We argue that tautological questions, such as those which concern the relationship of inclusion between a class and its sub-class, violate conversational rules because they are not informative. In this sense, it is not understood that the question in a Linda-type problem involves a comparison between an inclusive and included class, but presumed that a different type of comparison is intended.

Tautological questions (and, consequently, also their answers) do not become a matter of discussion except under certain specific conditions. Our second experiment showed that, providing the context was adequately marked (such as in the case of a rhetorical question), the conjunction fallacy disappears.

In two further experiments, the implications of our view were compared with those of the other critical approaches to the heuristic programme: the classical pragmatic view (which we call logical complementarity) and the frequentist view.