Аннотация:This paper is addressed the problems of modeling a rule-based approach to free word order languages. Grammar-based parsing for free word languages is problematic. Functionalist models of free word order are based on context-sensitive rules which cannot be parsed automatically. Generative models based on Chomskyan and Stablerian minimalist grammars undergenerate word orders licensed by natural free order languages. I argue that word order alternations in Russian can be predicted by Linear-Accent Transformations (LAT) linking together pairs of sentences with the same numeration but different information structure. LAT theory is compatible with the postulate on basic word order. Each LAT rule amounts to a pair of operations , which makes it possible to reset LAT rules as mildly context-sensitive. Deaccenting of sentence categories in Russian can be analyzed as remnant movement.Scopus eiD 2-s2.0-85068517553