This study explored relations between academic performance, cognition, cognitive self-evaluation and self-representation. We examined 408 participants, from 10 to 16 years, by a cognitive battery addressed to several reasoning domains (mathematical, causal, spatial, and social reasoning), self-evaluation of performance in each reasoning domain, and domain-specific and general cognitive self-representation. School grades in mathematics, science, and language indexed academic performance. Reasoning highly predicted school performance in primary and secondary school. Self-representations and self-evaluations were highly related to cognitive performance in secondary but not in primary school. Self-representation significantly predicted academic performance if used alone in the model; it is completely absorbed by cognitive ability, when used together. Self-evaluation predicted school performance additionally to cognitive ability in secondary but not in primary school. Effects of SES on academic performance were both direct and indirect, mediated by cognitive ability. The implications for cognitive developmental theory and educational implications are discussed. •Cognitive predictors of school performance vary with development•Cognitive ability is the strongest predictor of school performance•Cognitive self-evaluation predicts school performance in secondary school•Socioeconomic status influences school performance directly and indirectly via cognitive ability