The nature of the selective pressures which have resulted in the conical-papillate shape of the cells of the adaxial epidermis of many petals has been a matter for considerable speculation. One suggestion is that this shape focuses light within epidermal cells resulting in an increase in the amount of light absorbed by the floral pigments, intensifying the colour of the petals and possibly enhancing their attractiveness to potential pollinators; another is that conical cells aid pollinator orientation on the flower, either visually or when touched. The recent identification of a mutation at the MIXTA locus of Antirrhinum majus (which blocks the formation of conical petal cells) has allowed us to test this hypothesis. We report the results of field experiments indicating that, where pollinator number limits seed-set, flowers with conical epidermal cells receive more pollinator attention than do those with flat cells. Through the use of double mutants we have examined whether preferences for flowers with conical cells operate through the perception of flavonoid pigments. We have also examined the appearance of flowers with and without conical cells under ultraviolet light to determine whether differences in absorption or reflectance of light at these wavelengths may influence pollinator preference.