This squib first sketches the state-of-the-art in diachronic construction grammar by tracing it back to two strands of research which it distinguishes as historical construction grammar and constructionist grammaticalization theory. It then differentiates between usage-based work in diachronic construction grammar that focuses on (frequency of) use and work that centres on knowledge. It is posited that, to arrive at truly (radically) usage-based models of change, one should separate individual knowledge, or internal systems/constructicons, from assumed-to-be-shared knowledge, or external systems/constructicons. Two us-age-based models of constructional change, “Traugott/Trousdale” and “Fischer”, are assessed against this criterion. While the former explicitly distinguishes between individual and “community” knowledge, it is judged to confuse these by assigning a central role to reanalysis/neoanalysis. The latter model revolves around the role of analogy and is less confined to a semasiological account of the linear developments dictated by an external outlook.