When asked to find the referent of a novel label, children typically select an object that they cannot already name (the “disambiguation effect”; Merriman & Bowman, 1998). However, when the task required cross-modal extension of a label, children did not show this effect (Scofield, Hernandez-Reif, & Keith, 2009). In Experiments 1 and 2, preschoolers learned a label for a visual object, then examined it and a novel object by touch. On the critical trials, children were asked to decide which tactile object was the referent of a novel label. Four-year-olds only showed the disambiguation effect if, prior to the label test, they had identified the tactile object that matched the visual training object. The results of Experiment 3 suggest that the 4-year-olds expected to be asked about the matching object, which interfered with their tendency to disambiguate. This discovery-based interference appears to attenuate the use of common word learning strategies.