Stylistic Conventions and Complex Group Collaboration

Abstract

Social etiquette, dress codes and culture-specific architectural features are undoubtedly stylistic conventions. Literature from anthropology, sociology and ecological psychology suggests a coordinative function of such conventions, without, however, offering a theoretical analysis of this function. The best-known philosophical theory of conventions—by David Lewis—does offer a theoretical analysis of the coordinative function of conventions, but stylistic conventions typically fall outside the purview of this theory. The present paper suggests a remedy for this situation by putting to use the notion of ‘correlation devices’, developed as an addition to the Lewisian framework. I argue that stylistic conventions function as markers for social categories without which these categories become cognitively intractable. Given that social categories are a precondition for complex coordinated role-divisions, and given that such role-divisions are a major part of the explanation for our evolutionary success, I argue that it is likely that the psychological proclivities that make us susceptible to stylistic conventions can be explained as the result of group-level selection pressures.

There is a particular class of cultural conventions, which I shall refer to as ‘stylistic conventions’, that is not often discussed in the philosophical literature. Some theories in cultural anthropology, intercultural communication, sociology and ecological psychology assign a crucial coordinative function to these conventions—albeit largely implicitly. Yet, they seem to fall completely outside the purview of the best-known philosophical theory of conventions, according to which conventions are meant to facilitate social coordination. The aim of this paper is to dissolve this tension in the literature, and by doing so, to argue that stylistic conventions are an overlooked precondition for complex group collaboration.

There is a large body of literature on what enabled humans to collaborate in groups of ever-increasing size over the course of our evolutionary development. There is wide agreement on the fact that relatively small-scale (i.e., hunter-gatherer) collaboration requires capacities such as those for joint intentionality [1,2], pro-sociality [3,4], mindreading and/or mindshaping1 [5,6], language and other forms of referential communication [7], cultural learning [8,9], prestige bias and conformity bias [10], and norm-psychology, including punitive emotions [11,12]. Perhaps capacities such as these suffice for complex collaboration as well [13]. But it is likely that religion [3,14] and shared beliefs or ‘fictions’ [15] play crucial roles in such collaborations too. I want to argue that a crucial psychological characteristic required for complex collaboration is overlooked in this literature: our sensitivity to what I will label ‘stylistic conventions’.

The paper is organized as follows. I will start by briefly introducing the notion of stylistic conventions, followed by giving a few examples of theories from the social sciences that implicitly assign a coordinative function to such conventions. In Section 3, I will discuss David Lewis’ well-known theory of conventions and argue that it fails to assign a coordinative function to stylistic conventions, unless some emendations to the theory are made. In Section 4, I will argue that the notion of ‘correlation devices’, developed within the Lewisian framework, is a step in the right direction, but further refinements are needed. In Section 5, I will argue that the most important correlation devices in complex collaborations are in fact abstract structures that must be ‘brought to life’ through stylistic conventions—much in the way that the concrete styling of the game of chess allows us to recognize pieces that are defined in terms in abstract rules. I will call this the ‘activating marker hypothesis’ (AMH) and elaborate on a few salient kinds of categories that stand in need of stylistic marking in Section 6. In Section 7, I will discuss a few evolved psychological tendencies that are typical for humans and that can be interpreted as enabling our susceptibility to stylistic conventions. If the AMH is on the right track, then selection pressure for the emergence of these tendencies may in part be explained through the success of large-scale collaborations2. In Section 8, finally, I will compare AMH to the widely held and well-established idea that stylistic conventions serve as ‘ethnic markers’ that help to recognize in-group members. I will argue that both views complement each other but do different explanatory work and make different predictions.

1. Stylistic Conventions

Let me characterize the notion of ‘stylistic conventions’ by saying something about each of its two components.

Style is a slippery notion, but a useful definition for our purposes, derived from combining a few approaches from archaeology [19], is the following:

‘Style’ is an often highly specific way of doing that involves a choice between functionally equivalent alternatives that is characteristic of a certain time and place, where ‘doing’ is very broadly conceived to include e.g., making, building, decorating, embellishing, ordering (e.g., of objects in space), dressing and acting.

In most instances, style, defined in this way, is repetitive in the sense that choices of doing things in a specific way (functionally equivalent to doing them in various other ways) are made consistently in the same way by people who belong to the same group: greeting, dressing, eating, building, decorating, etcetera. These choices are stylistic conventions: culture-specific collective habits, rules, customs and regularities pertaining to the aesthetic design of our social and physical environment. These include, most notably, social etiquette, dress codes, personal adornments and the design and decoration of artefacts, public spaces, buildings and interiors.

I will use the term ‘conventions’ in a broader sense than is customary in some academic domains. Cultural psychologists, for example, distinguish between ‘conventions’ and ‘arrangements’ [20]. An ‘arrangement’ “is best understood as a harmonious set of elements in the cultivated environment and the practices triggered within that environment. It entails speech, attire, cars, buildings, unspoken expectations and mannerisms (…). In conjunction these elements form a kind of mold for or template of behavior” ([21], pp. 103–104). Arrangements are distinguished, in cultural psychology, from conventions. These are tacit or unwritten rules (broadly conceived) that are recognizable in the behavior of members of a given social group. Unlike with arrangements, we can make many of them explicit with some effort. Without wishing to ignore the differences between these concepts, the term ‘stylistic conventions’ as I will use it covers not only conventions in the sense outlined, but also arrangements. Many stylistic conventions are implicit and hard or even impossible to fully articulate.

2. Social Coordination Through Stylistic Conventions: Three Intuitive Accounts

A number of concepts developed in cultural anthropology, sociology and psychology highlight, mainly in a somewhat implicit way, the fact that stylistic conventions are used abundantly in social coordination. Let me mention three of these concepts as an illustration of the intuitive sense many researchers have of the coordinative function of stylistic conventions.

The first concept is ‘culture shock’, the often chronic affective reaction of alienation, powerlessness and depression experienced by people who spend more than a few months in a culture they were not brought up in [22,23,24]. Kalervo Oberg, who popularized the term (originally coined by anthropologist Cora Dubois), describes culture shock as a condition that results from losing all our familiar signs and symbols of social intercourse. These signs or cues include the thousand and one ways in which we orient ourselves to the situations of daily life: when to shake hands and what to say when we meet people, when and how to give tips, how to give orders to servants, how to make purchases, when to accept and when to refuse invitations, when to take statements seriously and when not. These cues, which may be words, gestures, facial expressions, customs or norms, are acquired by all of us over the course of growing up and are as much a part of our culture as the language we speak or the beliefs we accept. All of us depend for our peace of mind and our efficiency on hundreds of these cues, most of which we do not carry on the level of conscious awareness ([25], p. 177).

Culture shock involves much more than a mismatch in convention-based action coordination3. But it is interesting to note that in the first text on the topic, stylistic conventions play a key role. Social interactions such as ordering, tipping, purchasing, accepting invitations and greeting differ between cultures not only in terms of when they are performed and with whom, but crucially also in terms of how such interactions are conducted. The ‘signs, symbols and cues’ that are used vary significantly among cultures. They include gestures, facial expressions and the generic category of ‘customs’—which involves such things as dress and etiquette. These ‘signs, symbols and cues’ are stylistic conventions. Not knowing how to employ these will negatively affect our coordination with others, and that, Oberg suggests, contributes to feelings of alienation.

A similar intuitive acknowledgement of the coordinative role of stylistic conventions can be found in Pierre Bourdieu’s notion of ‘the habitus’ [29]. A habitus is a set of dispositions, attitudes and habits, both in behavior and in perception, that individuals develop through their socialization process. This process is highly dependent on social context. One’s habitus reflects one’s family, education, culture and broader social structures. A habitus encompasses both conscious and unconscious elements, but automatic, implicit preferences, habits, interactions, judgments and ways of perceiving things dominate. Think of the way in which a person’s upper-class upbringing reflects itself in demeanor, dress, accent and aesthetic preferences. Think also of the way in which specific items in the social and physical environment are salient to them while other items remain invisible. Academics, clergy, construction workers, etcetera each have their own set of stylistic conventions that are typical of their profession. Class, culture and profession are major factors that shape a person’s habitus.

Because class, culture and profession are important in shaping a habitus on the one hand, and because a habitus is largely automatic on the other, it helps to perpetuate and reinforce the social structures that have shaped it. For example, sharing the tastes, styles, accent, codes, mannerisms and norms of the upper class are preconditions for functioning well within the more privileged circles of society. Being brought up with the preferences, cultural background knowledge and codes of the academic world are a major contribution to a person’s academic success. A habitus can therefore endow some of us with ‘cultural capital’ that others lack [30]. As such, it is an important contributor to the reproduction of social inequality.

Stylistic conventions are an important part of the habitus, and the habitus plays a major role in social coordination. It is not only the case that one’s habitus allows one to interact smoothly with people who share one’s class, profession and/or culture. It also discloses a lot of relevant information about one’s background to others, required to assess the possibilities for interacting. Thus, the notion of the habitus fits well with the social coordination account of stylistic conventions implied by the idea of culture shock, but it adds a layer of subtlety to Oberg’s observations. A habitus contains codes and cues that are only recognizable to insiders, and that are often mainly processed unconsciously. Such codes and cues are examples of the social coordinative role of stylistic conventions.

As a third example, let me mention an interesting sub-branch of ecological psychology developed by Roger Barker, known as ‘behavior setting’ [31,32,33,34]. This example focusses on stylistic conventions pertaining to architecture and the design of public spaces. Offices, shops, classrooms, roads and pavements are examples of behavior settings: they provide opportunities for and constraints on specific kinds of behavior and by doing so serve a coordinative function. The fit between specific behavior and the environment it is conducted in—quietly reading in a library, swimming in a pool, teaching in a classroom—is called ‘synomorphy’. The pattern of behaviours that a given space is designed for is called its ‘setting program’. Behavior settings are highly context-dependent; economic, geographic, political and cultural contexts co-determine the kinds of actions that are elicited (a small pool in a museum of modern art, for example, will not invite swimming but must be interpreted as (part of) a work of art). In daily life, behavior settings provide an inconspicuous but hugely important scaffolding for coordinating the various interconnected tasks we carry out. Like the other examples, the idea of behavior settings highlights, implicitly, the coordinative use we make of stylistic conventions.

3. Lewis on Conventions and Coordination

Not everyone thinks that it is necessarily in the nature of conventions to serve coordinative purposes. But in what is arguably one of the most influential theories of convention, David Lewis claims exactly this. And yet the coordinative use of stylistic conventions implied by the accounts of Oberg, Bourdieu and Barker does not fit the mold of Lewis’ framework—or at least, not without further additions to it. I will argue that it is worthwhile to figure out which additions are required, for this will provide more clarity on the coordinative use of stylistic conventions that the intuitive accounts sketched in the previous section highlight. Moreover, it will underscore the importance of such conventions in complex collaborations such as whole societies. For this, we need to start with the basics of Lewis’ theory.

Lewis defines coordination problems as “situations of interdependent decision by two or more agents in which coincidence of interest predominates and in which there are two or more proper coordination equilibria. (…) [T]hey are situations in which, relative to some classification of actions, the agents have a common interest in all doing the same one of several alternative actions” ([35], p. 24). Suppose a phone call is interrupted due to loss of reception. Who calls back? If both sides start calling back immediately, they will not reach each other; nor will they reconnect if both sides wait for the other to call. This is a coordination problem, Lewis-style. Another example: Two persons want to have dinner together tonight. They each have different food preferences, but the preference for each other’s company is stronger. There is no communication. Each goes to the restaurant where she thinks the other will go to. Conventions solve such coordination problems. When it is a convention that the caller calls back (or the receiver of the phone call), or when the convention is to go to the only Italian restaurant in town, we coordinate.

Problems such as these can be analyzed in terms of simple game-theoretic matrices that plot the payoffs of all possible action combinations. In the phone call case, there is a positive pay off for both if one calls and the other waits; if both call or wait, the pay-off is zero. In the restaurant case, the pay-off for both participants is positive when both go to the same restaurant and zero when they choose a different one. It may be the case, though, that both coordinators have different preferences for restaurants—one for Greek, the other for Italian, say. In that case, when they do succeed in meeting at the Italian restaurant, the pay-off for one is higher than the pay-off for the other, even though none has a pay-off of zero. Once coordination is reached, however, there will be no reason for any of the players to change strategies unilaterally when the coordination game is repeated (given that there are no changes in preference), both in the restaurant case and in the telephone case. A convention can thus be modelled as a Nash equilibrium in a coordination game.

There is much more to Lewis’ theory, but this is the basic principle by means of which conventions are understood. There are certainly alternatives, but the elegance of this theory and the direct connection between conventions and coordination make it a prima facie attractive point of departure for attempting to get clearer on the coordinative use of stylistic conventions. On second glance, however, things do not look promising. Lewis’ theory has been criticized for not being able to accommodate a number of convention-types, precisely because these types do not—or not obviously—solve coordination problems as defined by his theory (i.e., in game-theoretical terms). The rules of games, in particular solitary games, or artistic conventions are frequently mentioned as examples. But crucially, social etiquette and dress codes are also mentioned.

As Andrei Marmor asks in a very insightful discussion of Lewis’ theory, what coordinative purpose can be served by the convention of holding our forks in our left hand? ([36], p. 364). It is clear that this is a convention—in some countries it does not matter in which hand the fork is held, and in many other countries forks are not even used at all. But the convention is hard to construe as a solution to a Lewis-style coordination problem. This does not mean that stylistic conventions do not figure in the many examples Lewis discusses. One such problem, for example, is about deciding what to wear to a party, given that one does not want to stand out as the only over- or underdressed person. But note that this coordination problem exists only against the background of pre-existing stylistic conventions about dress codes. The dress codes are required for the coordination problem to exist at all, but they are not discussed by Lewis as solutions to coordination problems.

Marmor seems to have a point, then. Examples such as holding one’s fork in one’s left hand are catchy and may move us to give up on the whole enterprise of trying to understand stylistic conventions in a Lewisian way. But against the background of the literature referred to in the previous section, we may want to be more careful. The fact that holding one’s fork in one’s left hand is part of the habitus of people in certain circles in society and not others, for instance, may point at a coordinative function that is overlooked by Marmor.

4. Correlation Devices—Primary and Secondary

One influential improvement on Lewis’ theory is the introduction of the idea of correlated equilibria [37,38]. Some equilibria can only be reached by means of some agreed on public signals, known as ‘correlation devices’.

Here is an instructive example ([39], pp. 44–51) based on what is known as a Hawk-Dove coordination game. Two nomadic tribes let their cattle graze on a patch of land. If both let their cattle graze, this will lead to a fight. That is the least desirable outcome for both parties. If they decide to avoid fighting by not letting their cattle graze, this is a slightly better outcome. If one tribe lets their cattle graze while the other does not, a fight is avoided too. This is the best outcome for one tribe; for the other, the outcome is as good as when both parties refrain from letting their cattle graze. This may seem like the optimal situation, if it were not for the fact that the arrangement is unfair. While in real life, unfair equilibria do exist and perpetuate, we may assume that in this fictitious example, the disadvantaged tribe will at some point decide to let their cattle graze too, which gets everybody back to square one. Then, due to a flood, the land is divided by a river that neither of the tribes can cross, and by accident, both end up on different sides of the river. At that stage, there is no coordination problem to be solved. But then, the river runs dry, leaving a mark on the land. This natural boundary allows for a set of strategies that were not available in the original grazing game: both tribes can choose to keep to the original ‘set up’, in which they let their cattle graze on one side of the boundary only. Thus, one tribe can keep to the rule ‘graze if North of the boundary, not-graze if South of the boundary’, while the other tribe can keep to the opposite rule. The correlated grazing game has a fair equilibrium; if both tribes settle on sticking to their implicit rules, there is little incentive for unilateral deviation. The dry river boundary is what is known as a ‘correlation device’.

Might stylistic conventions not also be understood as correlation devices? If so, they can be understood in a more or less Lewisian way. But there are still a few obstacles. One of these is the fact that the kinds of coordination that plays a role in the examples and theories of Oberg, Bourdieu and Barker are connected with the variety of roles that people play in daily life and with their differences. How does a host address their guests, a client a shopkeeper, or a pupil a teacher? How do we distinguish between those who do and those who do not know the implicit rules of the upper class, between real and would-be artists, or between religious and worldly authorities? Coordination problems such as in the case of the two tribes are different. They are not about social roles and distinctions, and they are solved by rules that apply to everyone equally.

A first step in tackling this problem is to distinguish between correlative and complementary coordination games. The situation with the two tribes and the situations Lewis discusses are instances of correlative coordination games. These are coordination games that can be solved by means of a convention that applies to every player in the same way. Complementary coordination problems, by contrast, can be solved when every player opts for a different course of action. Division of labor is a typical case in point. Suppose that players 1 and 2 must, together, perform two complementary tasks, A and B. Again, the outcomes of these players’ actions can be plotted in a matrix. Each player has three options: perform task A, perform task B or perform both tasks. Given that both tasks must be performed, the nine resulting options have pay-offs as follows. When both players perform the same single task, the outcome is 0 for both. When both players perform both tasks, the outcome is better for both (say, 2). But because none of the players can specialize, neither of the two tasks are performed optimally. When each specializes in a different task, both tasks are performed optimally, and the overall outcome is highest for both (say, 3). Division of labor is the equilibrium here.

Complementary coordination games involve different roles and tasks, like the situations referred to by Oberg, Bourdieu and Barker. But the problem now is that so far, there are no ‘signs, symbols and cues’ involved. Cailin O’ Conner has argued, however, that ‘markers’ such as sex, race and age function, crucially, as correlation devices that allow us to solve complementary coordination games such as division of labor [40]. The point is that although a complementary coordination game such as the above has an equilibrium when labor is divided, it is not clear yet whether player 1 should specialize in task A and player 2 in task B, or vice versa. When both players must choose their chore simultaneously without communicating, it helps to use a rule such as ‘the tallest person specializes in task 1 and the shortest person specializes in task 2’. While body length is used as a correlation device in this example, O’ Connor argues that similarly, and historically, easily observable markers such as sex, race and age have been used as correlation devices in more complex and real-life situations of dividing labor. This is the basis of what is referred to as gendered division of labor, which can be found in most human cultures, and which is often the basis for serious social inequality ([41], pp. 341–375; 493–507, [42]).

Interestingly, O’ Connor mentions that to amplify markers such as age, race and sex, most cultures use various additional markers. Many of these are alterable markers, such as dress codes, jewelry, demeanor, make-up and hairdos. But more permanent markers such as tattoos, piercing and other body ornamentations are also used. We might refer to race, age and sex as primary markers, and dress codes, etcetera, as secondary markers that both function as correlation devices—public signals that enable an equilibrium in a game-theoretical modelling of a coordination game. It is with the notion of secondary markers that stylistic conventions enter the debate on Lewis’ theory of conventions and coordination. Here, stylistic conventions are given a function in solving a Lewis-style complementary coordination problem that, like the examples by Oberg, Bourdieu and Barker, involves different roles and tasks.

5. Abstract Categories and Activating Markers

There is, however, one important difference between the coordinative function Oberg, Bourdieu and Barker ascribe to stylistic conventions and O ‘Connor’s example. The social categories in O’ Connor’s example are linked to observable markers; even without secondary markers, we can discern race, sex and age. The social categories involved in a person’s habitus, behavior-settings or in situations leading to culture shock, by contrast, are for the most part not observable without secondary markings—think of categories such as ‘employer’, ‘waiter’, ‘museum of modern art’, ‘upper class’. Other, more coarse-grained but instructive examples include categories such as ‘family’, ‘clan’ and ‘tribe’; categories that are used to structure all human societies up until roughly 1500 years ago and which still structure many (perhaps most) contemporary cultures [43,44]. Such social categories serve to divide roles and tasks just as well. But they are, crucially, not readily discernible. Rather than concrete categories that naturally come with primary markers, they are abstract categories that stand in need of secondary markers, not as optional extras, but as means to ‘bring these abstract categories to life’, so to speak. I will call such essential secondary correlation devices ‘activating markers’.

Here is a parallel example with a similar logical structure. Take the game of chess. We usually do not realize this, but playing a game of chess means using two separate sets of rules. The first set comprises what we think of as ‘the rules of chess’. These rules set out in terms of what pieces the game consists of, what moves each of these pieces are allowed to make, what the game board looks like, what the initial positions of the pieces are, when a game ends, etcetera. Very few people can play a game of chess using only these rules, for that would require them to keep track of which piece is which during the game without additional help. Perhaps some people can do this. But most of us are aided by a second set of rules that tells us what the different pieces look like—rules prescribing the styling of the pieces. That this second set of rules is indeed separate from the first set follows from the fact that apart from the traditional chess pieces, there are very many nonconventional games of chess. The same rules of chess can be used in games consisting of different beer glasses, fantasy games with dragons and elves, games using nuts and bolts and many more (a quick google search is instructive). Conversely, we can use traditional chess pieces and use them to devise a completely new game—i.e., not chess.

Being a member of a certain clan and the outward signs associated with this are related in the same way as the recognizable styling of a rook is associated with the rook’s prerogatives and duties in a chess game. The rook’s looks on the one hand, and the dress, demeanor, accent and ornaments of the clan member on the other, are activating markers that bring a certain chess piece and a social category to life. Like with chess pieces, the activating markers of social categories are only contingently connected with them. Being a member of clan C may involve different looks, demeanor and accent than it did 100 years ago. The same goes for being a member of the upper classes in most societies: the social category still applies, but the activating markers are subject to change.

Social categories like clans or classes are relatively coarse-grained. What I want to argue is that many more fine-grained social categories we use to coordinate the complex division of labor in human societies have a similar structure in which abstract categories are brought to life by means of contingently related stylistic conventions that serve (more or less—this is an idealization) as correlation devices. I will briefly discuss four types of social category that have this structure: (i) roles, (ii) transactions, (iii) locations and (iv) occasions.

(i) The first type of category—roles—should be taken to include professions (e.g., farmer, shaman, chief), functions (e.g., the leader of a group of hunters or the captain of a sports team) and statuses (e.g., being married or being an adult). Roles are associated with sets of tasks, but also with rights and obligations. In the example of the complementary coordination game of the previous section, permanent markers serve as a coordination device to determine who does what. Knowing which tasks others have taken on may help to determine one’s own tasks in situations, and for that, a rule such as ‘task A is done by seniors’ comes in handy. But in more complex situations, when roles have already been divided or taken on, coordination also requires that we have sufficient ideas about who does what. Dividing roles makes sense when roles complement each other. The execution of complementing roles must, inevitably, involve exchanging goods, information or services with people carrying out other roles. Easy recognition of roles through marking will thus facilitate coordination. Clothing is a salient form of such marking. A few examples from Western cultures should suffice. The clearest example is probably the use of uniforms that allow us to immediately recognize policemen, doctors, garage personnel, janitors, builders, military, firemen, school children in some countries, etcetera. But we can also think of the typical styles of dress that are associated with certain professions. These are usually not as standardized as uniforms, but in most cases, we have no problem telling a rock guitarist from a corporate lawyer or a butcher.

Children are socialized from very early on to be susceptible to such role-marking [45]. When we grow up, we are quickly able to discern the dress codes that exist within specific sub-groups in society. Think of the subtle cues that are used by teenagers to signal membership (or desired membership) of various sub-cultures. And within sub-cultures, even more subtle cues are used to signal hierarchy or dominance positions. It is very hard to think of any style of clothing that does not codify roles within groups in any way. Even such everyday garments as jeans and t-shirts will send messages in different situations: at a street barbecue, they will signal blending in and being part of the neighborhood, but at formal occasions, they will signal playing the role of a rebel or someone who is deliberately aiming to offend.

Apart from clothing, there are other conventional role-markers. For the trained perceiver, the use of gestures, posture, table manners, vocabulary, accents and in general our elaborate social etiquette can reveal a host of information about social status, upbringing and cultural roles. Many conventional cues that reveal roles are context dependent. Think of such things as the spatial location of people. A person standing in front of a classroom is probably a teacher, a person heading a military parade is a higher-ranking officer, a person standing behind a shop counter is personnel, a person in the middle of a party is probably an influential figure in that specific social group, etcetera.

(ii) I will use the term ‘transaction’ in a broader sense than in cases of buying and selling items. But such cases are useful to explain the concept. When one person buys something from another, both agree that there is a situation before and after the transaction that is, in relevant respects, different: before the transaction, person A was in possession of some item and person B was in possession of a sum of money; after the transaction, B owns the item and A the money. The transaction is the moment at which the situation changes. Understood in this way, our everyday lives contain many transactions that function like abstract categories for social coordination. Openings and closings of meetings are cases in point. Shaking hands, bowing, protruding one’s tongue, curtseying or putting one’s hands together with the fingers facing upwards are examples of public signals used to mark such transactions. When used as openings, such signals mark a special kind of transition in a ‘public space’: after shaking hands (say), we have opened a social space between ‘us’, an entre nous ([46], p. 266) that was not there before. In it, we feel obliged to attend to what the other is saying and doing and to respond appropriately. Shaking hands marks a mutual understanding to confer. The same gesture can be used to close this social space or dissolve the ‘us’.

Another standard example of a transaction is a (non-verbal) agreement. Again, this notion can be taken in a broad sense to include such instances as raising one’s glass in a toast, signing a contract to clinch a deal or giving a high five as a sign of understanding a joke or affirming a particularly good action in a sports match. Yet, another kind of transaction is the conferring of statuses. Think of handing over a diploma in a ceremony or celebrating a birthday. Activating markers, that is, stylistic conventions associated with such events, make the conferring of statuses public knowledge.

(iii-iv) Roles and tasks in a society with a complex division of labor interact in complex ways. For most roles, it is necessary to exchange with other roles (to keep things simple: a baker must interact with a miller to buy flour and with customers to sell bread; a miller must interact with a baker to sell flour and with a farmer to buy grain). Role-exchanges are coordinated by letting them occur at specific locations and occasions. Locations for specific transactions are often marked through architectural conventions. We have no trouble distinguishing office buildings from gas stations, supermarkets, temples (churches, mosques), houses and train stations. Occasions are often marked only by times, rather than stylistic conventions, but celebrations, religious services, official meetings and gatherings are among the exceptions. There, flags, ornaments, special dress codes, etcetera also serve as secondary ‘activating’ markers.

6. Complex Role-Division and Augmented Reality

I do not claim that all abstract social categories that function in the organization of societies always require stylistic activating markers. In many cases, we can infer people’s role or societal sub-group, for example, from the choices they make [47]. The claim I wish to make is that stylistic markers dramatically diminish the need for such inferences in daily life. By doing so, they allow us to wield very many social categories more or less simultaneously—just like they do in a game of chess. This, I claim, makes them inconspicuous but likely preconditions for complex division of roles in the economic division of labor and in cultural institutions.

The benefits of division of labor are hard to overestimate and widely acknowledged since Adam Smith’s Wealth of Nations. The same applies to role-divisions in institutional contexts. But the cognitive costs of role-divisions are not often mentioned4. These include the increased demand for coordination and signaling, and the transaction costs and investment in social knowledge involved [50,51]. A finer grained division of labor implies more cognitive effort aimed at coordination. This is demonstrated, for example, by a study that showed that a finer grained division of labor in companies leads to an increased demand for clerical office workers who process information relevant for the coordination of tasks [52]. In another study, it was shown that stimulating trust among workers—which lowers transaction costs because there is less need to assess the good or bad intentions of co-workers—tends to lead to increased division of labor [53]. What such studies show is that the extent to which humans can divide labor is a function of the availability of cognitive resources (offloaded or not) aimed at coordination. I want to argue that stylistic markers are a very important but overlooked cognitive resource of this kind.

In daily life, social structure—that is, the structure of divided roles, including professions and social statuses—is present and accessible to us to a large extent via stylistic marking. To describe the nature of such access, it is useful to draw a parallel with the way Edwin Hutchins describes social practices in what he calls ‘cognitive ecosystems’. Here is his description of the practice of queueing:

Seeing a line as a queue is an example of the mapping of a conceptual structure, what in cognitive grammar is called a trajector (…), onto a physical array. This mapping of imagined structure onto perceived structure produces a conceptual blend (…) which gives rise to a particular emergent property: a sequential ordering of the bodies of the individuals in the queue. (…) Seeing the line as a queue is a cognitive practice because it makes possible a set of inferences. Who is next in line? Who arrived before whom? How far am I in space from (and how long must I wait before) getting service?

([54], p. 39)

A similar kind of mapping of an imagined structure onto a perceived structure is what happens in the case of stylistic conventions that serve as activating markers. The perceived structure consists of, say, people shaking hands, hammering on a table, wearing uniforms or clothes in a specific style, giving high fives, purposefully entering certain buildings while ignoring others, etcetera. The imagined structure consists of people greeting or taking leave, opening a formal meeting, executing their roles as doctors, police, lawyers, neighbors at a barbecue, friends at a weekend out, acknowledging the wittiness of a remark or the cleverness of an action in a sports game, going to the office, ignoring offices of companies or institutions that one is not involved with, coming home, etcetera, etcetera.

The word ‘imagined’ is probably not entirely right, but it is useful. It portrays the perception of social structure in one’s cultural environment (including the behavior of others) as a kind of augmented reality. Nowadays, we can look through our smartphones and tablets at, say, a city scene and see the whereabouts of shops, restaurants and other services projected onto the visual array. Or we can peer through such a device at a large storage hall and see, projected onto the picture, which items are on which shelf in which aisle. Stylistic conventions function as a kind of augmented reality device: by looking ‘through’ them, we discern the structural organization of the society we live in. Just like our smartphone can project a layer of information over a picture of our surroundings, the ornamentation, decoration and conventionalization of behavior project a layer of information about possibilities and impossibilities for action and social interaction onto the physical array by marking social roles, occupations, statuses, transactions etcetera. I take Bourdieu and Barker to be talking about this ‘layer’.

We may, to use another metaphor, refer to this layer as our ‘cultural umwelt’. Between 1909 and 1940, the German biologist Jacob von Uexküll developed the notion of umwelt as part of a theoretical approach to the biology of animals [55,56,57]. To understand animal behavior, he argued, we must understand how animals perceive the world around them and are incentivized to act in it. The world as experienced by an animal is determined by the sensory apparatus and the bodily features of each animal. This is what he referred to as that animal’s umwelt. An umwelt contains exactly those elements of the physical environment that are relevant to an animal’s action possibilities. The same physical environment is thus a different umwelt for animals that differ in their bodily make-up. Humans do not differ in their bodily make-up; they differ in their familiarity with rather diverse sets of stylistic conventions. Given that in many instances, these conventions mark social categories that are relevant for social interaction, this difference leads to variation in the perception of possibilities for action and interaction just as well. Stylistic conventions as activating markers co-determine our cultural umwelt. The same physical environment can be a different cultural umwelt to people from different cultures. This is what Oberg describes.

7. Evolutionary Adaptations

A number of evolutionary adaptations that have been highlighted as contributing to human abilities to collaborate in large groups can be connected with the theory of the coordinative function of stylistic conventions sketched in the previous sections—which I will label the ‘activating marker hypothesis’ (AMH). Let me discuss three of these: (i) conformity bias, (ii) rule-sensitivity, and (iii) overimitation.

Conformity bias is the tendency to adopt those beliefs, customs, styles, norms and practices that dominate in one’s cultural environment. The more widely shared a custom or practice is, the more likely conformity biased-people will tend to copy these [10,58]. Humans generally have a good deal of conformity bias. The evolutionary explanation for this is that in many situations where groups have found good ways to survive and thrive in their specific environment, the most adaptive behaviours can be found in the majority of group members (if not, the group will not do well and will, in the long run, either vanish or adapt). Thus, a conformist tendency will (ceteris paribus) be good for individuals. If this tendency is genetically or culturally heritable, it will be favored by natural selection.

This explanation pertains specifically to knowledge and skills that are tailor-made for specific environments, but not to stylistic conventions. And yet, we do see at least as much stylistic conformity as conformity in the realm of knowledge and skills (I do not have the space to prove this point, but I believe it is obvious to anyone who pays attention to fashion and style, architecture, art, etcetera. If not, it is instructive to observe high school dress codes and the concomitant peer pressure). The existence of conformity bias in the realm of style can be given a similar explanation as above when we take seriously the coordinative role of stylistic conventions, as outlined in the previous two sections. Groups with many individuals that tend to align their dress codes, etiquette, architecture, etcetera will be able to coordinate their roles and tasks more efficiently than groups in which there are less or no stylistic conventions. Cultural group-level selection will hence favor conformity bias in individuals.

The idea of group-level selection has a somewhat problematic history. Darwin considered and rejected the idea that pro-social traits may be the result of group-level selection because these would produce more successful groups. For most of the second half of the previous century, the idea of kin-selection and the idea of strategic reciprocity were advocated as alternatives that could avoid the problem of explaining the within-group advantage of egoistic behavior [59,60,61]. Later on, multilevel selection theory reinstated group-level selection [3,62,63,64,65]. With sufficient between-group differences and sufficiently low within-group variety, between-group selection pressure may outcompete within-group selection pressure. The standard counterargument that such conditions are unlikely to obtain does not apply to the case of cultural multi-level selection, precisely because culture strongly enhances between-group variety and within-group similarity [66,67,68,69,70].

(ii) Rule-sensitivity is the tendency of human infants not only to adopt social rules extremely quickly when they are instructed (e.g., to play a pointless game in a specific way), but also, crucially, to enforce these rules in their peer group [71,72,73]. Such enforcement also occurs when infractions have no bearing on the enforcer—like instances of third-party punishment. This tendency is stronger toward in-group rule-breakers than toward out-group rule-breakers [74]. Rule-sensitivity is plausibly explained as an adaptation that allows for efficient cooperation and coordination [1]. Although such explanations do not specifically mention stylistic conventions, let alone their coordinative function, the theory outlined above can be plugged into these accounts without problems.

(iii) Overimitation is the tendency, specific to human infants, to imitate not just behavior with clear instrumental value, but also behavior whose purpose is opaque to those who copy it [75]. In the various experiments that have clearly demonstrated this tendency (in humans, but not in e.g., chimps), these non-instrumental behaviours have a ritualistic flavor. Explanations for overimitation vary. On the one hand, there are explanations that consider overimitation to be a spill-over effect of the otherwise clearly beneficial tendency of infants to copy adults [76,77]. On the other hand, there are explanations in terms of social communication. Imitation, even (or perhaps especially) imitation of nonsensical behavior, can be intended as a signal of the imitator that they belong to the same group as the model [78,79].

In the first paper reporting on overimitation, the authors explicitly mentioned the possibility that the phenomenon might serve to transmit cultural conventions ([73], p. 164). But in the absence of a clear function of such conventions that can explain selection pressure, this explanation was dropped in subsequent discussions. The theory of the coordinative role of stylistic conventions outlined in this paper may provide such a function and reinstate the original proposal. This option is attractive because in some overimitation experiments, children only copy the causally irrelevant bits of behavior when the demonstrator is present [80], while in other experiments, children stop imitating non-instrumental behavior when they see a few potential models not exhibiting the non-instrumental behavior that an initial model exhibited [81]. The first phenomenon does not sit well with the explanation of overimitation as a spill-over effect of the generally beneficial nature of imitation in general (why does copying the non-instrumental behavior cease to be rational when the experimenter is not present?). The second phenomenon does not sit well with the idea that overimitation serves a social-communicative function (what do potential models have to do with the group-signaling between an infant and an experimenter?). But crucially, both phenomena can be expected when overimitation evolved because it promotes stylistic conventionality for coordinative purposes: when copying non-instrumental behavior is meant for learning stylistic conventions, such copying is pointless when the experimenter is not present. And when other potential models fail to behave as the experimenter, it is rational for the infant to conclude that the non-instrumental behavior is not a convention—so that they can stop copying.

These three examples show that it is plausible that humans have evolved a set of psychological tendencies that facilitate the spread of stylistic conventions in collaborating groups, and that the coordinative function of these conventions, as outlined above, explains the selection pressure for these tendencies.

8. Activating Markers Versus Ethnic Markers

How does AMH compare to the widely accepted idea that people are highly sensitive to so-called ‘ethnic markers’? It is a well-established idea that one function of stylistic conventions such as dress, accent, and social etiquette is to distinguish in-group from out-group [1,43,82]. This function is supported by experimental and modelling evidence [83,84,85]. Some researchers even hypothesize that humans have an evolved cognitive system specifically aimed at recognizing in-group members (an ethnic concepts acquisition device (ECAD), modelled on Chomsky’s language acquisition device [86,87,88]. So how does the function of stylistic conventions as ethnic markers compare and relate to their proposed function as activating markers?

For this we need to look at evolutionary explanations for the emergence of ethnic marking. Some researchers connect it with our general ‘coalitional psychology’, that is, the fact that we form groups for strategic reasons [89,90]. A more elaborate explanation, the one that underlies computer models that support the notion of ethnic marking, is this:

In-group marking serves two purposes. First, the ability to identify in-group members allows selective imitation. When cultural adaptation is rapid, the local population becomes a valuable source of information about what is adaptive in the local environment. It’s important to imitate locals and avoid learning from immigrants who bring ideas from elsewhere. Second, the ability to identify in-group members allows selective social interaction. (…) [R]apid cultural adaptation can preserve differences in moral norms between groups. Best to interact with people who share the same beliefs about what is right and wrong, what is fair, and what is valuable so as to avoid punishment and reap the rewards of social life.

([10], p. 212)

I will refer to this as the ethnic marking thesis (EMT). Modelling shows that ethnic marking emerges under three assumptions that all enjoy empirical support: (i) people with shared beliefs and shared norms interact more succesfully; (ii) people interact preferentially with people whom they share markers with; and (iii) people tend to imitate successful others.

A cursory comparison may depict EMT as a coarse-grained version of AMT with only one social category—the in-group—made recognizable by a large number of markers. But this does not sufficiently highlight a subtle but important difference. The social category of the in-group stands for shared norms and beliefs that are detectable through shared markers. This is what makes marking worthwhile, according to EMT, because interactions between people who share beliefs and norms are more successful. AMH is not about beliefs and norms as such, but more specifically about the detectability of roles, transactions, locations, and occasions (for specific transactions). It stresses the function of stylistic markers in enabling and facilitating complex role-divisions in which peoples’ tasks complement each other. Thus, like EMT, AMH is also about what makes interactions successful, and what makes a group of people interact and collaborate successfully. But it focusses specifically on success based on having a shared recognizable organizational structure—in economic division of labor and cultural institutions.

With this difference in mind, we can formulate AMH’s own set of assumptions that parallel those of EMT but are nevertheless different. Assumption (i)’ of AMH is that people whose roles interlock in an organized way interact and collaborate more successfully than people who divide labor on the fly or not at all. Assumption (ii)’ of AMH would be that people preferentially interact with those who perform complementary tasks in a complex role-division when roles, transactions, locations and occasions are easily detectable. Assumption (iii)’ is that people tend to adopt stylistic conventions as a matter of automatic imitation from a very young age onwards—in the style of the conformity bias, rule-sensitivity and overimitation discussed in the previous section. The prediction of AMH is that under these conditions, groups will develop a fine-grained set of highly specific markers for role, transactions, locations and occasions.

EMT and AMH are fully compatible and complementary. Moreover, stylistic markers for e.g., roles or transactions may also serve as ethnic markers. Even so, EMT and AMH are different theses with slightly different predictions. The main differences are about the amount and the saliency of the markers involved. Group marking requires less and more salient (i.e., less subtle) markers. This is noted, for example, by paleoanthropologists in a study comparing different kinds of Paleolithic body ornamentation. Ochre body ornamentations can easily be made highly visible and can hence be compared to the behavioral display used by other animals, such as chimpanzees and gorilla’s “using stereotypical physical attributes, movements and postures to increase their visual impact on conspecifics as potential mates, allies or adversaries” ([91], p. 45) These are the kinds of signs that are eminently suited for group-marking. The use of more subtle and fine-grained body ornamentations, such as the use of beads, however, is fundamentally different. Beads are small and “poorly suited to simple display” ([91], 2007, p. 46). And yet they are widely used by humans from the Upper Paleolithic onwards. Why?

As Kim Sterelny [92] notes, beads have a number of interesting properties: they can be standardized easily, they are durable, and they can be used in composition to express complex within-group signals, like symbols on an army uniform. They thus provide an ideal format for conveying information about “achievements, age, ancestry, caste, class, club, gender, marriage, status, occupation, parenthood, social position, religion, sex and wealth. (…) Such adornments (…) are lubricants that oil the tracks on which social interactions run, helping them go smoothly and without misunderstanding” ([93], p. 202). Hence, they can be used to serve a subtler and richer purpose than merely conveying group membership. As Sterelny notes: “Beads and similar low-amplitude, short range signals appear (…) as groups become more internally complex, more differentiated, and perhaps more hierarchical” ([92], p. 55).

In short, the kinds of markers needed to indicate profession, status, or availability for specific transactions are subtly different from the kinds of markers needed to convey the more generic message of belonging to a certain group. Activating markers are diverse, fine-grained, and ubiquitous given the degree of complexity with which most human societies divide roles. They need to be recognizable for potential role-dependent interaction partners only. Ethnic markers, by contrast, need not be very diverse or ubiquitous, but they must be easily recognizable by all group members. Given the fact that stylistic conventions are ubiquitous, diverse and finely grained in all human societies, my claim is that ATM is a plausible complement to EMT.

9. Conclusions

Different theories in the social sciences assign an intuitive coordinative role to stylistic conventions, such as etiquette, dress codes, and architectural conventions. The best-known theory connection conventions to coordination, David Lewis’ theory, however, does not seem able to account for these conventions. I have argued that with a few additions, it can. We can interpret stylistic conventions as secondary ‘activating’ markers that make abstract social categories tractable so that these can function as correlation devices in complementary coordination games. Stylistic conventions bring social categories to life in the same way agreed-on styling can bring chess pieces to life. This allows us to track social categories, like clan- or tribe-membership, or social class, but also categories like cultural roles or roles defined by economic division of labor, transactions in the socio-cultural sphere, as well as locations and occasions for such transactions. This coordinative function differs from the function of ‘group-marking’ that is usually assigned to stylistic conventions in the literature on the evolution of culture. It also portrays the evolution of a number of psychological tendencies that make humans particularly susceptible to stylistic conventions as adaptations, because the coordinative success of groups that employ stylistic conventions can explain selection pressure on these tendencies.