The variance of groups of traps sampling the blowfly populations of areas of different degrees of complexity is examined from data for three representative types of pasture—semi‐arable, reclaimed peat bog, and upland sheepwalk.For mean catches of under twenty flies the coefficient of variation was less for limited areas of homogeneous vegetational cover than for larger areas and greater diversity of vegetation; at intermediate mean levels the difference was negligible, hut it reappeared at mean levels of 40 and upwards.Throughout the range of catches examined the variance was much greater than the mean. For very low catches even the standard deviation exceeded the mean, and at no level of mean was it less than 40%.Even for uniform areas such as association fragments the high variance made results with less than six traps difficult to interpret, and ten or more traps are advised.Two methods of comparative sampling of populations are discussed—stratification of the area to be sampled into subunits of relatively uniform vegetational or edaphic facies, and random distribution of the traps over the whole area. For the mean catches by these methods standard regression lines are suggested, from which variance estimates may be made when comparing observed means. Both regressions consist of two straight lines, the lower in each case having a steeper slope. The lines are:Sections S.D. =x̄+ 1 and 0·5x̄+7.Subsection blocks S.D. = 0·8x̄+ 1