Bumble-bees foraging an Hound's-tongue return to the same plants again and again for a period of 10 days or more. When the plants are close together (6' or less apart), they visit them in a rather irregular fashion, and experiments show that, while they have some knowledge of the position of individual plants with respect to other landmarks, bees are responding largely to the general form of the plants. They become conditioned to the form very rapidly after a few visits to plants, when they are responding largely to the flowers. On the other hand, when bees are visiting more distant plants, more than (12') 4 m apart, they are reacting almost entirely to the site of each. This, it has been shown, they fix with great accuracy during one or more brief orientation flights upon leaving the plant after their first few visits to it. They often find distant and partly hidden plants during the course of rambling 'searching flights', which usually start from the central group of plants. After a short time they develop a distinct route among the distant plants, and their flight is usually very direct, in contrast to the wavering flight within the central group, where they frequently react to and search for flowers, on other flowerless plants which resemble Hound's-tongue in general form. It is suggested that much of this behaviour is a direct adaptation to the Hound's-tongue plant because its flowers, though yielding abundant nectar, are so small and inconspicuous. This is confirmed by a comparison with the behaviour of bees foraging on Foxglove, which has very large, conspicuous flowers. Here bees rely almost entirely on direct stimuli from the flowers and never become conditioned to the form of the plant or visit plants without flowers. They learn practically nothing of the positions of plants which are close together, but do learn the sites of individual plants when these are more than about 12 feet from the next nearest. Unlike bees visiting Hound's-tongue, Foxglove bees become firmly conditioned to the colour of the flowers, and do not visit paper models of any colour other than a purple, closely resembling that of the real flowers. These results show that the foraging behaviour of bumble-bees is most flexible and that they can adapt themselves to suit very different features of the plants they visit. Hound's-tongue and Foxglove probably represent two extremes, and between them there may be many plants where both the flowers and general form play a part in guiding bees. The evolutionary implications of the work are briefly discussed.