Intralocus sexual conflict, which occurs when a trait is selected in opposite directions in the two sexes, is a taxonomically widespread phenomenon. The strongest genetic evidence for a gender load due to intralocus sexual conflict comes from the Drosophila melanogaster laboratory model system, in which a negative genetic correlation between male and female lifetime fitness has been observed. Here, using a D. melanogaster model system, we utilize a novel modification of the ‘middle class neighbourhood’ design to relax selection in one sex, while maintaining selection in the other. After 26 generations of asymmetrical selection, we observed the expected drop in fitness of the non-selected sex compared to that of the selected sex, consistent with previous studies of intralocus sexual conflict in this species. However, the fitness of the selected sex also dropped compared to the base population. The overall decline in fitness of both the selected and the unselected sex indicates that most new mutations are harmful to both sexes, causing recurrent mutation to build a positive genetic correlation for fitness between the sexes. However, the steeper decay in the fitness of the unselected sex indicates that a substantial number of mutations are gender-limited in expression or sexually antagonistic. Our experiment cannot definitively resolve these two possibilities, but we use recent genomic data and results from previous studies to argue that sexually antagonistic alleles are the more likely explanation.