We study whether gender differences in university major choices result from anticipated labor market discrimination. First, we document two novel facts using administrative transcript records from a large Midwestern university: women are less likely to study science, technology, engineering, and mathematics (STEM) as well as business and economics, but (1) those who do are positively selected on ability, and (2) obtain higher grades conditional on ability. Second, we show that these facts are consistent with a signaling model in which women anticipate greater labor market discrimination in STEM, business, and economics than in other fields. Third, we provide direct empirical evidence of anticipated discrimination using a student survey. The survey reveals striking patterns of anticipated discrimination by women, particularly in STEM, business, and economics, affecting both expected economic outcomes such as wages as well as expected workplace conditions. We conclude by showing that anticipated discrimination explains women's course taking and intended major choices, but not men's. [Financial support was provided by Riksbankens Jubileumsfond.]