We tested 3- to 5-year-old English- and Mandarin-speaking children on their interpretation of sentences like The elephant didn't eat both the carrot and the capsicum. These sentences are scopally ambiguous. Adult English speakers favor a weak interpretation of such sentences, with negation taking scope over conjunction (i.e., the elephant probably ate one of the vegetables, but not both). In contrast, adult Mandarin speakers favor a strong interpretation of the corresponding Mandarin sentences, with conjunction taking scope over negation (i.e., the elephant ate neither vegetable). The semantic subset maxim (Notley, Zhou, Jensen, & Crain, 2012) predicts that children acquiring all human languages should initially prefer the strong (subset) reading of such sentences. In contrast, the question–answer requirement model (Gualmini, Hulsey, Hacquard, & Fox, 2008; Hulsey, Hacquard, Fox, & Gualmini, 2004) predicts that children should initially prefer the scope reading that constitutes a good true answer to a question under discussion in the context. We designed a task in which the weak reading of our sentences corresponded to a good true answer to the question under discussion. We found that children across languages nonetheless preferred to assign a strong interpretation to our test sentences, providing empirical support for the semantic subset maxim.