Growth faltering and loss of development potential are common in low- and middle-income countries. We aimed to study whether linear growth before and after 2 years, height-for-age Z-score (HAZ) from 1 month until 13 years and morbidity during the first 3 years predict adolescent cognitive ability.

Cognitive assessment was done between 2018 and 2019 using Raven's coloured progressive matrices (N = 997), a measure of inductive reasoning and eye-tracking measures of saccadic speed (N = 760) and saccadic control (N = 618) among children whose mothers originally participated in a randomised clinical trial in rural Malawi. Linear regression was used to predict cognitive ability. The primary model was adjusted for age and sex, and the covariate-adjusted model for other prespecified variables.

Saccadic control was predicted by a change in HAZ between 2 and 13 years in the adjusted model (coef −0.03, p = 0.04). Raven's score was predicted by change in HAZ between 1 month and 2 years (coef. 0.47, p < 0.05), and HAZ at 2, 5 and 13 years (coefs. 0.27–0.38, p < 0.05). Morbidity did not predict adolescent cognition.

Linear growth before 2 years of age and single HAZ measurements from 2 years onwards associated with later cognitive ability measured with Raven's test, but not consistently with eye-tracking assessment.