The Linked Dual Representation model (Hutchins & Moreno, 2013) was designed to provide an account for the broad pattern of relationships between vocal perception and production, including both correlations and dissociations between the two. This model makes a unique prediction that musicians with absolute pitch (AP) should be biased towards compensating for objectively mistuned notes in a single note imitation task. In this paper, we tested this prediction by asking musicians with and without AP to imitate vocal notes that are either well-tuned or mistuned. We found that AP musicians were more likely to bias their responses to compensate for mistunings, and that this effect was stronger after longer response delays. We also showed evidence for some implicit AP-like abilities among non-AP musicians. Our findings were predicted by the Linked Dual Representation model, but not other models, providing further evidence for this model.