[M]y question is how these patients while competent might be able to give their own informed consent to treatment, despite being both unwilling and incompetent to do so when treatment is to be begun, thereby reducing the need to relax the dangerousness criteria for involuntary commitment. It is uncontroversial that the dangerousness requirement would be too restrictive for all treatment of mental illness. When competent patients voluntarily seek and/or accept treatment for their mental illness, neither public policy nor medical practice restricts treatment to those patients judged to be dangerous. Instead, criteria should be and generally are comparable to those for the treatment of physical illness -- whether the patient is ill, in this case mentally ill, and likely to benefit from treatment. Through use of advance directives, it would be possible for mentally ill persons who are currently refusing treatment to give prior consent, while competent and with their disease in remission, to treatment at a later time when they are incompetent, have become noncompliant, and are refusing treatment. My proposal is certainly not entirely novel, since others have made similar proposals under the heading of Ulysses contracts and voluntary commitment contracts. Addressing briefly some of the criticisms of these earlier proposals will bring out one fundamental difference between them and my proposal here for a new use of advance directives -- whether the patient must then be incompetent when the contract or directive made earlier is later invoked -- a difference I shall argue strongly favors my proposal.