Uncoupling proteins 2 and 3 (UCP2 and UCP3) may negatively regulate mitochondrial ATP synthesis and, through this, influence human physical performance. However, human data relating to both these issues remain sparse. Examining the association of common variants in the UCP3/2 locus with performance phenotypes offers one means of investigation. The efficiency of skeletal muscle contraction, delta efficiency (DE), was assessed by cycle ergometry in 85 young, healthy, sedentary adults both before and after a period of endurance training. Of these, 58 were successfully genotyped for the UCP3-55C>T (rs1800849) and 61 for the UCP2-866G>A (rs659366) variant. At baseline, UCP genotype was unrelated to any physical characteristic, including DE. However, the UCP2-866G>A variant was independently and strongly associated with the DE response to physical training, with UCP2-866A allele carriers exhibiting a greater increase in DE with training (absolute change in DE of -0.2 ± 3.6% vs. 1.7 ± 2.8% vs. 2.3 ± 3.7% for GG vs. GA vs. AA, respectively; P = 0.02 for A allele carriers vs. GG homozygotes). In multivariate analysis, there was a significant interaction between UCP2-866G>A and UCP3-55C>T genotypes in determining changes in DE (adjusted R(2) = 0.137; P value for interaction = 0.003), which was independent of the effect of either single polymorphism or baseline characteristics. In conclusion, common genetic variation at the UCP3/2 gene locus is associated with training-related improvements in DE, an index of skeletal muscle performance. Such effects may be mediated through differences in the coupling of mitochondrial energy transduction in human skeletal muscle, but further mechanistic studies are required to delineate this potential role.