On the paradigm shift of the author's invariant

One of the urgent tasks in philology is the attribution of texts. The quantitative indicator by which one can distinguish between the works of different authors should be called the author's invariant. The paper describes a number of studies (the method of G. Kjetsaa, the method of evaluating the pair connection of grammatical classes, the method of "decision trees", the Delta method), the results of which confirm that the initial definition of the author's invariant should be corrected. In particular, this applies to the time interval during which the attribution parameter should keep “constant value”. It does not necessarily coincide with the entire period of the writer's work. Also due to the lack of a universal criterion that uniquely distinguishes a particular writer from others, one should use a set of characteristics of author's invariants at different levels of the language. The performed analysis shows that the term "author's invariant" should be divided into two categories – "global author's invariant" and "local author's invariant" – which can be consistently studied independently of each other.