This paper examines an increase in upper elementary class sizes in California associated with the K-3 class size reduction program. I also use the variation in fourth and fifth grade class size generated by idiosyncratic first and second grade reductions required to meet program rules to identify a negative impact of larger class sizes on achievement in upper elementary grades. These quasi-experimental findings generally agree with results of prior foreign and fixed-effects studies on upper grade levels, though my estimates are larger in some cases. The findings also imply that any evaluation strategy for class size reduction that uses upper grades as implicit or explicit controls is likely flawed. (Contains 2 figures and 6 tables.)