Three-Dimensional Power: A Discussion of Steven Lukes’ Power: A Radical View



Lukes’ third dimension of power exists where people are subject to domination and acquiesce in that domination. The intentional stance allows us to predict and explain others’ behaviour in ways that those agents may not recognise. It denies agents’ privileged access to their own reasons for actions. Using the intentional stance we can understand how agents may acquiesce in their own domination. We can also make distinctions between those who dominate knowingly and those who dominate without realising they do so. It allows us to distinguish morally such cases and to understand the power structure without falling into the Foucaultian trap of seeing all social relationships in the same relativistic light and where all – dominant and dominated alike – are subject to the same power relations and moral responsibility.

Steven Lukes’Power: A Radical View was enormously influential for such a short book. As well as spawning a large debate among conceptual theorists it also led to a number of empirical studies attempting to measure the impact of power's third dimension on people's lives. Its re-issue with two new long essays is to be much welcomed.1 The first new essay locates Lukes’ original book in the context of the ‘community power debate’ and importantly distinguishes his third-dimensional view from Foucault's position. Here Lukes nicely demonstrates the thinking behind Foucault's writing and argues that studies which show that people willingly comply in patterns of normative control do not sustain the radical Foucault view that there can be no freedom since we are all constituted of power. His second essay –‘Three-Dimensional Power’– attempts to locate the sense of freedom or autonomy in relationship to the patterns of power that exist. It is upon this second essay that I will concentrate my attention.

I will suggest that we can defend Lukes’ third dimension of power by taking up the intentional stance (Dennett, 1987). Lukes wants to be able to identify and criticise values that lead dominated people to acquiesce and even celebrate their own domination. At the same time he does not want to fall into the Foucaultian trap where all social relationships are seen in the same relativistic light and where all – dominated and dominant alike – are subject to the same power of structural relations and so all subject to the same moral opprobrium. I argue that agents’ values might be naturally rationalised, even though the reasons agents have for acting might be neither conscious nor recognisable to them. I argue that this allows us to be critical of values, but only if we carefully demarcate the beliefs and desires that constitute value systems.2 The intentional stance allows us to interpret and efficiently predict the actions of people in ways that may or may not coincide with their own rationalisation of their behaviour. Both their and our rationalisation of that behaviour is subject to critique both in terms of the truth of the beliefs in it and in the character of the desires underlying it.

For Lukes the most insidious and important form of power is domination.3 His third dimension occurs not only where there is domination, but where the dominated acquiesce in their domination. Such acquiescence may happen in both a thick and a thin sense (Scott, 1990): the thick sense where people actively believe the values which oppress them and the thin where they are merely resigned to them. Scott endorses the thin view; Lukes is correct that both occur. Much of his second essay is about the acquiescence of the dominated: its forms and its causes. Here I will support Lukes against his critics, but argue that Lukes leaves too much unexplained in the black box of socialisation. I believe the collective action problem – which I argued can explain how people can be powerless on their own without action from the powerful (Dowding, 1991) – can also help illuminate part of the black box of acquiescence.

Taking on the values that justify domination is only one way to acquiesce. In the oft-quoted Nussbaum and Sen example, the Indian woman who gives up food and her health for her husband and male children, does so willingly (and she may do so resignedly or gladly). She is herself a willing part of the fabric of her own domination. Such claims lead Benton (1981), Clegg (1989, p. 95), (Hay, 1997) and others to suggest that Lukes is condescending towards people who are merely endorsing a set of values not shared by Lukes. Who is Steven Lukes to criticise the values of others? I think the answer is straightforward: Lukes, just like anyone else, can analyse and evaluate the situation of others. To suggest that people are always the best judge of their own interests and have privileged moral status over their own preferences is to deny any sort of normative social analysis. Nevertheless, the charge is one to be taken seriously and Lukes examines it carefully.

Several aspects of domination in the third dimension are important. First is the status of values, preferences, interests, beliefs and desires. Second, given their status, in what form can we criticise these objects? Third, there are queries over the dominant–dominated relationship. Is everyone who gains at the expense of others dominant? Is everyone who loses dominated? Fourth, there is the question of responsibility. Must the dominant need to know what they are doing or can their privilege be a by-product of forces they do not understand?

Lukes thinks his account of power – and therefore all accounts because of their juxtaposition to his – is essentially contested. He believes his account of power is essentially contested because it requires a notion of objective interests. I do not believe any account of power is essentially contested even if the application of any particular conception of power may be open to judgement and dispute. Before examining the claim of essential contestability towards the end of this essay let us consider the relationship between values, preferences, interests, beliefs and desires; and also how we might interpret ‘intentions’.

Boudon (1998) argues that intentional explanation is the first step in any social inquiry. If one can explain why people behave as they do by giving their reasons for action then we have a fully-fledged explanation of those actions. Here there is no ‘black box’ since the reasons we give to explain someone's behaviour must be explanatory for us (otherwise they would not constitute reasons). We might think the reasons that some people have for acting as they do are crazy. We might think they are based on false beliefs or perverted desires, but we understand the action when the agent's behaviour is rationalised. It is a mistake however to overly privilege the agent's own rationalisation. We do not have to accept only the reasons constructed by the individual herself as the most efficient predictive and explanatory reasons. The best intentional explanation of someone's behaviour may not be those reasons offered by the individual herself. She may deny having the reasons we impute to her. I may claim, quite honestly, that I am acting for your best interests, but a careful examination of all my actions may show that I only act so, when it suits my own interests too. Aggregate-data analysis can demonstrate why people act in various ways, even though those people would not specify so precisely those reasons themselves. Intentions do not have to find conscious expression in the mind of those to whom the intentional explanation of their actions is applied (Dennett, 1987).

Given someone's desires, their false beliefs may lead them to act in ways which do not further those desires. We can say here, because of those false beliefs, that the actions they think are in their interests are not objectively so (Dowding, 1991, pp. 30–46). We can also go further and suggest that people might find themselves in a situation where they know the actions they take will not lead them to attain what they most desire, but are forced by circumstances to make the best of a bad job. A one-shot Prisoners’ Dilemma is such a situation. We might say here that it is not in their interests to be in a Prisoners’ Dilemma game, especially if there is a feasible world in which they do not have to face the game (Dowding, 1991, pp. 39–41). So we have two ways of explaining objective interests, one based on false belief, and one based upon the (contingent) structure of the situation someone faces. However, Lukes wants a further notion of objective interest than those suggested here. He not only wants to look at the situation of people and their beliefs: he also wants to criticise their desires.

I use the term ‘desires’ advisably. In fact, Lukes wants to criticise ‘values’, but I take someone's values to be a complex mixture of beliefs and what I am here calling desires. If the criticism of values can be broken down into the component parts of beliefs and desires, and it is the beliefs that are criticised, then I see criticism of values to be no more controversial than claiming that false beliefs can lead people astray, given their desires. Take the Indian woman sacrificing her health for her husband. I take it that our judgement on that action – whether or not we consider it to be acquiescence in dominance – would depend upon her reasons for so doing. If her desire was to do the best for her family, and her family income – the food in everyone's belly – depended entirely upon her husband's ability to work in the field, then giving priority to her husband's nutrition may well be rational.4 If it were the case that should he die, she (and her children) would starve anyway – then it is perfectly reasonable for her to feed him first. In the Nussbaum and Sen story that is not the case. Indeed, in parts of modern India, women often take on primary employment responsibilities as well as primary childcare responsibilities. If women still prioritise their husband's nutrition (as evidence suggest many do) then this reasoned explanation cannot hold. What reasons might a woman have? One that is often suggested is that such behaviour ‘is part of our culture’ or ‘that is what is expected of us’. One way of considering this matter is to suggest that the expected cultural behaviour was once rational. It was once the optimal equilibrium of a survival game. Now, however, the equilibrium strategy is only so because it is still played: it is no longer optimal. The cultural hold-over can be criticised because it is no longer reasonable.5

Note that in this story we justify her behaviour even if the reasons she gives are not the ones underlying our explanation of the equilibrium behaviour. We might find, for example, that in societies where families do depend entirely on male income women do indeed feed their husbands first. However, their reasons for doing so have nothing to do with the family survival story. The reasons they give may be the same whether or not the actions are socially optimal – in either case the women may say ‘it is the way we do things around here’, ‘it is our custom’ and so on. As I have specified, where the behaviour can be externally rationalised in the current environment we still have an intentional explanation – even if those intentions are not recognised as such by the actors engaging in them. This means that the distinction Jon Elster (1983) makes between intentional and functional explanation cannot be so sharply drawn. Here the behaviour is functional and intentional, though not consciously so, and in this example we can easily hypothesise the feedback loop that Elster requires of good functional explanation.6 Where the behaviour is a cultural hold-over, intentional explanation remains in the reason offered by those engaging in the behaviour, although it is no longer functional. We still rationalise the behaviour as a cultural hold-over in terms of the external intentional explanation.

I think it is via these means that we can answer Elster's critique of Lukes’ claim that A may exercise power over B by ‘influencing, shaping or determining his very wants’ (p. 27; compare with pp. 134–7). Elster asks whether this is a purported purposive or functional explanation. If it is intentional, then we should be able to find evidence that A is manipulating B in this manner. Elster further suggests that it is unrealistic to suppose that leaders deliberately try to induce beliefs and desires in their subjects. I cannot think why. Perhaps Elster inhabits a different planet to me, and in this context makes the extraordinary claim that mental or social states can never be brought about intentionally ‘because the very attempt to do so precludes the state one is trying to bring about’. This may be true for self-referential attempts; certainly not for A attempting to do this for B.7 Lukes is surely correct that ‘Power can be at work, inducing compliance by influencing desires and beliefs without being “intelligent and intentional” ’ (p. 136). Not consciously intentional, but subject to intentional explanation in Dennett's sense. And explanation may be functional too. After all, all that is really required is that a person continues a pattern of behaviour that brings about desired results. I need not realise that I am wheedling or even that by wheedling I am getting what I want. All that is required is that I do x and results occur which lead me to do x again.8

It seems to me that this is the way to defend Lukes against Elsterian criticism. It is not quite Lukes’ own, I think, though I am not sure. One reason for my doubt is Lukes’ style of writing, in which he tends to ask a lot of questions and then leaves the answers open or somewhat cryptic. Lukes clearly wants to tie non-domination to some form of rational behaviour – though the rationality here is never specified beyond some Spinoza hand-waving about authenticity and freedom. But the reference to Spinoza and freedom does not really address the problem we have with the formation of preferences. Autonomy is easy to define. We can say that a person is autonomous to the extent that she acts through her own self-will; she is not controlled by anyone or anything. When we give an intentional explanation of an action that is one element of autonomy – but of course we may autonomously act in the face of incentives set up by others. We can still be autonomous even when our freedom is strictly contained. Similarly, the beliefs and desires that constitute our reasons for action may also be developed because of the incentives set by others. Here, however, we might baulk at claiming we are autonomous no matter how our preferences are formed by others’ actions. I think again that intentionality must enter our account of how to distinguish autonomous and non-autonomous ways in which our preferences are formed.

All our beliefs, values and actions are caused by aspects of the world outside of us. This does not make us less autonomous. If any and every influence outside of a person were thought to be autonomy reducing, then there could be no autonomy since everyone is constantly assailed by influences from the world around them. Someone uninfluenced by the world around them – someone living in some type of vacuum – would not be more autonomous, since she would not have been able to develop beliefs and values that those affected by other influences would have. Indeed, what sorts of beliefs, desires or values could she have? In other words, the random and systematic influences of the world around us cannot reduce our autonomy, such as it is. Our autonomy can only be reduced by influences that have some other character. I suggest we reintroduce intentions. If we are systematically affected by aspects of the world that are intended by others to affect us in some ways, then these influences may be autonomy reducing, especially if we are unaware of these intended influences. Furthermore, these influences reduce autonomy even if those systematic influences are not consciously intended by those who affect us.9

If we do not introduce intention then the account of power as domination will slide into Foucault's hole. After all it is clear from numerous social choice results that rules and procedures are not neutral selectors of outcomes and that the benefits and costs will not be allocated equally (Arrow, 1963[1951]; Baron and Ferejohn 1989; Gibbard, 1973; McKelvey and Schofield, 1986; Riker, 1982; Satterthwaite, 1975; Shepsle, 1979). Furthermore, since people's interests are at least partly exogenously defined by their place in society, the interests they will defend must at least partially be constructed by the institutions through which they lose or gain. Of course, given that rules and procedures – and institutions can be read as ways of behaviour here too – are not neutral will almost certainly mean that the beneficiaries will try to defend them. But what if their defence is in terms of values that have no direct relationship to their benefit? What if the beneficiaries buy into the ideology too?

On the one hand, if we say that whoever benefits is dominant – that is their power – then it is hard not to stop the slide into the Foucaultian trap. ‘Social choice proofs’ + ‘a normative claim that gainers are necessarily dominators equally subject to the forces of the institutions which create them’=‘Foucault on power’. A surprising result for Foucaultians and rational choicers alike perhaps, but pretty straightforward. But why should we accept the claim that simply because someone benefits they dominate others (Polsby, 1980)? The Indian husband might honestly say, ‘I try to get my wife to eat more, but she says she is full. I know she is lying but what can I do?’. Must we reply that his concerns are irrelevant, he dominates her and that is all there is to it? Is it really my fault if I cannot persuade someone that their behaviour is not in their best interests and they should stop acting simply for me?

I believe that intentions must come into this story. At the first pass we should say that if the beneficiary acts in ignorance of the effects of his actions, then we should not claim that he is dominating anyone. If he is not consciously intending the result he is not to blame. At the second pass, we should say that if he is ignorant but should not be, then he is dominating. He may not be conscious of his dominance – but he should be. And as we learn more about how economic and social life works then responsibility grows. At the third pass we should point out that once we realise how our institutions affect the interests of ourselves and others, then anyone who does not act to change those institutions for the better is part of the structure of domination. However, in the third pass we should not claim that those so implicated in the structure of domination are necessarily themselves dominators. After all, the class of such people includes those who are dominated. If we define dominance too readily at the third pass we fall into the Foucault trap.10

The Foucault trap awaits other accounts of domination too. For Pettit (1997) and other republicans a person is free to the extent that they are not subject to the arbitrary will of others. A slave may be allowed all the freedoms of a free man, but what he is allowed can be taken away at the whim of his master. This is both broader and narrower than some standard accounts of negative freedom. Negative accounts that claim that a person is free to do what he can do consider a slave allowed the liberties of a free person to be as free as a free person. Republican freedom is narrower in that case. But, it is broader in that it does not consider a person to be unfree when she is constrained by the non-arbitrary common will. But republican freedom – and the account of domination it relies upon – suffers from the same problem that assails the broader side of negative freedom. One negative view of freedom – mentioned by Lukes on p. 114 – says a person is free to do x only to the extent that no one can stop that person from doing x. But, virtually everything we can do, we could be stopped from doing by someone or some group. Thus, freedom does not exist or is ‘vanishingly small’ (Dowding and van Hees, 2003, p. 287). We could all be stopped from doing what we do by the arbitrary will of others. After all, each of us could be murdered. Thus, we are all subject to the arbitrary will of others or collections of others. So, the republican notion of dominance also falls into Foucault's trap. The fact that we could each oppress others does not mean that we do oppress them. The fact that we are all implicated in the structure of domination by not attempting to change wrongful institutions does not itself make us all dominant.

Being implicated in the structure of power relations –‘the system’– does not make us all dominators. But, we still have to query why we do not do more to break free from the modes of domination that we are subject to. Why do the oppressed, at times, go along with the institutions that oppress them? The answer is simple: because of collective action problems.

Consider how collective action can lead to types of domination. Lukes discusses an example derived from Bourdieu (2001, p. 35): ‘the vision that many women have of their bodies not conforming to the aesthetic canon imposed by fashion, and, more generally, in their adherence to a demeaning image of women’. Among the many other things that women do, they find themselves in competition with each other for men. And, indeed, their genes may develop their body in certain ways precisely to capture the attention of men when they are first ready to bear children (Ridley, 1993). It may not be in the collective interest of women to take on whatever fashionable shape they have to adopt in order to keep ahead or abreast of the crowd. They might all prefer not to be in the competition, but once the competition is there they either play the game or drop out. There might be domination in there somewhere. Men do tend to dominate women. Men are larger, they have louder voices, are more violent in deed and voice, they earn more and so on. But these facts about domination within households do not mean that it is men who consciously determine the ideal female form. Indeed, these days most of the cheering comes from women's magazines largely written and edited by women. It is the collective action problem again.11

We should note here that the second pass includes a moral judgement. We say someone is dominant if they act intentionally but non-consciously in ways which oppress others, but which they should realise does so. We cannot deny that such judgements are controversial and subject to dispute. This is precisely why Lukes thinks any account of power that includes an account of domination at its heart is essentially contested. I deny this. Why? Our views about how we judge a given issue might be contested, but not the conception of power as laid out here. That is not of course to say that everyone must agree that such a Lukesian account of power as domination must be accepted by all. Of course it will be contested, but not essentially so.12 How we judge what blame is to be laid against people for the way the world works is one thing. Whether we can see how a Lukesian account of domination can be empirically applied to those cases is another. In my view Lukes needs to restrict his account of domination and power to the second pass, and leave the third to Foucault. The third dimension of power is only power when our belief structures are intentionally caused by others but we must understand that such intentionality has the broader externalist interpretation that I have placed upon it and not the internalist conception implied by Lukes’ squashy liberal critics.

My defence of Lukes’ third dimension of power as domination relies heavily upon taking up the intentional stance. The intentional stance requires an externalist account of the human mind. The intentional stance is to invoke reasons for action although they need not be reasons invoked by the actor herself. Furthermore, there is no overriding reason to privilege any reasons she does invoke. Nevertheless, when using deeply normative notions such as domination we need to be careful about negatively judging people's actions. To dominate unconsciously is surely not as bad as to dominate in full consciousness of what one is doing. And, to be implicated in the structure of domination is less bad still although surely we all ought to take some responsibility for those structures – even though our lack of forthright action is due to subtle collective action problems that are hard to solve.

About the Author

Keith Dowding, Department of Government, London School of Economics and Political Science, Houghton Street, London WC2A 2AE; email:


  • 1

    Lukes (2005). All page references for unattributed quotations in this article are to this new edition of Lukes’ book.

  • 2

    The belief–desire distinction I use is not quite the familiar Humean one as I make clear below. I see desire as a form of belief and a distinction between beliefs and desires is made only for heuristic reasons.

  • 3

    In that sense his account of power is a sub-set of many others such as Dowding (1991; 1996).

  • 4

    One has to work hard to critique the desire to ‘do the best for her family’: such a desire is basic to the explanation of why she behaves as she does, but such a desire is itself a complex of beliefs and further desires. So, in any explanation I see a hierarchy. Values are based on beliefs and desires which rationalise them and actions. But, desires are themselves a complex of beliefs and more basic desires. To chase desires all the way down I believe you will eventually reach the biologists ‘four “Fs”: food, flight, fight and sex’.

  • 5

    I would suggest that such examples are the only cases where ‘cultural explanations’ are valid – and even here the behaviour can be rationalised from previous environments.

  • 6

    The social advantage given to males or females in various ways can be mapped for social environments both across societies and social classes within those societies in terms of self- or gene-interest; for example, in feudal times inheritance was the norm for sons among the Lords; inheritance for daughters among the peasants – for the Lords to keep the fortune in family; for the peasant to marry into higher social classes (Betzig, 1992). Although, basic and unexplained here is why it was acceptable for females to marry socially upward but not for males. There might be a dominance story here – one seen often in the animal kingdom. Functional intentional stories are easy to find.

  • 7

    Since the mind is now known to be a complex of competing interests there is no longer any reason for thinking Elster is entirely correct for self-referential attempts either.

  • 8

    Such algorithms can be used to explain various patterns of behaviour, even voting (Bendor et al., 2003), though it cannot always be rationalised as intentional (Dowding, 2005, p. 11).

  • 9

    I do not have the room here to develop this account fully. But what I have in mind are systematic attempts to influence us that try to hide any evidence which would be inconsistent with beliefs, desires and values that the agent is trying to induce in us. Tracking truths about the universe would not normally be autonomy reducing, since the attempt to track such truths generally requires openness about potentially contrary evidence.

  • 10

    In other words, not doing anything to stop those who dominate does not make us dominant, merely morally implicated. I am not a bully if I watch someone else bullying, but I bear some moral responsibility for the results of that bullying.

  • 11

    For a stunning example of how a simple collective action problem can create domination and massive welfare problems see Mackie (1996). What is also stunning is how a simple compact can all but solve them.

  • 12

    That is, many views about how the world is might be contested, but it does not follow that these views have to be so contested. There is no necessary contestation (Dowding, 1991, pp. 167–73).