Using Italian data on the universe of mandatory tests conducted in a low-stakes setting without air conditioning, we investigate the effect of temperature on student performance, with a focus on how manipulation distorts causal estimates of temperature effects on test scores. While high temperatures adversely affect students' performance, we find that score manipulation also increases with temperature within a specific range. Leveraging the random assignment of inspectors to schools as a natural experiment, we estimate the effect of temperature on test scores net of manipulation. We find that achievement declines at lower temperature thresholds when manipulation is accounted for, implying a larger number of affected students than previously estimated. Additionally, individual survey responses collected during the tests indicate that very high temperatures induce shifts in students' emotional states, affecting self-esteem and anxiety levels.