There is a growing appreciation that child functioning in different domains, levels, or systems are interrelated over time. Here, we investigate links between internalizing symptoms, externalizing problems, and academic attainment during middle childhood and early adolescence, drawing on two large data sets (child: mean age 8.7 at enrolment, n = 5,878; adolescent: mean age 11.7, n = 6,388). Using a 2-year cross-lag design, we test three hypotheses - adjustment erosion, academic incompetence, and shared risk - while also examining the moderating influence of gender. Multilevel structural equation models provided consistent evidence of the deleterious effect of externalizing problems on later academic achievement in both cohorts, supporting the adjustment-erosion hypothesis. Evidence supporting the academic-incompetence hypothesis was restricted to the middle childhood cohort, revealing links between early academic failure and later internalizing symptoms. In both cohorts, inclusion of shared-risk variables improved model fit and rendered some previously established cross-lag pathways non-significant. Implications of these findings are discussed, and study strengths and limitations noted. Statement of contribution What is already known on this subject? Longitudinal research and in particular developmental cascades literature make the case for weaker associations between internalizing symptoms and academic performance than between externalizing problems and academic performance. Findings vary in terms of the magnitude and inferred direction of effects. Inconsistencies may be explained by different age ranges, prevalence of small-to-modest sample sizes, and large time lags between measurement points. Gender differences remain underexamined. What does this study add? The present study used cross-lagged models to examine longitudinal associations in age groups (middle child and adolescence) in a large-scale British sample. The large sample size not only allows for improvements on previous measurement models (e.g., allowing the analysis to account for nesting, and estimation of latent variables) but also allows for examination of gender differences. The findings clarify the role of shared-risk factors in accounting for associations between internalizing, externalizing, and academic performance, by demonstrating that shared-risk factors do not fully account for relationships between internalizing, externalizing, and academic achievement. Specifically, some pathways between mental health and academic attainment consistently remain, even after shared-risk variables have been accounted for. Findings also present consistent support for the potential impact of behavioural problems on children's academic attainment. The negative relationship between low academic attainment and subsequent internalizing symptoms for younger children is also noteworthy.