When presented with an artificial flower patch of blue and yellow pedicellate flowers, individual honey bees, Apis mellifera L., became constant to one of the two flower colours, rarely even sampling the alternative colour. Some bees visited only blue flowers while others visited only yellow flowers. This paper describes the onset of constancy for bees that had had no experience with the experimental apparatus. In 3020 visits, bees failed to land on or drink from the flower colour on which they first landed only 17 times. This behaviour was not modified by quality or quantity of reward, training to the experimental site, group effects or presence of odour during trials. However, when we trained bees to a target painted with two colours and then forced them to sample monomorphic flower patches in sequence, all bees visited the only colour present: yellow or blue. When we subsequently offered these same bees yellow and blue flowers simultaneously (rewarded choices), they became constant. Eleven of 23 bees showed constancy to the less rewarding flower morph without even sampling the alternative. Those bees failed to sample even though they had previously been forced to visit the alternative flower morph, which offered a reward with twice the calories/volume. Constancy is thus spontaneous in honey bees, but it can be hidden by some experimental protocols designed to study learning.1997The Association for the Study of Animal Behaviour