There is a wealth of published molecular population genetic studies, however, most do not include historic samples and thus implicitly assume temporal genetic stability. We tested for changes in genetic diversity and structure in three populations of steelhead trout (Oncorhynchus mykiss) from a northern British Columbia watershed using seven microsatellite loci over 40 years. We found little change in genetic diversity (mean allele numbers and observed and expected heterozygosity), despite large variation in the estimated numbers of steelhead returning to the watershed over the same time period. However, the temporal stability in genetic diversity is not reflected in population structure, which appears to be high among populations, yet significantly variable over time. The neighbour-joining tree showed that, overall, two of the populations (Zymoetz and Kispiox) clustered separately from the third (Babine); a finding which was not consistent with their geographical separation. The clustering pattern was also not temporally consistent. We used the temporal method to estimate the effective number of breeders (Nb) for the three populations; our values (Nb = 17–102) were low for the large and presumed vigorous populations of steelhead trout sampled. The low Nb values were also not consistent with the generally high genetic diversity estimates, suggesting the possibility of intermittent gene flow among the three populations. The use of temporal analyses in population genetic samples should be a priority; first, to verify observed patterns in contemporary data, and second, to build a dataset of temporal analyses to allow generalizations to be made concerning temporal genetic stability and effective population size in natural populations.