Based on the recent developments of the distributional calculus of Harary and Paper, a method is proposed to discover the constructional mechanism underlying the phonemic distribution of a language. The phonemes are classified hierarchically and the mutual interdependence of their distributional characteristics is studied. To this end, the technics of correlation and regression are being used. In this way, a 'typology' of phonemes in the sense of Hempel and Oppenheim is arrived at.