Choice feeding could be more widely applied with pigs if the initial variability between individuals in diet selection was overcome and there was a better understanding of the factors affecting the learning process in groups. In six replicates, two groups of four pigs (live weight 35·5±0·95 kg) were formed containing either an individual trained to select between two foods or an untrained control animal. The groups were offered food L (130 g crude protein (CP) and 5·6 g lysine per kg) and food H (260 g CP and 15·5 g lysine per kg) in two troughs as a choice for a period of 14 days. Food intake was measured twice daily for the first 3 days and then daily. The pigs were weighed three times per week and feeding behaviour was video recorded on the 1st and 3rd days of grouping. There was no effect of the trained pig on the performance of the group over the 2-week period (daily live-weight gain: 1·19 kg and 1·21 kg for the pigs in control or trained groups; s.e.d. 0·057). Within the first 8-h period of food recording pigs in groups with a trained pig selected a diet similar to that of the trained pig, whereas those with a control pig showed initial variation in selection before adopting a pattern in favour of one food, usually after about 3 days. In the first 24 h the pattern of visits to the troughs by the naïve pigs followed that of the trained pig whereas the pigs in the control groups initially visited the troughs at random. After 3 days the difference between the groups was small and largely non-significant. In five of the six replicates the groups selected a diet comprising 0·65H: 0·35L, which was similar to that expected from their lysine requirements. The two groups in the other replicate selected more of food L and this and the other findings are discussed in relation to the influence of social learning, previous nutritional experience and food composition on diet selection by pigs.