Negative effects of habitat fragmentation are thought to be diminished when habitat patches are joined by a corridor. A key assumption is that corridors facilitate exchange rates of organisms between otherwise isolated patches. If the organisms are pollinators, corridors may be important for maintaining genetically viable populations of the plants that they pollinate. We tested the hypothesis that corridors increase the movement of insect pollinators into patches of habitat and thereby increase pollen transfer for two species of plants, one pollinated by butterflies (Lantana camara) and the other by bees and wasps (Rudbeckia hirta). We worked in an experimental landscape consisting of 40 ≥1-ha patches of early-successional habitat in a matrix of forest. Within each of eight experimental units, two patches were connected by a corridor (150 × 25 m), and three were not. Patch shape varied to control for the area added by the presence of a corridor. Differences in patch shape also allowed us to test alternative hypotheses of how corridors might function. The Traditional Corridor Hypothesis posits that corridors increase immigration and emigration by functioning as movement conduits between patches. The Drift Fence Hypothesis posits that corridors function by “capturing” organisms dispersing through the matrix, redirecting them into associated habitat patches. Using fluorescent powder to track pollen, we found that pollen transfer by butterflies between patches connected by a corridor was significantly higher than between unconnected patches (all values mean ± 1 se: 59% ± 9.2% vs. 25% ± 5.2% of flowers receiving pollen). Likewise, pollen transfer by bees and wasps was significantly higher between connected patches than between unconnected patches (30% ± 4.2% vs. 14.5% ± 2.2%). These results support the Traditional Corridor Hypothesis. There was little support, however, for the Drift Fence Hypothesis. To generalize our results to a larger scale, we measured the probability of pollen transfer by butterflies as a function of distance along a 2000 × 75 m corridor. Pollen transfer probability exponentially declined with respect to distance and successfully predicted pollen transfer probability on the scale of our previous experiment. These results suggest that corridors facilitate pollen transfer in fragmented landscapes.