this paper I discuss a corpus-based approach to the analysis of some phenomena of lexical semantics using empirical data drawn from a Russian-German paraUel corpus ofDostoevskij's Idiot together with its German translations. This parallel corpus is part of the Austrian Academy Corpus (AAC) at the Austrian Academy of Sciences in Vienna. The subject of investigation is lexical co-occurrences which determine the combinatorial profile ofa word. A corpus-based analysis oflexical co-occurrences contributes to both monolingual and buingual lexicography by providing new and more detailed insights into the contextual behaviour of a word. From the diachronic perspective the semantic change comes about at the periphery of the combinatorial profile of a given word. Some of the peripheral co-occurrences can become so frequent that they drift from periphery to centre while others fall out of use and start to be perceived as norm violations. The comparison ofcombinatorial profiles ofthe same word in the 1860s and in present day Russian proves to be an efficient instrument for defining the combinatorial norms of a given word against the background of its near-synonyms. The comparison of a given word with all possible translation equivalents has a similar function.