Purpose Few longitudinal studies have investigated the interwoven longitudinal dynamics between physical activity (PA), motor performance, and academic skills in middle childhood. Therefore, we investigated the cross-lagged associations between PA, motor performance, and academic skills from Grade 1 to Grade 3 in Finnish primary-school children.

Methods A total of 189 children aged 6–9 years at baseline comprised the study sample. Total PA was assessed using a questionnaire filled out by parents, moderate-to-vigorous PA by combined heart rate and body movement monitor, motor performance by 10x5-metre shuttle run test, and academic skills by arithmetic fluency and reading comprehension tests in Grade 1 and Grade 3. Data were analysed using structural equation modelling adjusted for gender, parental education, and household income.

Results The final model fitted the data very well [χ2(37) = 68.516, p = 0.0012, RMSEA = 0.067, CFI = 0.95, TLI = 0.89], and explained 91% of variance in the latent academic skills variable, 41% of the variance in the latent PA variable, and 32% of variance in motor performance in Grade 3. Better motor performance in Grade 1 was associated with higher academic skills in Grade 3, but it did not predict PA. PA was not directly or indirectly associated with academic skills. However, higher levels of PA in Grade 1 predicted better motor performance in Grade 3. Academic skills did not predict PA or motor performance.

Conclusions These results suggest that better motor performance, but not PA, predicts later academic skills. Academic skills in Grade 1 do not contribute to PA or motor performance in the early school years.