Actuarial risk assessments yield valid numerical information about violence risk, but research suggests that forensic clinicians prefer to communicate risk using nonnumerical information (i.e., verbal terms such as high risk). In an experimental questionnaire study, 60 forensic clinicians disagreed on the interpretation of nonnumerical terms, and their nonnumerical risk estimates for one group of violent offenders were influenced by comparison with another group. Adding nonnumerical terms to numerical probability statements had no effect on hypothetical forensic decisions. These findings suggest that nonnumerical descriptive terms do not aid effective communication of violence risk and that contextual information might artificially affect estimated risk. (Published Abstract)