In many audit tasks, auditors evaluate multiple hypotheses to diagnose the situation. Research suggests this is a complex task that individuals have difficulty performing. Further, there is little guidance in professional standards or literature dealing with the many complexities present in the audit environment. Using probability theory, this study derives the appropriate revision of likelihoods for multiple hypotheses given different realistic audit conditions. The analysis shows that the relationships among the hypotheses dramatically impact the use of audit evidence and the resulting pattern of probability revisions. We also identify testable hypotheses to guide future research and discuss practice implications regarding ways to improve the effectiveness of analytical procedures.