The paper examines the Russian construction Num + classifier + Ngen formed specifically with the numeral classifier shtuka (Num shtuk Ngen); this classifier occurs more frequently than other classifiers (chelovek, dush, edinic). I argue that the classifier shtuka undergoes a process of semantic bleaching: it occurs in the context of approximation (despite the literal meaning of the word shtuka ‘a single thing’) and animate nouns appear in this construction in colloquial speech. The paper focuses on the usage of the counting construction with the classifier shtuka with animate nouns in spoken Russian. Some restrictions on the usage of the numeral classifier shtuka with animate nouns are revealed. There is also substantial evidence in favor of a gradated semantic hierarchy of animate nouns.