Methodological developments in the social sciences in general and political science in particular have in recent decades been dominated by two major trends. The first came as a reaction to the functionalist-structuralist paradigm prevailed in the 19605. It was argued that the emphasis on social and political systems and other unreduced wholes, such as social classes and social norms, was scientifically unsound, since these theories could neither produce empirically testable hypotheses, nor provide plausible explanations of how the assertedly beneficial functions came about. Instead, the critics stressed the importance of founding scientific theories in the social and political sphere on methodological individualism and the behavioral assumption of individual rationality. We should, according to this line of thought, be 'Bringing Men Back In' (Homans 1964). This methodological approach accounts for such important developments as the logic of collective formal social choice theory, game theory, public choice theory and various other rationalistic explanations of political and social phenomena.

However, beginning in the 1980s many social scientists came to feel that this kind of model was not enough. It was argued that the prevailing models and theories lacked an understanding of, or even totally disregarded, the independent role played by different institutions in society. In order to understand and explain why the individual actors acted as they did and what the aggregate consequences of their actions would be. we also had to include institutions such as constitutional constraints and property rights in our explanations. Moreover, we should also provide plausible explanations of the characteristics of such institutions themselves. In particular, it was

Resumé

This article gives a rationalistic, individualistic explanation of social norms. Two different explanatory models for aggregate social phenomena are presented. It is argued that the emergence of such norms can be given a kind of socio-cultural evolutionary explanation, with boundedly rational actors, in social interaction situations having the characteristics of community. Their maintenance in larger settings, however, require that they are internalized into behavioral habits. These results, it is argued, are relevant for several fields of political science.

argued that even many political scientists had forgotten that the most important institution of all, the state, could play an independent role. We should, in short, be "Bringing the State Back In' (Skocpol 1985). Articles and books with titles including terms like 'institutionaF or 'neo-institutional' are at the moment flooding the market in the disciplines of social science.

At this point, it is time to take a further step; it is time to close the circle. And this pertains to political science as much as it does to other social sciences. We can no longer study politics without at the same time examining social norms and other non-political, cultural phenomena work in 'civil society'. We should, to paraphrase the popular mottos quoted above, start 'Bringing Social Norms Back In. By this, however, I neither mean that we should give up the firmly established individualistic, rationalistic nor that we should no longer regard the state as a particular kind of actor or institution that plays an independent and important in the development of society. Rather, my argument here is that we should try to integrate these approaches with methodologically similar theories or models that take account of social norms, since these norms may be regarded as a sort of institution in their own right.

The general purpose of this article is thus methodological. The underlying tenet is that we should still use individualistic, rationalistic models and theories when explaining social norms, even though we may have to modify or extend them a bit. In particular, I shall argue that we have to adopt a kind of evolutionary model. To my knowledge, social norms have not been studied from such a rationalistic perspective before.

This approach to social norms and institutions corresponds to what certain neo-institutional economists have called the behavioral view of social institutions (Langlois 1986, 15-21; Schotter 1981, 9-12; 1986, 117-118). contrast to the predominate view among political scientists in which institutions are seen simply as a set of rules or structures that constrain the actors' behavior, institutions are here viewed as unplanned and unintended regularities of social behavior. A major inspiration for this approach is F. A. Hayek's analysis of 'spontaneous orders' (see e.g. Hayek 1973), as well as Carl Menger's distinction between those social institutions that have a 'pragmatic' origin and those that have an 'organic' origin. The former are 'the result of a common will directed toward their establishment (agreement, legislation, etc.)' - i.e. intended products, while the latter 'are the unintended result of human efforts aimed at attaining essentially individual goals' (Menger 1985/1883, 132-134).

The two views of institutions are, of course, not incompatible with one another. Institutions with a 'pragmatic', most often political, origin will in various ways affect the institutions with an 'organic' origin, and vice versa. That is also the main reason why political scientists should care about social norms. It should also be noticed, however, that behavioral institutions,

spontaneous orders or regularities in behavior, once they have evolved, can turn into structures or 'rules' themselves, and thus be regarded as constraints in later time periods. This may occur because they are hardened' internalization, social inertia and cultural transmission, or even through positive legislation.

The article is organized as follows. First, two different rationalistic approaches that may serve as the starting-point for explanations of behavioral in general are presented. In the subsequent section social norms are the object of such a rationalistic explantation. This analysis constitutes the major part of the article. Then in the final section the relevance of social norms for political science is briefly discussed.

Intendedly Rational Actors and Situational Logics

At the most general level, rationality simply means that the actors are capable of relating their actions to their desires or goals. This is usually taken to correspond to Max Weber's zweckrational, 'rational in the sense of employing appropriate means to a given end'. He distinguishes this from wertrational, 'rational in the sense that it is an attempt to realize some absolute value', and also from social action that is 'affective' or 'traditional' (Weber 1978, 28-29). Usually this type of rationality is interpreted to imply that the actors are global maximize_rs. This means that they must have perfect information in the sense that the problem-structure is clear, the alternatives are well-defined and the means-ends relationship is unambiguous. the actors are even assumed to have perfect information each other's preferences. This is the standard rationality concept from social choice theory.

However, in situations of uncertainty, rationality in this interpretation breaks down. Fortunately, there is an alternative. The actors can instead be assumed to be boundedly rational or adaptive, they act to increase the satisfaction of their desires in a way they believe to be appropriate or reasonable in given situations. This means that the actors are assumed to be intendedly rational, but only limitedly so. Their behavior 'is adaptive within the constraints imposed both by the external situation and by the capacities of the decision-maker' (Simon 1985, 294). They are 'satisficers' rather than maximizers.

It should be noted that in both interpretations of rationality it is commonly that the agents are selfish or egoistic, i.e. that their desires are primarily self-regarding, directed towards the actors' own well-being. This assumption shall be retained here. To do otherwise would be to assume away what should be explained, especially since my interest in the next section is to explain social norms.

Corresponding to these two concepts of rationality, it is possible to construct two different explanatory models for aggregate social phenomena. The basic argument in both cases is that it is primarily the characteristics of the situation or local environment within which the interdependent individual actors act that determine what kind of behavioral regularities will prevail. In other words, when the actors are acting in similar situations of strategic interdependence, or social interaction situations, they will tend to adjust their actions to each other in a way that, in aggregation, may give rise to or maintain different behavioral institutions. This means that different types of situational logics (Langlois 1986, 252; Popper 1957, 149; 1945/1966, 97; Ullmann-Margalit 1977, 3-4) will produce different types of behavioral patterns.

In the first kind of processes, which I shall call processes of aggregation, it is assumed that the actors are fully rational. This means that if there is a rational way to act in a certain social interaction situation, all the actors will act in the same way, and, thus, in aggregation produce some kind of spontaneous pattern. It is, in other words, the characteristics of the relevant situation, in terms of various constraints, structures, the number of other actors, etc., together with the individual actors' rationality that will determine kind of behavior that arises. For example, if the actors are interacting in the well-known n-person Prisoners' Dilemma type of social interaction situation, they will all defect.

In the second kind of processes, which I shall call evolutionary selection, the actors are assumed to act under structural uncertainty. They are, thus, only limitedly rational. The regularities in behavior are supposed to arise spontaneously because some kind of evolutionary mechanism 'selects' certain forms of behavior. Since the social interaction situations now are complex or uncertain, the actors are assumed to form beliefs about it that are preliminary or hypothetical, beliefs that are kept as long as they are reinforced or until a better belief is found. Also in this case, however, it will in the end be the characteristics of the environment that determine which behavioral regularities can arise and be maintained. Two such evolutionary - 'natural selection' and 'diffused reinforcement' - can be distinguished. Since these mechanisms are less developed and also, presumably, less well-known, it will be useful to elaborate them briefly.

The defining characteristic of natural selection¹ in the social realm is analogous to the biological one, i.e. that regularities in behavior are selected through the differential survival or reproduction of the entities or agents who follow regularities in behavior (cf. Van Parijs 1981, 59). Hence, it is restricted to actors who are capable of dying or surviving and reproducing themselves. In a modern world, at least if we restrict ourselves to Western democratic market economies, this most often rules out individual human beings as the basic entities in processes of natural selection.

However, organizational entities such as firms, religious or political sects, voluntary organizations, interest groups, clubs, and even political parties, can often literally be said to fight for their existence. They have to attract and compete for customers, members, donations, votes, etc., in order to survive. In this context reproduction occurs in the form of opening of new plants; branch shops, establishments or 'sister'-organizations which then share the basic behavioral characteristics of their parent organizations. Survival is affected, furthermore, by the rate of 'prereproductive mortality' (for example in the form of bankruptcies) which thus serve as an additional factor in the selection process (Trivers 1985, ch. I). It will help to eliminate error.

As for variation, it arises mainly in the form of new behaviors introduced by new entrants or competitors to the established organizations. But it may also take place as 'copying errors' in the transference of behavioral regularities to the organizations' 'offspring'. Still, as soon as the organizations adopted their particular regularity in behavior, it must largely be fixed or 'tied' in order to ensure the necessary constancy and inheritance. Examples of this could be the use of brand names and patents. In natural selection, in other words, I neither allow the strategies to be deliberately changed by actors once they have adopted them, nor that strategies can be imitated by other established actors. These assumptions may seem quite unrealistic, but they are necessary in order to make natural selection distinct from diffused reinforcement. And they can surely be fulfilled in situations of profound uncertainty.

For the present purposes, however, I do not, as it may be reasonable to do in biology (Elster 1983, 51; 1984, 9), rule out the use of indirect strategies, such as 'one step backwards and two steps forward', exemplified by saving and investment decisions. This is due to the basic assumption for evolutionary explanations, i.e. that the actors are boundedly rational and have reached limits on information. This does not make the behavior of the actors totally blind, random or myopic. It only makes it impossible for them to act in the globally maximizing manner assumed in the strict rational choice model. They can still act reasonably in a satisficing way. This implies that the variations will be close to the existing regularities in behavior, that is, they will largely be based on local information. The consequence of this is that the selection process, through the causal feedback of differential survival and reproduction of different variations, can at the most be said to be adaptive to the environment in a locally maximizing way.

Diffused reinforcement,² by comparison, is the additional selection mechanism is restricted to socio-cultural evolution. In contrast to natural selection, its defining characteristic is that the regularities of behavior no longer are selected through the selection of entities that follow such regularities, but rather directly within those entities (cf. Van Parijs 1981.

95). The behavioral regularities are, in other words, no longer assumed to be fixed. They can be changed if the agents learn how to do better. Hence, variations can also occur more frequently. The mechanism is thus restricted to actors who are capable of registering, feeling and learning, consciously or unconsciously. Individual human beings are therefore not ruled out in this case. Inheritance, or more accurately constancy, now occurs mainly as social inertia, deliberately taught tradition, internalization and inter-agent imitation. Imitation also plays, as we shall see, a decisive role in the diffusion of the behavioral regularities to the whole population.

It is evolutionary explanations of this kind that are most common in the social sciences.³ For example, Robert Sugden's The Economics of Rights, Co-operation and Welfare (1986) belongs to this category. So also does Adam Smith's classical work The Theory of Moral Sentiments, where he analyzes the emergence of social norms and moral behavior (Smith 1982/ 1759). Another example is Weber's (1978, ch. 7) explanation of the rise of Western capitalism. The calculative, rational and innovative business practices of the hard-working members of the small puritan, primarily Calvinst, sects were reinforced in terms of sales and profits. And the more relaxed and traditional behavioral habits of their competitors in time became completely unfavorable. In order to stay in business, the entrepreneurs other confessions had to imitate the behavioral regularities of their puritan rivals.

That cultural traits or regularities in behavior are diffused or spread through imitation from person to person or agent to agent is common to all evolutionary models of this kind.⁴ Yet views about reinforcement, or the selection process itself, are more diverse. The general definition I employ is the same as John Finley Scott's (substituting 'agent' for organism'):

reinforcement is any event or process possessing stimulus properties for a particular organism
which, when it follows the occurrence of a particular repeatable act, increases the rate at
which the act subsequently occurs. This corresponds to the more popular term "reward. . . .
'Negative' reinforcement, where the rate of the reinforced act is decreased rather than
increased, consists of withdrawing positive reinforcement or presenting negative ones. . . .
This corresponds to the common term 'punishment' (Scott 1971, 46).

Certain implications of this definition should be noted. First, the criterion of selection is no longer survival or reproductive capacity, but something like the 'perceived' satisfaction or utility of different variations (Van Parijs 1981, 97). Second, just as in natural selection, what matters for diffused reinforcement as a selection mechanism, and what provides the causal feedback, is the actual consequences that come after the action in question. With respect to the intentionality or 'rationality' of the process involved, however, the definition above can be given a stronger or weaker interpretation.

In the stronger interpretation, argued for by Philippe van Parijs, and

simply stipulated in a critique of reinforcement processes by Jon Elster (Elster 1989a, ch. IX; Van Parijs 1981, ch. 4), no intentionality should be involved, at least not in the initial stage. That is, the reinforcing consequences not be deliberately sought, in the case of positive reinforcement, deliberately avoided, in the case of negative reinforcement. This means that this selection mechanism could only work through some kind of operant conditioning, as studied by Skinnerian behaviorist psychology 1953).

Even though it is possible, and Van Parijs argues at length with examples from linguistics and anthropology, that this kind of process could explain important instances of socio-cultural evolution, a weaker interpretation of reinforcement seems a lot more plausible, and also more compatible with the rationalistic approach argued here. This is the view advocated, for instance, by Axelrod and Sugden, even though they do not use this terminology. it is argued that the agents, in situations of uncertainty, in a deliberate process of trial-and-error, 'learn by experience" (Axelrod 1984, 50; Sudgen 1986, 16 and 26). In other words, the boundedly rational actors try different behavioral regularities, and then stick to what seems to work best. Plausibly, some actors 'are quicker than others at recognizing patterns in their experience and at learning how to profit from this knowledge; but if a pattern exists, and if there are benefits to be had from recognizing it, we should expect a tendency over time for more and more people to recognize it' (Sugden 1986, 40). This is especially so if actors can imitate the ones who already have established a reinforced regularity in their behavior.

Just as for natural selection, diffused reinforcement can nevertheless only be expected to select regularities in behavior that are locally maximizing. Because of uncertainty, variations that occur through trial-and-error and may be reinforced by higher levels of satisfaction for the actors, cannot be very far off from the existing ones. There may be other forms of behavior that would be more efficient, but these may be too remote to be considered.

This situation also has implications for the possibility that the reinforced behavior will be imitated and diffused to the whole group of actors. In an empirical survey of such diffusion processes conducted by Everett M. Rogers, the main conclusions are that the diffusion of a new type of behavioral regularity is more rapid the larger the perceived (economic) advantage, the more compatible with other established behaviors and values, and the less complex and easier to observe (Rogers 1983. 238-240). conclusions have also been drawn from more theoretical considerations by Sugden, (1986, 42-43 and 49-50). Of particular importance the argument that certain types of behavioral regularities are more fertile than others because they can be spread to or reproduce themselves in new situations through analogy.

Communities and Social Norms

By 'community' is often meant particular kinds of sentiments or values - such as solidarity, fellowship or a sense of belonging - prevailing among a population of individuals (Miller 1989, 229-231). And it may well be, as argued by Robert Nisbet (1970), that human beings in general have in this sense a 'quest for community'. However, given an assumption about selfish motivations, it should be an explanatory issue if and how such feelings and corresponding behavioral regularities arise and may be maintained.

As an alternative, therefore, I shall define communities as a specific kind of environment, or social interaction situation, in which the agents act. Its most important characteristics are that the relations between a not too large group of actors are informal, intensive, direct, long-lasting and multidimensional Taylor 1982, 26-27; 1984, 185). Typical examples are the work-place, the neighborhood, the family, social networks and voluntary associations. Such communities fulfill numerous functions in every society: they make cooperation on an informal basis possible outside the market, they stimulate responsible behavior and, not the least, they provide, I shall argue, the foundation for the emergence and maintenance of social norms. It is this latter aspect on which our attention should be focused.

What then should be meant by 'social norms'? Broadly speaking, a norm is often regarded as a prescribed guide or rule for how to act in certain types of situations (Ullmann-Margalit 1977, 12; Vanberg 1988, 148). Moreover, norm-guided conduct, while benefiting others, is usually not considered to be in the immediate self-interest of the individual actor in each particular instance. The benefits to himself accrue only indirectly through the responses of the other actors (Ullmann-Margalit 1977, 13; Vanberg 1988, 159). It is this property together with the requirement that the norms are shared with other people which makes them 'social'. And as shall be discussed below, this implies or requires that if these norms are going to emerge and be upheld, then people are going to be punished or given sanctions if they do not follow them.

Some authors have made this last feature a defining characteristic of social norms. Scott, for example, simply defines norms as 'a pattern of sanctions' (Scott 1971, 72; see also Axelrod 1986, 1097), and Elster goes even further by including the psychological expectations resulting from such sanctions: he defines norms as 'the propensity to feel shame and to anticipate sanctions by others at the thought of behaving in a certain, forbidden way' (Elster 1989b, 105).

I also use a kind of 'positive' behavioral definition, but one which fits more closely into my previously developed framework. By social norms is here meant cooperative behavioral regularities within a population of actors in situations in which it would not always be in the actors' immediate self-

interest to cooperate. Typical examples of the kinds of norms I am thinking of are the behavioral regularities of not lying to other people (i.e. telling the truth or being honest), of doing one's fair share in common ventures, of keeping promises, of work ethics, of respecting other people's property, of not littering in the streets, of helping other people in need, of not hurting others physically, of not stealing, and so forth. These norms can often be expressed as negative rules that prescribe that people should not do certain things.

It should be emphasized, however, that according to this approach 'social norms' are not necessarily normative in any ethical sense. The relevant behavioral regularities may be both too broad or too narrow compared to some 'common sense' view about social norms. Moreover, for clarity, social norms should be distinguished from both 'legal' norms and 'personal' norms (even though each of them, as will be argued later, may play an indirect role for social norms). Legal norms are directly enforced by the state whereas personal or private norms - such as when to get up in the morning or not to consume too many calories - are self-imposed rules primarily directed towards one's own behavior in isolation to others.

The question now is whether social norms can arise from a normless
'state of nature'? Can any of our previously discussed explanatory models
for aggregate social phenomena provide a plausible explanation?

Processes of Aggregation

Among many sociologists and idealists the standard approach is to take norms as given, that is, they are treated as exogenous, they exist totally independently of the social interaction situation in which the actors act. As indicated in the preceding discussion, I do not agree with this. Rather, as argued by Scott, whose analysis I shall try to integrate with the methodological and assumptions presented, I purport that 'human behavior and morality are much more dependent on social environment than conventional wisdom and the naive phenomenology of moral men indicate' (Scott 1971, 207).⁵ I shall try to demonstrate this in three steps: first, the relevant social interaction situations will be presented; second, a plausible explanation, based on individual rationality, of the emergence of social norms will be discussed; and third, the conditions necessary for maintaining these norms in larger settings will be analyzed.

The fact that norm-guided behavior, while benefiting others, is not usually in the immediate self-interest of the individual actor in any particular instance (as in the case of the habit of not lying) suggests that there are important elements of public goods interactions involved in social norms. In other words, it seems as if it most often would be irrational for the individual to act according to the norms. This means that the relevant social

interaction situations are of the Prisoners' Dilemma or Assurance types. It
is in these situations, to a larger or lesser degree, that it is problematic to
establish cooperative behavioral regularities.

Certain 'solutions' to this problem have been suggested. Gordon Tullock, for example, has argued that in practice this conflict between individual self-interest and some kind of public good almost never occurs. Most importantly, since the individual has to be cautious about his 'reputation' as an honest and trustworthy man in order to reap future benefits in the course of 'continuous dealings' and cooperation with other people, it is actually in his self-interest to act cooperatively (Tullock 1985). These arguments carry considerable weight, but at the same time it cannot be a sufficient argument since in many situations, especially with fully rational actors, the conflict is real.

Edna Ullman-Margalit and Kenneth Arrow, by comparison, admit that the conflict in situations of the types mentioned above is real. And from this fact they infer that this is the reason for the emergence and maintenance of social norms - i.e. they occur to avoid the potential suboptimality in public goods interactions (Arrow 1971, 22; Ullmann-Margalit 1977, 60). Social norms exist because they serve a 'function'. That, however, cannot count as an explanation of social norms. Rather, it is a typical example of the functionalist fallacy, as discussed by Elster (1983, 57-60; cf. Hardin 1980). Similarly, Amartya Sen's argument that cooperative behavioral regularities will arise in Prisoners' Dilemma situations if everyone behaved 'as if they instead were in an assurance-type of situation and had the assurance of cooperative behavior of the other actors, equally begs the question for the same reason (Sen 1974, 60).

A third, more promising, proposal given by Victor Vanberg and James Buchanan is that we should distinguish between a person's 'constitutional' interest and his 'action' or 'compliance' interest, which occur at two different of choice. The former is reflected in the preferences for alternative 'rules of the game' for all the interdependent actors with which a person interacts, the other is 'reflected in preferences over potential alternative courses of action under given situational constraints, including constraints that pertain to a given structure of rules and institutions' (Vanberg & Buchanan 1988, 140; see also Vanberg 1986; 1988).

Social norms can certainly be regarded as being in a person's constitutional everyone would be better off if each actor conformed to cooperative behavioral regularities, particularly in Prisoners' Dilemma situations. The problem, which Vanberg and Buchanan are well aware of, however, is that this interest does not automatically correspond to each person's action interest. The best for the individual would be, for example, that everyone else was honest while he himself was dishonest, that he could

act as a free-rider. The crucial question then turns into when and how a
correspondence between these two interests can appear.

It should be noted here that the view of social norms as being in an individual's interest conflict with a common presumption about such norms. Elster, for example, has argued at length that norms cannot be outcomeoriented. contrast to intentional rational action which is directed towards the future consequence of an act, norm-guided behavior purportedly does not have this property. Elster's main argument in this case is 'the lack of a plausible mechanism that could explain how norms appear and disappear according to the expected payoffs associated with them' (Elster 1989b, 99). However, it is exactly such a mechanism that we are about to discuss.

Starting out with processes of aggregation, with fully rational actors, the major lesson from the by now quite extensive literature on the subject is that cooperative behavioral regularities cannot arise or be maintained in bilateral, or two-person, Prisoners' Dilemma types of social interaction situations unless certain quite demanding conditions are fulfilled. In particular, is required that the game is iterated an uncertainly large number of times, that the actors use conditionally cooperative strategies, such as the 'tit-for-tat', that they are certain that the other actor will also use this strategy, that they give a high present value to expected future benefits and that they immediately can identify a change in the other's behavior and react to it. (For extensive treatments, see e.g. Axelrod 1984; Hardin 1982; Taylor 1976; 1987.)

In essence these conditions imply that the situation has changed, typically into a bilateral Assurance game, where it actually, through the disappearance the free-rider problem, is rational to cooperate, at least as long as the two actors can assure each other that they will do so. Moreover, these possibilities for cooperative behavioral regularities to arise and be maintained are enhanced if communication is possible. Following Vanberg and Buchanan, I prefer to call the social norms that may arise in such bilateral interaction situations trust norms (Vanberg & Buchanan 1988, 147).⁶

In multilateral, or n-person Prisoners' Dilemma types of social interaction situations, however, the prospects for cooperation arising in processes of aggregation are even slimmer. Not only do the conditions noted above for the bilateral situation have to be fulfilled, but it is also necessary that the conditional cooperators are able to monitor the behavior of all other actors. Or alternatively, that a cluster of conditional and unconditional cooperators can be formed. All this implies that the size of the interacting group has to be quite small (Taylor 1987, 85-105). And such a size effect is also present, although less severely, in the multilateral assurance-type of social interaction Again following Vanberg and Buchanan, I shall call

the social norms that may arise and be maintained in such multilateral
interaction situations solidarity norms.

The interesting thing in both cases is that the conditions necessary for cooperative regularities to emerge in situations relevant to social norms correspond to those that are characteristic for communities, namely that the relations within a not too large group of actors are informal, intensive, direct, long-lasting and multi-dimensional.⁷ It is exactly in such environments demanding informational requirements, the 'shadow of the future' and monitoring possibilities regarding the behavior of other actors are likely to be fulfilled. Another way of putting all this is to say that the characteristics of communities enable social sanctions to be effective, that is, the threat or the actual carrying through of retaliation by also defecting if the other actors unilaterally break the cooperative regularities is in many cases sufficiently credible, or costly, to make such regularities rational.

Notice also that the 'tit-for-tat' strategy, the strategy of conditional cooperation, exactly corresponds to what Robert Trivers has called reciprocal which is a kind of combination of short-term altruism and long-term self-interest (Trivers 1971). Only in situations approximating community will systems of reciprocity be able to emerge.⁸ And at the limit, at least for trust norms, such a community may only consist of two people.

Apart from the reasons already mentioned, communities also enhance the credibility of such sanctions through multi-dimensional relations. The reason is that this opens up the possibility of using numerous different sorts of sanctions, perhaps relating only indirectly to the particular interaction in which someone has defected. For example, if we find out that a friend of ours has lied about something that was important to us, we may 'punish' him by not inviting him for dinner, saying bad things about him to other friends, by ceasing to call him, etc. We all know, and practice, lots of such subtle and unsubtle ways of sanctioning other people.

It may be questioned whether I really have been discussing social norms here. As I have defined them, social norms were cooperative behavioral regularities in situations in which it would not always be in the individual's self-interest to cooperate. Have I not, so far, basically argued that if individuals are interacting in communities it will often be in their selfinterest cooperate? I must admit that that is essentially true; processes of aggregation can only take us this far. For fully rational actors the constitutional and compliance interest will only coincide in quite specific and limited cases.⁹

Evolutionary Selection

If instead, then, the actors are acting under structural uncertainty and are

only boundedly rational, adaptional satisficers, what will happen? Can social norms be selected in some sort of evolutionary process? Beginning with natural selection, the only possibility, since I here have ruled out individual human beings as the unit of selection, is that some types of organizational entities deliberately adopt cooperative behavioral regularities prove to be successful, and, then are spread and reproduced through the opening of new organizations with the same behavioral characteristics their parent organizations. In my view, however, even these kinds of cooperative regularities would not count as social norms since it would still be in each individual's immediate interest to adhere to them because of the system of enforcement within the organization.

What about processes of diffused reinforcement? Could it be that social norms as cooperative regularities are reinforced and spread among the population of actors through inter-individual imitation, even into situations in which it would not always be in the individual's immediate self-interest to do so? I think this could be the case, provided that two conditions are fulfilled: first, that the situation in which the behavioral regularities initially become established have the characteristics of community; and second, that these regularities are internalized into behavioral habits.

As argued above, in communities it was sometimes even fully rational to cooperate if certain conditions, in particular the efficient use of social sanctions, were fulfilled. This means that if cooperative behavioral regularities be reinforced, they would give a higher utility or satisfaction to the actor(s). Most probably this first occurs deliberately, the individuals consciously iearn by experience' and seek reinforcement. However, through time these cooperative regularities may turn into habits. People stop calculating the pros and cons in every interaction. And even more importantly, other actors start imitating the behavior of the successful cooperators without really knowing why or how their behavior became reinforced in the first place.

Scott summarizes his analysis of this type of process in the following way:

The events involved in the process of moral commitment and social control can be viewed as a sequence and may be outlined as follows: first, an ad. a part of a behavioral repertory of persons and usually amenable to operant conditioning; second, reinforcements, mainly as sanctions (social reinforcements) but often including nonsocial primary reinforcers: thirdly, consequences in interaction, so that the act is a stimulus and reinforcer to others, whose subsequent activity in turn stimulates the original actor (Scott 1971. 154).

In this way social norms, such as being honest, of doing one's fair share in common ventures, of keeping promises, of work ethics, of not littering in the streets, of respecting other people's property, of helping other people in need, etc., all ultimately based on reciprocity, may emerge and be maintained even if it is not in the immediate self-interest of the individual actor in each particular instance. Through the bounded rationality of the

actors, their constitutional interest and their compliance interest tend to
coincide.

It is also this uncertainty which explains the importance of a person's 'reputation'. It would be absurd to think that the imperfect rationality of the actors meant that they did not try to use any information at all about their potential partners in different cooperative ventures. Rather, it is likely that information about a person's past performances, for example his 'trustworthiness', will serve as a substitute for the kind of information a fully rational actor would use. Likewise, this also shows how the personal norms of an individual and his conventional behavior indirectly may play a role for social norms. His character and style in general may, 'irrationally', serve as a 'signal' for how he would act in other settings (cf. Axelrod 1986).

It should be noted, moreover, that in this imitational learning process of boundedly rational actors, it seems likely for a number of reasons that social sanctions will be a lot more efficient than they are for fully rational actors. Not only is it possible for the individuals to learn by the evidence of the sanctioning of others, which implies that they may think that what happens to others may happen to them, too (Scott 1971, 55 and 109), but it is also likely that various 'symbolic' sanctions, provided by the use of language and human culture, will serve as reinforcers and influence the behavior of the actors (Scott 1971, 56-57 and 65). We may even envision that the belief in 'super-empirical' sanctions, such as a judgement day in a life after death,¹⁰ will influence the behavior of the actors.

Finally, if we admit (and this is the crucial step in my argument) that these social norms may be internalized into a person's personality, they can even be maintained in larger settings where the conditions necessary for their emergence are not fulfilled. The term 'internalization' implies that something comes from the outside of the personality to a place within it. Basically, following Scott (1971, 88), I choose to define 'internalization' as a propensity to conform to a behavioral regularity at a spatial or temporal remove from its reinforcing consequences. " In other words, it is a particular kind of learning. Recall that on several occasions I have said that regularities in behavior may turn into a habit, and this means that they have been internalized - they become part of a person's personality. And if this can happen to social norms, they can be maintained even in situations where the characteristics of community are no longer present.

There is evidence that supports this view. The formation of social norms, personal habits and even political attitudes to a large extent seems to occur at the early stages in a person's life and are then kept as behavioral habits later on in life.¹² This means that the conditions prevailing in the individual's childhood and adolescence will have important consequences for his or her future behavior. If these have the characteristics of community, she is likely to learn and internalize social norms, which she will continue to act in

accordance with, even when the relevant social interaction situation has changed. The individual may even eventually come to feel that there is something inherently 'right' about these norms, that they constitute some absolutevalue, and we may thus have explained both what Weber called 'traditionaP and 'value-rational' social action.

Nevertheless, as argued by Scott, and this is important, '(s)uch learning is ... never complete; it remains ultimately dependent on subsequent sanctions' (Scott 1971, 92). Some 'subsequent "maintenance dose" of reinforcement is necessary to sustain the learned rate of activity' (Scott 1971,101,183 and 179).¹³ And it is necessary here to introduce a distinction between those sanctions that derive from the fact that the norm-guided behavior has become established and those which do not. Following Van Parijs, I shall call the former 'superstructural' and the latter 'structural'.

Superstructural sanctions, such as the symbolic sanctions mentioned above, provided by language and culture, 'constitute a protective shield for a social practice. The expectation of them is essential in securing the agents' conformity to the practice in the short run' (Van Parijs 1981, 132). And the internalization of social norms will certainly enhance this conformity. Structural sanctions, however, are only made possible by the real underlying social interaction situation, and in the long run they are the only ones that may sustain and promote the emergence of social norms.

This means that if the social interaction situations change sufficiently, and the cooperative forms of behavior are not positively reinforced by structural sanctions, then the social norms will also in the end disappear. People are not complete cultural dopes. Through a cultural lag, however, this adjustment may take quite some time, even generations if the social norms are sufficiently internalized. The maintenance of both trust norms and solidarity norms will in particular be possible in larger settings if the communities overlap each other or are connected so that the individuals at the same time belong to several different communities. In this way a relatively large number of actors can indirectly belong to a sort of network of communities, in which the conditions for social norms to be upheld are quite good. This corresponds to a society which sometimes is said to be characterized by cross-cutting cleavages (cf. Hardin 1982, ch. 11). Individual and individual B in Figure 1 are thus likely to conform to the same types of norms, if they would encounter each other in some interaction, even though they do not know each other; they are just a 'friend of a friend of a friend . . .'.

In any event we should not rule out the possibility that there may be competing norm systems existing at the same time. If society is clearly divided along, for example, religious, ethnical, regional or class lines, without much contact between these groups, differing and even conflicting social norms may emerge and be maintained within a given political terri-

tory. In addition, we of course cannot expect any of the social norms to be 'optimal' in the globally maximizing sense, since boundedly rational actors can only be expected to adjust their behavior to existing behavioral regularities local information.

Notice also that the internalization of social norms is likely to have important consequences for the prospects of establishing new voluntary organizations. According to Mancur Olson's classical analysis, rational actors would normally not contribute to, or participate, in some common venture in which they all have an interest, since such collective action most often has the characteristics of public goods (Olson 1965). And this conclusion also remains intact if the problem is reinterpreted as a Prisoners' Dilemma social interaction situation. Nevertheless, what if the actors are not fully rational, if they are boundedly rational adaptional satisficers rather than optimal maximizers, and have internalized social norms while interacting in small communities? People do come from somewhere before they start thinking about new organizations. It is apparent that the prospects for voluntary organization to emerge even among larger groups then are enhanced. And especially so in the short run when the collective action is about to get started.

Before concluding this section, we may return to the arguments of Elster. The major reason why I think they go wrong and why he cannot find a positive explanation of norms is that he is too fixated upon 'optimality explanations' (see in particular Elster 1989b, 125). When he purports that there is no plausible mechanism that could explain how norms appear and disappear according to the expected payoffs associated with them, he

disregards evolutionary processes of diffused reinforcement with boundedly rational actors, who are adaptional satisficers rather that optimal maximizers. he does not see, moreover, what for my argument is an important distinction between the conditions necessary for the emergence of social norms - i.e. the characteristics of community - and the somewhat less demanding conditions necessary for their maintenance.

In addition, when Elster argues against the view that norms are socially useful, his two major arguments, apart from the asserted absence of a 'feedback mechanism that specifies how the good consequences of the norm contribute to its maintenance', are, first, that some norms are not socially useful and, second, that some norms, such as a norm to use public transportation crowded cities or a norm to have small families in developing countries, which really would be socially useful, do not exist. These are hardly sufficient arguments, however. From the perspective developed here it is quite possible that social norms may exist even though they are not socially useful. This may happen either because they have emerged in a different kind of society and now are kept through social inertia and internalization, or because people in situations of uncertainty tend to generalize, through analogy, too much and too soon, and adopt behavioral regularities which are somewhat arbitrary and even socially harmful. Moreover, is also clear that if the conditions of community are not fulfilled, then social norms will not exist, and particularly not solidarity norms of the kind that Elster is asking for in his examples about the use of public transportation and the need for family planning in poor countries.

To summarize, my conclusion is that social norms can in fact arise from a normless 'state of nature' as the unintended consequences of human action in a process primarily relying on the evolutionary mechanism of diffused reinforcement. The original emergence of such norms, however, requires that the environment in which the individuals act have the characteristics community, but social norms may also be maintained in larger settings if they are internalized into behavioral habits. The likelihood for this to happen is greater for trust norms than it is for solidarity norms, which could not arise unless the size of the interacting group was quite small.

The Relevance for Political Science

Political scientists, at least those working within the individualistic, rationalistic have not given social norms much attention in the past couple of decades. I think this is a mistake. The study of social norms, possibly based on the analysis given above, should be relevant for political science for at least four different reasons:

First, it would provide new perspectives in the discussion within liberal political philosophy about the limits of the state. At least since Hobbes, Locke and Rousseau, the appropriate role and size of the state have been analyzed with an eye on the possible existence of various self-regulating structures and behavioral patterns in the rest of society. If, as argued here, social norms in certain situations may diminish the notorious free-riding behavior, the size of the state should presumably, given most normative theories, be smaller, than if no such norms were to exist.

Second, in the study of public policy it would be of interest to investigate how different kinds of political interventions affect these social norms. For example, a policy that weakened the role of communities in society would probably in the longer run also have effects on both the maintenance and emergence of social norms. Social norms, in other words, could be a highly relevant dependent variable for public policy.

Third, social norms may also be regarded as an independent variable, i.e. as something that has an impact on the input-side of the political process. As argued, for instance, the existence of social norms would enhance the prospects of establishing new voluntary organizations, which in turn may affect demands on the public sectors.

Finally, social norms may surely arise within the political sphere itself.
This will have consequences for the study of parties, coalitions, coalitions
and decision rules in parliaments, etc.

Investigations along each of these paths, however, remain largely to be
undertaken.

---

NOTES

1. I have been trying, unsuccessfully, to find a better term than 'natural'.

2. Just as for natural selection, 1 have been trying in vain to find a better term than 'diffused reinforcement". As we shall see, it might have been better to call it something like 'imitational diffusion of reinforced behavior' or 'adaptional imitation'.

3. According to John Langton (1979, 297), 'The struggle for reinforcement is the central explanatory concept of the behavioral theory of sociocultural evolution'.

4. For some other examples, see Hayek 1967, 78; 1988, 23-28.

5. The most important differences between the present approach and Scott's are that I shall incorporate game theory, explicit behavioral assumptions and evolutionary models. Moreover, Scott lacks an understanding of the difference between personal norms and social norms, and he also, at certain points, tends to commit functionalist fallacies (e.g. p. 195).

6. Substituting 'norms' for 'rules'.

7. This is one of my major arguments in this section, largely inspired by Michael Taylor (1982), even though he, in my view, confuses the distinctions between conditions, behaviors and values. In addition, he does not see the, to my argument, importance of bounded rationality and some kind of evolutionary selection.

8. Cf. Alvin W. Gouldner (1960), who has argued that the norm of reciprocity is one of the universal 'principal components' of moral codes. See also Taylor (1982, 28), who makes reciprocity one of the defining characteristics of community.

9. It is unclear to me whether Vanberg and Buchanan would agree to this, but I think it is at this point that I part with them, even though they definitely hint in the direction that I will be going.

---

9. It is unclear to me whether Vanberg and Buchanan would agree to this, but I think it is at this point that I part with them, even though they definitely hint in the direction that I will be going.

10. Incidentally, this will really make the 'shadow of the future' infinitely long.

11. Most importantly I have substituted 'behavioral regularity' for 'norms'.

12. Apart from all the examples given in Scott (1971), see also the theories of moral development by Piaget (1965), which emphasizes the interaction of cognitive and emphatic factors in given social settings. Concerning the formation and internalization of political attitudes and in particular the influence of the views held by the parents, see Westholm (1991), for an utterly thorough empirical investigation.

13. See also Hayek (1967, 78), who argues that learnt rules 'will need some continuous outside pressure to secure that individuals will continue to observe them'.

REFERENCES

Arrow, K. 1971. 'Political and Economic Evaluation of Social Effects and Externalities", in
Intriligator, M., ed., Frontiers of Quantitative Economics. Amsterdam: North-Holland.

Axelrod, R. 1984. The Evolution of Cooperation. New York: Basic Books.

Axelrod, R. 1986. 'An Evolutionary Approach to Norms'. American Political Science Review
80, 1095-1111.

Elster, J. 1983. Explaining Technical Change. Cambridge: Cambridge University Press.

Elster, J. 1984. Ulysses and the Sirens: Studies in Rationality and Irrationality (revised ed.).
Cambridge: Cambridge University Press.

Elster, J. 1989a. Nuts and Bolts for the Social Sciences. Cambridge: Cambridge University
Press.

Elster, J. 1989b. The Cement of Society. Cambridge: Cambridge University Press.

Gouldner, A. W. 1960. 'The Norm of Reciprocity: A Preliminary Statement", American
Sociological Review 25, 161-178.

Hardin, R. 1980. 'The Emergence of Norms', Ethics 90. 575-587.

Hardin, R. 1982. Collective Action. Baltimore: Johns Hopkins University Press.

Hayek, F. A. 1967. Studies in Philosophy, Politics, and Economics. Chicago: The University
of Chicago Press.

Hayek, F. A. 1973. Law, Legislation and Liberty, Vol. 1, Rules and Order. London: Routledge
& Kegan Paul.

Hayek, F. A. 1988. The Fatal Conceit: The Errors of Socialism. London: Routledge.

Langlois. R. N. 1986. Economics as a Process: Essays in the New Institutional Economics.
Cambridge: Cambridge University Press.

Langton, J. 1979. 'Darwinism and the Behavioral Theory of Sociocultural Evolution: An
Analysis", American Journal of Sociology 85.

Menger. C. 1985/1883. Investigations into the Method of the Social Sciences with Special
Reference to Economics. New York: New York University Press.

Miller, D. 1989. Market, State und Communitx: Theoreticul Foundations of Market Socialism.
Oxford: Clarendon Press.

Nisbet, R. A. 1970. The Quest for Community (2nd ed.). New York: Oxford University Press.

Popper, K. R. 1945/1966. The Open Society and Its Enemies, Volume 11. The High Tide of
Prophecv: Hegel, Marx, and the Aftermath (fifth, revised ed.). London: Routledee & Kegan
Paul.

Rogers, E. M. 1983. Diffusion of Innovations. New York: The Free Press.

Schotter, A. 1981. The Economic Theory of Social Institutions. Cambridge: Cambridge
University Press.

Schotter, A. 1986. 'The Evolution of Rules'. in Langlois. R.. ed.. Economics as a Process:
Essays in the New Institutional Economics. Cambridge: Cambridge University Press.

Scott. J. F. 1971. lnternalization of Norms: A Sociological Theory of Moral Commitment.
Englewood Cliffs: Prentice Hall.

Sen, A. 1974. Choice, Ordering and Morality", in Korner, S., ed. Practical Reason. New
Haven: Yale University Press.

Simon. H. 1985. Human Nature in Politics: The Dialogue of Psychology with Political
Science". American Political Science Reoiew 79, 293-304.

Skinner. B. F. 1953. Science and Human Behavior. New York: The MacMillan Company.

Skocpol. T. 1985. 'Bringing the State Back In: Strategies of Analysis in Current Research", in
Reuschemeyer, E. & Skocpol, T., eds., Bringing the State Back In. Cambridge: Cambridge
University Press.

Smith. A. 1982 1759. The Theory of Moral Sentiments. Indianapolis: Liberty Classics.

Taylor. M. 1982. Community, Anarchy and Liberty. Cambridge: Cambridge University Press.

Taylor. M. 1984. 'Social Order Without the State', in Hofferman, T. & van Maarseveen, H.,
eds.. Law and Anarchism. Montreal: Black Rose Books.

Taylor. M. 1987. The Possibility of Cooperation. Cambridge: Cambridge University Press.

Trivers. R. 1971. 'The Evolution of Reciprocal Altruism", Quarlerlv Reoiew of Biology 46,
35-57.

Trivers. R. 1985. Social Evolution. Menlo Park: Benjamin/Cummings.

Tullock, G. 1985. 'Adam Smith and the Prisoners' Dilemma", Quarterly Journal of Economics
100, 1073-1081.

1 iirnuin-Margalit. E. 1977. The Emergence of Norms. Oxford: Clarendon Press.

Van Parijs, P. 1981. Evolutionary Explanations in the Social Sciences: An Emerging Paradigm.
Totowa. New Jersey: Rowman and Littlefield.

Vanherg. V. 1986. De Moribus Est Disputandum: Morality in Economic Explanations. Fairfax,
Va.: Center for Study of Public Choice, George Mason University.

Vanberg, V. 1988. "Rules and Choice in Economics and Sociology', Jahrbuch fur Neue
Polidsche Ekonomie. vol. V//Tubingen: Paul Siebeck.

Vanberg. V. & Buchanan, J. 1988. Rational Choice and Moral Order". Analyse & Kritik 10,
138-160.

Wcber. VI. 1978. Weber: Selections in Translation. Cambridge: Cambridge University Press.

Westholm. A. 1991. The Poiitical Heritage: Testing Theories of Family Socialization and
Generational Change. Uppsala: Department of Government, University of Uppsala. Ph.D.
thesis.