Grades provide students with their primary performance feedback: signals which aect academic
choices. Variations in grading practice among courses impose grade penalties (and bonuses) on
students who take them. These grade penalties are sometimes gendered. Using extensive data
from the University of Michigan, we report on patterns of grade penalty and gendered performance
dierence across 116 large courses. We nd that signicant gendered performance dierences are
ubiquitous in large introductory STEM lecture courses. They are largely absent in both STEM labs
and in lecture courses in other disciplines. Exploring the features of these courses, we hypothesize
that evaluation methods used in STEM lecture courses interact with stereotype threat to create
these gendered performance dierences.