We investigated the dimensionality of inference making in samples of 4- to 9-year-olds (Ns = 416-783) to determine if local and global coherence inferences could be distinguished. In addition, we examined the validity of our experimenter-developed inference measure by comparing with three additional measures of listening comprehension. Multitrait, multimethod modeling determined that the best fitting model included both text and inference factors, but the factor loadings of these final models showed that local and global inference factors could not be measured reliably. The Inference Task as a whole was reliable and showed good validity at all grade levels.