In response to an absence of size-at-age information for juvenile Chelonia mydas in the South Atlantic Ocean, but the need for such regional data to support global ecological monitoring and conservation actions, skeletochronological analyses (using humeri) were undertaken for 63 stranded dead specimens (30.0¿58.0 cm curved carapace length¿CCL) collected between 2008 and 2013 from 40 km of beaches off southern Brazil. Specimens were defined as being 2¿8 years old and from various nesting sites (based on mitochondrial haplotypes). Parametric mixed-effect models were used to examine variation among CCL, maximum humeral length (MHL) and medial humeral width (MHW) due to age (fitted with a natural cubic smoothing spline), genetic haplotype and various random factors. Genetic haplotype was not significant and subsequently removed from analyses. All three size-at-age parameters had random variability, but changes in somatic C. mydas growth described by MHW were less representative than the others, remaining constant through a linear relationship with age. By contrast, both CCL and the preferred MHL similarly displayed positive exponential-shaped splines, with age-specific growth rates (3- to 7-year-olds) predicted at 1.4¿3.4 cm year¿1 CCL or 0.2¿0.7 cm year¿1 MHL. The observed CCL growth was comparable to recent estimates for other global populations, but with southern Brazilian C. mydas generally having larger sizes-at-age. Such intra-population differences can be attributed not only to regional prerecruitment environmental parameters and associated ecological implications (e.g. foraging ecology), but also divergent analytical methods. The choice of the latter to accurately model variable growth rates has implications for comparing size-at-age among other sea-turtle populations.