These models must not be too general, for in that case not only would their theoretical study prove difficult, but above all the adaptation of such models to an observed system would lead to an estimation of a great number of parameters on the basis of a necessarily restricted sample.