This thesis uses evolutionary computation to study Prisoner's Dilemma strategies. The first part concerns the theory of a tool used to analyze strategies, called 'fingerprints'. Fingerprints are calculated for some important strategies, and two lemmas, five theorems, and four corollaries are proven. The second part analyzes simulations in which strategies are given information about their opponent's fingerprints. The strategies evolve to utilize this information. Some populations develop strategies that cooperate with strategies with fingerprints different from their own and compete with strategies with similar fingerprints; others consist entirely of strategies that play tit-for-tat unless their opponent's fingerprint differs from tit-for-tat, in which case they compete in a variety of different ways. Some strategies also evolve a sort of dishonest signaling that causes their behaviour to differ from their fingerprint signal.