Theories of counterfactuals agree on appealing to a relation of comparative
similarity, but disagree on the quantificational force of counterfactuals.
We report on three experiments testing the predictions of three main approaches:
universal theories, homogeneity theories, and single-world selection theories (plus supervaluations over selection functions). To disentangle the predictions of these theories, we contrasted counterfactual sentences embedded under the scope of various quantifiers and tested speakers’ intuitions about such sentences using graded truth value judgment tasks. Our results provide empirical support for selectional theories, while challenging the other two approaches. We end by discussing how a more recent alternative implicature-based approach to counterfactuals is also not in line with our results, and the connection between counterfactuals and related phenomena like plural definites.