A key question in many genetic studies on marine organisms is how to interpret a low but statistically significant level of genetic differentiation. Do such observations reflect a real phenomenon, or are they caused by confounding factors such as unrepresentative sampling or selective forces acting on the marker loci? Further, are low levels of differentiation biologically trivial, or can they represent a meaningful and perhaps important finding? We explored these issues in an empirical study on coastal Atlantic cod, combining temporally replicated genetic samples over a 10-year period with an extensive capture–mark–recapture study of individual mobility and population size. The genetic analyses revealed a pattern of differentiation between the inner part of the fjord and the open skerries area at the fjord entrance. Overall, genetic differentiation was weak (average FST = 0.0037), but nevertheless highly statistical significant and did not depend on particular loci that could be subject to selection. This spatial component dominated over temporal change, and temporal replicates clustered together throughout the 10-year period. Consistent with genetic results, the majority of the recaptured fish were found close to the point of release, with <1% of recaptured individuals dispersing between the inner fjord and outer skerries. We conclude that low levels of genetic differentiation in this marine fish can indeed be biologically meaningful, corresponding to separate, temporally persistent, local populations. We estimated the genetically effective sizes (Ne) of the two coastal cod populations to 198 and 542 and found a Ne/N (spawner) ratio of 0.14.