It has been suggested that the absence of floral rewards in many orchid species causes pollinators to probe fewer flowers on a plant, and thus reduces geitonogamy, i.e. self–pollination between flowers, which may result in inbreeding depression and reduced pollen export. We examined the effects of nectar addition on pollinator visitation and pollen transfer by tracking the fate of colour–labelled pollen in Anacamptis morio, a non–rewarding orchid species pollinated primarily by queen bumble–bees. Addition of nectar to spurs of A. morio significantly increased the number of flowers probed by bumble–bees, the time spent on an inflorescence, pollinarium removal and the proportion of removed pollen involved in self–pollination through geitonogamy, but did not affect pollen carryover (the fraction of a pollinarium carried over from one flower to the next). Only visits that exceeded 18 s resulted in geitonogamy, as this is the time taken for removed pollinaria to bend into a position to strike the stigma. A mutation for nectar production in A. morio would result in an initial 3.8–fold increase in pollinarium removal per visit, but also increase geitonogamous self–pollination from less than 10% of pollen depositions to ca. 40%. Greater efficiency of pollen export will favour deceptive plants when pollinators are relatively common and most pollinaria are removed from flowers or when inbreeding depression is severe. These findings provide empirical support both for Darwin's contention that pollinarium bending is an anti–selfing mechanism in orchids and for the idea that floral deception serves to maximize the efficiency of pollen export.