One of the most consistent findings in the literature on teacher quality is that teachers improve with experience, especially in the first several years. This study extends this research by separately identifying the effect of general teaching experience and grade-specific experience. Using within-teacher variation, I find that both general experience and grade-specific experience improve teacher performance. For math scores, the magnitude implies that teachers who always repeat grade assignments improve approximately 35% faster than teachers who never repeat grade assignments. In addition to furthering our understanding of how teachers improve with experience, this paper contributes to a literature on task-specific human capital.