It is now generally agreed that we have to rely on value judgments to distinguish mental disorders from other conditions, but it is not quite clear how. To clarify this, we need to know more than to what extent attributions of disorder are dependent on values. We also have to know (1) what kind of evaluations we have to rely on to identify the class of mental disorder; (2) whether attributions of disorder contain any implicit reference to some specific evaluative standard; and (3) whether the concept of mental disorder is value laden in the definitional or in the epistemic sense. I will argue that the evaluations we have to rely on are mainly considerations of harm, but that we also need to rely on other evaluations; that there should be no references to specific evaluative standards; and that even though mental disorders are necessarily undesirable, "mental disorder" may well be a descriptive phrase.