The so-called unique items hypothesis claims that translations tend to contain fewer 'unique items' than comparable non-translated texts. This is proposed as a potential translation universal, or at least a general tendency. A unique item is one that is in some sense specific to the target language and is presumably not so easily triggered by a source-language item that is formally different; it thus tends to be under-represented in translations. The concept of a unique item is not well-defined, however. Drawing on some earlier work on transfer, contrastive and error analysis, this article offers a critical analysis of the concept, and raises a number of methodological issues concerning research on the topic.