1. Most models of animal choice behaviour assume that desirable but unavailable options, such as a high quality, but inhabited nest sites, do not influence an individual’s preferences for the remaining options. However, experiments suggest that in mammals, the mere presence of such ‘phantom’ alternatives can alter, and even reverse, an individual’s preferences for other items in a choice set. 

2. Phantom alternatives may be widespread in nature, as they occur whenever a resource is visible, but unavailable at the time of choice. They are particularly relevant for nectarforaging animals, where previously rewarding flowers may sometimes be empty. Here, we investigate the effect of phantom alternatives on feeder preferences in the eastern honeybee, Apis cerana. 

3. First, we tested the effects of unattractive and attractive phantom alternatives by presenting individual bees with either a binary choice set containing two feeders that differed strongly in two qualities, but were equally preferred overall (‘option 1’ and ‘option 2’), or a ternary choice set containing option 1, option 2 and one of two phantom types (unattractive and attractive). Secondly, we determined whether phantoms increase (similarity effect) or decrease (dissimilarity effect) preference for phantom-similar choices. 

4. In binary trials, bees had no significant preference for option 1 or option 2. However, after encountering an attractive phantom alternative, individual bees preferred option 2. The unattractive phantom did not influence bee preferences. Phantoms consistently changed individual bee preferences in favour of the phantom-similar choice. This means that the presence of an attractive food source, even if it is unavailable, can influence preference relationships between remaining items in the choice set. 

5. Our findings highlight the importance of considering the potential for phantom effects when studying the foraging behaviour of animals. Our results are particularly relevant for nectarivores, where empty but previously rewarding flowers are a common occurrence. Since an increase in pollinator visits can result in higher seed set, our results open up the possibility that by shifting pollinator preferences, empty flowers could have therwise-unpredicted influences on community composition, plant–pollinator interactions and pollinator behaviour.