Few studies have critically investigated the genetic composition of wild fish schools. Yet, such investigations may have profound implications for the understanding of social organization and population differentiation in both fundamental and applied research. Using 20 microsatellite loci, we investigated the composition of 53 schools (total n = 211) of adult and subadult migratory brook charr (Salvelinus fontinalis) sampled from the known feeding areas of two populations inhabiting Mistassini Lake (Québec, Canada). We specifically tested whether (i) school members originated from the same population, (ii) individuals from the same population within schools were kin (half- or full-siblings), and (iii) kin schooling relationships differed between sexes. Randomization tests revealed a tendency for most schools to be population specific, although some schools were population mixtures. Significantly more kin were found within schools than expected at random for both populations (≈ 21–34% of the total number of school members). This result, combined with the observed size range of individuals, indicated that stable associations between kin may occur beyond juvenile stages for up to 4 years. Nevertheless, a high proportion of school members were non-kin (≈ 66–79%). No differences were detected between sexes in the propensity to school with kin. We discuss the hypothesis that the stable kin groups, rather than arising from kin selection, may instead be a by-product of familiarity based on individual selection for the maintenance of local adaptations related to migration (natal and feeding area philopatry). Our results are noteworthy because they suggest that there is some degree of permanence in the composition of wild fish schools. Additionally, they support the hypothesis that schools can be hierarchically structured (from population members down to family groups) and are thus nonrandom genetic entities.