This paper argues that corpora containing translations of the same source text into two or more target languages represent a sounder base for comparing linguistic expressions in several languages than do corpora containing data from just two languages, whether they be unidirectional or bidirectional. The argument is illustrated with comparisons of some prepositional constructions in English and French using data from the Nor. – Eng. – Fr. – Ger. part of the Oslo Multilingual Corpus, and some constructions in Swedish and Norwegian using the overlapping original texts in the ENPC and ESPC.