Four methods have been tested of dividing cultures consisting of wheat grains containing larvae and pupae of Sitophilus granarius (L.) into samples weighing just under 10 g. and suitable for use in fumigation experiments. Two mechanical methods, each dividing a culture into two equal parts, the division being successively repeated to yield 32 samples, were shown to be satisfactory in every trial. One apparatus divided the culture on the chess-board principle, the other by means of radial fins equally spaced round the circumference of a conical divider. The standard method used at the Pest Infestation Laboratory also proved satisfactory; it consists of scooping up samples by hand radially, at random, from a culture dispersed symmetrically by pouring it over the bottom of a spherical flask, but it was evident that personal bias could be introduced by its use. The fourth method, consisting of ‘ random scooping ’, is widely used but yielded samples having considerable heterogeneity from all cultures containing larvae older than the late third instar and having a density exceeding one insect in 30 grains of wheat. This method was the only one that divided standard cultures into samples showing heterogeneity in their age structure. When cultures with an age range of only four days were divided, all four methods yielded samples for which the emergence curves were heterogeneous, apparently as a result of heating induced in samples as small as 9 g. as a result of the metabolism of the larger larvae.