when communication is permitted but subjects share output within groups different from the groups within which they communicate. Communication also tends to enhance the negative effect of the partnered group assignment on the equality of individual payoffs. We use detailed content analysis to evaluate the impact of communication messages on behavior across treatments.