The blocking effect, canonical in the study of associative learning, is often explained as a failure of the blocked cue to become associated with the outcome. However, this perspective fails to explain recent findings that suggest learning about a blocked cue is superior to a different type of redundant cue. We report an experiment designed to test the proposal that blocking is not a failure of association, but a performance effect arising from a comparator process (Denniston, Savastano, & Miller, 2001). Participants received A+ AX+ BY+ CY- training containing a blocked cue X and another redundant cue Y, before rating outcome expectancies for individual cues. These ratings were inconsistent with the association-failure view. After subsequent A- Y+ training, participants rated cues again. Ratings in the second test were inconsistent with the comparator theory. Our data suggest that neither perspective is likely to provide a complete account of causal learning.