Summary This paper considers a number of questions concerning a central principle of modern liberal medical ethics – that of respect for patient autonomy. This principle is generally interpreted as a consequence of the more fundamental ethical requirement of respect for personhood, defined in terms of the ‘autonomy’ of a rational being. In the context of medical ethics, this is taken to imply that patients, as persons, have the right to make decisions about their own medical treatment or non-treatment. This, it is argued, involves a misuse of the essentially Kantian concept of ‘autonomy’. For Kant, to be autonomous is to be capable of making universal moral judgements, based on impersonal reason alone. In liberal medical ethics, however, autonomy means the capacity to make decisions about non-moral matters such as one's own interests, on the basis of values which one has arrived at by free and independent thought. These are clearly different conceptions, so that a definition of personhood in terms of autonomy in the Kantian sense will be different from one in terms of liberal autonomy. This casts doubt on whether respect for patient autonomy in the liberal sense is an implication of respect for personhood. It is not clear why a patient's decisions about treatment ought to be respected simply because they are based on the patient's own (non-moral) values. There is a further difficulty. ‘Autonomy’, in the liberal conception, seems to be equivalent to ‘decision-making competence’. The elements of this are said to be: 1. The capacity to understand; 2. The capacity to reason; and 3. Possession of one's own developed structure of values. Decision-making competence in this sense is not found in all human beings, or in any human being all the time. For example, small children, adults with learning difficulties, and people with brain damage seem to lack all three elements, or to have them only to a limited extent. People with a serious illness may lack all three temporarily. And people with certain kinds of mental disorder may, just because of their disorder, deviate from their own normal structure of values. It seems to follow that not all human beings are fully persons worthy of respect in the liberal sense of personhood. This has morally undesirable implications, which suggests problems in the liberal conceptions both of personhood and of respect. It is argued that a satisfactory conception of personhood can be arrived at only by considering the use of terms such as ‘person’ in everyday, non-technical, contexts. It is argued that ‘person’, in such contexts, is almost synonymous with ‘human being’, though with particular reference to the non-biological characteristics of human beings – those in virtue of which we recognize each other as fellow-humans (as ‘one of us’). A human being does not necessarily possess decision-making competence as defined above: so respect for persons cannot depend on such competence, or require us always to accept a person's own presently expressed decisions about what is best for themselves. It does, however, require us to recognize others as separate selves, with interests of their own, which may differ from ours. A doctor is therefore morally required to give the patient's interests priority over his or her own, or those of the health care system. If doctors consider the patient's expressed decision about treatment to be not in the patient's own interests, they are morally required to engage in dialogue with the patient, in an effort to arrive at a decision which will be acceptable both to the doctor and to the patient. If the patient is unable to engage in such dialogue, because of impaired decision-making competence, then other methods of establishing the patient's interests must be tried: examples would be proxy decision-making, by someone who is close enough to the patient to be aware of, and concerned for, the patient's interests; or appeal to the patient's own advance directives, where available. The most difficult case is that of patients with certain kinds of mental disorder, where the patient's own sense of his or her interests has become distorted by the very disorder which needs treatment. Only by following an appropriate course of treatment, which the patient may currently reject, is it possible to restore the patient to a condition in which he or she can genuinely engage in dialogue. This, it is argued, makes it morally justifiable in such cases simply to override the patient's expressed wishes.