One of the main voice activity modeling problems for the properly linguistic and interdisciplinary purposes is to identify the normal (standard) and deviations (distortion), when the speaker enters into communion, among their number where the speaker creates text of particular type in conditions of natural or constructed communicative situation. These standards and deviations may occur both for the selection and combination of specific linguistic resources, and at the level of tactics and strategy of the author of the text. The article deals with the variation in the description of complex real color images of the normal speaker (without cognitive and speech disorders), and discussed the possibility of using some variative parametres for psychiatric research (in particular, schizophrenia spectrum disorders). Operation of «selection and combination» as applied to the description of the image include, on the one hand, the choice of referenced objects and combining them into a scene, on the other – the choice of nominations for these objects, and the whole scene and syntactic methods of combination of these nominations for the purpose of a various categorization and prioritization these objects in descriptions. Comparison with using these two parameters allows to distinguish the central and peripheral zone in the descriptions of a single image and typify the significant «deviations».