This paper analyzes the relationship between students grades and their evaluations of teaching. We exploit an original data set from almost 100 courses during 7 academic years in a French higher education institution. We use teacher fixed effects to rule out any simultaneity or endogeneity bias. We find that students take their exam grade into account when they evaluate teaching. A better grade is associated with a better evaluation of a teacher's pedagogy, although the size of the effect is relatively small. A one-point increase in by-course mean grade corresponds to a less than one percentage point decrease in the proportion of students giving bad evaluations. These results suggest that it is possible to manipulate evaluations through grade or exam leniency.