Individuals of some species of pollinating insects tend to restrict their visits to only a few of the available plant species, in the process bypassing valuable food sources. The question of why this flower constancy exists is a rich and important one with implications for the organization of natural communities of plants, floral evolution, and our understanding of the learning processes involved in finding food. Some scientists have assumed that flower constancy is adaptive per se. Others argued that constancy occurs because memory capacity for floral features in insects is limited, but attempts to identify the limitations often remained rather simplistic. We elucidate now different sensory and motor memories from natural foraging tasks are stored and retrieved, using concepts from modern learning science and visual search, and conclude that flower constancy is likely to have multiple causes. Possible constraints favoring constancy are interference sensitivity of short-term memory, and temporal limitations on retrieving information from long-term memory as rapidly as from short-term memory, but further empirical evidence is needed to substantiate these possibilities. In addition, retrieving memories may be slower and more prone to errors when there are several options than when an insect copes with only a single task. In addition to memory limitations, we also point out alternative explanations for flower constancy. We then consider the way in which floral parameters, such as interplant distances, nectar rewards, flower morphology, and floral color (as seen through bees' eyes) affect constancy. Finally, we discuss the implications of pollinator constancy for plant evolution. To date there is no evidence that flowers have diverged to favor constancy, although the appropriate tests may not have yet been conducted. However, there is good evidence against the notion that pollinator constancy is involved in speciation or maintenance of plant species integrity.