This paper contributes to the cross-linguistic study of perfect auxiliary selection and to the corpus-based study of English historical syntax. The authors present evidence showing that past counterfactual perfects, from their first appearance in Middle English, force the use of auxiliary have, and that such examples represent the first examples of auxiliary have with the verb come. The authors then propose an analysis of this effect in terms of Iatridou's Exclusion analysis of counterfactuals, arguing that this can account for certain cross-linguistic patterns. Finally, they point out problems posed by the ME data for theories of auxiliary selection based on unaccusativity, hierarchies of semantic verb classes or a mutative intransitive prototype.