Sedimentologists often find such depositional sequences, which visually, but even for microscopic examinations look homogeneous, and their fossil-content is very low. These sequences can be subdivided into genetic units only with great difficulties or not at all, and their correlation may be problematical.

An experiment is presented here, aiming at a subdivision of loess profiles by a combined use of sedimentological, palaeontological and mathematical methods. The loess samples selected for this examination were characterised by their grain size parameters as suggested by Folk and Ward (1957) and by their carbonate content. Using a Q-type cluster analysis these samples were classified into different groups by their similarity. After arranging the samples with their cluster symbols into their real stratigraphic position, the genetic units of the depositional sequence were delimited by an analysis of the vertical series of the previously established cluster-groups. Finally, a method of discriminant analysis was applied to establish the stability, ie the probability, of the existence of the vertical genetic units obtained by this zonation method.

Parallel to it the sequence was subdivided from a faunistic point of view, too. These units correlate well with the results of the numeric approach.