Oral and written language in late adolescence are influenced by many pre- and postnatal factors, including cognitive performance at earlier ages. We investigated whether the association between birth weight and lexical knowledge and reading comprehension in late adolescence (14-16 years) is mediated by verbal cognition during early adolescence (10-11 years).We conducted a mediation analysis via a potential outcomes approach to data from three United Kingdom (UK) prospective birth cohorts - The National Child Development Study (NCDS; year of birth (Y.B.) = 1,958; analytic sample size (A.N.) = 9,399; original sample size (O.N.) = 18,558), British Cohort Study (BCS70; Y.B. = 1,970; A.N. = 6,591; O.N. = 17,196), and Millennium Cohort Study (MCS; Y.B. = 2,000-2,001; A.N. = 3,950; O.N. = 18,552) - to evaluate the indirect effects of birth weight on lexical knowledge (BCS and MCS) and reading comprehension measures (NCDS) in adolescence.We found an indirect effect but no statistically significant direct effects for the BCS and MCS cohorts. The proportion of the effect of birth weight on oral and written language in late adolescence mediated by early adolescence verbal cognition was 59.19% (BCS) and 8.41% (MCS) for lexical knowledge and 61.00% when the outcome was reading comprehension (NCDS). Sensitivity analyses, used to assess whether unmeasured variables could have affected our mediation estimates, showed that for reading comprehension, in NCDS, the indirect effect is robust; only unmeasured confounders highly correlated with the mediator and outcome (ρ = .68) would explain away the indirect effect. For lexical knowledge, smaller correlations with hypothetical confounders (ρ = .33 for BCS) would suffice to render the indirect effect non-significant; the indirect effect for MCS non statistical significant.Birth weight affects oral and written language skills (lexical knowledge and reading comprehension) in late adolescence via verbal cognition in early adolescence in two cohorts born in 1958 and 1970, but not in a cohort born at the turn of the millennium. These indirect effects were stronger than the direct effects and are unlikely to be explained by unmeasured confounders when the outcome involves complex skills such as reading comprehension.