The objective of the experiment was threefold: (1) to test whether an ‘unlearned appetite’ for dietary protein exists in sheep, (2) whether such an appetite and subsequent diet selection depend on the degree of previous protein deprivation, and (3) whether the N-source in the foods offered as choice influences diet selection. Differences in protein deprivation were achieved by feeding sheep with food either with high (HP) or low (LP) protein. Sheep were fed on Food LP, either for the same period of time as sheep on Food HP or until they reached the same live-weight (LW) as sheep on Food HP. Following the feeding regimes that induced differences in animal state, sheep were given a choice between a novel low-protein food (T) or Food T supplemented with isonitrogenous amounts with one of three nitrogen sources: urea (U), casein (C), or formaldehyde-treated casein (TFC). Diet selection measured in the short-term or over the first few days did not provide any evidence in support of an unlearned appetite for protein by sheep of different states. In fact, diet selection by all animals was characterized by an avoidance of the foods supplemented with the three nitrogen sources. This avoidance was strongest for the food supplemented with formaldehyde-treated casein. Selection of considerable amounts of supplemented foods was gradual and consistent only after animals gained experience of them, i.e. were allowed to consume a single supplemented food for a period of 7 days. Following this period, animals that had previously consumed Food LP for the same period of time as animals on Food HP selected a higher proportion of the supplemented food than the other sheep. The results support the view that there is no unlearned appetite for protein in sheep, and that control over diet selection is learned.