Authorship profiling, which is a process of the extraction of information about the unknown author of а text (demographics, psychological traits, et al.) using the methods of mathematical statistics based on the analysis of quantitative linguistic parameters, is a problem of great importance. The English language materials have been mainly focused on, which is mainly due to no special research text corpora in place. A similar study of the Russian language has not been performed as of now. The article looks at the results of experimental studies conducted using a corpus of Russian-language texts “RusPersonality” including the authors' information (gender, age, psychological testing results, lateral profile, etc.) to identify self-destructive tendencies in authors A mathematical model which predicts the probability of self-destructive behaviour has been obtained.