Many studies of speciation rely critically on estimates of sexual isolation obtained in the laboratory. Here we examine the sensitivity of sexual isolation to alterations in experimental design and mating environment in two sister species of Drosophila, D. santomea and D. yakuba. We use a newly devised measure of mating frequencies that is able to disentangle sexual isolation from species differences in mating propensity. Variation in fly density, presence or absence of a quasi-natural environment, degree of starvation, and relative frequency of species had little or no effect on sexual isolation, but one factor did have a significant effect: the possibility of choice. Designs that allowed flies to choose between conspecific and heterospecific mates showed significantly more sexual isolation than other designs that did not allow choice. These experiments suggest that sexual isolation between these species (whose ranges overlap on the island of São Tomé) is due largely to discrimination against D. yakuba males by D. santomea females. This suggestion was confirmed by direct observations of mating behavior. Drosophila santomea males also court D. yakuba females less ardently than conspecific females, whereas neither males nor females of D. yakuba show strong mate discrimination. Thus, sexual isolation appears to be a result of evolutionary changes in the derived island endemic D. santomea. Surprisingly, as reported in a companion paper (Llopart et al. 2005), the genotypes of hybrids found in nature do not accord with expectations from these laboratory studies: all F₁ hybrids in nature come from matings between D. santomea females and D. yakuba males, matings that occur only rarely in the laboratory.