Plant-pollinator interactions have molded the stunning diversity of flower forms in nature. The flowers’ morphology determines their accessibility for foraging pollinators. In some plant species, flower parts fuse to form elaborate structures. Consequently, insects require intricate motor routines to access their nectar and pollen. Learning to handle such ‘complex’ flowers efficiently requires much practice in the short term, but successful foragers often reap high long-term rewards. How may the short-term barriers to exploiting complex flowers be overcome, allowing the evolution and radiation of such flowers? Despite much research on pollinator responses to specific floral features, an integrative view of flower accessibility to insects is still missing. Laboratory experiments with real and artificial flowers suggest that bumblebees innately prefer morphologically complex shapes over simple ones and that success with one complex morphology increases their tendency to forage on flowers with other complex shapes. A game-theoretic evolutionary model predicts a stable coexistence of simple and complex flowers in plant communities. The expected steady-state frequency of complex flowers is influenced by the flowers’ spatial aggregation and by the pollinators’ flying distances and learning abilities. We are currently testing these predictions using a large database of published pollination networks. In conclusion, our research suggests pathways to the evolution of morphologically complex flowers by considering the behavior and life history of their pollination partners.