The mechanism of appearance of combination tones is not yet well explained. It is the subject of numerous theoretical hypotheses. This paper presents a new approach to the description of combination tones based on the theory of finite groups. It indicates the existence of such a mechanism within the perception process, which is only slightly sensitive to the nonlinear distortion. This mechanism should directly follow spectral analysis in the human ear. A simplified model is constructed which explains many peculiarities associated with combination tones. For example a surprising result is obtained that cubic difference tone 2f1-f2 originates already from nonlinear distortion of the second order.