Even when both use and cognition are incorporated in its theorizing about grammatical change, research in the budding discipline of diachronic construction grammar which explicitly subscribes to a “usage-based” approach does not always distinguish between abstraction from the observed usage of a linguistic community and individual linguistic knowledge. Given that language change starts with innovations by individuals and that experientially constructed knowledge is necessarily diverse, such a distinction crucially needs to be made to arrive at a realistic usage-based account of grammatical change. This paper first assesses the extent to which the conflicting models of Elizabeth Traugott and Olga Fischer succeed in teasing apart internal and external systems, concluding that while the former’s reanalysis model results from an external semasiological perspective, the latter’s analogy model is more radically usage-based in that it does not inherently entangle intra- and extra-individual knowledge. By way of an illustration of a radically usage-based approach, the second part of the paper proposes an onomasiological account of how the pattern be bound to came to be used as a non-deontic/epistemic necessity marker, offering an alternative to viewing it as a development from the historically prior deontic be bound to construction.