The analysis of short-term feeding behaviour may give insights into how food intake is regulated in farm animals. Food intake is often recorded in terms of feeding events, e.g. visits to feeders, which can be clustered into meals. This enables calculation of the probability of cows starting a meal in relation to time since the last meal, which is thought to give insight into intake regulation. Starting probabilities are often calculated after data have been pooled, e.g. across day and night or across individuals. Recent work suggested that such pooling might have strongly affected previously published conclusions. We therefore constructed simulation models to investigate how such pooling affects interpretation of feeding behaviour and consequently the biological significance attached to results.