摘要：A usual criticism of the theory of infinitely repeated games is that it does not provide
sharp predictions since there may be a multiplicity of equilibria. To address this issue we
present experimental evidence on the evolution of cooperation in infinitely repeated
prisoners¡¯ dilemma games as subjects gain experience. We find that cooperation
decreases with experience when it cannot be supported as an equilibrium outcome. More
interestingly, the converse is not necessarily true: cooperation does not always increase
with experience when it can be supported as an equilibrium outcome. Nor is a more
stringent condition, risk dominance, sufficient for cooperation to arise. However, subjects
do learn to cooperate when the payoff to cooperation and the importance of the future is
high enough. These results have important implications for the theory of infinitely
repeated games. While we show that cooperation may prevail in infinitely repeated
games, the conditions under which this occurs are more stringent than the sub-game
perfect conditions usually considered.
