To interpret an interlocutor’s use of a novel word (e.g., “give me the papaya”), children typically exclude referents that they already have labels for (like an “apple”), and expect the word to refer to something they do not have a label for (like the papaya). The goal of the present studies was to test whether such mutual exclusivity inferences require children to reason about the words their interlocutors know and could have chosen to say: e.g., If she had wanted the “apple” she would have asked for it (since she knows the word “apple”), so she must want the papaya. Across four studies, we document that both children and adults will make mutual exclusivity inferences even when they believe that their interlocutor does not share their knowledge of relevant, alternative words, suggesting that such inferences do not require reasoning about an interlocutor’s epistemic states. Instead, our findings suggest that children’s own knowledge of an object’s label, together with their belief that this is the conventional label for the object in their language, and that this convention applies to their interlocutor, is sufficient to support their mutual exclusivity inferences. Additionally, and contrary to the claims of previous studies that have used mutual exclusivity as a proxy for children’s beliefs that others share their knowledge, we found that children — especially those with stronger theory of mind ability — are quite conservative about attributing their knowledge of object labels to others. Together, our findings hold implications for theories of word learning, and for how children learn about the scope of shared conventional knowledge.
Linguistic Conventionality and the Role of Epistemic Reasoning in Children’s Mutual Exclusivity Inferences
