2. Social practices

We are concerned with social practices in the broad sense of regularized patterns of behavior that are the result of people conforming to behavioral standards or norms (Bicchieri Reference Bicchieri2016). This expansive domain is appropriate given the apparently wide range of contexts where legitimacy makes sense. Although details vary, starting from social practices in this approximate sense is common in social theory (e.g., North Reference North1991; Tuomela Reference Tuomela2002; Searle Reference Searle2010).Footnote ⁴ The basic model of social practices I sketch in this section shows how normative orders function and emphasizes the place of role-kinds.

Social practices are normative orders: they coordinate behavior by applying various norms to entities within the practice.Footnote ⁵ Primary among these entities are practitioners, i.e., persons who follow the norms that apply to them. Practitioners put practices into practice. When practitioners share a significantly overlapping sense of what the norms require of them and accept that they should abide by the norms, they will generally act according to the norms in similar ways. (Here I leave aside practitioners’ motivations for following the norms, which will vary widely, including internal norm acceptance and external incentives.) The resulting consonant behavior produces some purported good and characteristically aims at some goal or end. In this sense, social practices are teleological, although not in the sense of being intentionally created to pursue an end (Miller Reference Miller2010).

When practitioners share norms, they have a common behavioral standard. Without shared norms, individual actions are unlikely to overlap in any interesting way; even if the actors all intentionally pursued a shared goal, conflicting individual judgments about how to proceed would lead toward haphazard and often contradictory behavior. The practice would not be able to succeed and would not exist in a significant sense without practitioners feeling bound by and sharing an understanding of the relevant norms. (Though the norms need not explicitly enjoin coordination.) Social practices are thus defined by behavior patterned by norms and only exist in virtue of their ability to establish and maintain such patterns.

Patterns are defined by their boundaries. A norm that people walk across a certain path in a field defines a boundary between acceptable and unacceptable behavior. If the norm is accepted within a social group, that normative boundary will result in a social boundary: when walking across the field, most people will follow the path most of the time. We can describe the pattern in terms of behavior that falls on one side or another of the boundary, either following the norm and walking along the path or violating the norm and walking elsewise. This allows us to predict action, to show how this pattern will interact with others, and so on. We can explain why the grass is worn there but nowhere else or why people (lightly) sanction people who stray off the path. These patterns of behavior shape the social world and we can identify them by their boundaries.

Complex social practices mostly coordinate norms and behavior by defining role kinds. Assigning unrelated bundles of norms to practitioners would be very unwieldy and inefficient for a variety of reasons. Instead, practitioners take up roles that bundle norms into a comprehensible whole. Roles are the functional units of most practices. They enable the higher-order coordination of behavior that results in a successful practice, for example by reducing the demands of interpersonal trust and cognitive load (Searle Reference Searle2010; Ritchie Reference Ritchie2020).Footnote ⁶ Roles are partly defined by deontic status, a collection of incidents including rights, duties, permissions, and powers (Hohfeld Reference Hohfeld1919). Roles also include important elements such as ideals and more general expectations. The status, ideals, and expectations that constitute a role must be understood in light of the function of the role in the practice.

All kinds of entities are assigned roles (and status). Here we are primarily focused on practitioner roles, which have agential status. Entity kinds also include mere subjects (dogs), objects (stop signs), events (recessions), ideas (theories), and actions (voting), among others. I may ask a student whether her raised hand was intended as a vote; I am asking whether the action falls within the role and so should count in the way votes do. To see how practitioner roles must be understood functionally, consider the familiar roles of professor and student in the university. The status that defines being a professor must be directed at enabling the university to pursue some overall purpose or telos—education, for example. Professors’ teaching obligations are then matched to students’ learning obligations. When kinds are defined and interrelated well, they enable us to pursue the practice’s telos.

So we have a basic model of a social practice. Practitioners pattern their behavior according to the norms of the practice. In any given moment, this takes the form of practitioners deciding how to act by applying the practice-defined kinds to assess their practical context. The routinized essence of social practices is practitioners determining whether an entity E is a kind K by applying K standards and then, if E meets the standards, treating E as a K by following the relevant K norms. For example, universities involve routinized classes. Practitioners gather with the expectation and intention that the standards for a class meeting will be met, take up coordinated roles, and engage in event specific behavior.Footnote ⁷ If the practice is constructed and practiced well, recursive events of this sort will lead to a professor having taught a course and students learning.

The basic model is a mechanical perfection. Internally, the normative structure is perfect. Roles are defined clearly, completely, and consistently such that any practical context a practitioner could face is subsumed into the practice. The norms are similarly clearly, completely, and consistently defined so that there is a univocal fact of the matter about what the practitioner should do. Externally, the normative structure is overriding. Practitioners are presented as practice automatons who do as directed without exception. They apply and follow norms as directed, they do not make mistakes, and they do not have other demands on their action. Considered as a whole, the practice appears autonomous with practitioners as cogs. This model illustrates how social practices function as role-based normative orders but it does not need legitimacy. Legitimacy discourses are solutions to problems it does not have.

3. Boundary work

Let’s introduce some reality. Practices do not appear or operate of their own accord. People enact the practice by taking up roles and patterning their behavior with norms. This is where boundary work enters the picture. Which norms people follow as practitioners and how they follow those norms are social questions that are only ever answered by particular people interpreting their practices in particular contexts. Social practices have to be put into practice collectively, and so always contestedly.

Consider the classic sociological example of John Tyndall’s nineteenth-century boundary work on the social practice of science (Gieryn Reference Gieryn1983). When a phrenologist was being considered for the chair of logic at the University of Edinburgh, a contest ensued over the boundaries of science. Tyndall advocated for a version of science that excluded phrenology while the phrenologists advocated for a more expansive version that would give them access to the social practice’s status and resources. No matter where the boundary of science lies according to our preferred philosophical solution to the demarcation problem, science as a realized social practice is shaped by where practitioners believe the boundary of science lies and so what they will treat as a science (cf. Rawls Reference Rawls1999, 48). Eventually Tyndall and his allies convinced decision-makers at the university as well as members of the community to adopt their understanding of science and so to treat phrenology as a pseudoscience. One version of the social practice of science was implemented and one was set aside.

Practitioners constantly face choices between versions of practices.Footnote ⁸ Normative orders necessarily underdetermine practitioner choice because they function by intervening in the practical decision-making of agents. Norms are directions for exercising agential discretion, not programming. The basic model oversimplifies insofar as it treats social practices as static objects rather than processes, as automatons’ mechanized behavior rather than as agents’ patterns of choosing. Social practices only ever exist in virtue of people choosing, again, to follow practice norms by playing their role in the practical contexts they face over time. Discretion is intrinsic to social practices.

Practitioners have both internal and external discretion. Internally, practitioners have discretion in the sense that they can follow practice norms in many different ways and so must choose how to follow. Internal discretion is ineliminable because being a practitioner is not exhausted by simply conforming to fully programmed directives. Norms are necessarily underdetermined: they can never fully specify how they are to be practiced in particular contexts. The command “Open the door” underdetermines which part of the body is used, how much force is used, how many nanoseconds the door is opened for, how wide the opening is, and so on. It partly determines many of those elements but cannot fully specify them. Underdetermination is especially prominent in systems of codified rules like law (Schauer Reference Schauer1991). Further, practitioners rarely face decisions that are determined by a single element of their role; norms overlap in application, ideals and expectations exert pressure, and so on. Acting out one’s role in a practice requires making the role’s practical demands determinate in a specific context in ways that the role’s norms cannot fully dictate. Call this ‘contextualization.’

Externally, practitioners have discretion in the sense that they can refuse to act as directed, instead acting against or outside the practice. External discretion is ineliminable because practitioners are not exhausted by their participation in any particular practice. Practitioners always act as whole persons whose actions implicate their whole selves and there are always intersections with other aspects of their identity. Due to the spatiotemporally local nature of agency and action, when I stay late for a work meeting, I cannot spend that time with my family. Practitioners are thus always exercising their discretion about which practice’s norms to follow in particular contexts and how to prioritize the various demands on their agency. Call this ‘situating.’

Contextualization and situating are ineliminable from practices. This is good; discretion is how we make social practices livable. For a social practice to actually be part of our lives, to be practiced by people like us in our circumstances, it must be flexible enough to be placed into our lives and into our contexts against other practices and demands. Discretion enables practices to be adaptable to changing contexts, to update norms that have become sclerotic, and to respond to the needs and desires of practitioners. However, discretion also goes hand in hand with hard cases.

Hard cases occur “when no settled rule dictates a decision either way” (Dworkin Reference Dworkin1978, 83). Such cases are at the core of general jurisprudence, asking how judges should exercise discretion. These are hard cases partly because of courts’ adjudicative function: not deciding is not an option and the contest is at least not trivial. Social practices are suffused with discretion but most discretionary choices are trivial. I get to choose at what decibel level I lecture; the university functions perfectly well as long as I choose within an understandable range. Neither the law nor the university takes a stance on this discretionary choice, and neither needs to. But some choices matter for practices in the sense that leaving them up to individual discretion would harm the practice’s functioning. Hard cases, then, are choices that the practice currently leaves open to practitioner discretion but where practitioner discretion is inadequate.Footnote ⁹

The obvious way to solve a hard case is to introduce a new practice norm, since norms reduce discretion.Footnote ¹⁰ Norms, and procedures for introducing norms, can be either formalized or nonformalized (cf. Brennan et al. Reference Brennan, Eriksson, Goodin and Southwood2013, chap. 3).Footnote ¹¹ Law stands at the maximally formalized end of this spectrum, solving hard cases by introducing formalized norms in a formalized way. Courts’ legal authority entails that their decisions bind, including by creating precedent. Law also has higher-order hard case settlement procedures; in the United States, for example, when federal appellate courts disagree, the Supreme Court takes over.

Reflection on this example demonstrates the limits of formalized norms. Notoriously, Supreme Court decisions are not immune from underdetermination. They render the law more determinate in some ways but also introduce new norms in new language. More importantly, Supreme Court decisions are not self-executing, so are open to pushback via contextualization and situating. The integration decision in Brown v. Board of Education resulted in decades of contested implementation. Legal decisions only pattern behavior in virtue of practitioners enacting them. Some legal practitioners have official roles and so play a special part in enacting law but, in the end, communities implement law by building legal norms into collective and individual decision-making. The relevant social practice is the legal order—a social order that is successfully organized via law. Supreme Court decisions change the law but how exactly that changes the legal order is an open question that can only be answered in practice by practitioners.

How do practitioners exercise coordinated discretion when formalized norms run out? A regress seems to threaten. Hard cases are solved by introducing new norms but all norms include discretion, so introducing a new norm also introduces new discretion. This can work by shifting the point of discretion, and perhaps reducing its scope. When the court decides, practitioners still have some discretion about how to implement the decision but at least there is an authoritative decision to implement. Norm introduction slows but does not halt this regression.

Social practices rely on nonformalized implementation norms to halt discretionary regression. To make practices collectively livable, practitioners coordinate discretion with what organizational theorists call the “logic of appropriateness” (March and Olsen Reference March, Olsen and Goodin2011). Professors teach classes not by mechanically applying the formalized norms but by situating and contextualizing the norms with, among other things, a shared sense of what is appropriate for professors. Nonformalized norms necessarily structure the implementation of formalized norms. These nonformalized norms arise from practitioners’ overlapping, more holistic understanding of what professors should be like and the role they should play in the university (and so, what the university should be like).

The faculty handbook doesn’t need to stipulate that I lecture at a certain decibel level because the nonformalized norms of appropriateness handle that. We understand that professors lecture to teach and so lectures need to constitute successful communication. The nonformalized norm that I teach at an understandable level is implicit in the very idea of the university and the role of professor. This norm is active; if I violate it, students or colleagues can bring it to bear. If I persistently violate this or many other implicit communicative norms, not only would I fail at being a professor, it is also doubtful that I even grasp what it is to be a professor. The broader social and linguistic context in which the practice exists provides resources for implementation that simultaneously channel the forms the practice can take. Competent practitioners necessarily possess a wide variety of appropriateness norms that guide choices when formalized norms run out. Instead of choosing randomly, practitioners can compare potential versions and choose to implement the better one. One way to identify which versions are better is to compare their kind-relative properties.

4. Kind-relative properties

Imagine I am at the store and offer what appears to be a twenty-dollar bill for my purchase. Sometimes people counterfeit money so cashiers have ways of testing putative dollars. These tests could be enabled with a new concept, schmollar. The associated schmollar discourse would include standards for distinguishing dollars from schmollars and behavioral norms for how to treat each. By enabling comparisons between putative dollars, this conceptual discourse functions as a solution to the problem of fake money.

But the problem of fakes arises in a wide variety of practices. Thus, we have more general concepts to enable our solutions. Instead of asking about dollars and schmollars, the cashier can ask about genuine dollars and fake dollars, and we can ask about genuine and fake handbags, or genuine and fake friends, or a wide variety of other entities. But it is also clear that what makes something authentic varies across kinds: fake dollars, fake handbags, and fake friends are very different entities in very different practices. Identifying each is a different project and we should also react to each in quite different ways. Authenticity discourses apply the same concepts of genuine and fake across practices but use different standards of authenticity and license different responses to authentic or inauthentic entities. This enables us to use the same discourse to perform the same function across very different practices.

This social function is identifiable in language: “genuine” is an essentially attributive adjective (Sibley Reference Sibley, Benson, Redfern and Cox2001, 169). Such adjectives ascribe properties that can only be assessed relative to membership in a kind. The claim that E is genuine must be understood as claiming that E is a genuine token of kind K. Famously, Geach (Reference Geach1956) argues that “good” is essentially attributive. Other essentially attributive adjectives include “real,” “tall,” and, “big.” As Sibley (Reference Sibley, Benson, Redfern and Cox2001, 175) puts it, “To ascertain that x is big, you must have a class of things in mind.” Bigness standards are relative to kinds. A big mouse is much smaller than a big elephant, which is much smaller than a big skyscraper. We do not always make the kind explicit because the context, including the entity itself, can make kind membership salient. But these comparisons necessarily assume that the entity is being evaluated as a member of a kind. Essentially attributive adjectives point us to kind-relative properties.

This ordinary usage calls for explanation, especially concerning the function of discourses that employ these concepts. Geach’s point about good is not about language but about how we assess properties of this sort (Thomson Reference Thomson2008, 17). He argues that because goodness must be relative to kinds, it is a grave mistake to search for a reified property of goodness that all good things share. When we use adjectives attributively, we draw attention to membership in some kind and its associated standards because there is no way to assess the truth value of the attributive use without those standards.

Cross-practice discourses to assess kind-relative properties like authenticity and size are useful because in-kind comparisons of the same sort are useful. Big and small mean the same thing across practices; bigness and smallness pick out the same comparative properties. But the standards for assessing big and small are relative to each kind. If we only have one standard of size, it would be very difficult, e.g., to compare mice to each other because they would all be very small. Crucially, then, it is the standards of kind-relative properties that are relative to each kind, not the properties themselves or the function of assessing those properties.

Big and small differentiate members of the same kind. Do genuine and fake? If we ask whether something is a genuine mouse and discover that it is a cleverly designed robot, we don’t seem to be differentiating members of the same kind. Mice and robots are different kinds. This shouldn’t lead us to conclude that genuine is not assessed according to kind-relative standards. Instead, it points us to the fact that different kind-relative evaluations play different functions. Size evaluations are useful for some things, genuineness for others.

We use genuine when fakes or forgeries are salient possibilities. When asking whether something is a genuine mouse, the standards for genuine are affixed by the kind mouse but the class of objects we are differentiating among are not mice but putative mice. There’s no sense in asking whether an apple is a genuine mouse because it’s not even putatively a mouse. But it does make sense to ask whether the thing that appears to be a mouse but is also sparking a little around its tail is a genuine mouse or a fake mouse. A fake mouse is, of course, not a mouse at all; a fake mouse is a putative mouse that does not meet the kind-standards for membership in the biological category. So an authenticity discourse must have a class of evaluands that includes entities that are not part of the practice. By provisionally incorporating external entities into this liminal part of the practice, the authenticity discourse regulates inclusion in the practice, enabling it to function under the threat of fakes.

Genuine is a conceptual tool that we use in authenticity discourses to solve the problem of fakes: the cross-practice property of genuineness is evaluated according to kind-relative standards. Fakes disrupt the routinized essence of practitioners evaluating whether Es are Ks by applying shared K standards. Convincingly fake Ks will lead practitioners to treat non-Ks as Ks, thus impairing the practice’s functioning, for example by diluting the resources and status of Ks. Fakes are a threat when the normal K standards could be inadequate for identifying genuine Ks. So, to solve the problem of fake Ks, genuine discourses must employ a more stringent set of K standards. While visual inspection of dollars may be a good enough dollar standard for most transactions, enhanced standards might be necessary for higher denominations where fakes are more common.

We have detoured into authenticity because it gives us a clue: legitimate meant legally genuine, as in a legitimate child who thus has the legal status of heir (OED). This points to the idea that “legitimate” is an essentially attributive adjective: “E is legitimate” means “E is a legitimate K.” If legitimacy is a kind-relative property, then we can only determine whether E is a legitimate K by reference to K standards. The meaning of legitimate and illegitimate is the same across practices; we can coherently talk about legitimate states, legitimate contracts, legitimate questions, or legitimate wins. But legitimacy standards must be kind-relative to play their distinctive function within each practice: gatekeeping.

5. Gatekeeping

Analogous to the problem of fakes, social practices face the problem of bad boundaries. Consider a version of football where the purpose of a season is to identify the best team. If the seasons consists only in a round-robin tournament, the team with the most wins is the best team. In this version of the social practice, the role of the kind “win” is to identify the best football team. But not all wins are created equal. If team A beat team B when B was missing their best players due to injury, we might say that this win shouldn’t count. Although A won, this win does not serve the purpose of counting wins in this context because this win is not a good indicator of A’s quality.Footnote ¹² The extant practice boundaries define the kind “win” in a manner that doesn’t always help us identify the best team. Well-drawn boundaries will define wins so that they indicate team quality. Bad boundaries will define wins so that they include some bad indicators and exclude some good indicators. So bad boundaries paradigmaticallyFootnote ¹³ result in bad Ks: putative Ks that shouldn’t count as Ks because they don’t function well as Ks and putative non-Ks that should count as Ks because they would function well as Ks. For simplicity, I’ll focus on the former kind of bad K (false positives or Type I errors).

The problem of bad boundaries is that badly drawn kinds result in bad Ks, which in turn result in a malfunctioning practice. Bad boundaries are bad from the internal perspective of the success of the practice. Other perspectives ground other problems of bad boundaries; how social practices handle such problems is an important topic, it is simply distinct from my concern here.

The problem of bad boundaries is closely related to practice underdetermination and so practitioner discretion. All social practices can be multiply realized and must be rerealized in changing contexts. Practitioners exercise discretion over which version to realize. This choice is often between (de re) better and worse versions. If practitioners implement bad boundaries, their routinized evaluation of Es as Ks will degrade the practice’s functioning. To recognize this as a problem, practitioners need to be able to take up a stance internal to the practice but distinct from the extant version of the practice. That is, they need to appeal to a better version of the same practice. This enables immanent criticism by giving practitioners a perspective internal to a shared understanding of and commitment to the practice from which to criticize the extant version of the practice.Footnote ¹⁴

Practitioners qua practitioners have a guide for evaluating better and worse versions ready to hand: teloses. Practitioners must have a practical understanding of the telos of the practice and of the kinds within the practice because such an understanding is presupposed when they employ a logic of appropriateness. When practitioners situate and contextualize practices, they necessarily interpret the practice to implement a version that they offer as appropriate (Dworkin Reference Dworkin2011). They present themselves as practitioners in good standing, so putting into practice a version that other practitioners should recognize as fitting with the ongoing practice. All such interpretations can only be understood as an appropriate version of the practice qua pursuit of its telos. We get telic standards from teloses. The claim that A’s win against B shouldn’t count draws on telic standards for wins. The telos of the kind “win” is to measure the quality of the teams, serving the telos of the season to determine the best team. Wins against teams who are shorthanded do not help us measure team quality, so they are degenerate instantiations of the kind: counting them does not serve the purpose of counting wins.

Legitimacy discourses are a solution to the problem of bad boundaries. Legitimacy discourses help us identify bad Ks and coordinate our response to them. Putative Ks that shouldn’t be treated as Ks are illegitimate Ks. (The most salient way to be a putative K is to be an extant K but there are many other ways.) Illegitimate wins are wins that shouldn’t be counted as wins. Illegitimate states are states that shouldn’t rule. Illegitimate children are children that shouldn’t inherit.Footnote ¹⁵ In each case, E is a putative K but should not be treated according to the status that a K normally has. Legitimacy discourses employ telic standards to identify those putative Ks that are inadequate as Ks. Footnote ¹⁶ As noted above, bad boundaries also give rise to false negatives. Putative non-Ks that should be treated as Ks are legitimate Ks. Legitimate wins count toward our assessment of team quality and legitimate states have the right to rule.Footnote ¹⁷

This explains what we add when we ask after an entity’s legitimacy. When we ask after an entity’s goodness, for example, we are admitting that they belong to the kind but evaluating whether they realize the virtues of the kind. When we ask after an entity’s legitimacy, by contrast, we have yet to admit them into the kind as a practical matter: we are considering whether we should treat them as a member of the kind at all. This presumes that there is some work to be done still. Simply asking whether they are a K is not enough; we imply that meeting the extant standards for K-ness doesn’t settle whether they should be treated as a K, that another set of standards should orient our behavior.

Legitimacy discourses therefore serve as a corrective on the extant practice. When a practitioner judges that E is an illegitimate K, then they will not treat E as a K. If practitioners share a legitimacy discourse, they will coordinate their treatment of E and will shift the boundary of the practice to exclude E from K-ness. Shifting the boundary in this way is an inelegant solution. It handles a particular case but does not address the underlying problem of how this E became putatively K. Presumably the processes that led to this result will continue to generate illegitimate Ks. Trying to correct this on a case-by-case basis is a kludge that depends on a sufficiently shared nonformalized legitimacy discourse rather than the core machinery of a practice.Footnote ¹⁸

This kludginess is mitigated by the internal logical of legitimacy judgments. Telic standards appeal to practitioners qua practitioners: they say the practice is malfunctioning. Calling E an illegitimate K says not to treat E as a K because treating it as a K does not serve the purpose of the practice. This implies that the extant boundaries are insufficient for properly orienting ourselves to E. It also implies that extant procedures for correcting boundaries are insufficient, since we are having to correct the practice by violating it.Footnote ¹⁹ When K -ness goes wrong in less worrying ways, social practices generally have other ways of handling bad Ks. Acting against the practice also risks destabilizing the practice. So the implication is that this bad K is such a bad K that we should fix the practice by shifting the boundaries ourselves but also that we should fix the underlying boundary-setting machinery of the practice. Due to the type of standards they employ and the discursive context in which employing these standards becomes useful, legitimacy judgments push reform both at the level of particular entities in the practice and at the level of the practice itself.

For all this to make sense, legitimacy standards must be quite minimal. Just asserting that the practice has diverged from some standard is not enough. Practitioners are always instantiating the practice by interpreting it, situating and contextualizing it according to standards they endorse. There will be constant divergence, most of it trivial, and much nontrivial divergence will be handled by extant internal correction procedures.Footnote ²⁰ If legitimacy standards were demanding, they would often be violated and practitioners would deviate so often from practice norms in an attempt at correction that the practice would be destabilized. Less demanding standards identify a line that cannot be crossed qua minimally acceptable version of the practice. Crossing this line warrants violating the extant practice norms from the perspective of the practice itself. Other telic standards will perform other functions. An excellent discourse gives practitioners a way of identifying Ks that realize the telos particularly well, giving us goals and other kinds of practical guides. The maximal telic standards associated with excellence enable us to improve and perfect our practices.

The minimal telic standards associated with legitimacy enable us to gatekeep our practices. Gatekeeping ensures minimally acceptable boundaries are being instantiated by correcting the practice when boundaries go too far astray. So here is the gatekeeping account of legitimacy. That A is a legitimate K means that A meets some set of minimal telic K standards and so practitioners should treat A according to the associated K status. That A is an illegitimate K means that A does not meet those standards, so practitioners should not treat it that way. When practitioners have shared legitimacy discourses and so make overlapping judgments, their actions are coordinated and thereby gatekeep social practices. This is a general account of the core meaning across any social practice; penumbral meanings of course follow. This generality explains the inclination to falsely ascribe a deep polysemy to legitimacy.

6. Law

As I’ve described it, gatekeeping is a way that practitioners exercise control over the version of the practice they are instantiating by choosing among possible boundaries. There are many ways to perform this general function since choosing which version to put into practice is a standard part of boundary work. Tyndall’s rhetorical boundary work appealed to a shared understanding of science and its telos without necessarily employing legitimacy. We eventually developed pseudoscience as another tool for drawing some boundaries of science.

Gatekeeping’s distinct usefulness becomes clear when considered against the background of all the other tools that we use to instantiate different versions of practices. Nonformalized telic norms are vague, not easily enforceable, and not guaranteed to be shared. They grant an insecure social grasp over the entities they apply to. In general, practices function by stably reproducing some default version, especially via what Bourdieu (Reference Bourdieu and Nice1977) calls “habitus”: the internalized, embodied way that we situate ourselves into the social world, enacting roles by default. But precisely because of the facts of discretion and the need to shape practices differentially over time, practices also often have ways of enabling contests over versions. Highly formalized rulemaking is the limit case.

Practitioners always situate social practices in and against other practices in ways that affect which versions get put into practice. In a particular course meeting, the professor and students are choosing which version of this course and courses in general to instantiate. But they cannot choose anything. They are constrained by the form that courses are given by their particular university and the broader cultural understanding of courses. They are also constrained by what they are willing to choose given their understanding of their roles in the university, their identities more generally, the other roles they inhabit, and much else. And, of course, by law.

Law enables formalized, enforced, public control over social practices. Modern law plays an integrative function between various social domains, limiting which versions of social practices can be put into practice in order to ably situate practices against each other (Habermas Reference Habermas and Rehg1996). For example, legal corporate forms are more readily put into practice than illegal forms, so legality exercises some collective control over economic life. Practitioners are also always legal subjects in a modern regime and we put a legal order into practice partly by situating all our other roles as legal subjects. When Amazon workers successfully unionized in New York in April 2022, they used their powers as legal subjects to force Amazon into a version that bargains with workers as a unit. Amazon was aware that such a version was a possibility, so engaged in (among other tactics) boundary work to persuade workers to vote against the union, presenting the union as undesirable. United States law mediates this contest within Amazon qua social practice in virtue of its effective sovereignty claims.

Within any given domestic social practice, then, we must understand the function of legitimacy discourses in relation to legality. Law claims to set the outer limits of permissibility and possibility when practitioners are choosing which version of a social practice to instantiate. In a functioning legal order, that claim will be mostly effective due to both law abidingness and enforcement. Thus, calling a potential version illegitimate when it is illegal is largely redundant; it may be true but it is mostly beside the point. Illegal versions will be difficult to operate in distinctive ways, including in light of practitioners’ (variable) commitment to legality.Footnote ²¹ So legitimacy discourses in this context are primarily useful to address telically bad but not illegal versions of the practice.

Of course, law is also a social practice. In a way, it is the social practice of social practices (cf. Rawls Reference Rawls1999, 462). Like all our social practices, practitioners always face an open question of which version to instantiate. But unlike all our other practices, law cannot stand outside this practice as arbiter of the outer limits of socially permissible versions. Law cannot stand outside itself. Law has highly formalized procedures for choosing which version to implement, up to and including a constitution and procedures of constitutional interpretation. There can be illegal versions of law but positive law is precisely changeable by the lawmaking regime, so illegality is not a very secure critical stance on that regime. Luckily, no social practice can close off practitioners’ external discretion, our ability to step outside any practice and ask how we should act.Footnote ²² So the legitimacy discourse that we employ to solve the problem of bad boundaries of law has its own distinctive form and function.Footnote ²³ We have discovered the legitimacy discourse that we call political legitimacy.

‘Political legitimacy’ is a term of art. It is not about all questions of legitimacy in the political domain, although such questions are coherent. Instead, it is about the regime-level claim to the right to rule. Political legitimacy discourses give citizens minimal telic standards for evaluating state rule. Lower-level questions of legitimacy within the regime take the same form as any other domestic social practice because they are also legally regulated. If someone questions Brian Kemp’s legitimacy as the governor of Georgia, respondents can appeal to the fact that he won the election; if the question is pressed further, they can appeal to the legality of the election under state and, eventually, federal law; further again, to the constitutionality of such law. At this point, notoriously, positive law exhausts itself. The only place for this exchange to go is the nature of law and a legal order—perhaps officials’ beliefs, or a basic norm, or an act of popular sovereignty become the object of discussion. That is, where the formalized system of political contestation exhausts itself, telic standards on the ruling regime take over. Political legitimacy is where we turn when we need critical standards for gatekeeping states.

If we take a functional approach to legitimacy discourses, we must take account of practitioners’ realistic alternatives. Precisely because they employ nonformalized telic standards, legitimacy discourses are an insecure solution to the problem of bad boundaries; in some ways they are last resorts. This is especially true of political legitimacy because it distinctively lacks legality as an external source of gatekeeping.Footnote ²⁴ As classical social contract theory emphasized, practitioners also do not have the option of abolishing the political institution because organized community is morally and practically necessary (Christiano Reference Christiano2008, 237). So, practitioners employ political legitimacy discourses to assess whether they should treat E as K (whether they should treat their state as having the right to rule) in a very particular context where they need a K and where they lack formalized external gatekeeping mechanisms.

The institutional form of the state further complicates gatekeeping. The sovereign state constitutively claims final and exclusive authority (Morris Reference Morris1998). The state’s monopoly on violence, in Weber’s (Reference Weber, Henderson and Parsons1947) classic characterization, ensures these claims are effective. Effectively enforced final and exclusive norms guarantee “average norm compliance” (Habermas Reference Habermas and Rehg1996, 31) and so result in a social order structured by law. Modern legal orders are characterized by successful integration of all the social practices in a territorial jurisdiction and the primacy of individuals as citizens above and beyond their participation in any particular practice.

For our purposes, the upshot is that state legitimacy is the limit case for legitimacy discourses. Attempting to fix the problem of bad boundaries by treating the sovereign state as if it lacks the right to rule is a fraught enterprise. By definition, there are no higher authorities to appeal to and no competitor institutions that are in place to carry out the same functions. As defenders of the status quo are quick to emphasize, correcting the state by disobeying it risks destabilizing the legal order. The alternatives are limited and by definition untested in that context—not to mention arrayed against whatever capacities for violence the state maintains and is willing to wield against challengers.

This does not mean that gatekeeping the state is impossible. It does mean that effective internal boundary repair procedures are more important for the state than any other social practice. In other words, the more a state enables citizens to legally contest and change the law, the less gatekeeping there is for a legitimacy discourse to perform (cf. Pettit Reference Pettit, Sadurski, Sevel and Walton2019). It follows that democratic procedures, division of powers, and rule of law all make states more legitimate in the sense that telic failures can be handled without treating the state as if it lacked the right to rule.Footnote ²⁵ Since these procedures give practitioners modes of effective boundary work within the practice, bad boundaries can be addressed without making the drastic judgment that the appropriate way forward is to treat this K as not a K.

On the gatekeeping account, then, a state is legitimate insofar as its laws should be treated as binding according to minimal telic standards, especially its higher-order laws concerning political contestation and change. We have alighted upon a familiar understanding of political legitimacy.Footnote ²⁶ My goal in articulating the gatekeeping account is to refine our understanding of state legitimacy, not to overthrow it.Footnote ²⁷ It is a virtue of the account that it can explain the language of legitimacy in a broad variety of contexts while remaining consistent with theories of legitimacy in the political context, where they have been most carefully considered.

7. Justice

A central question for political philosophy is how political legitimacy and justice relate (Larmore Reference Larmore2020). On a standard, Rawlsian understanding, legitimacy is useful because permanent reasonable disagreement about justice makes justice norms unsuitable as an object of stable agreement. On my view, by contrast, legitimacy norms are tools for implementing any kind of social practice, which take on particular importance in the context of the modern state. In short, even Rawls’s well-ordered society would need political legitimacy discourses.

We can start with Rawls’s (Reference Rawls1999) appeal to the circumstances of justice. This presents norms of justice as solutions to problems that arise under certain conditions, following Hume. Different kinds of norms will be useful as solutions to different problems that arise under different conditions. The problem of justice as Rawls conceives it is contestation over moderately scarce resources. Norms of justice solve this by articulating acceptable public standards for distributing benefits and burdens. If people don’t accept justice norms, then they won’t accept the public authority’s distributional decisions as binding, i.e., they won’t reliably abide by law. This society will be deeply unstable because political decisions do not settle social contests. Rawls’s worry that any liberal society is fatally unstable is answered by the possibility of the well-ordered society. Eventually, however, Rawls recognizes that norms of justice are insufficient grounds for a stable society because of their deep embeddedness in broader ideologies. Among others, Valentini (Reference Valentini2012) narrates Rawls’s famous political turn from justice to legitimacy as recognizing the condition of permanent reasonable disagreement about justice (cf. Buchanan and Keohane Reference Buchanan and Keohane2006). This circumstance of legitimacy presents a different problem and so demands a different solution: the liberal principle of legitimacy, public reason, and the overlapping consensus.

For Rawls and many following him, legitimacy norms are weaker justice norms (Rawls Reference Rawls2005, 428). They solve the same problem as justice norms but under different, especially nonideal, conditions (Freeman Reference Freeman2007; Langvatn Reference Langvatn2016). Their function is to stabilize a political community by providing standards for accepting political decisions. They must be less demanding than justice norms to be agreed upon even amidst disagreement about justice but they must also make sense as standards for accepting political decisions. Justice norms are preferable standards—after all, they tell us how people deserve to be treated—but disagreement rules them out as a workable solution, so we settle for second best.

The gatekeeping approach makes clear how this story mistakes the nature of legitimacy. As Waldron (Reference Waldron1993) argues, justice must be administered, not merely meted out. Justice requires formalized rules of justice: institutionalized administration over a particular jurisdiction with publicly defined statuses and mechanisms for contestation (Rawls Reference Rawls1999, sec. 1; Christiano Reference Christiano2008, chap. 2). This formalized social practice, as all formalized social practices, can only be put into practice with the help of nonformalized implementation norms. Legitimacy discourses help practitioners implement social practices by solving the problem of bad boundaries.

The problem of bad boundaries is not a problem of disagreement. Consider Rawls’s wellordered society. Insofar as it is a human society, its law is underdetermined and practitioners have a great deal of discretion to implement their legal order in their changing contexts as they see fit. How they do so cannot be fully determined by justice norms since justice must be institutionalized. Formalizing implementation norms simply creates new underdetermined points of discretion. But implementing law via discretion can lead to versions of the practice that drift, including the parts of the practice that serve as internal correction mechanisms. The possibility of sufficiently bad boundaries cannot be ruled out by stipulating compliance with principles of justice.

Consider civil disobedience as part of our broader theory of political legitimacy. Which injustices justify civil disobedience is underdetermined by the theory of justice because it is partly a strategic question of effective remedy. Citizens need nonformalized political norms to guide illegality as a way of riding herd on the state (Habermas Reference Habermas1985). It is precisely a question of exercising control over the form of our political institutions when formalized mechanisms have run out. As Rawls (Reference Rawls1999, 342) puts it, “The final court of appeal is not the court, nor the executive, nor the legislature, but the electorate as a whole. The civilly disobedient appeal in a special way to this body.” This appeal must be made to nonformalized telic standards. Appealing to formalized standards in this context means appealing to legalized standards, but making legal space for civil disobedience is conceptually incoherent since disobedience implies illegality in this context. The same dynamic of ineffective internalization haunts legitimacy standards. Formalizing them fundamentally changes them.Footnote ²⁸

A human society that fully agreed on a theory of justice would still need political legitimacy norms.Footnote ²⁹ The theory of justice cannot fully specify how to implement a legal order. Legal orders are social practices that function via the enactment of shared, public norms, so are inevitably underspecified and require discretionary implementation. The function of political legitimacy norms is to ride herd on that practice by providing minimal telic standards for assessing versions and correcting sufficiently bad boundaries. Political legitimacy discourses provide the nonformalized norms that are necessary to implement the formalized administration of justice norms.

In summary, human social practices only exist in virtue of people exercising their discretion about which version of the practice to implement in their practical context. Practices guide this choice in many ways and so result in patterned behavior defined by boundaries. But explicit guidance always runs out and practitioners need ways to coordinate the nonformalized implementation of boundaries. One place such nonformalized guidance is especially important is in the correction of bad boundaries. Bad boundaries are a problem from the internal perspective of the practice but are also resistant to correction by extant procedures. Legitimacy discourses solve the problem of bad boundaries by identifying minimal telic standards that practitioners can employ in immanent, critical boundary work. Political legitimacy discourses enable citizens to gatekeep their states, coordinating corrective measures for when the state pulls away from our collective aims.