Selection in commercial populations on aspects of output, such as for growth rate in poultry. against fatness and for growth rate in pigs, and for milk yield in cows, has had very barge effects on such outputs over the past 50 years. Partly because of the cost of recording intake, there has been little or no selection for food intake or feeding behaviour. In order to predict the effects of such past, and future, selection on intake it is necessary to have some suitable theoretical framework. Intake needs to be predicted in order to make rational feeding and environmental decisions. The idea that an animal will eat 'to meet its requirements' has proved useful and continues to be fruitful. An important part of the idea is that the animal (genotype) can be described in a way that is sufficient for the accurate prediction of its outputs over time. Such descriptions can be combined with a set of nutritional constants to calculate requirements. There appears to have been no change in the nutritional constants under selection for output. Under such selection it is simplest to assume that changes in intake follow from the changes in output rates, so that intake changes become entirely predictable. It is suggested that other ways that have been proposed for predicting intake cannot be successful in predicting the effects of selection. Feeding behaviour is seen as being the means that the animal uses to attain its intake rather than being the means by which that intake can be predicted. Thus, the organisation of feeding behaviour can be used to predict neither intake nor the effects of selection on it.