2 For a survey of the various conceptions of autonomy in the recent philosophical literature, cf. Christman, John ‘Constructing the Inner Citadel: Recent Work on Autonomy,’ Ethics 99(1988) 109-24.CrossRefGoogle Scholar

3 This is, of course, the sense of ‘freedom’ relevant to moral and political debates. For a discussion of the various conceptions of freedom in this sense, see Parent, W.D. ‘Some Recent Work on the Concept of Liberty,’ American Philosophical Quarterly 11 (1974) 149-67.Google Scholar The notion of internal and external restraints is explained in Feinberg, Joel Social Philosophy (Englewood Cliffs, NJ: Prentice Hall 1973), ch. 1.Google Scholar There Feinberg uses the terms ‘compulsion’ and ‘constraint’ interchangeably, and in doing so, I think he confuses the very issue I discuss.

4 Even if ‘internal restraints’ are included among the kinds of things that diminish freedom, things like hypnosis or subliminal messages do not, internally or externally, stop a person from doing something. Quite the opposite: they force a person to (prefer to) do something.

5 The Theory and Practice of Autonomy (Cambridge: Cambridge University Press 1988), 16

6 Cf. Wright Neely: ‘freedom is not just a matter of doing as one desires, but requires, in addition, that we should have something to say about what we desire’ (‘Freedom and Desire,’ Philosophical Review 83 [1974] 37). This brings the account of autonomy into line with the debate in the philosophy of the social sciences over the problem of ‘endogeneity of preferences.’ Cf., for example, Elster, Jon Sour Grapes (Cambridge: Cambridge University Press 1983).CrossRefGoogle Scholar Also, for an account of autonomy that resists the reduction to reference only to preferences, see Robert Young,’ Autonomy and the Inner Self’ reprinted in Christman, John ed., The Inner Citadel: Essays on Individual Autonomy (New York: Oxford University Press 1989), 77-90.Google Scholar

7 This does not preclude any of a variety of claims about the contingent relation between autonomy and freedom, i.e., that freedom is contingently necessary for the further development of a person’s autonomy: cf. Young, Robert Personal Autonomy: Beyond Negative and Positive Liberty (New York: St. Martin’s Press 1986), ch. 2.Google Scholar Similarly, this will not rule out (in fact it complements) the use of the term ‘autonomy’ to refer to a right to these conditions, or the right to be treated with respect for one’s capacity to act under such conditions.

8 I say ‘belief here, but for the most part my discussion concerns autonomous preferences. Whether or not this model can be applied to belief formation is left open (though see my discussion of this in section IV [1] below). Also, the terms ‘desire’ and ‘preference’ are used broadly to mean ‘motive,’ ‘value,’ or any ‘positively valenced’ attitude a person might have vis-à-vis a state of affairs.

9 Dworkin, ‘The Concept of Autonomy,’ in Haller, R. ed., Science and Ethics (Amsterdam: Rodopi 1981), 212.Google Scholar Dworkin has since published an updated version of this essay where he reverses certain aspects of his view: cf. The Theory and Practice of Autonomy, ch. 1. His revisions are treated below.

10 Harry Frankfurt puts forth a similar account in his well known work concerning freedom of the will. Although Frankfurt introduces the notion of ‘volitions’ to capture those desires that an agent actually wants to issue in action (to be one’s will), the components of the theory are essentially the same. And, as we will see, the problems are also the same. Cf. ‘Freedom of the Will and the Concept of a Person,’ reprinted in John Christman, ed., The Inner Citadel, 63-76.

11 Strawson, Galen Freedom and Belief (Oxford: Oxford University Press 1986), 245Google Scholar

12 This is a problem that Gary Watson and others have raised about models such as this: some of the desires we identify with we do so not simply because we desire them but because we view them as desirable, objectively. This suggests at least that the requirement of identification is ambiguous. Cf. Watson, ‘Free Agency,’ in The Inner Citadel, 109-22. Cf. also Irving Thalberg, ‘Hierarchical Analyses of Unfree Action,’ in The Inner Citadel, 123-36; and Marilyn Friedman, ‘Autonomy and the Split-Level Self.’ For another (albeit unsuccessful, I think) response to the charge that the condition of identification is implausibly vague, cf. Young, Personal Autonomy, 43-7.

13 Cf. The Theory and Practice of Autonomy, ch. 1. The problems Dworkin adduces with the condition of identification, however, differ from those discussed here.

14 Dworkin comments on disabilities such as these and says that they are compatible with autonomy (ibid., 17). But if that is true, it is not clear what the requirements for the ability in question really are.

15 For a discussion of this charge, d. Thalberg, 129-35 and Friedman, 22-3. For development of the reply given in the text, cf. Christman, John ‘Autonomy: A Defense of the Split-Level Self,’ Southern Journal of Philosophy 25 (1987) 281-93.CrossRefGoogle Scholar

16 Strawson argues along these lines in ch. 2. He concludes, however, that this shows that free action is conceptually impossible. If my arguments below are successful it will be clear that this is not so.

17 ‘Identification and Wholeheartedness,’ in Schoeman, F. ed., Responsibility, Character, and the Emotions (Cambridge: Cambridge University Press 1987) 27-45Google Scholar, at 37.

18 Another reading of ‘decisive,’ however, might be this: in light of an ‘objective’ analysis of all relevant information having to do with the agent’s desire, the agent need go no further in investigating and evaluating her desire. That is, no matter what the agent herself knows or wants, there are objectively good reasons for her to identify with the desire in question. This amounts to adding an ‘external’ rationality condition as a requirement of autonomy. But this move effectively separates the property of autonomy from the actual decisions and judgments of real people; hence, this component of freedom would fail to capture the idea of self-government that is the motivating concept behind autonomy. I develop this point in section IV (1) below. Dworkin also discusses these criticisms and responds to them in The Theory and Practice of Autonomy, 18-20. However, his response does not, in my view, succeed in galvanizing his theory against the incompleteness problem.

19 Integration views are ones where, for example, ‘autonomy is achieved in virtue of a … process of integration within a person’s hierarchy of motivations, intermediate standards and values, and highest principles’ (Friedman, 34). Only when this complete integration takes place — with no level commanding special status in the conferring of autonomy— can we say that a person is autonomous. The major defect in integration views of autonomy, however, is this: a person can be manipulated and conditioned to such an extent that she gains a coherent and integrated set of desires as a result, but one which is totally the result of external manipulation. And it will not do to spread the integration conditions over a person’s entire life, saying that a person is autonomous (over her whole life) only if each desire she has conforms to an over-all coherent life plan, or in relation to her entire character. Again, a person might have been so severely conditioned to want to do what her brainwashers say she must that she will continue to develop her life along those very directives if not otherwise interfered with. An entire life can be ‘coherently’ manipulated just as a single set of desires can.

20 Fingarette, Herbert in Self Deception (London: Routledge and Kegan Paul 1969Google Scholar) describes the act of ‘spelling out’ to oneself one’s beliefs. This is similar to what I have in mind here. For a critical discussion of Fingarette, cf. Haight, M.R. A Study of Self Deception(Atlantic Highlands, NJ: Humanities Press 1980Google Scholar), Ch. 7. This notion of transparency might be resisted by those influenced by Freud who are convinced that much of our motivational structure is not immediately transparent to us (without therapy, dream interpretation or the like). I do not wish to dispute this. I only claim here that insofar as a person’s motives are subject to reflective consideration by the agent under normal conditions, she is autonomous. If therapy is necessary to make this possible, then therapy is necessary to make autonomy possible.

21 Jon Elster, in Sour Grapes, provides a topology of changes in belief and desire that is informative here. Some of the changes he labels ‘autonomous’ and others not.

22 ‘Consistent beliefs and desires’ is ambiguous in several ways. By ‘consistency of beliefs’ I mean that the set of beliefs could all be true in a single possible world (though as I explain in the text, the consistency requirement I adopt is that there can be no manifest inconsistencies in the belief set). In the case of consistency of preferences, it is common to require that they be transitive, complete, and continuous. These are very stringent requirements, though, because most people have not compiled a complete ranking of all the available objects of preference. Hence, by ‘consistency of desires’ I will mean simple transitivity of those desires plus consistency of the beliefs upon which they rest (if any).

23 The distinction I am describing closely parallels Richard Brandt’s distinction between ‘subjective’ and ‘objective’ rationality. Cf. Brandt, A Theory of the Good and the Right (Oxford: Oxford University Press 1979) 72ff.

24 For consideration of this question, cf. Young, Personal Autonomy, ch. 2; Lindley, Richard Autonomy (New York: Macmillan 1986),CrossRefGoogle Scholar and Haworth, Lawrence Autonomy: An Essay in Philosophical Psychology and Ethics (New Haven, CT: Yale University Press 1986)CrossRefGoogle Scholar Part I.

25 ‘Two Concepts of Liberty,’ in Four Essays on Liberty (Oxford: Oxford University Press 1969)133

26 For example, Lawrence Haworth (Autonomy, ch. 2) argues that ‘full rationality’ is a necessary condition for what he calls ‘normal autonomy.’ Full rationality involves critically analyzing one’s ends (as well as beliefs about means) in accordance with one’s other beliefs, higher principles and values, future preferences one expects to have, and present preferences one has about the future. Wholesale avoidance of manifest inconsistency of the sort I defend would ensure that this kind of condition is also met.

27 Cf. Elster, Sour Grapes, for a discussion of this. I am grateful to Richmond Campbell for calling my attention to the need for this qualification.

28 Cf. Haight, A Study of Self-Deception, ch. 6.

29 For a more detailed analysis of these conditions and their relation to autonomy, cf. Young, Robert ‘Autonomy and the Inner Self,’ The Inner Citadel, 81-9Google Scholar.

30 David Zimmerman notices the necessity of this premise and also denies it (though for reasons different from mine). Cf. ‘Hierarchical Motivation and Freedom of the Will,’ Pacific Philosophical Quarterly 62 (1981) 354-68.

31 For a discussion of this phenomenon within a general theory of self deception, cf. Robert Audi, ‘Self Deception and Rationality,’ in Martin, M. ed., Self Deception and Self Understanding (Lawrence, KS: Kansas University Press 1985) 169-94Google Scholar.

32 A full defense of my approach to autonomy would involve an exhaustive account and criticism of alternative ‘non-subjective’ views which this is intended to replace. I cannot give such an exhaustive argument here, though the simple answer to the non-subjectivist could be put as follows: imagine any process which you stipulate as resulting in the subversion of autonomy, and I can imagine a person who (given a fantastic enough hypothetical situation) would want her choices to be formed that way, and this desire is itself an autonomous one. What follows is that, since for that person, the choice generation procedure is an expression of ‘self-government’ of the crucial sort, she should not be labeled non-autonomous because she chooses unconventional means for getting what she wants.

33 It may illuminate this suggestion to contrast it with a similar view put forth by Young, Robert in ‘Autonomy and Socialization’ (Mind 89 [1980] 565-76)CrossRefGoogle Scholar. Young argues that we display autonomy if, after ‘the processes of our socialization are brought to the level of consciousness … the possibility is there for us either to get free of desires we’d prefer not to have or to identify with previously unrecognized motivations’ (573-4). In other words, if our identification with our desires survives bringing to consciousness an awareness of the processes that brought about that desire, the person is autonomous vis-à -vis that desire. This account has much in common with my view, though I avoid the notion of identification and I add conditions to respond to the ‘regress’ and ‘incompleteness’ problems. Young’s view would still run afoul of the kinds of counterexamples we have discussed if the conditioning has been so fierce as to manipulate identification with the desires even after the person gains knowledge of the source of the preference.

34 Cf. Gerald Dworkin’s updated view (The Theory and Practice of Autonomy, ch. 2) where he also defends a ‘content neutral’ conception.