The objective of the experiment was to test whether the rules that appear to underlie diet selection by pigs given choices between two foods of different crude protein (CP) content, persist when a rapeseed meal with possible goitrogenic effects is included in one or both foods. Forty-four pigs were given access to one food, or to two foods as a choice, from 12-30 kg liveweight. The foods were the combinations of two levels of CP (140 and 300 g CP/kg food) and two levels of inclusion (0 and 180 g/kg food) of a rapeseed meal, which resulted in foods L, H, LR, HR; in addition the equal parts mixture of foods L and H was also made (food M). The treatments were: access to a pair of foods as a choice, pairs LH, LR, H, LHR and LRHR (n = 6 per treatment) or access to a single food L, LR, M, H and HR (n = 4 per treatment). The inclusion of the rapeseed meal did not significantly affect the rates of food intake and liveweight gain of the pigs given access to one food. Pigs given a choice between a rapeseed-based food and another one (pairs LRH and LHR), showed a significant preference for the food without it. The rejection of the foods with rapeseed meal was such that it overpowered the physiological need of the pig to select a diet that would meet its CP requirement. When the two foods differed only in CP content (pairs LH and LRHR) pigs were able to select a diet whose composition changed systematically as they grew, and supported rapid growth. The diet selection of the choice-fed pigs is consistent with the idea that pigs, given a choice between a potentially harmful food and a non-harmful one, will favour the latter irrespective of its nutritional properties, only if it is assumed that the inclusion of rapeseed meal had a detrimental effect on the single fed pigs, which could not be seen in their performance as measured by the rates of food intake and liveweight gain.