How is a person's freedom related to his or her preferences? Liberal theorists of negative freedom have generally taken the view that the desire of a person to do or not do something is irrelevant to the question of whether he is free to do it. Supporters of the “pure negative” conception of freedom have advocated this view in its starkest form: they maintain that a person is unfree to Φ if and only if he is prevented from Φ-ing by the conduct or dispositions of some other person(s) (Steiner, Reference Steiner1994; Carter, Reference Carter1999; Kramer, Reference Kramer2003). This definition of freedom is value-neutral in the sense that no reference is made to preferences over options or indeed to any other indicators of the values of options, either in the characterization of “Φ-ing” itself (any conduct fits the bill) or in the characterization of the way in which Φ-ing can be constrained (any prevention counts as a constraint on freedom).

In a stimulating article recently published in this journal, Keith Dowding and Martin van Hees have attacked the pure negative-liberty theorists’ claim to have constructed a genuinely “preference-independent” conception of freedom (Dowding and van Hees, Reference Dowding and van Hees2007). In this reply, we shall defend the pure negative conception against their attack. We shall first impugn their central claim that pure negative-liberty theories have to allow that a person's preferences in themselves can be determinants of her own freedom. To this end, we shall distinguish between two ways in which a person's freedom can be said to depend on her preferences. We shall then argue that only one of these kinds of “preference-dependence” is unacceptable, whereas the other is both unavoidable and unobjectionable.

1. TWO KINDS OF PREFERENCE-DEPENDENCE

The central claim in Dowding's and van Hees's critique of pure negative-liberty theories is that, although the proponents of such theories purport to separate freedoms from preferences, they are unable to sustain such a separation. Dowding and van Hees set out to demonstrate two theses: (1) “a person's freedom may depend on the preferences of others”; and (2) “a person's [pure negative] freedom may depend on her own preferences” (Dowding and van Hees, Reference Dowding and van Hees2007, 152). It is not entirely clear whether Dowding and van Hees regard the first of these theses as incompatible with pure negative-liberty theories. Their discussion of the matter seems to be headed in that direction, but they carefully draw back from affirming such a position. At any rate, it would be exceedingly odd for them to suppose that their first thesis is inimical to the pure negative-liberty analyses advanced by either of us, given that each of us has propounded arguments that either presuppose or explicitly champion that first thesis (Carter Reference Carter1999: 228–34, 246–68; Kramer Reference Kramer2003: 209–38). An insistence on the preference-independence of freedoms and unfreedoms is focused principally on the preferences of each person to whom any freedoms or unfreedoms are ascribed, rather than principally on the preferences of other people. (To be sure, a pure negative-liberty theorist will firmly deny that there are constitutive connections – as opposed to causal connections – between anybody's freedoms and the preferences of other people. However, very few philosophers of any stripe have ever posited such connections. Constitutive links have usually been drawn instead between each person's freedoms or unfreedoms and her own preferences. Consequently, the insistence of pure negative-liberty theorists on the preference-independence of freedoms and unfreedoms is focused primarily on each person's own preferences. Such theorists have wisely concentrated their efforts on rebutting a view that has been espoused with some frequency, rather than on rebutting a view that has very seldom been articulated.)

Unlike the first of the quoted theses, the second thesis appears at first sight to be more damaging for the pure negative-liberty conception. What Dowding and van Hees are asserting is that a person's own preferences can augment or diminish that same person's array of pure negative freedoms. Construed in a certain fashion, such an assertion is precisely what the pure negative-liberty theorists deny when they insist on the preference-independence of freedoms and unfreedoms. As will be seen, however, Dowding and van Hees do not genuinely engage with the pure negative-liberty theorists’ position; their second thesis turns out to assert something quite different from what those theorists deny.

A pure negative-liberty theorist asks the following question. If a person's preferences have undergone some change, and if nothing else in the world has altered, can the person's freedoms have been affected by the change in her preferences? Some philosophers such as the Stoics have answered this question in the affirmative. According to the Stoical account of freedom,Footnote ¹ a person will acquire new liberties if she loses a desire to do something which she is unable to do, even if nothing else in the world has changed. A similar conclusion follows from the view that an agent is unfree when and only when he is prevented by somebody else from doing what he wants. If Joe cannot defeat Harry in a race R, and if Joe wants to defeat Harry in R, then Joe is unfree to defeat Harry in R. If Joe now loses his desire to defeat Harry in R, then – according to the “doing what he wants” conception of liberty – he has relieved himself of an unfreedom even though nothing in the world apart from his desire has changed.

The key question for the pure negative-liberty theorists, then, is whether the preferences of a person P in isolation from all other factors can affect the range of P's freedoms. Can P expand/reduce his number of freedoms simply by altering his attitude toward his situation? Now, Dowding and van Hees are not Stoics. They want to avoid any affirmative answer to the pure negative-liberty theorists’ question. However, they believe that those theorists themselves cannot avoid such an answer. Dowding and van Hees write that, under the pure negative conception of liberty, “one can lose a freedom simply because one's preferences have changed” (p 155, our emphasis). They seek to substantiate their indictment of that conception with the following example, which should be quoted at some length:

Far from showing that the pure negative conception of liberty entails the conclusion that a person's desires in themselves can be determinants of the person's freedoms and unfreedoms, the example adduced by Dowding and van Hees tends to show the opposite. What is determinative of Fred's freedom in their example is not his preferences in themselves – his preferences in isolation from other factors – but instead Ami's beliefs and concomitant intentions. Only because her beliefs about his preferences have changed, and only because her concrete intentions have changed correspondingly, is there any alteration in his freedoms. Had there been no actual change in his preferences, and had her beliefs about his preferences shifted in the specified manner along with her accompanying intentions, his freedom to obtain the book would have been lost in the same way as in the quoted example. Conversely, if Fred's preferences had altered without any corresponding shift in Ami's beliefs and her concomitant intentions, his freedom would have remained unaffected. In short, Dowding and van Hees have failed to demonstrate that a person's pure negative liberties are susceptible to being removed or restored by a change in the person's own preferences in isolation from other factors. What they have demonstrated is that one's pure negative liberties are susceptible to being removed or restored by changes in the intentions of other people, and that those intentions can themselves be affected by one's own preferences. Pure negative-liberty theorists are certainly alert to these latter points: it is indeed obvious that what a person decides to do will depend partly on what she thinks other people are inclined to do.

From the foregoing discussion, it is evident that the idea of “the dependence of a person's freedoms on her own preferences” can be interpreted in two different ways. Under the Stoical conception of liberty, changes in the preferences of a person P can be sufficient as well as necessary conditions for changes in P's freedoms. Under the pure negative-liberty conception, by contrast, changes in P's preferences are sometimes necessary but never sufficient conditions for changes in P's freedoms. Some concise though slightly imprecise labels for these two kinds of preference-dependence will be useful. Let us designate the first (which can occur under the Stoical conception) as “direct preference-dependence”, and the second (which can occur under the pure negative-liberty conception) as “indirect preference-dependence”.

Note that our present discussion – with its distinction between direct and indirect preference-dependence – has not invoked at all the distinction between causal connections and conceptual connections, to which Dowding and van Hees devote attention (Dowding and van Hees, Reference Dowding and van Hees2007: 155–6). Though the connections between freedoms and preferences posited by Stoics and others are indeed constitutive rather than causal, our rejoinder has not focused on that dichotomy. Instead, we have concentrated solely on the distinction between the freedom-affecting efficacy of preferences in themselves and the freedom-affecting efficacy of preferences in combination with other factors. By highlighting that latter distinction, we have singled out the central mistake which a pure negative-liberty theory avoids: the mistake of thinking that “one can lose [or gain] a freedom simply because one's preferences have changed”.

Mysteriously, Dowding and van Hees state that an earlier discussion of freedoms and preferences by Kramer invoked semantic considerations in support of the pure negative-liberty position (Dowding and van Hees, Reference Dowding and van Hees2007: 155). In fact, Kramer makes no mention whatsoever of semantic considerations in the pages cited by Dowding and van Hees, and he elsewhere indicates that such considerations are inadequate to settle any of the interesting debates about the nature of freedom (Kramer Reference Kramer2003: 152–3). An appropriate justification for the pure negative-liberty theorists’ distinction between direct and indirect preference-dependence must lie elsewhere. Such a justification can be found in the normative arguments against preference-dependence cited by Dowding and van Hees in their critique of the pure negative conception.

Before we present that justification, however, we shall briefly assess Dowding's and van Hees's preferred conception of freedom by reference to their own normative and methodological standards.

2. WHY INDIRECT PREFERENCE-DEPENDENCE IS UNAVOIDABLE

Because Dowding and van Hees consider indirect preference-dependence to be no less damaging for a conception of negative liberty than direct preference-dependence, they seek to formulate a conception of freedom that avoids both. This alternative conception of freedom they call the “intention-dependent” conception (although it should be emphasized that the ‘dependence’ referred to here is a dependence of the freedom of a person P on the intentions of a preventing agent Q, and not a dependence of P's freedom on the intentions of P himself). According to the intention-dependent conception, a person's freedom to Φ is removed if and only if, in the event that he attempts to Φ, he will be intentionally prevented by some other person or persons from Φ-ing. Whereas Dowding and van Hees attribute to the pure negative conception of liberty the consequence that a person may shed a particular unfreedom by suppressing her own desire to do something, they believe that their intention-dependent conception avoids such a consequence. We shall argue, against Dowding and van Hees, that both the pure-negative conception and the intention-dependent conception leave room for the indirect preference-dependence of freedoms and unfreedoms.

Consider the story of Fred and Ami that was quoted earlier. That story contains two alternative scenarios, which Dowding and van Hees call “Game 3” and “Game 4” and which have the following features:

Under a non-intention-dependent conception of freedom (such as the pure negative conception), Fred is free to obtain the book in Game 3 but unfree to obtain the book in Game 4. His freedoms are therefore (indirectly) dependent on his preferences. Under an intention-dependent conception, according to Dowding and van Hees, the two games will instead produce identical results concerning Fred's freedom or unfreedom to get the book. Dowding and van Hees thus believe that, under the latter conception of liberty, Fred's freedom to get the book is never (even indirectly) dependent on his preferences. Their argument can be reconstructed along the following lines.

In Game 4, Ami's behaviour might be explained by reference to any one of four possible preference profiles over outcomes:

1. (a) Ami desires only to do the same thing as Fred;

2. (b) Ami desires both to do the same thing as Fred and to prevent Fred from getting the book, but the former desire dominates;

3. (c) Ami desires both to do the same thing as Fred and to prevent Fred from getting the book, but the latter desire dominates;

4. (d) Ami desires only to prevent Fred.

Let us take as given (i) the intention-dependent conception of freedom and (ii) any one of these four preference profiles on the part of Ami. Does Fred's freedom change as a result of his losing his desire to get the book, as happens in the move from Game 4 to Game 3? Here are the answers:

• • Under preference profile (a): Fred is free to get the book in both games, for although Ami prevents Fred from doing so in Game 4, she does not intend to do so either in Game 3 or in Game 4.

• • Under preference profile (b): Fred is free to get the book in both games, for although Ami prevents Fred from doing so in Game 4, and although she intends to do so in Game 4, her intention to prevent Fred is not what explains her preventive action in Game 4.

• • Under preference profile (c) or preference profile (d): Fred is unfree to get the book in Game 4, but Game 3 cannot occur, for Ami's preferences exclude its possibility.

Accordingly, it will never be the case that, in the absence of a change in Ami's preferences over possible outcomes, Fred can gain any freedom by moving from Game 4 to Game 3 – that is, by eliminating his preference for obtaining the book. The intention-dependent conception of freedom appears indeed to rule out the indirect preference-dependence of Fred's freedoms.

This reasoning is sound, but it does not exhaust all the possible preferences for Ami nor all the possible games in which Ami's propensity to perform preventive actions might be affected by Fred's preferences. Consider, for example, a situation in which Ami desires above all that the outcome not include Fred's having the book, but in which she also has other desires that (for practical reasons) she is able to satisfy only if she does not act to prevent Fred from getting the book. In this case, Ami will act to prevent Fred from getting the book unless she believes that he will not attempt to get it. In the light of this observation, the following variations on Games 3 and 4 suggest themselves:

Now, in Game 6 the following two claims are true: (a) Fred desires to get the book; (b) if Fred attempts to get the book, he will be intentionally prevented by Ami from doing so. In Game 5, on the other hand, the following two claims are true: (a*) Fred does not desire to get the book; (b*) if Fred attempts to get the book, he will not be intentionally prevented by Ami from doing so. The change in Ami's intentional behaviour (the change from b to b*) is not brought about by any change in her own preference profile over the possible outcomes of the game, but is a rational change of strategy effected in the light of a change in Fred's preferences (the change from a to a*). The change from Game 5 to Game 6 involves a change in Fred's freedoms that is (indirectly) preference-dependent. Under the intention-dependent conception of freedom no less than under the pure negative conception, then, a person can bring about changes in the freedom-restricting behaviour of others by changing his or her own preferences. Contrary to what Dowding and van Hees assert, the intention-dependent conception allows for the indirect preference-dependence of freedoms just as surely as does the pure negative conception.

Dowding and van Hees might retort that (indirect) preference-dependence will occur less often with intention-dependent liberty than with pure negative liberty. But this is an empirical claim, the truth of which is far from obvious. After all, rulers do have a greater propensity to legislate against certain forms of behaviour when they think citizens are inclined to engage in them, and a similar point applies to choices about the use of police time in enforcing existing laws. Moreover, even if the empirical retort just mentioned is correct, its importance is doubtful. That retort presupposes that a theory's allowance of the possibility of indirect preference-dependence is a shortcoming that should be minimized. We reject such a presupposition. Having emphasized that every credible theory of liberty leaves room for the potential occurrence of indirect preference-dependence, we shall presently argue that that feature of every such theory is wholly innocuous.

How does the intention-dependent conception fare under Dowding's and van Hees's methodological criterion? It seems to us that the evidence here is inconclusive. Dowding and van Hees point out that intentional preventions appear to be more significant politically than non-intentional ones: being prevented from voting by a policeman intentionally blocking the entrance to a polling station is more objectionable than being prevented from voting by a traffic jam. Hence, we are told, the conception is especially suitable for a political theory of freedom. Even if we grant this point, however, there are countervailing considerations on the other side. People commonly wish to condemn incompetent governments as well as malign or paternalist ones on the grounds that they restrict our freedom. Accordingly, it would be unwise for a political theorist to dismiss the category of “unintentional restrictions of freedom” as an empty set. There is no plausible reason for denying, by definitional fiat, that the unintended side-effects of well-meaning but short-sighted government measures can themselves constitute politically relevant restrictions of freedom.

In discussing the objection that their conception of freedom is skewed towards Hayekian libertarianism (and thus that it is normatively appealing only in the eyes of libertarians), Dowding and van Hees suggest that their conception is able to accommodate the view that unintended side-effects can constitute restrictions of freedom. “Intentional prevention” can be interpreted strongly or weakly, they say. On a weak interpretation, for a person to have done something intentionally it is sufficient that that person “could have known” that his action would have the effect of preventing some conduct by another person (Dowding and van Hees, Reference Dowding and van Hees2007: 158). Now, in the context of an intention-dependent conception of freedoms and unfreedoms, this weak interpretation of “intentional prevention” may indeed prove more attractive than the strong interpretation (as is argued in Miller, Reference Miller1983, where such prevention is labelled as “responsibility-triggering”). Unfortunately, the weak interpretation of ‘intentional prevention’ also provides the coup de grâce for Dowding's and van Hees's already ailing thesis that their intention-dependent conception of freedom allows us to escape indirect preference-dependence. For the weak interpretation of ‘intentional prevention’ directly undermines their own argument to the effect that Fred does not gain any intention-dependent freedom as a result of the move from Game 4 to Game 3. As we have seen, a necessary premise of that argument is that if Ami prevents Fred from getting the book in Game 4 merely as a result of her intention to do the same thing as Fred, then Ami is not restricting Fred's freedom. The prevention is merely an unintended side-effect and therefore does not constitute a constraint on Fred's freedom. However, on the weak interpretation of intentional prevention, even where Ami's dominant intention is to do the same thing as Fred, the latter will be unfree to get the book in Game 4. The move from Game 4 to Game 3 will then be analogous to the move from Game 6 to Game 5: in the former case, no less than in the latter, Fred will gain a particular freedom as a result of the suppression of his desire to obtain the book.

The failure of Dowding and van Hees to come up with a conception of freedom that successfully avoids indirect preference-dependence is symptomatic of a more general truth: as long as human agents interact strategically, indirect preference-dependence will constitute an inevitable feature of negative freedom, on any remotely plausible conception of such freedom. Pure negative-liberty theorists are not worried by this fact, for they content themselves with the avoidance of direct preference-dependence. Is their position tenable? Let us now turn to the question of its normative justification.

3. WHY INDIRECT PREFERENCE-DEPENDENCE IS UNOBJECTIONABLE

Dowding and van Hees implicitly hold that, if direct preference-dependence is damaging for a conception of negative freedom, then indirect preference-dependence is damaging too. They think that the normative arguments presented by negative-liberty theorists against conceptions that make freedoms and unfreedoms directly preference-dependent also tell against conceptions that make freedoms and unfreedoms indirectly preference-dependent. Of these normative arguments, the two that feature most prominently in their paper are Berlin's argument against the Stoic conception of freedom and Carter's argument grounding the definition of overall freedom on freedom's non-specific value.Footnote ² Let us consider each of those two arguments in turn.

One cannot deny the initial plausibility of Dowding's and van Hees's appeal to Berlin's argument against preference-dependence. If freedom is defined as the ability to do what one wants, Berlin says, “I need only contract or extinguish my wishes, and I am made free” (Berlin, Reference Berlin1969: 139). This implication is counterintuitive. Yet, as a result of indirect preference-dependence, a similar implication might appear to follow from the pure negative conception of freedoms and unfreedoms. For although we have seen that the presence of indirect preference-dependence does not imply that a change in my preferences is a sufficient condition for my gaining or losing any freedoms, the possibility of indirect preference-dependence does sometimes make it the case that the only step that has to be taken by me in order to gain or lose freedoms is to change my preferences. I myself “need only contract or extinguish my wishes” as a means to acquiring a particular freedom, even if the resultant acquisition depends additionally on others' behaving in certain ways. It is indeed an implication of indirect preference-dependence that, in certain circumstances, a change in my preferences is both (i) a necessary condition for gaining or losing some freedom and (ii) the only condition that depends on me.

However, given that Berlin's argument against preference-dependence is explicitly aimed only at the direct preference-dependence of freedoms that occurs under the Stoic conception, and given that direct preference-dependence is clearly distinguishable from indirect preference-dependence, one should not assume automatically that Berlin's normative worry also applies to indirect preference-dependence. In order to see whether Berlin's normative argument against the Stoic conception also affects indirect preference-dependence, one needs to examine the reasons Berlin gives for wanting to avoid preference-dependence. It is not enough to point out that, for Berlin, preference-dependence has “unwanted consequences” (Dowding and van Hees, Reference Dowding and van Hees2007: 155). One should look more closely at those consequences as they emerge in cases of direct preference-dependence, and then ask whether those same consequences emerge in cases of indirect preference-dependence.

What, then, are the “unwanted consequences” of preference-dependence in Berlin's view? As Berlin makes clear both in the main text of Two Concepts of Liberty and in his introduction to the second edition of that essay (1969), his central objection to the Stoic conception of freedom is that it confuses being free with feeling free. On the Stoic conception, “[i]nstead of resisting or removing the pressures that bear down upon me, I can ‘internalize’ them. This is what Epictetus achieves when he claims that he, a slave, is freer than his master”. What Berlin is objecting to here is the idea that a person can become free by “ignoring obstacles, forgetting, ‘rising above’ them, becoming unconscious of them” (Berlin, Reference Berlin1969: xxxviii). Teaching a person that “if he cannot get what he wants, he must learn to want only what he can get” may increase his welfare, but to say that it increases his freedom is to confuse the distinct goods of freedom and welfare. Most importantly, Berlin is worried that such a conflation may have damaging political consequences: the danger is of “confusion in theory and justification of oppression in practice” (1969: xxxix). Why should the theoretical confusion of ‘being free’ with ‘feeling free’ carry with it the possibility of justifying oppression? As Berlin says (and as Dowding and van Hees quote him as saying), if a tyrant manages to condition his subjects in such a way as to bring it about that they no longer desire to do any of the things that he is actually preventing them from doing, “he will, on this definition [namely, the definition of freedom as the ability to do what one wants], have succeeded in liberating them” (1969: 140). Through the conditioning of subjects’ desires – whether by those subjects themselves or by their oppressors – a state of subjugation may masquerade as one of freedom.

Consider now the difference between the ways in which a person may liberate herself by ridding herself of certain desires, first on the Stoic conception of freedom and then on the pure negative conception. The Stoic conception permits the following scenario: a person P is oppressed in the sense of being prevented by another person Q from performing a certain set of actions, but P can nevertheless liberate herself by ridding herself of the desires to perform those actions. What is counterintuitive about this idea of liberation – in Berlin's view and in that of most liberals – is that P remains, in her ‘liberated’ state, an oppressed individual (that is, prevented from performing the specified actions). The pure negative conception, on the other hand, permits the following scenario: P is oppressed in the same sense as above, but P can nevertheless liberate herself by ridding herself of the desires to perform the actions in question, owing to the fact that Q will cease preventing those actions if and only if P ceases to desire to perform them. This second case of liberation does not have the counterintuitive consequence noted above. It is not the sham liberation witnessed in the first scenario. In the second scenario the obstacles to P's actions really are lifted by Q, and the aspect of the first scenario to which Berlin objects is therefore not present. From Berlin's normative point of view – and from that of the pure negative-liberty theorists – what should concern us is not freedom as “the absence of frustration” but freedom as “the absence of obstacles to possible choices and activities – absence of obstructions on roads along which a man can decide to walk” (1969: xxxix). A diminution of such obstacles really does occur when a person gains a freedom in an indirectly preference-dependent sense, whereas it does not really occur when a person gains a freedom in a directly preference-dependent sense. Paradoxes like that of the “contented slave”, who achieves freedom while remaining a slave, simply cannot arise as a result of merely indirect preference-dependence.Footnote ³

Berlin's normative argument does not imply that there is something wrong in itself with saying that a person has gained some freedom by changing her preferences (an interpretation that presumably would reduce Berlin's point to a merely semantic one). What it implies is that there is something wrong with saying such a thing of a person who remains subject, no less than before, to “obstacles to possible choices and activities”, to “obstructions on roads along which a man can decide to walk”. A corollary of this last thought is that there is something wrong with holding a person to have gained some freedom by changing her preferences when nothing in the world apart from her preferences has changed. The conclusion to be drawn from Berlin's normative argument, then, is that direct preference-dependence is normatively unacceptable, while indirect preference-dependence is normatively unobjectionable.

Let us now consider the compatibility of indirect preference-dependence with Carter's normative claim that freedom has “non-specific” or “content-independent” value.Footnote ⁴ We should begin here by clarifying what is meant in saying that freedom has non-specific or content-independent value, for the account given by Dowding and van Hees is somewhat misleading. In the course of advancing a perceptive observation about the instrumental value of citizens’ freedoms of expression (2007: 150), Dowding and van Hees mistakenly suggest that the distinction between the value of one's freedom-to-Φ and the value of Φ-ing is tantamount to the distinction between the non-specific (or content-independent) value and the specific (or content-dependent) value of each of one's freedoms. As Kramer has argued elsewhere, however, those two distinctions are not equivalent (Kramer Reference Kramer2003: 240–5). In thinking about the values of freedoms, we need a tripartite classification rather than a dichotomy: (1) the non-specific (content-independent) value of any freedom-to-Φ; (2) the specific (content-dependent) value of any freedom-to-Φ; and (3) the value of Φ-ing.Footnote ⁵ Dowding and van Hees purport to be differentiating between the first and the third of these elements in their discussion of the freedom to express oneself, when in fact they are differentiating between the second and the third. They show that, because one's freedom of expression is the type of freedom that it is, it carries considerable value irrespective of whether any typical person attaches much value to the exercise of that freedom. Their discussion of freedom of expression does not genuinely address – and does not need to address – the non-specific value of freedoms of expression at all.

The way to capture the idea of freedom's non-specific value is not through a discussion of freedom of expression (or of any other specific freedom), but through a more abstract consideration of the mere fact of having freedom – that is, through a perception of the value of freedom as such. Any freedom's non-specific value prescinds from its content as the freedom to express this or that view, to travel to this or that destination, or to attend this or that meeting. Indeed, the non-specific value of freedom even prescinds from more general contents such as those of freedom of expression and freedom of movement and freedom of association. (Freedom possesses non-specific value if freedom itself is intrinsically valuable, but it may also be held to possess non-specific value for various instrumental or constitutive reasons [Carter, Reference Carter1999: 43-60]. For example, if freedom is good because it tends to produce economic or social or moral progress, and if our ignorance of the direction of progress itself is such that we are unaware which particular freedoms are useful for this purpose and which are not, then freedom will have non-specific instrumental value.)

In Carter's view, a person's overall freedom is a property on which freedom's non-specific value supervenes: ceteris paribus, the more freedom one has (in this overall sense), the greater the value of one's situation in terms of that particular value called “the non-specific value of freedom”.Footnote ⁶ Freedom's specific value supervenes instead on the presence of particular freedoms, understood as freedoms to do specific things or specific kinds of things. If specific value were the only kind of value that freedoms possessed, then a normative political theory could content itself with the prescription of certain specific freedoms or kinds of freedom for each person, and could dispense with the measurement of each person's overall freedom. The whole point of measuring a person's overall freedom, in Carter's view, is to gauge how much of the non-specific value of freedom there is in her situation.

Now, as Dowding and van Hees rightly point out, the property on which the non-specific value of freedom supervenes must itself be value-neutral in the following sense: we cannot say that the level of that property for any person is a function of how much she desires to be free to perform the specific actions she is free to perform. After all, to say so would be to construe that property as a function of the specific values of her particular freedoms. We would thereby be failing to single out the property on which freedom's non-specific value exclusively supervenes. Therefore, the measurement of that property for any person should make no reference to the specific values of her freedoms, and should instead refer only to the ‘sheer quantity’ of the combinations of options that are open to her. (We do not have the space here to expound the precise meaning of ‘sheer quantity’. What matters is that the measurement of the sheer quantity of freedom makes no reference to the values of particular actions or types of actions or to the values of the specific freedoms to perform those actions or types of actions, whether on the basis of the agent's preferences or of less subjective values.)

Clearly, the above line of reasoning rules out the direct preference-dependence of the property on which the non-specific value of freedom supervenes. Dowding and van Hees believe, however, that it also rules out indirect preference-dependence. On Carter's normative argument, the non-specific value of a person's set of freedoms is solely a function of the degree of the overall freedom that is constituted by that set. Yet, if indirect preference-dependence occurs, it will nevertheless be true that “the preferences of the individual are important in the measurement of that individual's [overall] freedom” (Dowding and van Hees, Reference Dowding and van Hees2007: 155). And this will mean that the measure of a person's overall freedom will fail to tell us the level of the non-specific value of her set of freedoms. Only a completely preference-independent measure of a person's freedom can perfectly capture freedom's non-specific value.

Like Berlin's argument, however, Carter's argument linking overall freedom and the non-specific value of freedom turns out on closer scrutiny to be compatible with the indirect preference-dependence of freedoms and incompatible only with their direct preference-dependence. As we have already seen, there is a relevant distinction to be drawn between (1) a putative change in a person's freedoms as a direct result of a change in her preferences even though nothing in the world apart from her preferences has altered; and (2) a change in a person's freedoms as a direct result of a change in the degree to which others subject her to obstacles, where that change in her subjection to obstacles is itself brought about by a change in her preferences. In the case of (1), there is not really any gain or loss of freedoms, whereas in the case of (2) there really is such a gain or loss. In (2), that gain or loss comes about partly because of a change in the person's own preferences. Nevertheless, the effect of the gain or loss on the sheer quantity of a person's available options remains a distinct phenomenon that is interesting in its own right. And the reason it is interesting in its own right is that freedom has non-specific value.

Now it is true that accepting the possibility of (2) entails accepting that the degree of a person's overall freedom is not wholly independent from the (subjective) specific values of her freedoms. As the example of Fred and Ami clearly shows, the latter variable can affect the former. However, the normative arguments presented by Carter in favour of the non-specific value of freedom neither assume nor imply that freedom's non-specific value varies independently of its specific value. Rather, what those arguments suggest is that the two types of value are distinct and important, and that liberals ought to be interested in measuring both of them. The distinctness of two value-variables does not entail that they are wholly independent variables. The non-independence of the two variables, then, does not undermine the claim that the property on which the non-specific value of Fred's set of freedoms supervenes is itself value-neutral.

If the empirical correlation between freedom's non-specific value and its specific value were tight enough, it might be suggested that one could content oneself with measuring only the latter. Would that degree of (indirect) preference-dependence not render the concept of overall freedom normatively redundant? It is extremely difficult to imagine such a tight correlation ever existing. However, even if it did, the answer to the question just posed is that the concept of overall freedom would remain just as important normatively as in the more realistic case of a merely spasmodic and rarely predictable dependence of the two variables. To be sure, the tight correlation would provide us with an alternative, indirect way of measuring the quantitative attribute on which the non-specific value of freedom supervenes. But that attribute, conceived in value-neutral terms, would continue to exist in the world, and freedom's non-specific value would continue to supervene on that attribute and only on that attribute.

Far from ruling out all cases of preference-dependence as normatively objectionable, then, the arguments of Berlin and Carter cited by Dowding and van Hees serve to ground a normatively significant distinction between direct and indirect preference-dependence. Direct preference-dependence is normatively unacceptable, whereas indirect preference-dependence is both unavoidable in practice and normatively unobjectionable.