I. Honeybees were trained to collect syrup from coloured discs and then presented with a choice of 'model' flowers. 2. The colour of a model was an important distinguishing feature, but its scent was even more important. A foreign odour made the models less attractive than no odour. 3. The size of a training model did not influence the size of model later chosen, but the bees preferred radially symmetrical to bilaterally symmetrical models, and models with a disruptive outline to circular models, even though trained to circular ones. 4. Adding nectar guides to a model increased its attractiveness, independently of conditioning ; dotted lines were more attractive than continuous lines, and a group of dots was more attractive than a black circle in the centre of a model. Adding a disruptive outline to a model similarly increased attractiveness and effects of a disruptive outline and nectar guide lines were additive. However, a limit was soon reached in which more guide lines or further segmentation failed to increase attractiveness. 5. Bees showed no preference to alight in the centres of circular models, and preferred the edges of the petaloid or star-shaped models. 6. Nectar guide lines had a directing function only when the bees had learned to seek food at a particular site in relation to them. Training to a point where nectar guide lines converged was quickly achieved, and could be transferred to models of other types. A nectar guide ring in the centre of a model sometimes slightly increased the proportion of visits to its centre, but conditioning was again necessary to obtain much effect. 7. Bees without previous training were attracted to a black central area, and this was still more effective after training. The bees' behaviour was not affected by attempts to give models an illusion of depth.