Conclusion

Cooperation is found in most taxa. One has to keep in mind, that cooperation never was a goal but rather a better solution than others to problems that undoubtedly occurred frequently during the evolution of life. Why?

On the foregoing pages I tried to draw a short sketch of the current state of research on the evolution of cooperation. As we have seen in the first four sections, this interest in the evolution of cooperation has led to independent theoretical work on the IPD for its own sake. An outstanding performance of theorists in the last decade has provided us with a large number and diversity of papers investigating or elaborating theoretical concepts within the IPD-framework.
Nevertheless, "our understanding of the evolution of cooperation is at such an elementary stage as to suggest that additional paradigms remain to be developed" (Bull and Rice 1991). That hitherto no biological system could be found in which all the conditions for an IPD were met, clearly demands new approaches, which will be more readily testable. In more general models, random mutations might provide its bearers with traits that could change environmental, physiological and developmental parameters. Environments, selection pressures or mutation rates are neither fixed nor imposed, inaccessible to evolution, but rather highly dynamic variables. I agree with Dugatkin et al. (1992) and Mesterton-Gibbons and Dugatkin (1992) that "it is time for the theoretical work to go beyond the [...] IPD" towards paradigms that include reciprocal altruism, by-product mutualism and other potential categories of cooperation between nonrelatives. The aim, of course, is to combine such new paradigms to a model, that would provide a powerful tool to investigate, under which precise conditions, which forms of cooperation could evolve.
Assume that an observed behaviour corresponds to one solution of the game. Then, one can try to find the conditions under which this observed behaviour is evolutionarily stable. In this context, the stability question can be put at several levels: the stability of single strategies (Lorberbaum 1994), the stability of relative frequencies in an ensemble of strategies (Feldmann and Thomas 1987), the stability of oscillations in these frequencies (Nowak and Sigmund 1993b), etc. The construction of such a general model would eliminate several weaknesses of the IPD as well (Dugatkin et al. 1992, Mesterton-Gibbons and Dugatkin 1992):

1. It would allow the exchange of information between the interactants. Only few higher species ignore signals (of willingness to cooperate or threat to break the partnership) from their opponents during interaction (see Noë 1990 for a model including information exchange).
2. It would allow to change the geometric probability distribution of THETA from Prob(THETA)= w^THETA-1 to a behaviour dependent distribution. The formation of preference between successful partners will certainly flatten the steep decline of Prob(THETA) with THETA (see Feldmann and Thomas 1987, Noë 1990).
3. While the IPD models the maintenance of cooperation fairly well, it still fails to elucidate the origin of cooperation from an asocial state (see Feldmann and Thomas 1987).
4. In the IPD, there are only two choices: C or D. Corresponding punishment of different nuances of defection seems more realistic (see Clutton-Brock and Parker 1995 for the role of punishment, and Leimar and Axén 1993 for a discussion on gradually reactive behaviour).
5. A new model would enlarge the scope to N-player games (see Dugatkin 1990, Noë 1990) in order to model the evolution of social groups.

However, evolution is a historical process (see also Quenette and Gerard 1993) acting on a dynamical landscape, producing unique solutions to each new problem. Simulations of evolution should also be expected to have different outcomes for every run.