Reciprocity and rats

Rutte and Taborsky report in PLoS Biology that their rats know how to be nice to others:

The evolution of cooperation is based on four general mechanisms: mutualism, where an action benefits all partners directly; kin selection, where related individuals are supported; "green beard" altruism that is based on a genetic correlation between altruism genes and respective markers; and reciprocal altruism, where helpful acts are contingent upon the likelihood of getting help in return. The latter mechanism is intriguing because it is prone to exploitation. In theory, reciprocal altruism may evolve by direct, indirect, "strong," and generalized reciprocity. Apart from direct reciprocity, where individuals base their behavior towards a partner on that partner's previous behavior towards themselves, and which works under only highly restrictive conditions, no other mechanism for reciprocity has been demonstrated among conspecifics in nonhuman animals. Here, we tested the propensity of wild-type Norway rats to help unknown conspecifics in response to help received from other unknown partners in an instrumental cooperative task. Anonymous receipt of help increased their propensity to help by more than 20%, revealing that nonhuman animals may indeed show generalized reciprocity. This mechanism causes altruistic behavior by previous social experience irrespective of partner identity. Generalized reciprocity is hence much simpler and therefore more likely to be important in nature than other reciprocity mechanisms.

In the discussion of the paper, the authors describe the results of other experiments that suggest rats are even more cooperative when repeated interactions occur:

In a follow-up study we tested whether the propensity to cooperate would be increased further when Norway rats interacted with a known partner who had helped them before [32]. As expected, this direct reciprocity caused even higher levels of cooperation than generalized reciprocity, i.e., a rat was 50.7% more likely to help a conspecific who had helped her before than an unknown rat after experiencing cooperation with anonymous partners. This is compatible with a "hierarchical information hypothesis" assuming that specific information about the helping propensity of a partner is used if available, but if not, anonymous social experience is used when deciding whether to cooperate or not [32], i.e., cooperation may ensue also when specific information is limited or costly to be obtained. A similar mechanism might operate in humans [29]. Theoretical models showed that the existence of direct reciprocity in a population will induce the evolution of generalized reciprocity [22], entailing much higher levels of cooperation overall.

Well, lab rats aren't necessarily the same as wild rats, so I suppose it's possible that these rats are doing unusual things compared to their natural habitat. But I'm not really surprised that rats would be capable of either general or direct reciprocity. The game theoretic basis of reciprocity is very simple, and can easily be implemented in a short algorithm. The hard part is when you try to be choosy about it -- determining exactly the right social situations in which a bias toward reciprocity pays off. Evolving the bias in a social species might be very easy. Finding that rats exhibit the behavior under some circumstances tends to confirm that this class of altruism may evolve readily, in even a modestly social species.

I take it as a suggestion that this probably isn't the main reason why humans evolved large brains...


Rutte C, Taborsky M. 2007. Generalized reciprocity in rats. PLoS Biol 5:e196. doi:10.1371/journal.pbio.0050196