Do children always conform to a majority’s testimony, or do the pragmatics of that testimony matter? We investigate children’s reasoning about mapping a novel word to a referent in an object-labeling task. Across four conditions, we modified the testimony in an object-labeling task, to account for pragmatic principles, so that the majority does and does not provide an explicit opinion about the alternative object chosen by the minority. Four- and 5-year-olds were given a choice between an object endorsed by a three-person majority, or one endorsed by a single minority informant. In the unendorsed condition, informants explicitly unendorsed the unchosen object. In the nothing condition, informants said nothing about the unchosen object. In the ignorance condition, informants explicitly expressed uncertainty about the unchosen object, and in the hidden condition, the chosen object was the only one present at the time of the endorsement. Children were most likely to endorse the majority object in the unendorsed condition, in which the majority explicitly stated that the label applied to only one referent, whereas in the hidden condition, where only one object at a time was present in the discourse, children chose objects endorsed by the majority and the minority equally, with the other two conditions intermediate. This suggests that children might not simply have a conformity bias; rather, they are sensitive to the majority’s implied intentions when learning from testimony.