To determine the population variation in the androgen receptor (AR) and its association with body composition in a subsistence population, we sampled 87 settled and 65 nomadic males ages 20+ among the Ariaal of northern Kenya. Anthropometric measures included height, body mass index, fat-free mass (FFM), upper arm muscle plus bone area (AMPBA), % body fat (%BF), suprailliac skinfold (SISF), and waist-to-hip ratio. Salivary testosterone (T) was determined from both morning (Am T) and afternoon (Pm T) samples. Hair roots were obtained for genotyping AR CAG repeat length. AR CAG repeat length did not vary between the two sub-groups (overall value = 22.6 ± 3.1). Multiple regression models, controlling for age and residence, indicate that Pm T was positively associated with all measures of body composition. AR CAG repeat length was a significant positive predictor of height, FFM, %BF, SISF and waist circumference. There was a significant negative Pm T by AR CAG repeat length interaction in predicting all anthropometric measures but AMPBA. These findings provide evidence for population variation in AR CAG repeat length and suggest that both T and AR CAG length play a role in body composition in this extremely lean population.