When archaeologists studying societies without texts began to organise their data in the 19th century, they created “boxes” called “cultures” (Kultur in German) or “civilisations”, defined by “type fossils” (Leitfossil) following an almost explicitly geological paradigm. These “cultures” were nevertheless considered also as biological entities, able to reproduce themselves, or even combine. Gustaf Kossinna took a decisive step by assimilating “archaeological cultures” with “peoples”. Today, one can distinguish two extreme positions, illustrated by the debate on the “Celts”: either cultures really exist as human groups that can be defined, which is the majority view in central and eastern Europe; or cultures are a pure fabrication on the part of the observer, and represent at best broad stylistic zones, which is a common view in the archaeology of English-speaking countries and their near neighbours. The latter is a perspective which was emphasised even more by so-called “post-processual” archaeology. Reality no doubt lies somewhere between these two extremes, but is also linked to what can be reconstructed of the underlying ancient societies and their degrees of political integration. Yet in recent years, palaeogenetics has joined the game, hoping to reconstruct population migrations, if not “peoples”, with the risk of reproducing, by biological determinism, the models of thinking of 19th century craniometry. The debate goes on!