When taught a label for an object, and later asked whether that object or a novel object is the referent of a novel label, preschoolers favor the novel object. This article examines whether this so-called disambiguation effect may be undermined by an expectation to communicate about a discovery. This expectation may explain why 4-year-olds do not show the disambiguation effect if a sense modality shift occurs between training and test. In Study 1, 3- and 4-year-olds learned a label for a visible object, then examined two hidden objects manually and predicted which one they would be asked about. Only the older group predicted that they would be asked about the object that matched the visible object. Study 1 also included a test of the standard disambiguation effect, where both the training and test objects were visible. Both 3- and 4-year-olds showed a weaker disambiguation effect in this test when the matching object was unexpected rather than expected. In Study 2, both age groups predicted they would be asked about this object when it was unexpected. In Study 3, both age groups showed a stronger disambiguation effect when allowed to communicate about this object before deciding which object was the referent of a novel label. Metacognitive ability predicted the strength of this disambiguation effect even after controlling for age and inhibitory control. The article discusses various explanations for why only 4-year-olds abided by the pragmatics of discovery in the test of the cross-modal disambiguation effect, but both 3- and 4-year-olds abided by it in the test of the standard disambiguation effect.