At issue in this article is the soundness of archaeological inferences which proceed stepwise from data about the material culture of Middle Stone Age humans, via assumptions about their symbolic behaviour, to the conclusion that they had modern language. Taking as paradigmatic the inference that the humans who inhabited Blombos Cave in South Africa some 75,000 years ago had fully syntactical language, the article argues that the inferential step from symbolic behaviour to modern language lacks the required warrant. This step, it is shown, is not underpinned by an adequate bridge theory of the putative links between symbolic behaviour and modern language. The bridge theories invoked to date to shore up the Blombos inference are flawed, for instance, in that they incorporate untenable assumptions about language, including an incorrect view of the expressive power of relatively simple linguistic means.