Effects of children's classroom behavior on school performance over a 4-year period are examined for a large, representative panel of beginning first graders. Scales developed from homeroom teachers' ratings of children in the spring of their first, second, and fourth years of school are used to predict spring marks in reading and math and spring scores on verbal and quantitative subtests from the CAT battery. The teachers' ratings cluster in three domains: Interest-Participation (I-P), Cooperation-Compliance (C-C), and Attention Span-Restlessness (A-R). The I-P and A-R ratings, but not C-C ratings, affect test score gains in first grade and marks in all 3 years. Behavior ratings from Year 1 also affect Year 2 and Year 4 performance, with indications that effects are understated over single-year periods. The importance of assessing classroom behavior in a longitudinal framework that allows for lagged and cumulative effects is discussed.