Previous game-theoretic models of reciprocity have assumed that popula
tions are large and organisms effectively sessile. This paper analyzes
an iterated prisoner's dilemma among non-sessile organisms in a finit
e population, on the assumption that an individual's chance of remaini
ng in one place is not influenced by a partner's behavior. This mode o
f interaction is suitable for analyzing potentially cooperative behavi
ors that are secondary to the advantage of group formation, e.g. allog
rooming among social mammals. The analysis yields necessary conditions
for stable reciprocity in terms of three parameters, namely, a benefi
t/cost ratio, the probability of further interaction and the probabili
ty of partner retention. The results suggest that, in highly mobile or
ganisms such as fish, birds and mammals, reciprocity may be stable onl
y if the population is small and the relative benefit and future inter
action probability are both large.