This research used data from the National Longitudinal Survey of Youth (NLSY) to describe and model developmental trajectories across middle childhood. Our sample consisted of approximately 1,000 children of NLSY women who were aged 6 to 7 years in either 1986 or 1988. Assessments of PIAT math and reading scores and the mother-reported Behavior Problem Index in 1986, 1988, 1990, and 1992 provided data for middle-child trajectories of children aged 6 to 7 in 1986. Assessments in 1988, 1990, 1992, and 1994 provided data for children aged 6 to 7 in 1988. We used the raw score form of these data to estimate LISREL-based models of their autoregressive structure. As with other samples, average math and reading achievement trajectories were parabolic for NLSY children, with scores increasing at a decreasing rate over this period. Average behavior-problem trajectories were flat. Behind these average shapes was extreme diversity in level (and in some cases, slopes), of individual trajectories and a pronounced tendency for above average changes between two adjacent assessments to be followed by opposite-signed changes in the subsequent period. Estimates from our structural models showed great heterogeneity in the average level of achievement and behavior for all three outcomes and heterogeneous slopes for reading scores as well. Boys but not girls were found to have heterogeneous slopes for math and behavior problems, whereas girls but not boys showed a significantly higher degree of persistence if “shocked” off of their expected trajectories.