Numerical data based on stomach content analysis of the zooplanktivorous freshwater fish, Retropinna semoni, were used to examine the effect of fish sample size on mean counts of dominant prey items. Fifty adult R. semoni were collected from throughout the open-water of Lake Benanee, Australia at each of five times over a diel period. Bootstrapping was used to generate confidence intervals around sample means, and markedly more accurate means were obtained from samples collected in the day than the night. High variation in night samples was the by-product of a diurnal feeding regime. Traditional sample sizes of 10 to 15 stomachs resulted in reasonable confidence intervals of sample means derived from collections on the first day, corresponding to uniform feeding patterns in the population. However, increased sample sizes were required to describe more complex feeding behaviour on the second day, when a proportion of the population switched to an alternative prey source.