Study participants were 95 women and 96 men who answered an ad in the University of Oregon student newspaper. Most were university students. They were shown hypothetical stimulus vignettes that described mental patients, and they were asked to assess the probability that the patient would harm someone else, whether the patient should be categorized as "dangerous," and whether coercion should be used to ensure treatment. Findings show that probability and dangerousness judgments were systematically related and were predictive of the judged necessity for coercion; however, judged probability was strongly dependent on the form of the response scale, suggesting that probability was not represented consistently and quantitatively in participants' minds. The second study replicated these findings with forensic clinicians as participants. These results underscore the importance of violence toward others in mental health law and have important implications for the manner in which risk assessments are formulated for use by the legal system. 7 tables, appended study instruments, and 33 references