Published online by Cambridge University Press: 14 February 2005
Four models commonly employed in sharing analyses (reciprocal altruism [RA], tolerated scrounging [TS], costly signaling [CS], and kin selection [KS]) have common features which render rigorous testing of unique predictions difficult. Relaxed versions of these models are discussed in an attempt to understand how the underlying principles of delayed returns, avoiding costs, building reputation, and aiding biological kin interact in systems of sharing. Special attention is given to the interpretation of contingency measures that critically define some form of reciprocal altruism.