By asking questions that can be understood and handled by mental health professionals, the court can receive useful information and force such witnesses to be more effective in testifying about a defendant's dangerousness. The court-clinic prediction issue cannot be clarified, however, until the principal parties use the same terminology. Common language, such as 'Do you consider Mr. Jones to be dangerous?' is not sufficient. A better question is 'Can you give an expression of probability concerning the likelihood that Mr. Jones, if given the opportunity, will commit further assaults on his employer?' Even better is the following question: 'Can you cast some light on the likelihood that Mr. Jones is likely to act in a harmful fashion in the future?' When performing as an expert witness, mental health professionals should reply in specific terms, citing empirical studies on predictability and clearly stating that the opinion is an estimate, not certainty. This approach does not alter the way forensic clinicians do assessments, but merely draws attention to the choice of words in testimony and written reports and avoids lay phrases that are subject to misinterpretation. The paper includes 14 references.