Violence and the materiality of power

ABSTRACT The issue of political violence is mostly absent from current debates about power. Many conceptions of power treat violence as wholly distinct from or even antithetical to power, or see it as a mere instrument whose effects are obvious and not in need of political analysis. In this paper, I explore what kind of ontology of power is necessary to properly take account of the various roles that violence can play in creating and maintaining power structures. I pursue this question by contrasting the views of Hannah Arendt and Michel Foucault. For Arendt, power is generated and maintained by communicative practices. She argues that power and violence are ‘opposites’ because violence can only destroy but not create these practices. In contrast, Foucault’s conception explicitly allows violence to play a constitutive role in generating power. I argue that while Arendt is right to insist that power and violence are not identical, it does not follow that violence cannot play any role in constituting power. Guided by Foucault’s approach, I formulate a non-dualist account of the relationship between power and violence that takes seriously the role that bodies, material things, and built infrastructures play in making social relations ‘more durable’ and constituting power.


Introduction
What are the political effects of violence in democratic societies? In the United States, police killings of unarmed Black men and women have recently reignited discussions about police brutality and state violence more broadly (see Cherry, 2017). Public debates about sexual violence and rape culture attest to the pervasiveness of informal violence. But the political implications of such forms of violence are not a central topic in contemporary political philosophy and theory. 1 Violence is routinely depoliticized: It is treated as a brute, pre-political phenomenon which is transformed in democratic politics, or as a mere instrument whose effects are obvious and in no need of political analysis (see Winter, 2018, pp. 1-7). Discussions of violence focus on terrorism, civil war, rebellion, tyranny, genocide, etc., which are taken to be outside of democratic politics or, in some cases, the political sphere altogether. This depoliticization of violence is also reflected in debates about power. Many conceptions of power treat violence as distinct from and sometimes even antithetical to power. Others concede that violence is an obvious instrument of power, at least as a last resort, but consider questions about its effects trivial. In any case, the issue of political violence is mostly absent from recent academic discussions about the concept of power (see for example Allen, 1999;Forst, 2015;Hayward, 2000;Lukes, 2005;Searle, 2010). 2 In this paper, I explore what kind of ontology of power is necessary to understand the various roles that violence can play in creating and maintaining social orders. How do power and violence fit into our understanding of the social world? Our social ontology is often the starting point for normative inquiries; it shapes what kind of normative questions we consider significant. For example, Charles Mills (2009) argues that John Rawls's understanding of society as a 'cooperative venture for mutual advantage' is not an illuminating starting point for normative inquiry in our societies, which are built on White racial domination and violence. It foregrounds questions about fair cooperation while giving less priority to normative questions about historic injustices, systematic violations of personhood, and the entrenched material structures responsible for social inequality. An analogous concern arises for views that leave little room for violence to play a complex role in the creation and maintenance of social order. How then do we formulate an ontology of power that is able to recognize the various political roles of violence? I pursue this question by contrasting two influential conceptions of power. Hannah Arendt's conception starts from the idea that power is generated and maintained by communicative practices. She argues that power and violence are 'opposites' because violence can only destroy but not create these practices. In contrast, Michel Foucault's conception allows violence to play a constitutive role in generating power. In the first chapter of Discipline and Punish, for example, he analyzes the role of public torture in constituting a concrete social order. This profound difference in their views is often overlooked, perhaps because both authors emphatically distinguish power and violence. But Arendt excludes from the generation of power not just violence, but indeed most material human activities and their products. While these may be prerequisites for power to emerge, she claims that they play no constitutive role. Her view makes it difficult, for instance, to account for the ongoing effects of racist violence on the power structures of modern democratic communities. Foucault, in contrast, rejects this dualist approach and emphasizes that power is constituted by a set of heterogeneous relations, including but by no means limited to violence.
A precise diagnosis of this disagreement will equip us with the resources to formulate a non-dualist account of the relationship between power and violence. This position acknowledges that we describe phenomena in different terms when we speak of violence and power, respectively: Violence talk generally focuses on the physical effects of actions on bodies and things, while power talk usually describes how actions affect the (potential) actions of other agents. 3 The position also recognizes that power and violence are not identical: Instances of violence are not necessarily exercises of power, and violence by itself is never sufficient to generate it. Nor do all forms of power necessarily require violence for their generation, maintenance, and/or exercise. However, violence can play a constitutive role in the production and maintenance of power, though its specific role depends on the form of violence at issue and the context in which it is used. More generally, this position takes seriously the role that material things, built infrastructures, and bodies play in making social relations 'more durable' and thereby creating structures of power.
The paper proceeds as follow: I start by outlining Arendt's view of power and violence (section 1). I explicate the view's dualist assumptions and show that it cannot account for the constructive role that violence can play in shaping power structures (section 2). I then argue that while Arendt is right to insist that power and violence are not identical, this does not support her dualist view. We can make a clear distinction between power and violence without denying that violence can contribute to the creation and maintenance of power (section 3). In the second part of the paper, I elaborate a nondualist account of power. Drawing on an argument by Michel Callon and Bruno Latour, I first clarify the role that material things and built structures play in constituting power (section 4). I then apply the argument to violent practices, using Foucault's analysis of public torture as an example (section 5). A brief discussion of disciplinary practices further applies the argument to non-violent forms of power and shows that it does not collapse power and violence (section 6).

Arendt's oppositional view of violence and power
In her essay On Violence, Arendt summarizes her account of the relationship between violence and power as follows: To sum up: politically speaking, it is insufficient to say that power and violence are not the same. Power and violence are opposites; where the one rules absolutely, the other is absent. Violence appears where power is in jeopardy, but left to its own course it ends in power's disappearance. This implies that it is not correct to think of the opposite of violence as nonviolence; to speak of nonviolent power is actually redundant. Violence can destroy power; it is utterly incapable of creating it. (1970, p. 56) As this summary makes clear, Arendt is not merely claiming that the terms 'violence' and 'power' mean different things or that their referents are not identical. She is making the stronger claim that power and violence are, in virtue of their nature, incompatible or at least in tension with one another. Power is essentially nonviolent since violence is by its very nature incapable of creating power.
Arendt characterizes power as 'the human ability not just to act but to act in concert ' (1970, p. 44). This ability emerges when agents cooperatively constitute a shared form of life through rules and institutions. While speech and action are not the same thing, collective action relies on speech: Those who engage in collective action have to communicate with one another in order to discuss their ideas and problems, persuade one another, jointly make plans, agree on the norms that govern their joint activities, and interpret those norms. The ability to act in concert is based on communicative relations between those who deliberate and act together (Arendt, 1970, p. 178; see also Habermas, 1977;Young, 2002, p. 266). Of course, exercises of power need not be purely verbal. For instance, Arendt considers the American War of Independence to be a forceful exercise of power with serious material consequences. 4 Indeed, the use of violence can be justified in limited circumstances, e.g. to protect a community from external threats (1970, pp. 51f.). But the power to collectively use violence is only 'kept in existence by the [. . .] means of promise and covenant ' (1990, p. 176). Power is generated communicatively and any continued exercise of power also relies on communicative relations. In Arendt's words, power can be 'actualized only where word and deed have not parted company ' (1958, p. 200).
Arendt does not provide a straightforward definition of violence in her essay, but her examples suggest that she is primarily concerned with actions that cause pain to, injure, or kill others, or that damage or destroy things that are significant to humans (1970, pp. 46-51; see also Young, 2002, p. 263). For Arendt, an essential characteristic of violence is its instrumental character. Violence relies on instruments that are 'designed and used for the purpose of multiplying natural strength ' (1970, p. 46). This also means that, at least in principle, violence does not require acting in concert. Moreover, violence is instrumental in the sense that it is always used as a means to an end. Its use is governed by means-end rationality and is justified, when it is, by the immediate end to which it is used (1970, p. 46). Since the use of tools is governed by means-end rationality, understanding violence as relying on the use of instruments implies its instrumentality in this second sense.
In contrast to the instrumental nature of violence, Arendt claims that power is 'an end in itself,' pursued for its own sake and not as a means to other ends (1970, p. 51). She concedes that governments employ power to pursue policies, so the exercise of power can involve means-end reasoning. But Arendt's focus is the 'power structure' that generates power, which she claims 'precedes and outlasts all aims' (p. 51). Power cannot be exercised without collectively deliberating about the ends for which it is to be used. This form of communication is a condition for the possibility of using power and in that sense, power is prior to any specific end for which it may be used (see Frazer & Hutchings, 2008, p. 101). The use of violence, on the other hand, does not require deliberation about ends. It takes ends as given and only requires determining whether the instrument is an effective means to an end.
In summary, Arendt appeals to two major considerations to support her claim that violence and power are 'opposites.' The first one is the contrast between power's communicative and violence's merely physical nature. Physical violence can destroy the communicative relationships that power is based on by preventing people from communicating and acting jointly. But since violence is 'incapable of speech ' (1990, p. 19), it cannot generate new communicative relations. Sometimes, violence can be used 'as a last resort to keep the power structure intact against individual challengers ' (1970, p. 47), or to destroy violent regimes that prevent such relationships from emerging. In other words, violence can sometime negatively or indirectly facilitate the emergence of power (see Finlay, 2009, p. 38). But it cannot itself create the communicative space in which agents deliberate and act together, i.e., it cannot generate power.
The second consideration is based on the claim that power and violence are, in virtue of their nature, each governed by different forms of practical reasoning. Arendt believes that the instrumental reasoning that governs the use of violence is likely to overwhelm the collective practice of deliberating about ends: 'Where violence is no longer backed and restrained by power, the well-known reversal in reckoning with means and ends has taken place. The means, the means of destruction, now determine the end-with the consequence that the end will be the destruction of all power ' (1970, p. 54). The use of violence can destroy power not just by making communication physically impossible, but by undermining the non-instrumental character of political speech and action. Indeed, as I will elaborate in the next section, Arendt's worry about the effect of instrumental reasoning on deliberative processes has implications for all human activities that are involved in fabricating durable material objects-activities which she characterizes as both instrumental and violent in nature.
While Arendt's oppositional view of power and violence may seem idiosyncratic (Bernstein, 2012;Lukes, 1974, p. 59), views that characterize power as wholly distinct from violence are not uncommon. Consider, for example, John Searle's collective acceptance view of power. Searle argues that political power should be understood as 'deontic power,' i.e. as a matter of rights, duties, obligations, authorizations, and permissions. Such deontic powers are social statuses that are instituted by collectively accepted status declarations (Searle, 2010, p. 164). So just like Arendt, Searle takes political power to be based primarily on speech; it is, he says, 'normally exercised through the performance of speech acts' (p. 151). Searle concedes that the threat of violence may be a background condition for collective acceptance, but this does mean that power is based on violence: '[. . .] collective recognition or acceptance, though typically not itself based on violence, can continue to function only if there is a permanent threat of violence in the form of the military and the police' (p. 163). In other words, violence or its threat can indirectly facilitate the communicative processes that are the basis of power, but it does not generate power. Indeed, since the 'police forces and armies are also systems of deontologies' (p. 88), organized violence depends on power, rather than the other way around.
Similar understandings of political power can arguably be found in contemporary liberal contract theories and theories of deliberative democracy, all of which try to 'expunge [violence] from the realm of politics' (Frazer & Hutchings, 2008, p. 92). According to such views, the distribution and exercise of power is, or at least ought to be, governed by deliberative processes; violence is limited to punishment and defense. Even within the so-called 'power debate' in political science, little attention has been paid to the role that violence plays in constituting power (see for example Lukes, 2005). Arendt's oppositional view is a paradigmatic member in a family of accounts of power that pay little attention to violence. Arendt articulates and defends the opposition between physical violence and power most explicitly. Thus, it is worth discussing the motivations for and implications of her view in more depth.

A dualist conception of action and power
Arendt argues that power and violence are opposites. While they 'usually appear together ' (1970, p. 52), violence is radically different in kind from power. In this section, I will make explicit the dualist character of this view. To do that, we need to situate Arendt's account of power in her general social theory. In The Human Condition, she argues that power does not depend on material mediation: While strength is the natural quality of an individual seen in isolation, power springs up between men when they act together and vanishes the moment they disperse. Because of this peculiarity, which power shares with all potentialities that can only be actualized but never fully materialized, power is to an astonishing degree independent of material factors, either of numbers or means. (1958, p. 200, emphasis added) This lack of material mediation is a general feature of action, which Arendt claims is 'the only activity that goes on directly between men without the intermediary of things or matter [. . .]' (1958, p. 7, emphasis added; see also p. 182). We will need to consider her account of action in order to fully understand the implications of her view. It makes it impossible, I argue, to understand the ways in which violent practices positively shape how people act in concert.
For Arendt (1958), action is distinct from the production of durable things, which she calls work, and the material activities necessary to sustain life, which she calls labor (see Canovan, 1992, pp. 122-141, for a helpful overview). Labor is governed by biological necessity and thus essentially unfree; when concerned with mere survival, humans do not set their own ends. Work produces a durable material world that helps humans to partly transcend the repetitive nature of labor. This world includes artificial things, including tools and weapons, as well as built structures such as houses, bridges, marketplaces, railroads, etc. Productive activities are instrumental in nature and inherently involve physical violence, which is required to transform nature into useful objects. Both labor and work can be done, at least in principle, by solitary individuals because they do not require the recognition of others. Action, in contrast, is essentially directed toward other persons. In action, individuals disclose their uniqueness to one another, which requires the presence of others who acknowledge and judge the action, or what Arendt calls a 'space of appearance.' Action and speech are closely related because speech is necessary to articulate the meaning of acts (see Canovan, 1992, p. 131). Since they are essentially directed toward other individuals and require their uptake, they go on 'directly between' people.
Arendt acknowledges that the ability to engage in action depends on work and labor, but she construes this dependence very narrowly. In order for citizens to come together to deliberate and act freely, life's necessities must already be taken care of. Arendt points out that throughout most of human history this liberation of some from life's necessities required the exploitation and violent domination of many others (see for example 1990, p. 114). Similarly, she recognizes that a durable world created by work is a necessary prerequisite for agents to engage in action; action can only appear in concrete and limited spaces (1990, p. 275). The ability to act in concert would not emerge without the material activities of work and labor. But Arendt carefully restricts this claim: 'The only indispensable material factor in the generation of power is the living together of people ' (1958, p. 201). Power depends on work and labor only in the sense that they create the preconditions for speech and action.
Despite this dependence, Arendt insists that action occurs in a distinct sphere with its own, distinctive character. While the material spaces created by work have a durability that outlasts the activities and intentions of their producers, the space of appearance 'does not survive the actuality of the movement which brought it into being ' (1958, p. 199). It comes into being when people 'are together in the manner of speech and action ' (p. 199) and it disappears when those activities cease. This being-together is disconnected from work and labor. Arendt insists that even public economic activities, such as exchanging the products of work in a marketplace, have the character of work, not action (p. 209). Arendt explicitly contrasts the character of the space of appearance with that of material spaces: [. . .] the physical, worldly in-between [...] is overlaid, and as it were, overgrown with an altogether different in-between which consists of deeds and words and owes its origin exclusively to men's acting and speaking directly to one another. This second, subjective in-between is not tangible, since there are no tangible objects into which it could solidify; the process of acting and speaking can leave behind no such results and ends products. (1958, pp. 182f., emphasis added) While the space of appearance is worldly (because it goes on between people), it is at the same time intangible and fleeting. Despite its dependence on labor and work, it is sui generis: it owes its emergence only to speech and action, which themselves only emerge from within this space (see Passerin d'Entrèves, 1989, p. 326).
As we can see now, Arendt draws the distinction between action and the material activities of labor and work, and by implication between power and violence, in a dualist manner. A dualism is a distinction that renders crucial relations between the distinguished items unintelligible (Brandom, 2009, p. 98). If action and power are sui generis phenomena, it makes no sense to ask how specific violent practices or built structures shape how people act in concert. Labor and work create the preconditions for people to be together in action and speech, but they have no bearing on how people relate to one another as they act and speak. It is in this sense that power is 'to an astonishing degree independent of material factors.' Let me elaborate this interpretation a little further. Arendt specifies that power is largely independent of 'either means or numbers.' She has the latter in mind when she points out that '[a] comparatively small but well-organized group of men can rule almost indefinitely over large and populous empires' (1958, p. 200). This reflects her belief that power is not an aggregation of natural strength but a matter of social organization. Larger groups are harder to organize, although if they do act in concert, they 'may engender an almost irresistible power [. . .] in the face of materially vastly superior forces ' (1958, pp. 200-201). This last point implies that a group's power is not fully determined by its access to material means such as wealth or weaponry. But while that is correct, it is much weaker than Arendt's claim that power is largely independent of material means. Consider again Arendt's first point: To rule over a large empire, a small group would need a robust material infrastructure in order to establish long-distance control. John Law (1984), for example, has argued that the 15th-century Portuguese imperial expansion became possible in part due to the availability of the right kinds of ships and navigational instruments, which together with appropriately trained people allowed for fast and reliable communication half way across the globe. Law would agree that Portuguese power was not fully determined by these material means, but he nonetheless emphasizes that they were crucial in creating the communicative relations on which it was based. By emphasizing the sui generis character of action and power, Arendt would have to deny Law's point.
How useful is a view that insists on power's sui generis character? One important motivation for Arendt's self-consciously dualist account is her concern with the possibility of creating genuinely non-violent political spaces (Finlay, 2009, p. 39). Arendt observes that revolutions, even as they set out to abolish domination, often reproduce violent forms of rule. She is worried that the use of violence even for laudable political goals, unless it is carefully circumscribed, will lead to the replacement of free deliberation about ends with mere means-end reasoning. Once politics becomes only about determining the most effective means to given ends, the instrumental effectiveness of violence will make it the focus of political action, leading into a '"fateful" cycle of violence' (Finlay, 2009, p. 4). For this reason, Arendt also notoriously rejects politicizing the social issue of poverty, which for her is merely an administrative issue of determining the best means to satisfy basic human needs (see Canovan, 1992, pp. 230-233;McGowan, 1998, pp. 45-52). I cannot do justice to Arendt's discussion of these questions here, but I mention them because they suggest that we could read Arendt's discussion of power and violence as an 'ameliorative' form of conceptual analysis. Such an analysis clarifies a concept in a way that best serves legitimate purposes, even if that clashes with everyday intuitions about the use of the concept (Haslanger, 2012). If we think of violence as a form of power, we are more likely to construe political action in instrumental terms. To avoid the resulting cycles of violence,we need to clearly delineate power from violence and means-end reasoning.
But while Arendt's concern about the hard-to-control dynamics of political violence is important, her dualist conception of power sidelines investigations into the roles that violence and the built environment can play in constituting the power structures of political communities. Consider, for example, the timely question of how policing and the built infrastructure in the United States help to perpetuate racialized social hierarchies (see, for example, Capers, 2009;Rolnick, 2019;Schindler, 2015). Because Arendt insists that power emerges from a distinct sphere of non-violent and noninstrumental communal practice, her account has little to say about this connection. She acknowledges that racial violence, in the form of slavery, established the conditions in which the political community was founded and recognizes the ongoing discrimination in housing, education, and employment. But on her account, these 'social' issues are pre-political and have no direct bearing on how people will engage with one another in collective action (once they do). For example, while Arendt rejected the legal enforcement of segregation, she held that people had a right to discriminate throughout the social sphere. Critics such as Robert Bernasconi conclude that Arendt's 'account of political community [. . .]) lacked the resources necessary to address the divisions sustained by racism' (Bernasconi, 1996, p. 4; for similar assessments see James, 2003;Gines, 2014;Kautzer, 2019).
A clear distinction between power and violence serves important purposes. But the dualist ontology of power that Arendt adopts to make sense of this distinction has significant drawbacks. In light of that, we should question whether we need to adopt a dualist ontology in order to take seriously the goal of a non-violent form of politics. In the following section, I argue that we can decouple the two issues: We can recognize a clear distinction between power and violence and yet acknowledge that violence can help constitute power.
Why power and violence are not identical I suggested that Arendt's oppositional view of power and violence can be understood as an ameliorative conceptual analysis that helps us pursue a non-violent form of politics. Achieving or even just imagining this goal arguably requires a clear distinction between power and violence. But there is room between making this distinction and Arendt's oppositional view, as I will illustrate in this section by comparing her view to Foucault's approach to power. It may seem that Foucault and Arendt have similar views about the relationship between power and violence. Johanna Oksala, for example, has suggested that Foucault follows Arendt in putting forward 'an oppositional view of the relationship between power and violence' (2012, p. 46). 5 But I will show that despite some shared starting points, they disagree considerably about the ontology of power.
Like Arendt, Foucault holds that power and violence are distinct concepts. Talk of violence refers to the direct imposition of physical effects on bodies and things: 'A relationship of violence acts upon a body or upon things; it forces, it bends, it breaks on the wheel, it destroys, or it closes the door on all possibilities ' (1983, p. 220). In contrast, power is the ability to have an effect on other agents' actions or on their dispositions to act: In effect, what defines a relationship of power is that it is a mode of action which does not act directly and immediately on others. Instead it acts upon their actions: an action upon an action, on existing actions or on those which may arise in the present or the future. (1983, p. 220) In contrast to violence, power only affects agents who act freely in the sense that they 'are faced with a field of possibilities in which several ways of behaving, several reactions and diverse comportments may be realized' (p. 221). Both Arendt and Foucault thus recognize an important connection between power and freedom. They construe this connection differently; whereas Arendt stresses the cooperative nature of power, Foucault primarily talks about power as something that is exercised over agents. But it is outside of the scope of this paper to discuss how deep this difference goes. 6 In any case, Foucault, like Arendt, stresses the relational nature of power: Power relations 'are rooted in the system of social networks ' (1983, p. 224; see also Allen, 2002, p. 141). 7 Based on these characterizations, Foucault concludes that power and violence are not identical: '[i]n itself, the exercise of power is not violence [. . .]' (1983, p. 220). The two authors thus share an important starting point.
Nonetheless, Foucault and Arendt diverge radically in their views about the substantive relationship between power and violence. For Foucault, power can and often does involve the use of violence, as he points out in the same text from which I quoted above: 'Obviously the bringing into play of power relations does not exclude the use of violence any more than it does the obtaining of consent; no doubt the exercise of power can never do without one or the other, often both at the same time' (1983, p. 220).
Far from being incompatible with violence, power often involves violence, Foucault claims. In the following sections I will further explicate this claim. First, however, I will consider exactly why Arendt's and Foucault's views on this issue diverge.
Both authors explicitly reject views that conflate violence and power. For example, Arendt mentions the sociologist C. Wright Mills, who holds that 'the ultimate power is violence' (quoted in Arendt, 1970, p. 35). I will assume that such views do not simply conflate the meaning of terms but make a substantive claim about the nature of power, viz., that physical violence is a genuine or maybe even the most paradigmatic form of power. 8 It will be helpful to consider why this view is mistaken, since Arendt draws stronger conclusions from its rejection than Foucault. Consider a standard case: Someone holds up a loaded gun and threatens to shoot me unless I give her my wallet. Suppose the threat is credible and I have sufficient evidence that the robber is a good shooter, her weapon is working, and she is willing to shoot if necessary. It may seem that the robber's ability to inflict physical harm just is what gives her the ability to affect my actions, i.e. power. Since violence is a particularly effective and context-independent means of threat, we might even arrive at the conclusion that violence is a paradigmatic form of power. Thus, while power and violence talk may be semantically different, they can be used, at least in many cases, to describe the same thing.
The central problem with this argument is that the ability to use physical force alone, without an appropriate alignment of other agents, does not yield a robust ability to affect the actions of others. 9 An agent who can, in a particular situation, affect others' actions only because of fortuitous circumstances does not have genuine social power. If the robber is successful only because she is lucky that no one happens to challenge her use of force in this particular case, we would not usually ascribe social power to her. She would have power if other agents back her use of force, tolerate it, or if other background conditions make it unlikely that they will challenge her. In those cases, however, her power would depend not only on her individual ability to use force but on the fact that other agents and their actions are suitably aligned with her. Hobbes's (1996) account of power provides an abstract illustration of this point, even though Arendt considers it an example of the tradition that conflates power and violence (1970, p. 38). While the use of force does play a role in this account, Hobbes actually argues that power is socially constituted (see Field, 2014). In the Hobbesian state of nature, individuals can rely only on their non-social abilities; they cannot make use of means that require social cooperation since in the state of nature no one can trust others to act their part within a social alignment. While agents may be able to use fortuitous opportunities to achieve their goals, that is not a sufficient basis to attribute power to them. In the state of nature, I might be able to coerce another agent in a single case, but I cannot stabilize the conditions of my superiority because I would not be able to prevent others from overpowering me should the circumstances change even slightly. Without stable social alignments, no one has power. Hobbes argues that such alignments come about only when everyone agrees to refrain from using force and obey the commands of the sovereign. But the sovereign herself has power only in virtue of a social alignment, i.e., insofar as most individuals refrain from challenging her power and actively support her in enforcing her commands. Her threat of using force can be used to reliably get other agents to do something, i.e., to exercise power, only insofar as those involved can expect that the use of force will not be effectively challenged by other agents. Violence is thus not identical to power.
Arendt makes a similar argument in support of her oppositional view. She points out that any effective use of violence already requires an appropriate social organization i.e. it requires a power base: No government exclusively based on the means of violence has ever existed. Even the totalitarian ruler, whose chief instrument of rule is torture, needs a power basis-the secret police and its net of informers. [. . .] Single men without others to support them never have enough power to use violence successfully. (1970, pp. 50f.) In other words, the government requires social support in order to effectively use violence. While their use may be temporarily decoupled from the possession of power, the creation, maintenance, and effective use of the implements of violence ultimately require social organization. Rule by violence alone, even if it is momentarily effective, will destroy the social bases of power: 'To substitute violence for power can bring victory, but the price is very high; for it is not only paid by the vanquished, it is also paid by the victor in terms of his own power' (1970, p. 53). Since the effective and reliable use of violence requires collective action, Arendt argues, violence by itself cannot constitute power.
This kind of argument shows what is wrong with views that conflate power and violence. But it fails to support Arendt's oppositional view. It does not show that violence cannot play any role in the constitution of power. Hobbes, for example, does not claim that sovereign power is based solely on the ability to use of force. Nonetheless, the social organization of violence arguably plays an important role in his account. Arendt does not carefully distinguish these two points. In the following passage from On Revolution, for example, she moves from one to the other without argument: Where violence rules absolutely, as for instance in the concentration camps of totalitarian regimes [. . .] everything and everybody must fall silent. It is because of this silence that violence is a marginal phenomenon in the political realm; for man, to the extent that he is a political being, is endowed with the power of speech. [. . .] The point here is that violence itself is incapable of speech, and not merely that speech is helpless when confronted with violence. (1990, pp. 18f.) Arendt is right to point out that social power cannot exist where the absolute rule of violence has destroyed all communicative relationships. But it does not follow, as Arendt suggests in the following sentence, that violence, when it does not 'rule absolutely,' is a marginal political phenomenon. Indeed, it does not follow that violence is always 'incapable of speech,' a point to which I will return in section 5 below. 10 Arendt is right to reject views that conflate power and violence, but this does not settle whether violence can help constitute power. We can insist on a clear distinction between violence and power without adopting Arendt's dualist ontology. In the following sections, I will elaborate an alternative, nondualist account that can capture how violence can, in some cases, help to create and maintain power structures. I start in the next section by arguing that appropriate material arrangements are not just necessary prerequisites for the emergence of power, but play a constitutive role. Recognizing this point will help us see how violence can, in some cases, help constitute power.

Durable connections and the materiality of power
When we speak of power, some stability is usually implied: Individuals or collectives are powerful only if their actions can have effects in a wide range of circumstances. Arendt recognizes this when she points out the need for a 'stable worldly structure' which can 'house, as it were, their combined power of action ' (1990, p. 175). This stable structure is not, in her view, made up of material things but based on the 'force of mutual contract or promise ' (1958, p. 245). Mutual promises create 'islands of predictability' that allow agents to collectively make plans without completely eliminating the open future that is essential to free action (1990, p. 175; see also, 1958, pp. 243-247). Yet, Arendt recognizes that this may not be sufficient: 'Neither compact nor promise upon which compacts rest are sufficient to assure perpetuity, that is, to bestow upon the affairs of men that measure of stability without which they would be unable to build a world for their posterity ' (1990, p. 182). It is not clear how the communicative act of promising could create the requisite stability.
In a similar vein, the sociologists Callon and Latour (1981) have argued that social ties, by which they mean connections that are based solely on direct interactions between agents, cannot by themselves provide the stability that is necessary for power structures to emerge. They illustrate the point by discussing the social organization of baboons, which is based solely on social ties. Baboons communicate with one another, behave in light of social expectations, and form hierarchies and alliances. But nothing in baboon society guarantees that these ties will endure beyond the social, temporal, and spatial confines of direct, face-to-face interactions. The ties are 'constantly decaying' and need to be rebuilt anew in each situation: A baboon's life is not easy. [. . .] He must constantly determine who is who, who is superior and who inferior, who leads the group and who follows, and who must stand back to let him pass. [. . .] Of course, many signs, growls and hints exist, but none of them is unambiguous enough. Only the context will tell, but simplifying and evaluating the context is a constant headache. (1981, pp. 283f.) Without a stable social structure, baboons are in a condition akin to the Hobbesian state of nature, unable to form any stable expectations regarding the behavior of others. In this state, stable social cooperation and power relations are unlikely to emerge.
The baboon example illustrates that power cannot be based solely on situationally negotiated expectations and agreements. Thus, when Arendt says that 'binding and promising, combining and covenanting are the means by which power is kept in existence ' (1990, p. 175), that can only be part of the story. Agreements have to be durable so that individuals do not have to remake them in each new interaction. But how can agreements become durable? Callon and Latour answer this question by pointing to something that baboons are missing: material culture. The durability of material things can help to extend social connections beyond face-to-face interactions: [. . .] if you transform the state of nature, replacing unsettled alliances as much as you can with walls and written contracts, the ranks with uniforms and tattoos and reversible friendship with names and signs, then you will obtain a Leviathan. [. . .] Hobbes [. . .] omits to say that what makes the sovereign formidable and the contract solemn are the palace from which he speaks, the well-equipped armies that surround him, the scribes and the recording equipment that serve him. (1981, p. 284) Walls and palaces, written contracts, uniforms, and weapons: Due to their own durability, such material things can be used to extend the reach of social expectations, roles, and hierarchies across a wide range of situations. This reduces the need to re-establish decaying social ties in each situation from scratch and thus allows individuals to engage in more stable forms of social cooperation.
According to this proposal, an artificial world of durable things is not simply a prerequisite for communication. Material things transform and mediate communicative relations. A wall, for example, can play a substantive social role. With the right material properties (height, thickness, opacity, etc.), it can set up a private space that is protected from the direct intervention or even observation of others. Once built, it assures agents of this expectation without the need to re-negotiate it anew in each situation. As a durable object, a wall can have such effects long after it is built, independently of and sometimes even against the specific intentions of its designers, users, and administrators. Of course, to have stable social effects, a wall needs to operate in concert with other agents and things, such as guards, maintenance workers, weapons, sensors and cameras, etc. Its social role cannot be reduced to its physical properties. But at the same time, it is also not merely a symbolic marker of prior social differences. Its concrete material properties contribute to the shape of social and communitive relations; they have an effect on who can participate, how, and under what conditions. Recognizing this role opens up productive lines of research about the heterogeneous set of relations that help to, in Bruno Latour's words, assemble common worlds. 11 Arendt's account does not ignore durable things, but her attention is limited by her dualist framework. She argues that city walls and national borders create a material space for free action to occur, but they do not themselves belong to the 'space of appearance' that emerges (1990, p. 186; see also, 1958, p. 195). If communicative relations are sui generis, it makes little sense to ask how this infrastructure would shape, mediate, or stabilize these relations. Arendt sometimes suggests instead that laws and constitutions, created through authoritative foundational acts, can secure stability. But insofar as the authority of laws emerges from mutual acts of promising, this suggestion does not sufficiently address the issue of stability. Mutual promises require an authoritative, relatively uncontested interpretation to serve as a stable basis for social interaction (see Keenan, 1994). It is not obvious that an authoritative interpretation can be established without relying on the durability of material infrastructures. In many contemporary societies, material law enforcement practices (which involve surveillance equipment, implements of violence, prisons etc.) make one interpretation of the law the stable and authoritative basis of social order. It is telling, in this context, that Arendt likens the law to a wall (1958, p. 63; see also, 1990, p. 186). She metaphorically draws on the durability of material objects but does not explain how this durability could arise from speech and action alone.
Callon and Latour's point puts pressure on the claim that power emerges from speech and action alone, with no direct role played by other material activities and the durable world they create. Material infrastructures shape communicative relations by making them more durable and extending them beyond the social, temporal, and spatial limits of direct social interactions. Since genuine power requires stable alignments, an adequate view of power needs to take seriously the material things that mediate and shape those interactions. In the following section, I will draw on this idea to explicate how violent practices can help constitute power.

The body in pain: violent torture as a manifestation of power
To spell out how violence can help constitute power, I discuss the example of public torture. My aim here is not to provide a comprehensive account of violence's political effects, which will differ depending on the form of violence at issue and the context in which it is used. I merely use this example to challenge the claim that violence can never positively contribute to the constitution of power. In the first part of Discipline and Punish, Foucault suggests that public torture can help reproduce sovereign power: 'The public execution, then, has a juridico-political function. It is a ceremonial by which a momentarily injured sovereignty is reconstituted. It restores that sovereignty by manifesting it at its most spectacular ' (1977, p. 48). This analysis illustrates Foucault's suggestion that power can involve violence. To undermine Arendt's claim that violence can only destroy but never create power, we need to explain how this particular form of violence can shape and stabilize alignments in which agents act in concert.
Elaine Scarry has argued that the spectacle of torture can produce a 'fiction of power ' (1985, p. 57). Her analysis starts with the observation that torture creates a stark contrast between the victim and the torturers. Fully vulnerable to the material effects of the torturers' violence, the victim is reduced to the mere physicality of her pain. In contrast, the torturers are able to present themselves as possessing a nearly absolute power. The purpose of demonstrating this superiority is not simply instrumental, i.e. to gather useful intelligence. Rather, the central purpose of torture is to make the victim submit or at least to make it seem as if she submits (1985, pp. 28-38; see also Foucault, 1977, p. 43). This presents the victim as doing something, as acting in at least a minimal sense, thereby transforming superior physical strength into an appearance of power. This transformation is possible because the torturers are in a much better position to shape the social significance of the situation and make their own interpretation authoritative, while the victim's pain prevents her from speaking for herself. The torturers make the victim's body speak for them, as it were. 12 In this way, to use Scarry's dramatic phrase, torture converts 'absolute pain into the fiction of absolute power ' (1985, p. 27).
While it may seem as though the torturers can create only an appearance of power, the appearance can serve as a catalyst for creating stable, powerconstituting alignments. As a public spectacle which, in Foucault's words, 'was to deploy its pomp in public' (1977, p. 49), torture publicly manifests power and thereby provides those in the audience with a reason to obey and support the powerful. If people in the audience go along with or at least do not challenge the display, they act in concert with those in power. Scarry's analysis emphasizes the role of the specific physical effects of torture: It is the imposition of extreme pain on the body that allows the torturers to make the victim's body speak for them. At the same time, however, torture is not completely 'incapable of speech.' Precisely because of those physical effects it can produce politically significant speech: The submission of the torture victim constitutes or reconstitutes a relationship of subordination, one that the audience strengthens by going along with the display. 13 This is what Foucault seems to have in mind when he suggests that power often involves violence and consent at the same time. Thus, the tortured body and the torture instruments are made to play the role that Callon and Latour suggested for material things: they create 'associations that last longer than the interactions that formed them ' (1981, p. 282).
The physical practice of torture does not constitute power all by itself; it has to be appropriately staged and taken up by its audience. To understand its constitutive effect, we have to take account of the broader social and material context. Public lynchings in the United States in the early 20 th century, for example, drew national attention because photographs and early motion pictures that depicted them circulated widely. These depictions gave visual substance to white supremacist beliefs and embodied habits. Through watching them, historian Amy Louise Wood argues, White spectators could gain a sense of racial dominance, bound up with claims to moral authority and social power (2013, p. 130). By bringing the audience into alignment, these depictions of violence contributed to the (re)production of a racialized social power structure. However, such effects are not guaranteed. Foucault points out, for example, that torture in 18 th -century France often induced in the audience pity, outrage, or solidarity with the victim, rather than fear and submission, and sometimes even led to revolt (1977, p. 63). In 20 th -century America, depictions of violence were also used to make secret anti-Black police violence more visible in order to prosecute it and to advocate for civil rights legislation (see Niedermeier, 2013). While violence can help constitute power, we need to pay attention to its specific social context to understand how (if at all) it does that.
An analysis of public torture that takes seriously its effects on social alignments challenges Arendt's claim that violence can only destroy communicative relationships but can never create them. While torture destroys the ability of the victim to communicate on her own terms, it at the same time allows the torturers to make the victim's body speak for them, as it were. This speech can open up a 'field of possibilities' in which other agents communicate and act in concert with one another. 14 Violent public torture is not merely an instrument to pursue an already given political end; it can help constitute the concrete social spaces in which the collective setting of ends becomes possible in the first place.

Docile bodies: the materiality of a non-violent form of power
Taking my cue from Foucault, I set out to formulate an ontology of power that acknowledges a role for violence in constituting power without thereby collapsing power and violence. Critics such as Axel Honneth, however, have argued that Foucault's approach cannot avoid the latter. Even Foucault's analysis of disciplinary power, which he explicitly characterizes as nonviolent (1977, p. 177), reduces power to the use of force, or so Honneth claims: Contrary to [Foucault's] own claims, the social-theoretic determination of the character of modern techniques of power contains nothing more than the conceptually differentiated but nonetheless fundamentally reductionistic idea of a one-sided rule of force. (1990, p. 175) This objection is instructive because Honneth relies on a distinction that is similar to Arendt's. Considering this objection will allow me to characterize more precisely the dualism at issue and elaborate some of its problematic consequences.
Foucault argues that disciplinary power functions by making individuals and their bodies available for assessment, regulation, and use, for example by shaping the material spaces in which these bodies move, by organizing the movements and interactions of these bodies, and by constantly surveying and documenting their performances. Disciplinary practices are primarily concerned with increasing the capabilities of bodies (1977, p. 11). Inducing pain or physically compelling or limiting bodily movements is not central to the functioning of disciplinary power. Consider the use of exercise to teach a student good handwriting. Honneth suggests that exercise consists in a 'compulsory standardization of motor and gestural motions' which forces bodies into a 'blind automatism of routinized acts ' (1990, p. 168). While that is not completely wrong, it focuses too narrowly on the physical aspects of exercise. Exercising handwriting as a pedagogical practice requires finegrained assessments of bodily performances: Is the student holding the pen appropriately? Are they putting the right amount of pressure on the pen? Do they have the right posture? etc. Such assessments are the basis for decisions about the future course of training. Disciplinary practices increase the capabilities of the body by way of making new dimensions of the body available to assessment. In other words, they open up new ways of judging and talking about bodily performances and create new forms of knowledge about individuals. It is in this sense that Foucault speaks of power's 'directly productive role ' (1978, p. 94). Disciplinary power is not exercised by physically restricting or compelling bodily movements, but by making new dimensions of the body available as objects for normative demands.
We cannot understand the normative contours of disciplinary practices without taking account of their material infrastructures. Discipline relies, among other things, on built structures to manage bodies and to shape the options available to agents. Specifically designed prisons, schools, factory floors, hospitals, etc. enable new forms of management, observation, classification, and assessment (1977, p. 147). For example, a classroom with individually assigned desks arranged in rows allows a teacher to keep track of the conduct and learning progress of individual students. This setup can be used to differentiate individuals by behavior, skill level, character, etc., to facilitate or interrupt communication, and to easily document student performance. Such spaces are at the same time '[. . .] real because they govern the disposition of buildings, rooms, furniture, but also ideal because they are projected over this arrangement of characterizations, assessments, hierarchies' (p. 148). In other words, these material spaces are normatively structured: They establish where an individual ought to be and how she is supposed to act, move her body, and interact with others. The standards that are embodied in these spaces directly shape the options available to agents. In many cases the subjects themselves evaluate and adapt their behavior in light of these standards; they are 'caught up in a power situation of which they are themselves the bearers' (p. 201). Like Arendt, Foucault believes that power opens up spaces for action and speech. But for Foucault, the material properties of such spaces are crucial for understanding their concrete shape.
To conclude that this analysis reduces power to violence, Honneth has to claim that the guidance of actions in disciplinary spaces involves nothing more than the use of force, however subtle. He arrives at this claim on the basis of the following disjunction: Each social stabilization of a position of power [. . .] presupposes the interruption of the struggle in the form of a normatively motivated agreement, or of a pragmatically aimed compromise, or of a permanently emplaced use of force. (1990, p. 174) Honneth contrasts the stabilizing role of communication with that of the use of force: The first two options he mentions are what he calls 'two-sided stabilizations of social power' based on speech, while the use of force is 'onesided' and merely physical (p. 174). On the one hand, it seems right to say that disciplinary power is not based on an agreement by those subject to it, either motivated by mutually accepted norms or the strategic aims of individuals. Individuals who are subject to disciplinary power simply find themselves in concrete material spaces that shape their abilities to act. This is not only the case in total institutions such as the prison, but in more everyday institutions such as schools, hospitals, and factories. Based on the quoted disjunction, Honneth concludes that 'there inevitably remains for [Foucault] only the possibility of interpreting the institutionalization of positions of power as a process of the constant use of force ' (p. 174).
But on the other hand, disciplinary power also does not function by physically compelling or causing pain to bodies. It guides actions by making specific dimensions of bodies available for normative assessment. It helps to constitute new ways of speaking about the bodies and characters of individuals, for example, in terms of criminality, sexuality, development, etc. Moreover, it often relies on the active involvement of its subjects, who assess themselves and act in light of disciplinary standards; this may not qualify as agreement but is surely a form of acting in concert. Honneth does not take these aspects sufficiently seriously. He assumes that power emerges either from communication or from the use of force, and on the background of this dualist assumption, he has to interpret discipline's reliance on material infrastructures as 'the exclusively physical process of an evermore-perfect directing of sequences of bodily motions ' (1990, p. 168, emphasis added). I argued above that even the political role of violent public torture cannot be reduced to its physical effects. Honneth's dualist ontology obscures the specific political effects not just of disciplinary practices, but of at least some forms of violence as well.
Arendt's and Honneth's accounts rely on similar dualist frameworks, despite differences in the details. For Arendt, power only emerges from communication; violence or the use of force cannot create power. Honneth allows power on both sides of the distinction: It can be based either on communicatively generated agreement or on the use of force. But since he construes these two bases of power as mutually exclusive and different in kind, his approach is nonetheless dualist in character. The target of my discussion has been the dualist strategy that these authors share, not Arendt's somewhat idiosyncratic terminological choice. Xavier Marquez, for example, has argued that it is an 'implausibly restrictive interpretation' to suggest Arendt claims that power can only emerge in a space of appearance (2012, p. 15, fn. 26). While I disagree with this interpretive claim for reasons I laid out in sections 2 and 3, Marquez is right to say that Arendt analyzes what we might reasonably call forms of power outside of the space of appearance. For example, her concern with the 'rise of the social,' the preoccupation of modern politics with managing productive activities and processes of life, converges with Foucault's concern with disciplinary and biopower (2012, p. 27). However, this reading still leaves us with Arendt's account of a sui generis form of communicatively generated power or Honneth's account of a form of power that is generated by 'normatively motivated agreement.' By definition, these forms of power cannot be shaped by violence or material practices. 15 Thus, both accounts leave us unable to understand how violence and material practices more generally can affect how agents act in concert.

Conclusion
I set out to formulate a non-dualist ontology of power that allows us to investigate the political effects of various forms of violence. Arendt's view excludes from the generation of power not just violence, but most material human activities and their products. If power is a sui generis communicative phenomenon, it makes no sense to ask how violent practices might shape how people act in concert. But we can insist that power and violence are not identical without claiming that violence cannot play any role in constituting power. Following Foucault as well as Callon and Latour, I argued that we need to attend to the role of bodies, material things, and built structures to understand how power relations are shaped and become relatively durable. I showed how this approach can be used to analyze violent and non-violent forms of power. A non-dualist approach to power allows us to inquire into how forms of violence and entrenched material infrastructures contribute to the emergence and reproduction of power, rather than ruling out such questions from the start. Notes 4. I would like to thank David Luban for prompting me to distinguish more clearly between the exercise and generation of power. 5. In her own analysis, Oksala (2012) does not follow Arendt's oppositional view.
She rejects views that take violence to be essential to politics but also argues that violence plays important, albeit historically contingent, political roles. Given the main targets of her criticism, it makes sense for Oksala to foreground the fact that Arendt and Foucault both reject views that conflate power and violence. In contrast, I engage with views like Arendt's that take violence to be external to power. Both kinds of views arguably share the assumption that violence is instrumental and coercive in nature and thus see no need to analyze the specific political effects or rationalities of different forms of violence (see Winter, 2018, pp. 2-7, Oksala, 2012, . 6. This includes but is broader than the capacity to get another agent to do something they would not have otherwise done. I assume that Arendt thinks that collective deliberations and actions make a difference to the potential actions of those participating in them. Consequently, I would suggest that the difference between Foucault's power-over and Arendt's power-with is a difference of emphasis, not a substantive disagreement about the nature of power (pace Allen, 1999). 7. Amy Allen argues that these similarities are 'ultimately rooted in a critique of one and the same understanding of power ' (2002, p. 132). 8. This formulation is supposed to distinguish the view from the stronger view that violence is essential to any form of power. I take it that the latter position is the target of Johanna Oksala's (2012) argument. 9. The notion of a social alignment is adapted from Wartenberg (1990, chapter 7).
As will become clearer in the following discussion, we need to look at 'socialmaterial alignments' that involve not just agents and their actions but also the material things and structures with which social agents interact (see also Rouse, 2002, pp. 177f.). 10. Many of Arendt's commentators do not carefully distinguish these two points and consequently cannot resolve the resulting ambiguity. Frazer and Hutchings, for example, suggest that Arendt aims her criticism at theories that 'treat violence as integral to politics ' (2008, p. 99). But 'integral' is ambiguous here: It could mean that violence is necessary for politics. But it could also mean that violence often plays an important role, without being essential. Similarly, Frazer and Hutchings suggest that Arendt's 'crucial point is that [violence] should never be conflated with politics itself' (2008, p. 102). But we can avoid conflating politics with violence while holding that violence can play a role in politics. Conversely, they argue against Arendt that 'a clear conceptual distinction between [violence and power] is problematic ' (2008, p. 103). In my view, it is not the distinction itself that is problematic, but Arendt's dualist account of it. See also Allen (2002, p. 137), Ayyash (2013, p. 351), Hanssen (2000, p. 25), Herzog (2017, p. 167). 11. Latour and Foucault both have pursued concrete investigations of this kind (see Foucault, 1978, pp. 92-95 for the notion of 'heterogeneous force relations'; see Latour, 2005, pp. 247-263 for the notion of 'assembling a common world'). Another example is the work of political theorist Timothy Mitchell (2011), which explores how the material properties of coal, oil, and the infrastructures required for their extraction, transport, and use have shaped capacities for political action.
12. See Luban (2014) for a similar, communicative analysis of torture. Like Scarry, Luban focuses on the communicative relationship between torturer and victim and pays less attention to the effects that torture has on the audience. 13. Similarly, George Sorel (whom Arendt criticizes for conflating power and violence) argues that violence can be 'acted out and dramatized theatrically' and thereby shape a political consciousness that can help generate a new political order (see Finlay, 2009, pp. 30-38). Ayyash (2013, p. 345) also recognizes a 'productive' role of violence. 14. Of course, this space may not always be as egalitarian and inclusive as Arendt seems to suggest (1990, pp. 33-31, 1958, p. 175; see also Bernstein, 2012, pp. 8-9). Even if power is a form of togetherness, this togetherness is usually partial, a matter of determinate groups of actors with specific interests interacting, and the power emerging from these interactions need not be fully inclusive or nonhierarchical (Breen, 2007, p. 364;see also zabermas, 1977;Allen, 1999, p. 103). 15. Marquez's own proposal, in particular his distinction between spaces of appearance and 'spaces of surveillance' and his analysis of power in terms of visibility, deserves more discussion than I can provide here. Whatever its merits, it deviates significantly from Arendt's account.