Previous work has shown that children under age 3 often perform very poorly on the model room task, in which they are asked to find a hidden toy based on its location in a scale model. One prominent theory for their failure is that they lack the ability to understand the model as both a physical object and as a symbolic representation of the larger room. A hypothesized additional component is that they need to overcome weak, competing representations of where the object was on a previous trial, and where it is in the present trial, in order to succeed in their search. Children aged 33–39 months were tested on the model room task, as well as on measures of cognitive inhibitory control, cognitive flexibility, and working memory. Results showed that performance on the model room task was not predicted by measures of inhibitory control or cognitive flexibility, but was predicted by performance on the Delayed Recognition Span Test (DRST), a measure of working memory. These findings lend support to the theory of competing representations and demonstrate the necessity of updating and maintaining strong representations in working memory to succeed in the search task. Success on the DeLoache Model Room Task was predicted by performance on the Delayed Recognition Span Test in children aged 33–39months. The data suggests that working memory updating is a key component in success on the dual representation Model Room Task.