AbstractAt least five studies report elevated schizotypy scores in cannabis users. The current research confirms higher schizotypy scores in regular cannabis users. Nevertheless, further analyses reveal that select items on the brief version of the Schizotypal Personality Questionnaire do not function comparably in current and former users. Multiple statistical approaches confirmed this problem, including the Mantel‐Haenszel statistic, Rasch difficulty estimates, a logistic regression approach, the Breslow‐Day (BD) statistic, and a combined decision rule using Mantel‐Haenszel and BD together. Cannabis users appear to misinterpret at least one item, “I sometimes use words in unusual ways” making them more likely to endorse it even if they are no more schizotypal than the former users. Users might consider cannabis‐related slang as an unusual use of words. Removing problematic items does not decrease the internal consistency of the measure, but eliminates the significant difference between current and former users. These results suggest that links between cannabis use and schizotypy require cautious interpretation and may arise from measurement problems. Copyright © 2006 John Wiley & Sons, Ltd.