Composites owe their high physicomechanical properties to the complex interaction of the many elements that compose the structure of the material. The irregular nature of the real structures requires solution of problems in prediction of the effective physicomechanical properties of the composites in statistical formulation. A new method for use in the statistical mechanics of composites, a generalized self-consistency method [1,2], is submitted for prediction of the effective elastic properties of composites; unlike known methods, for example the singular approximation of random-function theory or the traditional self-consistency methods [3-6], it permits direct correction for such “subtle” structural peculiarities as the probabilistic laws specified for the random relative positions and statistical particle-size scatter and the inhomogeneity of inclusions in the composite.