Abstract: Responding to Allen and Bekoff's (this issue) critique of Heyes and Dickinson's (1990) analysis of the intentionality of animal action, we reiterate that our approach does not assume that a hypothesis can be definitively falsified by the results of a single experiment, and argue that the evolutionary analysis favoured by Allen and Bekoff insulates intentional accounts of animal behaviour from rejection in the usual‘holistic’process of scientific evaluation. Specifically, we present data showing that the maintenance of behaviour on an omission schedule cannot be construed as rational because on these schedules it is reward for withdrawal that restores approach. In addition, we argue that, since behaviour can be affected by the non-intentional properties of representations such as search images and cognitive maps, whether or not these representations have intentional properties can be assessed empirically only through research on instrumental behaviour.