This paper examines whether hiring discrimination against transgender women arises from gender identity, non-cisnormative appearance, or their interaction. In a German field experiment, I send 10,022 fictitious applications in matched pairs to real job postings. Applications randomly vary whether the applicant visually looks like a cisgender woman–that is, whether she "passes"–and whether she indirectly discloses a transgender identity through a male-to-female name change. Relative to otherwise similar cisgender women, applicants receive 4.1 percentage points (pp) fewer positive responses (9%) when they disclose a name change but pass and 4.6 pp fewer (10%) when they do not disclose but do not pass. Applicants with both signals receive 15.2 pp fewer responses (32%), significantly larger than the sum of individual penalties. The non-passing penalty is larger in postings with a male contact person and in customer-facing occupations; discrimination overall is larger when employers face less competition for workers. A follow-up survey experiment elicits perceptions of otherwise similar transgender and cisgender applicants to examine potential mechanisms and heterogeneity by characteristics unobservable in the field experiment.