This paper elaborates on the rule concerning the phonological history of Russian worded by Andrey Zaliznyak in his classic 1992 work. The rule states that an optional elimination of the word-final unaccented vowels operated if: 1) the word was not proclitic; 2) the final vowel of the phonological word was not a separate morpheme; and 3) the vowel was preceded by a single consonant or /st/. However, Zaliznyak postulates inconsistent application of (1) due to
a few irregular forms. In what follows I propose a possible solution of this problem (re)considering individual cases. Thus, for the conjunction ili ‘or’ (originally end-accented, then proclitic), I put forward a novel explanation (blending of ili and the synonymous enclitic =l’). The words skvozě́ and čtobý appear to undergo apocopation only after leftward accent shift had occured. The lexemes kolí ‘if, when’ and kol’ ‘how many, how much’ are simply separate
words which merged in usage only during the 18th century, i.e. after the end of Zaliznyak’s rule operation. I also examine the trend for eliminating end-stressed infinitives which seems to be related with this rule. In conclusion, I argue that Zaliznyak’s rule affects neither proclitics, nor end-stressed words.