Methods for estimating ‘typical’ asymptotic‐ body size in populations of the North African agamid Agma impalearis using individuals of undetermined age were investigated, and the pattern of geographic variation in this characteristic described and tested against simple models based on putative causes. Resampling of a large single‐site data set indicated that use of the sample mean led to substantial underestimation of asymptotic size while use of the largest animal per sample led to overestimation, on average. The 80th percentile provided the most accurate estimate of asymptotic size and, compared with the largest individual per sample, had low variance, showed low dependence on sample size and was amenable to bootstrap estimation of confidence intervals. A clear pattern of geographic variation in asymptotic size was evident across Morocco: animals were smallest in the north and east and largest in the southwest. This pattern was discordant with previously‐described patterns of variation in scalation and size‐free body dimensions and, unlike these characteristics, could not be explained by a model based on secondary contact of previously vicariant populations. Among‐site differentiation was strongly associated with a model based on a proxy of current potential gene flow between sites, supporting an alternative in situ differentiation hypothesis based on the combined effects of isolation‐by‐distance and the dispersal barrier provided by the higher parts of the Atlas mountain range. Possible reasons for the discordance between the pattern of variation in body size and other morphological characteristics are discussed