Several category verification models predict a positive error-reaction time correlation. A speed emphasis category verification task was used to test the hypothesis that disagreement among subjects' category membership judgments accounts for a significant part of that correlation. Two commonly investigated categories were used: vehicle and furniture. Results showed that controlling for intersubject disagreement significantly reduced the correlation (α = .05). The exact way by which disagreement affected the correlation was also shown. Words that admitted intersubject differences in category membership judgments produced false negative and false positive errors, with the latter having a significantly higher probability (α = .01). Consequences for general memory and categorization models are discussed.