One of the central methodological problems within contrastive linguistics concerns the nature of the tertium comparationis (TC). On the whole, we can distinguish two approaches. The first one induces the TC directly from the language data and is universalistic. The second deduces the TC from a metatheory about the intrinsic characteristics of natural language and is relativistic. The present contribution serves a strictly methodological purpose, namely to present some new arguments in favour of the latter approach.