Journal of Economic Behavior & OrganizationVol. 47 (2002) 275–290Identifying cooperative behavior: some experimentalresults in a prisoner’s dilemma gameJeannette BrosigFaculty of Economics and Management, Otto-von-Guericke-Universität Magdeburg,P.O. Box 4120, D-39106 Magdeburg, GermanyReceived 30 March 1999; received in revised form 18 May 2001; accepted 29 May 2001AbstractIn recent years, experimental economists have discovered that people exhibit different patternsof cooperative behavior. This paper presents findings from a face-to-face experiment that analyzedwhether individuals who possess a willingness to cooperate can credibly signal it and whether it isrecognizable by the partner. Results revealed that both capabilities, signaling and recognizing, de-pendupontheindividual’spropensitytocooperate.©2002ElsevierScienceB.V.Allrightsreserved.JEL classification: C91; C78Keywords: Experimental economics; Communication; Cooperation; Fairness1. IntroductionThe investigation into the cooperative behavior exhibited by individuals constitutes a ma-jor research topic in experimental economics. Cooperativebehavior is manifested wheneverindividuals maximize the joint payoff, in lieu of their own payoff, in prisoner’s dilemmaand public good games. When observed in these games, this behavior attracts attentionfor it runs contrary to the more self-centered behavior predicted by standard game theory.For this reason, an extensive literature has developed attempting to explain this behavioralpattern. Various divergent theoretical explanations have been proposed, including “pure”as well as “warm glow” altruism (Andreoni, 1993; Andreoni and Miller, 1995; Offermanet al., 1996), the relative payoff position of individuals (Fehr and Schmidt, 1999; Bolton andOckenfels, 2000), “cooperative gain seeking” (Brandts and Schram, 1996), different sortsof reciprocity (Rabin, 1993; Dufwenberg and Kirchsteiger, 1998; Falk and Fischbacher,1998), “quasi-maximin preferences” (Charness and Rabin, 2000), and the combination ofaltruism with error effects (Palfrey and Prisbrey, 1997; Anderson et al., 1998).E-mail address: [email protected] (J. Brosig).0167-2681/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved.PII: S0167-2681(01)00211-6276 J. Brosig / J. of Economic Behavior & Org. 47 (2002) 275–290The majority of theories that seek to explain cooperative behavior postulate the exis-tence of different types of individuals, each type exhibiting a distinctive pattern of behavior.The existence of these different behavioral types has been confirmed by experimental find-ings. For example, in a finitely repeated one-shot prisoner’s dilemma game, Andreoni andMiller (1993) observed that some subjects always chose to cooperate with their partnerswhile others either chose not to cooperate or played a mixed strategy. Similar results werereported by Cooper et al. (1996). In public good experiments, Weimann (1994) system-atically identified three types of subjects. Each of these types made differing decisionsregarding the quantity and timing of the contributions made. The same types of individualswere observed in the public good experiments conducted by Isaac et al. (1994).1Analyzingbehavior in a centipede game, McKelvey and Palfrey (1992) also concluded that there ex-isted a smallfractionofsubjects who decided to cooperate throughout the gamewhileothersdid not.Recently, evolutionary game theory has been developed into a helpful tool for explainingthe existence of these different types of individuals. According to this theory, only thosebehavioral patterns that produce a relative fitness advantage for an individual will survivethe selection process (i.e. those behavioral patterns that are reproductively successful).2Based on this theory, it seems very unlikely that cooperation would be a selected behavioralpattern, because it enhances the fitness of non-cooperative individuals rather than assistingthe cooperative ones. This is only true, however, as long as no information is availableconcerning the types of individuals being confronted with that would allow for selectedinteractions to occur.3In order to justify the existence of cooperative types, evolutionarymodels assume that relevant information can be communicated in different ways. In par-ticular, some of these models postulate the existence of signals indicating a partner’s type,and thus, offer explanations for cooperative behavior even in one-shot situations (Robson,1990; Yang, 1993; Amann and Yang, 1998).4These signals are only useful, however, ifthey can be communicated successfully. This makes communication a necessary conditionfor the evolution of one-shot cooperation.51This conclusion is based on my own calculations utilizing data graciously provided by James M. Walker.2Evolutionarygametheory usesthe equilibrium conceptof evolutionarystability thatdefinesonly thosestrategiesas evolutionary stable that have a higher reproductive success rate than any other mutant strategy that might arise.Fora more detailed description, see Weibull (1995). Hammerstein and Selten (1994) provide an excellent overviewof the biological applications.An introductioninto therapidly developingfield ofevolutionarypsychology isgivenby Barkow et al. (1992), and Buss (1999).3In a bargaining experiment, Charness (2000) showed that providing information about the allocations chosenby subjects in a dictator game increased the payoff to those, who were more generous dictators. Cain (1998)obtained similar results in a prisoner’s dilemma game after subjects were informed about their partner’s chosendictator allocation. Interestingly, in our research, fair behavior in the dictator game is per se not related to jointpayoff maximization.4Other models that explain the evolution of cooperation are based upon the assumption of repeated interactionusing some form of reciprocity (see for e.g. Kreps et al., 1982; Guttman, 1996, as well as the learning modelproposed by Vogt, 2000).5The fact that communication has a considerable influence on cooperative behavior has been reported in variousexperimental studies. In prisoner’s dilemma games, Dawes et al. (1977), Orbell et al. (1988), as well as Bohnetand Frey (1995) found that pre-play communication enhanced the number of decisions to cooperate. Similarobservations were reported in experiments on common pool resource games (Ostrom and Walker, 1991; Ostromet al., 1994; Hackett et al.,