Two recent studies—the first by MacDonald & Roebroeks (2013), the second by Tallerman (2013)—draw inferences about the social use of language by premodern hominins from data about the linguistic behaviour of modern hunter-gatherers and other modern people with traditional cultures. Such inferences cannot be sound, though, unless they meet a particular requirement: they need appropriate warrants. These have to serve as conceptual bridges that span the ontological gap between the behaviours and capacities of modern humans and those of the premodern hominins concerned. Interestingly, both MacDonald & Roebroeks and Tallerman make a serious attempt to support their respective inferences with the aid of such conceptual bridges. The present article inquires whether these bridges are strong enough to serve this purpose, and argues that both bridges have components that are harmful to their solidity. In the process of arguing this, the article pursues the question of the conditions under which uniformitarian assumptions can be used as components of the substructure of the conceptual bridges needed for underpinning inferences about the use of language in the teaching and learning of subsistence skills by premodern hominins. More generally, the article elucidates an important limitation of the ethnographic record as a putative window on the evolution of language.