For the modal meanings of conditionals with the form if p then q, there are three existing psychological accounts: the original suppositional theory, the original and revised mental models theory. They imply different modal meanings of if p then q. Alternatively, we propose a revised suppositional theory with the unique prediction that the set from which the instance referred to by a true conditional was randomly drawn, necessarily includes pq cases, and possibly includes ¬pq / ¬p¬q cases, but impossibly includes p¬q cases. One experiment investigated whether category modal inferences from a conditional would be consistent with preceding instance modal inferences from it. The results revealed that (1) previous instance inferences did not affect subsequent category inferences; (2) relevant cases pq and p¬q tended to elicit consistent response patterns, but irrelevant cases ¬pq and ¬p¬q tended to elicit inconsistent response patterns; (3) the overall response pattern of category inferences is consistent with only the prediction of the revised suppositional theory. The latter implies that only the pq possibility is required by a true conditional, but the ¬pq / ¬p¬q possibility is unrequired. On the whole, these findings favor the revised suppositional theory over the other accounts.