The presence of radioactive hot particles in environmental samples (e.g., soil, vegetation, sediments) is frequently detected by observing significant differences in the activities of sub-samples, which are otherwise alike. The probabilities for detecting hot particles in this way were calculated by using Monte Carlo methods as a function of the number of hot particles in the original sample, the number of sub-samples used, the frequency distribution of the activities of the hot particles, and the precision with which the activities of the sub-samples are determined. Assuming, for example, (i) a log-normal distribution of the activities of the hot particles with a relative standard deviation eta > or = 1, and (ii) that a difference of > 30% between the activities of the sub-sample with the largest and that with the smallest activity can be detected, splitting the original sample into three sub-samples will be sufficient to detect the presence of up to five hot particles with a probability of > 95%. If four sub-samples are used, the presence of up to 20 hot particles can be detected with this probability. In general, it will not be effective to increase the precision of the activity measurements of the sub-samples at the expense of the number of sub-samples investigated.