Empirical studies of natural populations have commonly reported departures from Hardy-Weinberg expected proportions of heterozygote individuals. Recent advances in statistical population genetics now offer the potential to exploit individual multilocus genotypic information to test more rigorously for possible sources of heterozygote deficiencies. In a previous study in lacustrine brook charr (Salvelinus fontinalis), we reported stronger deficits in small than in large lakes. In the present paper, we propose a methodology for empirically testing alternative hypotheses to identify the cause of the deficits observed in three of the smallest lakes (85, 109 and 182 ha) analysed. First, as in several salmonid species, brook charr may exhibit a trophic polymorphism in north temperate lakes. If morphs are genetically divergent, indiscriminate sampling of both forms would result in less heterozygote individuals than expected in a randomly mating population (Wahlund effect). Using an individual-based method aiming at detecting cryptic population structure, we can reject this explanation as the sole source of deficits for all three lakes. Secondly, mating among relatives could also be frequent in small lakes and lead to heterozygote deficiencies. Significantly more fish than expected at random had low individual multilocus heterozygosity in two of the lakes, suggesting that inbred fish may have been present. Thirdly, sampling of genetically related fish would also lead to departures from Hardy-Weinberg proportions. In the same two lakes, the distribution of pairwise relatedness coefficients departed from its random expectation, suggesting that non-random sampling of kin may have occurred.