Two experiments explored the communicative bases of preschoolers' object appearance-reality (AR) errors. In Experiment 1, 3-, 4-, and 5-year-olds (N = 36) completed the AR test (with high- and low-deceptive objects), a control test with the same discourse structure but nondeceptive stimuli, and stimulus naming and memory tests. AR performance correlated positively with control (discourse) and naming test performance. Object deceptiveness had little effect. In Experiment 2, 3- and 4-year-olds (N = 64) completed AR tests that experimentally varied question phrasing and use of exemplar objects. Children also completed memory, vocabulary, and control tests (of verbal perseveration). AR performance variance was predicted by a composite perseveration score from three non-AR tasks, vocabulary, and exemplars. The results indicate that the discourse structure of the AR test elicits a perseverative tendency that is mediated by children's verbal knowledge.