1. The theory of food-regulated self-thinning (FST) for mobile animals predicts population density (N) to be an inverse function of mean body mass (W) scaled to an exponent (b), such that N = k W−b, where k is a constant. FST also predicts energy requirements (or energy flow) to remain constant over time (termed energetic equivalence) as losses to cohorts (e.g. emigration and mortality) are balanced by increased growth of surviving individuals.

2. To test these predictions, we analysed the dynamics of six experimental minnow cohorts. Replicate populations of fish were held under identical conditions with a constant and limited supply of food over a 126-day period. Half of the cohorts were open to emigration, and half were closed so that fish could only be lost through starvation mortality.

3. Patterns of self-thinning indicated non-linear changes in population density and energy flow in relation to changes in mean body mass and time, respectively. Non-linear patterns of self-thinning were probably due to a delayed growth response to changes in population density effected through mortality and/or emigration. Contrary to results of similar experiments on other fish, emigration did not have a significant influence on the pattern of self-thinning.

4. These results may be attributed to trophic interactions within cohorts and the importance of social behaviour to cohort dynamics. Both population density and energy flow in our experimental populations appeared to cycle, with episodes of starvation and mortality alternating with food recovery and weight gain, as predicted by recent models of stepwise die-off and stunted growth in animal cohorts.

5. Most of the support for FST in mobile animals comes from observational data on mean body mass and population density. Potentially important mechanisms, including the manner in which individuals are lost or retained in populations, are usually not investigated directly. Such tests of FST can only provide equivocal support. Detailed observational study and controlled experiments are needed to understand casual mechanisms.