Understanding discrimination is key for designing policy interventions that promote equality in society. Economists have studied the topic intensively, typically taxonomizing discrimination as either taste-based or (accurate) statistical discrimination. To reveal the limitations of this taxonomy and enrich it psychologically, we design a hiring experiment that rules out (by design) both of these sources of discrimination with respect to gender. Yet, we still detect substantial discrimination against women. We provide evidence of two forms of discrimination, explicit and implicit belief-based discrimination. Both rely on statistically inaccurate beliefs but differ in how clearly they reveal that the choice was based on gender. Our analysis highlights the central role played by contextual features of the choice setting in determining whether and how discrimination will manifest. We conclude by discussing how policy makers may design effective regulation to address the specific forms of discrimination identified in our experiment.