The cross-sex-shift hypothesis predicts that homosexual men and women will be similar in certain neurobehavioral traits to their opposite-sex counterparts. Accordingly, it predicts that homosexual men should perform in the direction of heterosexual women, and homosexual women in the direction of heterosexual men, on neurocognitive tests that show normative sex differences. We conducted a meta-analysis on the relationship between sexual orientation and cognitive performance, and tested the effects of potential moderating variables separately by sex. A total of 106 samples and 254,231 participants were included. The meta-analysis revealed that homosexual men performed like heterosexual women in both male-favouring (e.g., spatial cognition) and female-favouring (e.g., verbal fluency) cognitive tests, while homosexual women performed like heterosexual men only in male-favouring tests. The magnitude of the sexual orientation difference varied across cognitive domains (larger for spatial abilities). It was also larger in studies comparing exclusive heterosexuals with exclusive homosexuals compared to studies comparing exclusive heterosexuals with non-exclusive homosexuals for both sexes. The results may narrow down potential sites for sexual orientation-related neural differences.