Some plant species produce deceptive flowers that attract pollinators without rewarding by mimicking rewarding flowers. Understanding the factors that affect the success of “deceptive flowers” and the suffering of “model flowers” are important to consider the conditions of their coexistence and the evolution of deceptive flower . It has been discussed that success of deceptive flowers depends on similarity to their model flowers and their density in the community. However, other factors that affect leaning of pollinators could also affect the success of deceptive flowers. In the field with several interspersed plant species in bloom, “third flowers” that coexist with deceptive and model flowers may also affect the learning. Here, we examined the effect of “third flower” on the learning of bumble bees by using artificial flowers: model rewarding flowers (Blue: Model), two types of non-rewarding flowers that differ in color similarity to the model (Sky-blue and Ink-blue: Mimics), and rewarding flowers that remarkably differ in color (Yellow: Outsider). Based on the bee color hexagon (Chittka et al. 1992: J Comp Physiol), color distance between Sky-blue and Blue is 0.058, and Ink-blue and Blue is 0.124, respectively. When bees foraged in the mixed array of Model and Mimic (Sky-blue or Ink-blue) flowers, they learned to avoid Mimic with their experience. However, in the array of Model, Mimic and Outsider, bees almost did not avoid Mimic even after they gained experience, when the color of the Mimic was Sky-blue. When the color of the Mimic was Ink-blue, bees learned to avoid Mimic independent of the existence of Outsider. In the array of Model and Outsider, bees equally visited both types of flower. Our results suggest that the third flower could affect the learning of deceptive flowers by bumble bees, while its effect may depend on similarity between deceptive and model flowers.