Flowering plants manipulate the behaviour of their pollinators in a wide range of ways, including exploiting associative learning by changing the colour of flowers as they age and at the same time withdrawing the nectar reward. This study investigated the adaptive function of the floral colour change from while to purple in Erysimum scoparium (Brouss. ex Willd) Wettst in relation to its main pollinator. Anthophora alluadii (Perez 1895). Both species are endemic to the Canary Islands. Field obser
vations of nectar production and bee behaviour, plus inflorescence and whole plant manipulation experiments at sites in Tenerife in 2003,2004 and 2006, provided support for three hypotheses. Hypothesis 1: The flower colour change in E scoparium serves as a visual signal, indicating to pollinators that the nectar reward of the flower has changed. Late phase purple flowers provide much less (and more frequently zero) nectar compared to early phase white flowers, and pollinators make fewer visits of shorter duration to purple flowers. Hypothesis 2: Removal of purple flowers from inflorescences reduced the rate of pollinator visitation to white flowers on those inflorescences. Retention of late phase, purple flowers therefore increased the visual attraction of individual inflorescences, and this short-distance (up to tens of centimetres) visual signal directs pollinators to newly opened, rewarding flowers on that inflorescence potentially reducing geitonogamy. Hypothesis 3: Finally, evidence was found that retention of late phase, purple flowers on the plant acted as a long-distance visual signal to pollinators, highlighting the presence of plants over scales of metres to tens of metres. Removal of all purple flowers from individual plants did not result in reduced overall visitation rates to those plants, but it did change the relationship between plant size and bee visitation rate, from a positive correlation to no significant relationship. In relation to experimentally testing the function of the floral colour change, previous studies have obtained contrasting results. This may be because whole flower colour changes have evolved as medium and long range signals, whereas colour change of isolated parts of flowers may more often be a short range signal.