The article presents an attempt to adopt a comparative-interpretive method to study non-equivalence at word level in translation. In the theory of translation, the problem of non-equivalence was mainly considered from the point of view of how we can transfer such words or word combinations. Less attention was paid to a comparative study of the semantic potential of such lexical units and their translation correspondences, especially within an interpretative approach in translation studies. This article discusses the method of comparative study of nonequivalence at word level by building a semantic model of a word and its translation correspondence as an associative verbal network to determine the relationship between the unit in the source text and the corresponding unit in the text of tranlsaiton.