In a recent series of papers, Pearce and colleagues (e.g., Pearce, Dopson, Haselgrove, & Esber, 2012) have demonstrated a so-called "redundancy effect" in Pavlovian conditioning, which is the finding of more conditioned responding to a redundant cue trained as part of a blocking procedure (A+AX+) than to a redundant cue trained as part of a simple discrimination procedure (BY+CY-). This phenomenon presents a serious challenge for those theories of conditioning that compute learning through a global error-term. In this paper, we use the Rescorla and Wagner (1972) model as a prototypical example to demonstrate that the redundancy effect can be accounted for by this class of theories if the experimental stimuli are assumed to share a common component. We also point to some domains in which this approach leads to novel predictions that may deserve empirical evaluation.