The article deals with the question, what new information about the semantics of
German idioms can be acquired from computer-readable text-corpora. Illustrative
material is taken from the dictionary articles, which were composed by the author. The analysis of corpora data gives an opportunity to detect meanings, which have not yet been fixed in dictionaries, to reveal, what meanings are the most frequent, and to order them accordingly to their frequency, as well to consider cases of non-idiomatic or weakly idiomatic usage.