Some plants have evolved flowers of extraordinary depth, a phenomenon which puzzled Darwin1. Darwin suggested that the evolution of deep flowers could be a response to a kind of 'race' with pollinating insects: the length of the tongues of pollinating insects could increase as a result of a general size increase, or because it increased their nectar foraging efficiency. As this occurred, plants with relatively shallow flowers could be disadvantaged since pollen transfer, which is effected by physical contact between the pollinator and the anthers or stigma of the plant, could be reduced when the insect tongue is long relative to flower depth. This could lead to the evolution of increasing flower depth which in turn could drive the evolution of a further increase in insect tongue length. Various predictions of Darwin's proposal were tested here for orchid species with deep flowers that are pollinated by moths. It was found that insects do indeed insert their probosces no further than necessary to obtain nectar; that an experimental reduction in flower depth reduces both the male and female components of fitness; and that in natural populations there is a correlation between flower depth and female fitness measured by fruit set. These results all support Darwin's hypothesis to explain the evolution of flower depth.