Abstract Although language acquisition is frequently invoked as a cause of syntactic change, there has been relatively little work applying a formal model of acquisition to an actual case of language change and testing its predictions empirically. Here we test the model of Yang (Language Variation and Change 12: 231–250, 2000) on the historical case of the loss of verb movement to Tense (V-to-T) in Faroese and Mainland Scandinavian, using quantitative data from a number of corpora. We show that the model straightforwardly predicts the historical data, given minimal and uncontroversial assumptions about Scandinavian syntax. In contrast to a number of previous attempts to explain this repeated pattern of change, it is not necessary to appeal to any bias against learning structures involving V-to-T—a welcome result, given current evidence from acquisition. The newer V-in-situ parameter setting overcomes the original V-to-T grammar because it is more learnable in a language that also has embedded verb-second (EV2). Finally, we argue that the course of the diachronic change is evidence against a strong version of the “Rich Agreement Hypothesis” (RAH), but that under this account the stability of V-to-T in Icelandic provides evidence for the weaker version (cf. Bobaljik, Journal of Comparative Germanic Linguistics 6: 129–167, 2002).