The paper proposes the research methodology aimed to study corpora of literary texts based on the synthesis of stylistic, semiotic, classification and statistical conceptions. The research has been carried out on the material of Computer Anthology of Russian Stories of the XX century, which is currently created at the Department of Mathematical Linguistics of St. Petersburg State University. The paper presents detailed illustrations of two typical tasks: 1) statistical taxonomization of fiction texts based on their syntactic similarity, and 2) reconstruction of semantics of the world-perception of writers by means of lexical investigations.