Three small Newfoundland lakes were fished to estimate the population sizes of Atlantic salmon, Salmo salar L., parr and brook trout, Salvelinus fontinalis (Mitchill), and to study within-lake movements of the species. The whole shoreline of the lakes was seined, the catches in different sections were recorded separately and the fish were given different marks. Estimates for the lakes derived from different subareas were constantly lower than the estimate derived by using the data from the whole shoreline. If estimates of the different sub-areas were extrapolated according to their proportion of the shoreline, they were positively biased, and showed higher variance. Significant correlations between the movement rates and the accuracy of sub-area population estimates and recapture rates were found. Validation of the assumption of random mixing of marked and unmarked fish, when a spatially stratified sampling is carried out, is emphasized.