4.1 Clarifications about Both Theories

First, note our ontological assumption that social norms are requirements, under both theories. This clarifies the type of objects social norms are in a (we believe) very natural way. For us, social, moral, rational and legal norms all are requirements – but requirements on different grounds. Requirements may or not be normative, as explained later.

Second, a requirement has a content: the thing being required. The content of a requirement on individuals typically concerns behaviour, and more rarely thoughts or attitudes. It can usually be expressed in the form ‘do (think, intend, …) X in circumstances Y’, for instance, ‘keep noise down after 10 pm’, ‘show up at pub meetings’ and ‘stand on the right on escalators’.Footnote ⁶ Working out what ontological type of object ‘requirements’ are is a problem in its own right.Footnote ⁷

Third, a conforming practice is simply an individual conduct matching (the content of) the requirement. For instance, working hard is a practice conforming to the requirement to work hard.

Fourth, what does it mean for something to ground or (as we also say) generate the social norm, i.e. that the norm is grounded in or generated by it? It means that the requirement exists in virtue of it – a relation that is distinct from causation or mere determination, and that captures what constitutes and/or explains the norm.Footnote ⁸

Fifth, groups can have a more or less well-developed identity and demarcation. Nations, ethnicities and religions might qualify as tight groups, whereas the users of London Underground escalators form a loose group. Social norms can also arise in loose groups, even if membership is fluctuating and fuzzy.

4.2 Clarifications about the Process-based Theory Specifically

First, a social communication process is a communication process between sufficiently many members. The communication could take many forms – verbal or behavioural, structured or chaotic. Many such processes ground or ‘generate’ no requirement – think of random dinner conversations.

Second, social communication processes can generate requirements in all sorts of ways. They do so through some expressive acts by some of the participants of the process, whether these expressive acts are verbal or behavioural.

Third, many generated requirements are not social norms. Only a few social communication processes live up to the standards of a SNP. Which ones do? On a first pass, the following condition captures the presence of a SNP, i.e. the generation of a social norm (with content C):

This informal condition will be made precise in sections 6 and 7, where we will clarify the notion of ‘support’. Many processes violate the condition. For instance, a process during which one member is elected as a dictator and orders everyone to dress in blue succeeds in generating a (dressing) requirement, but not a social norm. Similarly, a tumultuous communication process in a crowded staircase in which some shout out ‘Walk left!’, others ‘Walk right!’, yet others ‘Stand still!’, generates three (conflicting) requirements, but no social norm, since these three expressions of will were made and perceived insufficiently often.

Fourth, as is now clear, the Process-based Theory gives a necessary condition for being a social norm, not a sufficient condition. The same is true of the Attitude- and/or Practice-based Theory. Providing a necessary and sufficient characterization is the ambition of a full-fledged account of social norms, to which we turn in sections 6 and 7.

Finally, do attitudes and practices play any role at all? They can only play an indirect role, by either driving the SNP or resulting from the SNP or from the social norm. They do not ground the norm.

5.1 Action-based vs. State-based Grounding

Our first argument for the process-based approach rests on the idea that social norms exist in virtue of being made (or re-affirmed), and hence in virtue of actions. We thus impose the:

Our process-based account of social norms satisfies AGC, since communication processes consist of (communication) acts. Practice-based accounts also satisfy AGC, since conforming practices are actions.Footnote ⁹ But attitude-based accounts, such as Brennan et al.’s and Valentini’s, violate AGC, since attitudes are states rather than actions, notwithstanding the importance of actions in Brennan et al.’s account, where attitudes respond to (presumed) practices. If the grounds are states, we will talk of state-grounding rather than action-grounding.

Our argument for AGC can be cut into smaller steps. First, social norms are by nature ‘constructed’ or ‘made’. Second, as such, they are grounded in a construction. Finally, this construction consists of actions.

To see the point, compare with promises. In virtue of what is a promise a promise? Surely not in virtue of the present attitudes of promiser and promisee, but of the act of promise-making (Norris Reference Norris2021). The promise would exist even in the absence of attitudes such as intentions or normative expectations. Similarly, what grounds laws and contracts is at least partly the act of law-making or contract-making. These are all examples of arrangements that are ‘made’ and thus action-grounded. It would be surprising and question-begging if social norms – another similar arrangement – fell out of the common pattern by being state-grounded rather than action-grounded. The actions in question can be acts of expressing an attitude – just as contracts arise by expressions of will. But expressions of attitudes are not attitudes.

5.2 Social vs. Individual Grounding

Our second argument for the process-based approach is, in a nutshell: if social norms are to be truly ‘social’, their grounds should be social. We thus require the:

Social features can be any features pertaining to the group, such as social processes or joint commitments. Some social features are ‘social’ only in a thin sense of being ultimately features of group members. Examples are the feature that every member expects everyone to speak French, or that most members dislike people chewing gum in public. These social features are the aggregation of individual features, specifically individual attitudes. We here speak of aggregative social features, as opposed to irreducibly social features, which go beyond any summary of individual features. Aggregatively social features are effectively a conjunction or “sum” of individual features: they are reducible to individual features. An example is the feature that, in a group of five individuals, a majority composed of Ann, Bob and Claire normatively expects everyone to wear blue shirts. It is effectively the conjunction ‘Ann expects the behaviour and Bob expects it and Claire expects it’. If a social norm to wear blue shirts were grounded in this feature, then SGC would be violated, since the grounds are not irreducibly social.

The Process-based Theory satisfies SGC. So do accounts that ground norms in irreducibly collective attitudes, such as those by Gilbert and the later Tuomela (see also Blomberg Reference Blomberg, Garcia-Godinez and Mellin2023). Grounding norms in collective attitudes, however, faces other problems. Social norms seem to exist in many groups that are too loose for having collective attitudes. For instance, the highly disconnected and fluctuating users of the London Underground have come up with their social norm about how to use escalators without being a well-structured group with collective attitudes. Our process-based account achieves social grounding without requiring any collective attitudes. By contrast, SGC is violated by accounts that ground norms exclusively in individual attitudes and/or practices.

We have required social grounding because otherwise social norms are not genuinely social constructs. But is this actually true? One objection might claim that social norms are social because of being norms for the group, not norms by the group, so that we need not worry about what generates the norm. This claim is, however, problematic. For one, being a norm for a group is not a unique feature of social norms. Moral norms can also be for a group – for instance, if wealthy individuals have a moral obligation to alleviate poverty, then we have a moral norm for wealthy individuals. For another, some groups develop social norms that aim to apply even to outsiders, perhaps to all humans whatsoever. Such ‘imperial social norms’ are not for the group, but only by the group – we discuss them later.

On another objection, social norms that are grounded in individual attitudes are still ‘indirectly social’, because the relevant attitudes have a social basis or a social cause. Some individual-attitude-based accounts indeed emphasize the social origins of the attitudes, presumably recognizing the force of SGC. In particular, for Brennan et al., the relevant attitudes are based on social practices. Social norms that are grounded in individual attitudes with a social basis or cause might be called indirectly socially grounded, but we would resist calling them socially grounded simpliciter.Footnote ¹⁰ Moreover, although it is often true that the relevant attitudes are caused by social interactions, there are troubling counter-examples, in which the individuals have developed their attitudes in utter isolation. For instance, assume an attitude-and-practice-based account whereby a social norm to do X is grounded in people doing X, (normatively) expecting everyone to do X and believing that others also expect this. Now assume these shared attitudes and practices came about as follows:

The alleged social norm is not even indirectly socially grounded: it is grounded in individual features (attitudes and practices) that were generated purely individually. The point is related to Andrei Marmor’s observation that a ‘happy coincidence’ of attitudes is not enough for social rules (Marmor Reference Marmor2023: 13). The Dream might create the illusion of a social norm, but not a social norm. This is not to deny that a social norm might arise later, once individuals have communicated support for X – but it is, yet again, the communication that grounds the norm, not the attitudes.

Hart’s account admits partially social grounds if and insofar as his ‘social practice’ includes genuinely interactive patterns that constitute irreducibly social features.Footnote ¹¹

Table 1 summarizes where we stand, by classifying different accounts in terms of their type of grounding. Thus, only process-based accounts satisfy both desiderata, AGC and SGC.

Table 1.

Classification of accounts of social norms

5.3 Synchronic vs. Diachronic Grounding

On most accounts, social norms are grounded synchronically, i.e. in something that exists presently. Most attitude- and/or practice-based accounts, such as Marmor’s (Reference Marmor2023) and, on our reading, Valentini’s (Reference Valentini2021, Reference Valentini2023), take the norm to depend on present attitudes and/or practices, and to disappear if those attitudes and/or practices vanish.

By contrast, on our process-based account, the norm is grounded diachronically, i.e. in something than can have stopped by the present moment (see, e.g. Wilson Reference Wilson, Glick, Darby and Marmodoro2020). For sure, the SNP can extend into the present and future. But it can equally well have stopped earlier, just as promises, laws and contracts are grounded in past actions of promise-making, law-making or contract-making. The SNP cannot be interrupted for too long, though: it cannot issue and later maintain a norm once and forever. To survive, the norm needs occasional re-confirmation by a reopened SNP, as discussed in section 8. Here social norms depart from promises, laws and contracts, which typically need no re-confirmation.

Diachronic grounding is a distinguishing feature of our account. Some instead advocate synchronic grounding by construing social norms as higher-level features that are grounded in present lower-level features, namely present attitudes and/or behaviours (see Marmor Reference Marmor2023: 2).

5.4 Social Norms vs. Social Pressures

Why is it so popular to ground social norms in attitudes? Perhaps partly because of the social pressure exercised by attitudes in society, such as expectations to behave in certain ways. Yet a social pressure to follow some behaviour is not a social norm: it is not a requirement, but some other causal force acting on people’s psychology. Social norms can exist without social pressure (if no pressurizing attitudes exist), and social pressures can exist without a social norm (if no SNP took place).

Nonetheless, a social norm requiring C often co-exists with a social pressure to follow C, be it because the norm causes pressurizing attitudes or, conversely, because these attitudes initially caused the SNP. Sometimes the pressurizing attitudes are revealed through the SNP. For instance, consider a social norm not to litter in a public park, grounded in a communication process in which people publicly utter snide remarks at litterers. These snide remarks reveal all sorts of attitudes that pressurize people to refrain from littering.

5.5 Social Norming Processes vs. Social Practices

Sometimes, a behaviour can communicate support for this very behaviour. Thus, a conforming practice can become a communication process, a non-verbal SNP. Here, practice-based and process-based accounts of social norms overlap superficially. But the overlap is slim. For one, a practice becomes an act of communication only if additional features hold. For instance, the practice must be suitably public to be perceived. And it must be ‘understandable’ as an act of communication, which in turn depends on the context – for instance, on the pre-existence of a convention, as in an example below. While practice-based accounts ground social norms in conforming practices per se, process-based accounts ground social norms in a SNP, and if this SNP happens to include practices, then the norm is grounded not in the practices per se, but in the practices as acts of communication.

Sometimes, practicing C is even a necessary part of communicating support for C. If people “preach” C without acting accordingly, they might overall fail to communicate support for C, so that the Norm Generation Condition fails. In other cases, however, someone can perfectly communicate support for C without practicing C. This is, for instance, the case if one’s lack of practice is unobservable – think of a norm not to vote for extremist parties in secret ballots. It is also the case if C is so difficult to follow that a lack of personal practice expresses a personal weakness rather than a lack of support – think of a norm requiring to love one’s enemies. Someone might also manage to communicate support for C without practicing C if her personal circumstances give her publicly known reasons against C – moral or rational, say – where those reasons outweigh any (pro tanto) obligation to C.

When Hart grounds social norms in a social practice, he uses the term ‘practice’ in a much broader sense that includes not only conforming practices, but also certain attitudes, and possibly many other behaviours, such as criticizing or sanctioning deviations. Insofar as such behaviours can represent communication acts, Hart’s account looks like a hybrid practice-, attitude- and process-based account. Our earlier reservations apply analogously.

5.6 Social Norms vs. Conventions

Social norms should also be distinguished from Lewisian conventions. Lewis (Reference Lewis1969) describes conventions in terms of game-theoretic equilibria, in which every player acts in rational pursuit of her preferences, based on correct beliefs about the other players. Unlike social norms, Lewisian conventions can evolve in the complete absence of communication. They require widespread conforming behaviour, again unlike social norms. While (our) social norms are based on a communication process rather than attitudes and practices, Lewisian conventions are based on attitudes and practices.Footnote ¹²

While a convention as such does not yield a social norm, let alone social normativity, it implies a rational requirement, and rational normativity. Indeed, a convention to C implies that people have a rational reason for C, since C-behaviour maximizes their preferences given their beliefs. This said, if a convention to C is in place for a while, a social norm requiring that same C often emerges on top of the convention, because the convention sparks a communication process that grounds a social norm. The typical sequence of events is this. First, people behave chaotically. Then their behaviours (and beliefs) coordinate such that a convention emerges. Then people start communicating support for this behaviour, giving rise to a social norm. As usual, the communication can be highly implicit – people might communicate their support merely through practicing the convention. If this is how they communicate their support, why did the social norm not emerge earlier, simultaneously with the convention? After all, the same practice was alive already earlier, giving rise to the convention. No doubt, the social norm could arise simultaneously with the convention. But it need not, because, as seen above, this practice might only later acquire the meaning of an act of communication. For example, consider the use of public staircases. Initially, everyone walks chaotically, leading to collisions. After a while, more people start walking on the left, pushing away people who block the flow. They start believing that others will do so, too. A convention is born. Suddenly, this same behaviour of walking on the left and pushing away obstructers, which used to express an aggressive attitude, becomes an act of communication: it now communicates support for this behaviour. This communication process, if wide enough, grounds a social norm.

7.1 Truthful and Credible Communication

Suppose the local mafia boss orders that all front doors be painted in blue, his favourite colour. Fearing repercussions, nearly everyone complies and shows various signs of approval, thereby publicly expressing the relevant will. This is widely perceived in the community. Thus, COM holds. But is there really a social norm to paint doors blue? Two things are peculiar. First, the expressions of will are not truthful: members have pretended to want this, out of fear. Second, those who perceive these expressions do presumably not believe what is being expressed: they realise that the others only pretended.

The account of social norms can be refined in two different ways to respond to the two concerns. We say that someone communicates X (e.g. a will) truthfully to someone else if she communicates X to him and X indeed holds.Footnote ²⁰ She communicates X credibly to him if she communicates X to him and he then believes X. The two refined conditions are:

On the t-COM-based account, no ‘door colour norm’ exists in our mafia example, as the widely communicated will is fake (except for the boss). On the c-COM-based account, the norm again fails to exist, now because people fail to believe that the widely communicated wills exist (except for the boss). While t-COM and c-COM are possible alternatives to COM, one might insist that truthfulness and credibility are not essential, though often present. Why? In a realistic version of the mafia story, members won’t manage to fake the will all the way: they will send mixed signals, including support (say, by complying with the order) and disapproval (say, by their facial expression or lack of enthusiasm for blue doors), overall not expressing the will. In consequence, COM fails, and this blocks a social norm. But if members have faked support consistently and these expressions of support have been widely perceived, nobody questions their truthfulness, so that COM holds, and then the norm did arguably emerge. For a legal analogy, note that a contract comes to exist once parties have expressed their will or agreement; speculations about ‘secret’ dispositions are irrelevant. Since, to us, social norms are also grounded in certain acts or processes rather than private attitudes or thoughts, social norms can plausibly emerge whenever the process succeeds – even if the attitudes involved are skilfully feigned.

7.2 Communicating Agreement

Taking the analogy to contracts and laws further, some might argue that social norms arise through communication of agreement, not of will. They would replace COM with:

Will and agreement are often linked, but agreement is more cognitive while will is more desire-based. One can want something without agreeing to it, and arguably also vice versa.

Expressions of agreement can again be highly implicit – more implicit than for creating contracts or laws. Being silent, complying or nodding can all express agreement, not just will.

7.3 Communicating Direct Attitudes

In the accounts above, members do not communicate that they want (or agree to) C simpliciter, but that they want (or agree to) an obligation requiring C. They communicate an attitude towards an obligation requiring C rather than towards C itself – an ‘indirect’ rather than ‘direct’ attitude towards C, as we shall say. By contrast, many attitude-based accounts of social norms tie social norms to direct attitudes, such as expectations, demands, preferences or wills. For many such accounts, one might consider a corresponding process-based account in which this direct attitude is communicated:

We are sceptical about a COM′-based account. Communicating mere approval of some behaviour without any obligation seems insufficient for a social norm. If everyone wants everyone to dress properly, and communicates this, but everyone also rejects an obligation to do so, then arguably no social norm was created, although COM′ holds. For comparison, to create a law (or contract), the legislators (or parties) must approve the law (or contract), not just what it requires. While COM* moves social norms closer to laws and contracts, COM′ moves social norms further away from the legal sphere, because expectations (demands, etc.) do not ground laws or contracts.

7.4 Higher-order Communication

Proponents of an attitude-based approach to social norms often do not stop with requiring that enough members (say) expect some behaviour: the expectations must be widely believed to exist (first-order beliefs), widely believed to be widely believed to exist (second-order beliefs), etc.Footnote ²² COM only guarantees the existence of first-order beliefs: the expressions of will must be widely believed to have occurred. In fact, COM guarantees something stronger: the expressions have been widely perceived (hence believed to exist). We have explained earlier why we require perceptions rather than mere beliefs. For analogous reasons, when going higher order, requiring higher-order perceptions is more adequate than requiring higher-order beliefs. We will talk of ‘higher-order communication’. We define communication of order 0, order 1, order 2, etc., as follows:

These conditions were stated informally but can be disambiguated.Footnote ²³ Two facts follow:

Fact 1: For any order k ≥ 1, ${\rm{CO}}{{\rm{M}}_k}$ implies ${\rm{CO}}{{\rm{M}}_0}$ , …, ${\rm{CO}}{{\rm{M}}_{k - 1}}$ , assuming our notion of perception (and a non-degenerate notion of enoughness whereby ‘enough members’ excludes ‘no members’).

Fact 2: ${\rm{CO}}{{\rm{M}}_1}$ is equivalent to COM, assuming our notions of perception and communication.

To prove Fact 1, it suffices to show that ${\rm{CO}}{{\rm{M}}_k}$ implies ${\rm{CO}}{{\rm{M}}_{k - 1}}$ for each k ≥ 1. This holds because if enough members perceive something, say X, then someone perceives X given a non-degenerate notion of enoughness, and so X must hold given our notion of perception. Fact 2 holds by a slightly more elaborate argument.Footnote ²⁴

By Fact 1, an account of social norms based on communication up to a certain order k (e.g. up to order 2) can be defined through a single unified condition, namely ${\rm{CO}}{{\rm{M}}_k}$ , which automatically subsumes ${\rm{CO}}{{\rm{M}}_0}$ , …, ${\rm{CO}}{{\rm{M}}_{k - 1}}$ .

We leave it open whether social norms require higher-order communication, and if so whether one needs communication up to a certain order k or of all orders k. If ${\rm{CO}}{{\rm{M}}_k}$ holds for all orders k, one might talk of ‘quasi-common perception’, in analogy with the notion of ‘common knowledge’ in logic. The qualification ‘quasi-’ reflects that the conditions ${\rm{CO}}{{\rm{M}}_k}$ are stated with ‘enough members’ rather than ‘all members’. Full-blown common perception holds if each condition holds in the stronger sense with ‘all members’.

8.1 Failures to Generate a Social Norm

Many social processes aim to generate a social norm, but fail, as COM is violated. COM can fail because of expression failures, or perception failures, or both. Examples will help.

Expression Failures. Expression failures happen, first, if too few members express approval, i.e. COM₀ fails. For instance, a lone supporter of orderly queuing is insufficient to instil a queuing norm. Similarly, in the mafia example, most members will likely send mixed signals, overall failing to express the will.Footnote ²⁵ Second, expression fails if it is not public enough. No norm arises if everyone expresses a will to herself. Similarly, if a social network is divided into cliques where users only contact members of their own clique, social norms cannot arise because wills are not expressed publicly across cliques – at most, norms within cliques could emerge.

Perception Failures. Modifying the social-network example, imagine a different clique effect: members rarely see messages or posts from people outside their clique, be it because they chose certain settings or because the network provider promotes intra-clique communication. Then COM fails since expressions of will are not widely perceived. The same happens if prejudices prevent members from perceiving wills of a certain type or wills expressed by certain members.

8.2 Disappearance of a Social Norm over Time

Many social norms disappear over time. Attitude- and practice-based accounts explain this as follows: the norm disappears once the grounding attitudes or practices have disappeared, assuming that grounding is synchronic, so that norms are grounded in present attitudes or practices. Our process-based approach explains norm disappearance by a lack of sufficient communication of support for the obligation. This idea is intuitively clear in examples.Footnote ²⁶ But is it coherent? One might at first think that the communication process that had initially put the norm in place grounds the norm forever, given that our notion of grounding is diachronic, i.e. that norms can be grounded by past communication. This reasoning is too simple. Let us inspect the condition COM carefully. This condition can initially hold and later fail, for essentially two reasons.

Group evolution: Groups gain and lose members over time, for instance through birth and death. The term ‘members’ in COM stands for ‘members of the present group’. If the communication process does not continue over time by taking the new members on board, then, eventually, insufficient communication among present members has taken place, so that COM now fails. As a drastic example, if the communication process stopped 200 years ago, when only past group members participated, then this past communication does not ground a present norm – COM fails today, relative to today’s group members.

Communicating retraction of support: If the communication process pauses for too long, then those who used to communicate support for the obligation will – by their silence – effectively communicate that they no longer care about the obligation. For instance, if the pub regulars stop criticizing the absence of non-attenders for a long while, then they signal indifference about attendance. COM is violated by now, since not enough people have communicated support – the norm is gone. The same happens if people start criticizing the obligation after a while. This might again seem surprising: can someone who has communicated something later lose the property of having communicated this? Yes, she can, because ‘having communicated X’ should be read as ‘having communicated X overall, up to the present moment’. As time progresses, the interval of time over which communication activity could happen grows. If the person never communicates X again or even starts communicating the opposite, then it is eventually no longer true that she has overall communicated X.

11.1 The Autonomy Reason

On one view, a social norm becomes normative if an autonomy condition holds (besides COM):

The conditions AUT and COM are independent. While both demand a decision process of whether to require C, AUT adds that this process meets the standards of an exercise of collective autonomy, while leaving open whether the requirement finds any support and thus whether the social norm emerges.

Why should AUT (jointly with COM) give reason for C? A broadly Kantian thesis says that an exercise of individual autonomy commands respect.Footnote ³⁴ Similarly, an exercise of collective autonomy in a SNP commands respect, by complying with the social norm. This is because there is value in groups setting their rules autonomously rather than being governed by external forces. The value of such autonomous self-governance could be intrinsic or instrumental. Indeed, it may firstly be valuable in itself that members jointly author their shared environment. Secondly, this may have valuable effects, including benefits of social cooperation, such as to allow members to pursue their goals and set terms of social interaction.Footnote ³⁵

The parallel with legal normativity strikes again: laws owe their normativity partly to an underlying process of autonomous democratic will formation. In Waldron’s words:

If Waldron is right about legislation, an analogous argument for social norms looks promising: they deserve respect and become normative because a collective action problem was solved autonomously. This reasoning also explains why social norms can have force on outsiders interacting with the group: they should also respect the group’s autonomous self-governance.

Which SNPs represent an exercise of collective autonomy is debatable. Arguably, SNPs in which members fake or misrepresent their will, so that t-COM fails, do not succeed in generating social normativity. Members may then at best believe to have reason to comply, particularly if people fake their wills credibly, so that c-COM holds – the case of pluralistic ignorance discussed earlier. Other counterexamples are arguably SNPs in which members are brainwashed: the wills they express are genuine, but not ‘free’. Here normativity fails although t-COM holds.

In general, properties of a SNP promoting the exercise of autonomy include: transparency, inclusiveness, open deliberation and positive responsiveness to individual inputs. Conversely, autonomy is undermined if the SNP is subject to malevolent or arbitrary influences from inside or outside.

In sum, the autonomy of the SNP and thus the social normativity stands and falls with the quality and integrity of individual will-formation, will-expression and will-aggregation.

11.2 The Public Reason Reason

Alternatively, a social norm becomes normative if a deliberative-democratic condition holds (besides COM):

REA requires an exchange in which members justify their wills publicly, by offering publicly acceptable reasons. A public reason exchange would then ground social normativity – just as public reason processes can justify democratic decisions, following a vast literature on public reason and deliberation that we cannot review here.Footnote ³⁶

On most accounts, public reason processes require high-level cognitive engagement. Few SNPs will clear this hurdle. Examples of SNPs violating REA are non-verbal SNPs, verbal SNPs where members express wills but not reasons, and SNPs where members express publicly unacceptable reasons, such as racist or dishonest reasons.

11.3 The Consent Reason

On an entirely different view, a social norm becomes normative if members give their unanimous consent:

The idea is that a requirement binds someone if she has given her consent – just as contracts bind each party because they have given consent, say by signing. Giving consent is an act, in fact often the same sort of act by which members also express their agreement in COM* or their will in COM, such as speech acts or acts of nodding. Arguably, giving consent always expresses agreement, truthfully or non-truthfully; and so CON implies COM*.

A CON-based account of social normativity faces two problems. First, CON rarely holds, and so few social norms would be normative. Second, explaining why the norm should bind outsiders interacting with the group becomes hard, unlike for an AUT-based or REA-based account. Two unsatisfactory possibilities present themselves. Either the norm does not bind outsiders – but this is ‘selective normativity’, which leaves outsiders and newcomers beyond the force of social normativity, against common views and intuitions about the normativity of social norms.Footnote ³⁷ Or the norm does bind outsiders – but this view is question-begging, if not incoherent. The problem is familiar from consent-based views on political obligation (e.g. Simmons Reference Simmons1981).

11.4 Additional Background Conditions

Normativity requires not only one of these substantive reasons, but also some background conditions. We propose two such conditions:

FEA excludes norms requiring infeasible actions, attitudes or thoughts. Its exact meaning depends on the notion of (in)feasibility, ranging from logical or physical notions to psychological, economic or political notions (e.g. Southwood Reference Southwood2018). One could defend FEA using the principle ‘ought implies can’, applied in a social rather than moral setting.

By ICO, the norm only speaks to people in, or interacting with, the group. C could say ‘we eat with knife and fork’ or ‘we and our guests eat with knife and fork’, but not ‘all humans eat with knife and fork’. Norms violating ICO will be called ‘imperial’, as they interfere with other groups. Imagine, say, a social norm created among UK residents prescribing cutlery norms for people living in Vietnam. This norm is exclusively imperial: it is only about outsiders. Exclusively imperial social norms are rare. Partially imperial ones, about insiders and outsiders, emerge in abundance, however. Why? Sometimes real-life SNPs are simply too unsophisticated to set scope-restricted requirements.Footnote ³⁸ But often the imperial nature of norms is intended: many groups strive for universal rules, valid for everyone, not just themselves.

Imperial social norms arguably lack normativity, at least for outsiders, because such norms are a form of unjust interference.Footnote ³⁹ In fact, ICO might follow from AUT, REA and CON, since interfering with others’ self-governance cannot be classified as an exercise of collective autonomy or defended by public reason. Moreover, the interference would not have been unanimously consented to.

Though lacking social normativity, imperial social norms can – unfortunately – still exist and be enforced through sanction mechanisms, creating rational rather than social reasons to comply.

11.5 Social Normativity Summarized

We have proposed three potential sources of social normativity, and some background conditions. A narrow but unified theory of social normativity would accept just one source. For instance, by accepting only the autonomy reason, a normative social norm would arise if and only if this combined condition holds:

$\matrix{{\underbrace {{\rm COM}\; \left ({\rm or\;a\;{\mathop{\rm variant}} } \right )}_{ \to social\;norm}\;} \hfill & \& & {\underbrace{{\rm{AUT}}}_{ \to normativity}\;} \qquad \qquad \ \ \hfill & \& & {\underbrace {{\rm{FEA}}\;\& \;{\rm{ICO}}}_{ \to background\;conditions}\;} \cr } $

Perhaps more plausibly, one could accept different sources of social normativity. For instance, by accepting all three sources discussed, a normative social norm would arise if and only if this combined condition holds:

$\matrix{{\underbrace {\rm COM\;\left( {or\;a\;{\mathop{\rm variant}}} \right)}_{ \to social\;norm}\;} \hfill & \& & {\underbrace {{\rm{AUT}}\;{\rm{or}}\,{\rm{REA}}\;{\rm{or}}\;{\rm{CON}}}_{ \to normativity}\;} \hfill & \& & {\underbrace {{\rm{FEA}}\;\& \;{\rm{ICO}}}_{ \to background\;conditions}\;} \cr } $

On a sufficiently wide understandings of ‘moral’, social normativity is inherited from moral normativity. Indeed, one might have moral reason to respect requirements created by SNP that meets AUT, REA or CON. Social normativity would then be inherited from moral normativity. We say ‘inherited’ because morality does not require C itself, but requires a conditional: if a process of such-and-such type happened then one follows C. This makes C a contingent moral requirement.