Deleterious mutations occur frequently in eukaryotes, resulting in individuals carrying multiple alleles that decrease their fitness. At a population level, if unchecked, accumulation of this mutation load can ultimately lead to extinction. How selection counters the accumulation of mutation load, limiting declines in population fitness, is not well understood. Here, we use manipulative experiments in zebrafish (Danio rerio) to investigate the opportunities for selection on mutation load. Inducing high mutation load through mutagenesis, we applied one generation of within-family selection on locomotor performance, and characterised both the direct response to this selection and the indirect response of reproductive success. Offspring of slow swimming parents exhibited age-dependent declines in swimming speed, while their cousins, with faster swimming parents, did not. This pattern mimics previously documented differences between high and low mutation load populations of zebrafish, suggesting that slow swimming siblings inherited (and transmitted) more mutations than their faster swimming siblings. Crosses among offspring of slow swimming fish had, on average, <75% of the reproductive success of crosses among offspring of fast swimming parents, or crosses of offspring of slow swimmers with offspring of fast swimmers. This evidence of mutationally correlated swimming speed and reproductive success reveals the potential for concordant selection on mutation load through different fitness components. There was no evidence that crosses within families (where parents potentially shared the same mutations inherited from their common ancestor) had lower reproductive success than crosses among families, suggesting that viability selection was not acting predominantly through lethal recessive homozygotes. Rather, patterns of reproductive success are suggestive of effects of mutation number per se on embryo viability. Overall, our results highlight the potential for early life mortality to remove deleterious mutations, and the need to account for this mortality when investigating the evolutionary dynamics of mutation load.