•Five experimental studies examine gender discrimination in art judgments based on different criteria•Participants chose between male and female-originated artworks with/without artist information•We find evidence for statistical but not taste-based gender discrimination•Participants quasi discriminate on gender in a bias towards a correlated trait: artist fame We conduct five experimental studies to examine whether and what kind of gender discrimination explains deep and persistent gender gaps in the art market. 1112 participants chose between male and female-originated artworks with and without artist information. Gender-specific artist names did not affect personal preferences or preference norms. They did however cause significant swings towards male artworks when participants were incentivised to guess the more pedigreed or more expensive artwork. When artist name information was controlled, manipulating artist fame information shifted preference norms towards artworks of males, who are more famous on average. Overall we find no taste-based but significant statistical gender discrimination. We also find quasi gender discrimination, in which discrimination based on a particular characteristic (fame) may be falsely attributed to a highly-correlated one (gender).