This paper argues that desinence allmorphy should be described in terms of distributional factors rather than in terms of lexical selection of declensional patterns. The approach taken in this paper captures the ranking relation among a variety of distributional factors, which ae suggested to be distinguished between lexical and non-lexical factors. Generally speaking, lexical factors outrank non-lexical. factors, as shown in the in the Russian declension. The distribution-based approach also confirms that what has been referred to "declensional pattern" in the Russian declension must apply only to the singular inflection, since the lexical classes corresponding to declension patterns are effective as a distributional factor only for the singular inflection. The desinence allomorphy in the plural inflection does not allow one to establish such declensional patterns as posited for the singular patterns; each case has its own distribution governed by differnent factors. Finally, on the basis of the claim that gender functions as a factor for substantive declension, whereas it functions as an element of feature-combinations for the adjectival declension, this article argues that desinence allomorphy in both adjectival and substantival declension should be referred to by gender, which is one of lexical categories defined solely by lexical characteristics. Gender is not a formal(i.e., syntactico-morphological) category that is defined in terms of agreement patterns manifested in the adjectival declension, as suggested in Zaliznjack(1967), Corbett(1991), and many others.