2 For a literary expression of this clash of intuitions, compare this passage of Eliot, George ‘No man can begin to mould himself on a faith or an idea without rising to a higher order of experience: a principle of subordination, of self-mastery has been introduced into his nature: he is no longer a mere bundle of impressions, desires, and impulses’ (Scenes of Clerical Life [New York: Penguin], 320)Google Scholar, with this passage of Shakespeare, ‘Thus conscience does make cowards of us all,I And thus the native hue of resolution I Is sicklied o'er with the pale cast of thought,I And enterprises of great pitch and moment I With this regard their currents turn awry I And lose the name of action’ (Hamlet III.l).

3 This is analogous to Dan-Cohen's, Meir heuristics in his ‘Conceptions of Choice and Conceptions of Autonomy,’ Ethics 102 (1992)CrossRefGoogle Scholar: ‘I view the two conceptions of autonomy that I distinguish mainly as heuristic devices, designed to facilitate the consideration of various strands in the discussions of autonomy that, given the vagueness of the underlying concept of choice employed in them, are often interwoven in the writings of a single author’ (233). I should perhaps emphasize that my heuristic distinction in no way implies that the creator of the split-level analysis of autonomy became himself a contradictory split personality.

4 For a good recent example of this standard analysis, see Mele, Alfred R. Autonomous Agents: From Self-Control to Autonomy (New York: Oxford University Press 1995)Google Scholar. I have discussed Mele's, view in my ‘Alfred Mele's Voluntaristic Conception of Autonomy,’ in Moral Responsibility and Ontology, ed. Beld, T. van den (Dordrecht: Kluwer 2000) 259–70.Google Scholar

5 Reprinted in Frankfurt, Harry G. The Importance of What We Care About (Cambridge: Cambridge University Press 1988) 80–94CrossRefGoogle Scholar. Hereafter I refer to this first anthology of Frankfurt's articles as ICA.

6 Cf. especially Frankfurt, ‘Identification and Wholeheartedness,’ ICA 159-76Google Scholar.

7 Throughout this paper personal pronouns are used in the generic rather than in any gendered sense.

8 For summaries of the relevant literature, see Shatz, David ‘Free Will and the Structure of Motivation,’ Midwest Studies in Philosophy 10 (1986) 451–82CrossRefGoogle Scholar; Watson, Gary ‘Free Action and Free Will,’ Mind 96 (1987) 145–72CrossRefGoogle Scholar; Christman, John ‘Constructing the Inner Citadel: Recent Work on Autonomy,’ Ethics 99 (1988) 109–24CrossRefGoogle Scholar. I myself expose in more detail the ‘Frankfurt-Dworkin’ model of personal autonomy in my ‘Is Personal Autonomy the First Principle of Education,’ Journal of Philosophy of Education 26 (1992) 5-17.

9 Here I work on the assumption that the regress and authority problem can be partitioned off from the manipulation problem. Accordingly, within the confined format of this paper, I shall not further discuss the controversial idea that autonomy is essentially externalistic or historical. I am aware of the fact, though, that a full defense of hierarchy must include a response to the question whether the hybrid theory has sufficient resources to tackle the manipulation problem. In another paper ‘The Trouble With Harry’ (unpublished manuscript) I try to deal with this immensely complicated issue of Frankfurt's free will internalism.

10 For the detailed statement of this problem, see Friedman, Marilyn A. ‘Autonomy and the Split-Level Self,’ The Southern Journal of Philosophy 24 (1986), 24CrossRefGoogle Scholar; Christman, John Introduction to The Inner Citadel: Essays on Individual Autonomy (Oxford: Oxford University Press 1989), 9–11Google Scholar. In his ‘Autonomy and Personal History,’ Canadian Journal of Philosophy 21 (1991), Christman further claims that whatever version of the hierarchical model will run into this major difficulty: ‘It is important to see how general the regress I ab initio I incompleteness problems are. Any account of rational action that presupposes that the desires that move an agent are “accepted” by her will invite an infinite regress of desires in the explanation of this acceptance’ (8). Note that the regress cluster-problem is sometimes stated in terms of internality instead of autonomy.

11 I hasten to add that, notwithstanding first appearance, it is perhaps not only a possibility but also a requisite that there can be autonomy without autonomous foundations. One could argue, for instance, that for a naturalist-compatibilist there is no ‘ab initio’ problem since compatibilist autonomy must-at least developmentally -be founded on a heteronomous bedrock. For lack of non-natural, transcendental foundations, the autonomous develops only on the foundation of the heteronomous. I owe this point to David Zimmerman. In this paper, however, in order not to complicate matters further, I work on the assumption that the foundations of autonomy must themselves be autonomous. For this standard assumption, see again Friedman, ‘Autonomy and the Split-Level Self’ and Christman, ‘Autonomy and Personal History.’

12 As the incompleteness problem would be solved within the enlarged hierarchical model, the problem could properly be renamed the ‘supplementary’ problem; but to avoid terminological confusion, I retain the standard label of ‘incompleteness’ problem.

13 For the detailed statement of this problem, see Watson, Gary ‘Free Agency,’ in Free Will, Watson, Gary ed. (Oxford: Oxford University Press 1982), 107–10Google Scholar; Thalberg, Irving ‘Hierarchical Analyses of Unfree Action,’ Canadian Journal of Philosophy 8 (1978), 219–20 and 223-5CrossRefGoogle Scholar. Note that this problem is sometimes stated in terms of normativity instead of authority, especially by critics of a Kantian inspiration. See, for example, Piper, Adrian M.S. ‘Two Conceptions of the Self,’ Philosophical Studies 48 (1985), 176–81CrossRefGoogle Scholar; Audi, Robert ‘Autonomy, Reason, and Desire,’ Pacific Philosophical Quarterly 72 (1991), 257–8 and 266CrossRefGoogle Scholar.

14 Frankfurt, ‘Freedom of the Will and the Concept of a Person,’ ICA, 21.Google Scholar

15 Cf. Frankfurt, ‘Identification and wholeheartedness,’ ICA, 168-9 and 172.Google Scholar

16 Frankfurt, ‘Concerning the Freedom and Limits of the Will,’ reprinted in Frankfurt, Harry G. Necessity, Volition, and Love (Cambridge: Cambridge University Press 1999), 79Google Scholar. Hereafter I refer to this second anthology of Frankfurt's articles as NVL.

17 Although this solution does not, of course, address the problem of manipulation, nor that of authority, it adequately addresses, according to these critics, the regresscum- incompleteness problem. See, for example, Fischer, John M. and Ravizza, Mark ‘Responsibility and History,’ Midwest Studies in Philosophy 19 (1994), 443–4, esp. n. 36CrossRefGoogle Scholar.

18 See also note 9 above. I have uncovered and discussed the incompatibilist intuitions behind Frankfurt's, conception of the active will in my ‘Harry Frankfurt on the Will, Autonomy and Necessity,’ Ethical Perspectives: Journal of the European Ethics Network 5 (1998) 44–52.Google Scholar

19 Frankfurt's hierarchical model, in which (effective) desire is the fundamental motivational notion and in which the conception of practical reason remains wholly instrumentalistic, is standardly interpreted as an enrichment and an improvement of the classical compatibilism of Hobbes and Hume. See, for example, Richard, Double The Non-Reality of Free Will (New York: Oxford University Press 1991), 27–31Google Scholar. See also Watson, ‘Free Action and Free Will,’ 147–9Google Scholar and ‘Free Agency,’ 107-9; Piper, ‘Two Conceptions of the Self,’ 173-6; and Audi, ‘Autonomy, Reason, and Desire,’ 259-61.

20 For this interpretation, see, for example, Audi, Robert Practical Reasoning (London: Routledge 1989) 39–59Google Scholar.

21 Although it brings in elements which are at variance with the Hobbesian-Humean theoretical framework, a ‘rationalistic’ revision of the hierarchical model might be perfectly feasible in its own terms and, moreover, completely successful in dealing with the standard problems. For such a revised Frankfurt account which appeals to the person's intellect, see Stump, Eleonore ‘Sanctification, Hardening of the Heart, and Frankfurt's Concept of Free Will,’ The Journal of Philosophy 85 (1988) 395–411.CrossRefGoogle Scholar

22 Frankfurt, ‘Identification and Wholeheartedness,’ ICA, 170CrossRefGoogle Scholar; see also 175. This voluntarism in the analysis of the concept of personal autonomy should not be confused with volitionalism in the analysis of the concept of intentional action (a bodily movement is an intentional action if and only if it is caused by a volition).

23 Here I employ a distinction between an act of deciding (act of will) and the issue or result of deciding (determination of the will) which we also call a ‘decision.’ Although an act of deciding cannot be passive in any way, it may turn out afterwards that the resultant decision is self-deceptive (does not correspond with the true structure of one's will) or akratic (one still does something else than the thing decided upon) and so remains idle (has no real effect on attitude and action). 1n my opinion, Frankfurt has these volitional limitations in mind when he says that the will is not wholehearted.

24 Frankfurt, ‘Identification and Wholeheartedness,’ ICA, 174Google Scholar

25 Frankfurt, ‘The Faintest Passion,’ NVL, 95-107Google Scholar

26 Cf. ibid., 99-105.

27 Ibid., 100

28 The main consequences of this presumption are that (i) the distinction in this paper between (voluntaristic) self-creation and (non-voluntaristic) self-discovery is too sharp and that (ii) the impact of indirect strategies of self-management or ‘character planning’ is neglected. Admittedly, matters are far more complicated than adumbrated here and a fuller discussion should make room for these complications. However, no matter how much ‘making’ and ‘finding’ ourselves are intertwined, states of mind like ‘being self-satisfied’ and ‘caring about something,’ in my view, will always remain at least what Elster, Jon calls ‘states that are essentially by-products: they are states that cannot be brought about intelligently and intentionally [voluntary],’ Sour Grapes: Studies in the Subversion of Rationality (Cambridge: Cambridge University Press 1983), 56CrossRefGoogle Scholar.

29 Cf. Frankfurt, ‘The Faintest Passion,’ NVL, 105.Google Scholar

30 Ibid., 104-5

31 In my judgment, at least three conceptions of the will are present in Frankfurt's writings. According to the first, the will is nothing but (i) effective desire (volition). This ‘appetitive’ conception is standardly associated with the hierarchical model (see section II). According to the second, the will is synonymous with (ii) act of will (decision, choice and intention). This ‘active’ conception is directly connected with what I called voluntarism (in section III). According to the third, the will is (iii) a faculty of the mind (the ‘motivational organ’). This ‘substantial’ conception is the one which emerged during the discussion of limits to voluntarism (in the preceding section IV) and continues to be prominent in what I immediately shall call Frankfurt's non-voluntarism (in this section V and the following ones). Talk about the nature, the identity, and the essential character of a person's will only makes sense on the latter conception, according to which the will has a substantial existence with an essence of its own independent from appetite and reason.

32 Frankfurt, ‘The Importance of What We Care About,’ ICA, 84. One reason why Frankfurt-2's alternative view has hardly attracted any attention is, to my mind, that most participants in the debate presume that a non-voluntaristic view on autonomy must already be a contradiction in terms. But, as I shall try to show, this is not necessarily so. It is true that the distinction between activity and passivity amounts to the same thing as the distinction between autonomy and heteronomy. However, the autonomous/heteronomous distinction does not coincide with the voluntaristic/ non-voluntaristic distinction -that is, the distinction between being under one's voluntary control and being outside such control. Activism or autonomism is not necessarily voluntarism, because a person may be in a state that is not wholly under his direct and immediate voluntary control yet at the same time be fully active, and hence autonomous.

33 Although I cannot possibly give a full analysis of the concept of care within the confines of this paper, it is important to clearly distinguish Frankfurt's construal of this concept from its construal in feminist educational theory or care-ethics as, for example, by Nel Noddings in her Caring: A Feminist Approach to Ethics (Berkeley: University of California Press 1984). According to Frankfurt, the concept of care is not so much ethical as it is anthropological and metaphysical. Furthermore, whereas educationalists and social scientists give an account of ‘caring for,’ Frankfurt gives an account of ‘caring about.’

34 Cf. Frankfurt, ‘The Importance of What We Care About,’ ICA, 82-5Google Scholar.

35 I owe this point to Frankfurt. While treating the authority problem in section VII, I shall point out another essential difference between caring and desiring.

36 Frankfurt, ‘The Importance of What We Care About,’ ICA, 86Google Scholar

37 Cf. Frankfurt, ‘Rationality and the Unthinkable,’ ICA, 181-4Google Scholar.

38 Cf. Frankfurt, ‘The Importance of What We Care About,’ ICA, 87-8Google Scholar.

39 Frankfurt, ‘Autonomy, Necessity and Love,’ NVL, 135Google Scholar. In light of the fact that the volitional necessity of caring about a certain object is both (a) voluntaristically self-imposed and (b) non-voluntaristically imposed by the cared-about object, it might be helpful to compare the somewhat paradoxical phenomenon of caring about something as restricted by volitional necessity to the equally complex phenomenon of reverence for the law as described by Kant, Immanuel in his Groundwork of the Metaphysics of Morals, translated by Paton, H.J. (New York: Harper & Row 1964), 16 (401)Google Scholar, note **: ‘The object of reverence is the law alone- the law which we impose on ourselves but yet as necessary in itself.’

40 In my discussion below of the relevance of the hybrid theory of autonomy and the asymmetrical dependency thesis to the (unsettled) incompleteness problem (in section VII) I touch upon two important qualifications of this claim. First, although the necessities of a person's will might not be diachronically necessary, they certainly are synchronically necessary for him. Second, although it is strictly speaking volitionally possible for a person to escape from the necessitating force of the cared-about object, such an escape will inevitably be tantamount to his self-betrayal. Notwithstanding the ‘contingent’ character of a person's volitional necessities, the necessitating force of the things he cares about is experienced by him as strong as — or perhaps even stronger than — the necessitating force of the Kantian moral law.

41 In spite of their distinctness, these two conceptions of autonomy, each in its own way, agree with the hierarchical model. If the subject/object distinction in epistemology can equally be applied in the metaphysics of autonomy, then one can say that Frankfurt-1 's view gives priority to the reflexive subject, whereas Frankfurt-2’ s view gives it to the cared-about object. This latter view is nevertheless in line with the hierarchical model because caring about something still remains a reflexive process in virtue of the self-imposition of its volitional necessity. Notwithstanding the non-voluntaristic and perhaps even (Freudian) unconscious way in which caring about something constitutes autonomy, the upshot of this process must be consciously accessible and reflexively endorsable. As caring about something is typically personal and therefore not merely wanton, hierarchy (reflexivity) has to make its appearance somewhere. Whereas in the voluntaristic conception of autonomy hierarchy comes in at the very beginning of the constitutive process due to the primacy of the subject, in the non-voluntaristic conception it only comes in at the very end-ex post factum, so to speak- due to the primacy of the object.

42 Additional evidence for hybridity as well as asymmetrical dependency in the conceptual analysis of personal autonomy is given by Charles Taylor in his The Ethics of Authenticity (Cambridge, MA: Harvard University Press 1991), where he contrasts (the ideal of) authenticity with (radical) self-determination and, furthermore, shows how the latter asymmetrically depends upon the former. I have discussed Taylor's, view in my ‘What Wittgenstein Would Have Said About Personal Autonomy,’ Studies in Philosophy and Education 14 (1995) 251–65Google Scholar.

43 I borrow the term ‘asymmetrical dependency’ from Fodor, Jerry A. who uses it while explaining his informational (psycho)semantics in A Theory of Content and Other Essays (Cambridge, MA: The MIT Press 1990), 91–5Google Scholar. In the present context, the relation of asymmetrical dependency is best construed as a relation among identifications. Identifications themselves are intentional and/or causal relations. The (second-order) relation of asymmetrical dependency can be expressed in terms of subjunctive conditionals (counterfactuals): if there were no (or had not been) identifications through caring, then there would not be (or would not have been) genuine and effective identifications through willing, but not the other way around. There being genuine and effective identifications through caring does not depend upon there being identifications through willing.

44 Frankfurt, ‘On the Necessity of Ideals,’ NVL, 110Google Scholar. See also his ‘Rationality and the Unthinkable,’ ICA, 177-8.

45 Frankfurt, ‘On the Necessity of Ideals,’ NVL, 108-16Google Scholar and ‘Autonomy, Necessity, and Love,’ NVL, 129-41

46 Frankfurt, ‘On the Necessity of Ideals,’ NVL, 114Google Scholar

47 By approximation, these notions can be understood as follows. Synchronic contingency: F is synchronically contingent at t, if and only if, F is exemplified at t and it is possible at t that not-F (or G) is exemplified at t. Diachronic contingency: F is diachronically contingent at t₁, if and only if, F is exemplified at t₁ and it is possible at t₁ that not-F (or G) is exemplified at t₂ (and t₂ ≠ t₁). To make sense of the ‘historicity’ of a person's volitional essence, the following condition must also be added: The person-phase including t₂ is only weakly, if at all, mentally connected with the person-phase including t₁

48 Let it be clear that, in my argument for the intrinsic internality of the will, I do not mean by ‘the essential volitional identity of a person’ his identity as one and the same person through time, but his identity as a character. Amélie Oksenberg Rorty and David Wong capture well the sense of identity I have in mind, in Flanagan, Owen and Rorty, Amélie Oksenberg eds., Identity, Character, and Morality: Essays in Moral Psychology (Cambridge, MA: The MIT Press 1990)Google Scholar: ‘A person's identity is constituted by a configuration of central traits’ (19). However, it should be observed that the destruction of the latter ‘qualitative’ type of identity (personal integrity and self-respect) sometimes leads to the destruction of the former ‘numerical’ type as, for example, in cases of transgressions of ‘codes of honor’ (e.g., Samurai, Mafia) where betrayal ipso facto means self-betrayal, which makes suicide (biological self-destruction) almost inevitable.

49 Cf. Frankfurt, ‘Autonomy, Necessity and Love,’ NVL, 134-8.

50 Ibid., 137-8

51 For Frankfurt's further development of this theory of value, see his ‘On the Usefulness of Final Ends,’ NVL, 82-94.

52 Cf. Frankfurt, ‘Rationality and the Unthinkable,’ ICA, 190. See also note 31 above.