This paper concerns the populations of 13 species of soil Orlbatel in two forest communities, studies by the analysis of frequency distributions of individuals in the samples. The data are based on a total of 1200 samples collected in the course of one year. A statistic G1 is used to discriminate between the different theoretical distributions which have been proposed to allow for overdisperslon, viz. Neyman type A, Polya-Aeppli, negative binomial and discrete log-normal. It is shown that, in general, the negative binomial is the type of distribution which most frequently occurs in our fauna. The application of the U and T test to each species at the different times of the year confirms this conclusion. Among different biological models which might generate a negative binomial distribution, it appears that, for our fauna, the most likely is the heterogeneous Poisson model, according to which the distribution is considered to be the summation of a set of Poisson series in which the means are distributed according to a Pearson type m distribution. In this case, the parameter k of the negative binomial can be considered as a characteristic of the heterogeneity of the distribution of a species in his habitat. For a given species, It appears that, when the density falls, the heterogeneity of the distribution generaUy increases, the animals probably being clumped in the most favourable sites and being, at this moment, indifferent to intraspecific competition. In spite of differences in density, it is possible, however, to detect patterns which are characteristic of the different species studied.