The effects of language on symbolic functioning were examined using the "boxes task," a new symbolic understanding task based on DeLoache's model task. Children ("N" = 32; ages 2;4--3;8) observed an object being hidden in a stack of four boxes and were then asked to retrieve a similar object in the same location from a set of four target boxes. Each box was identified with a different object sitting on a small ledge in front of it. Language use was manipulated by providing linguistic scaffolding (naming vs. standard) and by using objects to identify the boxes whose names were either known or unknown to the children (familiar vs. language control). Home language environment (monolingual vs. bilingual) and verbal age (measured via the Peabody Picture Vocabulary Test-Third Edition) were also examined. Main effects were found for type of object (with familiar objects yielding better performance), home linguistic environment (with bilingualism resulting in better performance), and verbal age. There was also a verbal age × linguistic scaffolds interaction: Verbal age was related to correct retrievals only in the naming condition. These results provide further evidence that language mediates children's symbolic development and indicate that the new boxes task is an effective way of evaluating young children's symbolic competencies.