We discuss here the influence of sample size (number of replicates) on the accuracy and precision of the results when sampling profundal benthos with an Ekman grab according to the Finnish standard, SFS 5076, which is equivalent to the Swedish and Norwegian standards. The aim was to find criteria for choosing a sample size which would avoid any powerful influence of chance on the results without entailing an unreasonable amount of work for monitoring purposes.Lake Haukivesi (area 620 km2, total phosphorus 13 pg 1-I and colour 35 Pt mg 1-I), Lake Paasivesi (1 16 km2, 5 pg 1-I and 35 Pt mg I-') and Lake Puruvesi (322 km2, 4 pg 1-' and 5 Pt mg I-') were sampled randomly in June and October 1991. 25 Replicate samples were taken on each occasion from the deep profundal area of each lake, defined here as 60-100% of the maximum depth. The sedimentation areas studied were fairly homogeneous, since the animal communities were not markedly affected by the variations in depth. Distribution estimates for the statistics studied, such as number of individuals, expected number of species, diversity and benthic quality indices, were calculated for a large set of random samples taken from the empirical data by computer (bootstrap sampling). The sample variance, s2, correlated with the mean animal density, m (ind. m-'), according to the equation s2 = 31.77 m1.247. The sample size required to achieve the desired precision in mean animal density (D, expressed as the ratio standard errorlmean) can thus be estimated as n = 31.77 m -0 . 7 5 9 0 -2 .The number of replicate samples needed to achieve a standard error of 20% of the mean density was 10 in Lake Haukivesi, seven in Lake Paasivesi and 1 1 in Lake Puruvesi. The accuracy and precision of the estimated number of species, Shannon's diversity and Benthic Quality Index improved markedly as the sample size was increased to 10 replicates. As a compromise between work load and statistical reliability, a figure of 10 replicate Ekman samples is proposed here for the monitoring of profundal benthos. The proposed sample size usually produces individual numbers which are high enough for practical purposes, probably at least 100 individuals, which is recommended as a minimum in the standard. The lower number of replicate samples recommended in recent Finnish handbook, 3-5, usually produces inadequate data, and this may detract from the comparability of the results and leave the changes in profundal communities undetected.