A group of fifteen-spined sticklebacks were given the choice of two food sources. The fish sampled both sources, but with decreasing frequency as each trial progressed. Sampling was more frequent when the fish were hungry than when partially satiated. The fish spent more time at the more profitable source, discrimination being marked when profitability was determined by delivery rate, but less pronounced when determined by prey size. Given food sources of equal mean delivery rate, hungry fish concentrated on the variable source and partially satiated fish on the constant source. The fish became more reluctant to feed and to visit a source where there had recently been a simulated threat of predation, but reluctance was less when the dangerous source was also more profitable, especially when the fish were hungry. The fish distributed themselves in proportion to the profitability of the food source (ideal free distribution). Competitive ability varied among the fish according to their size. The two largest, competitively superior fish sampled the food sources more frequently than their inferior competitors. They quickly intercepted food and could efficiently track any short-term changes in delivery rate. Inferior competitors sampled frequently at first, but then settled for one food source, concentrating on prey missed by their competitive superiors and reducing travel costs. Fifteen-spined sticklebacks, therefore, behave in accordance with the 'energy maximization premise', subject to the constraints that risks of starvation and predation are minimized and that adjustments must be made towards competitors. Sampling is a conspicuous feature of the foraging behaviour and is appropriate to the temporally and spatially heterogeneous, natural food supply of these fish. © 1991 The Association for the Study of Animal Behaviour.