Faced with a sentence like Every horse didn't jump over the fence as a description of a scenario in which one out of two horses jumped, adults readily endorse the utterance as a good description, while children overwhelmingly reject it. However, systematic changes to the task setup lead to marked increases in children's endorsement rates (Musolino & Lidz 2006; Viau et al. 2010). Savinelli et al.(2017) use a computational cognitive model of utterance endorsement in truth-value judgment tasks to analytically demonstrate that both children and adults' interpretation behavior is affected by pragmatic manipulations. We test a clear prediction of these models: manipulating the conversational goal (or Question Under Discussion) should lead to clear effects on utterance endorsement. In addition to investigating the predictions for English, we also investigate Spanish and Mandarin, where the status of the relevant ambiguity may be less clear.