Anim. Behav., 1997, 53, 1363–1364 Reply to Roberts: cooperation is an outcome, not a mechanism D. W. STEPHENS & J. P. ANDERSON Nebraska Behavioral Biology Group, School of Biological Sciences, University of Nebraska at Lincoln ( R eceived 20 M ay 1996; initial acceptance 24 A ugust 1996; final acceptance 15 October 1996; M S. number: -1187) Roberts’ (1997) commentary raises an important question about the definition of cooperation. We see his comments about ‘mutualism’ to be periph- eral to the main point, and his statement that Clements & Stephens’ (1995) evidence for mutu- alism is inadequate because it ‘. . . does not involve cooperation’ (page 1361) shows that cooperation and not mutualism per se is at issue. Students of animal behaviour use the word ‘cooperation’ in two distinct ways. When students of behaviour speak of ‘cooperative hunting’, ‘cooperative breeding’ or ‘cooperative territorial- ity’, they are usually talking about purely behav- ioural phenomena. For example, we observe ‘cooperative breeders’ breeding together and exhibiting some coordination in their breeding behaviour. In this use of the word ‘cooperation’, we are more likely to label actions as cooperative if they seem to be elaborately coordinated or if individuals seem especially sensitive to the actions of others. Since 1981 (Axelrod & Hamilton 1981), game theoretical models have been a large part of the literature of animal cooperation. The main ana- lytical tool of the game theorist is the game matrix. The game matrix shows how the com- bined actions of a focal player and his opponent a ffect the focal player’s ‘payoffs’. The game matrix focuses not on the nature of the behaviours involved, but on their economic consequences. Game theory’s significance as a tool for modelling cooperation hinges on the idea that one can determine which combinations of actions are ‘cooperative’ by examining the game matrix. Hence, students of game theory prefer a view of cooperation that de-emphasizes the behavioural properties of an interaction, and focuses on the economic consequences of an interaction. Clements & Stephens (1995) followed this game theoretical tradition in offering their operational definition of cooperation as ‘joint action for mutual benefit’ (Dugatkin et al. 1992; Mesterton- Gibbons & Dugatkin 1992; Dugatkin, in press). We see Roberts’ commentary as a special case of the more general conflict between the ‘econ- omic’ and ‘behavioural’ uses of ‘cooperation’. As Roberts points out, ‘mutual benefit’ can be achieved quite coincidentally. Roberts, however, does not point out that elaborate coordination may occur without mutual benefit (for example, in agonistic and coercive interactions). We are not claiming that Roberts advocates the ‘behavioural’ definition over the ‘economic’ one. We simply believe that it is important to view the issues that Roberts raises in a larger context. How should students of cooperation proceed? Should they adopt one use over the other? Should they adopt a hybrid definition? We argue that the economic definition of co- operation is the best option because it can be objectively applied, and it offers us the powerful tools of game theory. The passage below shows the subjectivity of Roberts’ behavioural view: Thus, two birds that have independently chosen to feed on a school of fish may happen to benefit from the flushing effect of each other’s activities but they are not mutualists. Two birds that have modified their behaviour to take advantage of the flushing effect of the other are mutualists (page 1361). If I eat a fish that would not have been there without your behaviour, I have modified my behaviour to take advantage of your behaviour (I chased and ate the fish that was there because you flushed it). Roberts’ example of a non- cooperative, but mutually beneficial, interaction, satisfies his own definition of mutualism! Correspondence: D. W. Stephens, School of Biological Sciences, University of Nebraska, Lincoln, NE 68588, U.S.A. (email: dws@forager.unl.edu). 0003–3472/97/061363+ 02 $25.00/0/ar960379  1997 The Association for the Study of Animal Behaviour 1363