Introduction

1Amongst the many philosophical controversies that have predated the social sciences, there is certainly one that still has tremendous importance: I am referring to the debate between methodological individualism and methodological holism (Boudon, 1977, 1984; Bunge, 1996; Elster, 1982, 1989; Hechter, 1983; Ph. Pettit, 1993; Coleman, 1990; Homans, 1967; Moessinger, 1996; Van Parijs, 1985). This issue is of such importance that most modern social theories have been committed to one or the other of these positions. The reason for this interest is not just philosophical. The debate around methodological individualism also has important consequences for the way empirical research is conducted and for the kind of explanations and theories that are considered worth looking for.

2The discussion between methodological individualists and methodological holists is far from being new. On the contrary, from the very beginning of the discipline these two positions have been hotly defended by outstanding sociologists. Nowadays however, the growing interest raised by rational choice theory has given a new twist to this old debate. It could be said that the general landscape of mainstream social science has been deeply modified by this new revival of methodological individualism. On one hand, methodological individualism has greatly contributed to the understanding of a range of complex and difficult phenomena: electoral competition (Downs, 1957), the emergence of political coalitions (Riker, 1962), the organization of collective action (Olson, 1965; Hardin, 1982), the behavior of bureaucracies (Niskanen, 1971) and social mobility (Boudon, 1973) are some examples of paramount importance. On the other hand, whole theoretical approaches that were thought to be allergic to methodological individualism—like Marxism—have been critically re-analyzed in order to provide them with more rigorous microfoundations (Elster, 1985; Roemer, 1982).

3The first objective of this text is to clarify some aspects of the debate concerning methodological individualism and methodological holism. I will argue that methodological individualism could be interpreted as holding at once two theses. The first thesis—that I will call “weak thesis”—concerns the fact that social phenomena is the outcome of aggregated individual properties. The second thesis—that I will call “strong thesis”—is that high level order should be explained in terms of microscopic order.

4The second objective of this paper is to assess the claims made by methodological individualism. I will argue that high-level order need not be explained in term of low-level order—this means that methodological individualism is not an unavoidable explanatory framework for the social sciences. I will explore the interesting idea that specific kinds of social explanations—selectional explanations—do not require the individualization of common patterns of behavior at the individual level. By relying only on outcomes, rather than on the process by which these outcomes are produced, selectional explanations show that the search for individual commonalities can be avoided. My objective is to unravel the logic of this problem. It is important to observe that my claim is rather moderate: I will not claim that methodological individualism is conceptually defective, because it is not. I will simply argue that it cannot have the monopoly of sociological explanations.

I

5The first problem that we should tackle is definitional. Methodological individualism is a somewhat vague notion, admitting more than one interpretation. The argument that I will develop in this paper is strongly dependent on the definition of methodological individualism that is taken into account. Methodological individualism is committed to two theses. According to the weak thesis, all social outcomes are produced by the aggregation of individual properties. Societies are made out of individuals and social phenomena should be understood as the product of complex interactions taking place at the individual level. The weak thesis provides a way to deal with the issue of social complexity—the fact that macro-sociological items are a composite set of microscopic elements. Societies are composed of individuals—social phenomena are exhaustively composed of individual properties. This does not mean, however, that all social phenomena can be explained in terms of the behavior of individuals. In this paper I will not dispute this weak thesis.

6According to strong thesis, what is at stake is not the compositional nature of macro-phenomena, but its explanation. The main claim is that in order to account for macro-level regularities we need to detect the underlying individual mechanism that is governing them. High-level order is explained by identifying patterns of low-level individual order.

7It is important to clearly understand the difference between the strong and the weak thesis. The strong thesis not only assumes that high-level social phenomena are the result of low-level properties, but also holds that these low-level properties have some kind of organization. It is this microscopic organization, combined with aggregation, that accounts for high-level convergence. In a nutshell: while the weak thesis holds that social phenomena consists of individuals, the strong thesis claims that macroscopic order may be explained in terms of individual behavior. In both cases, macroscopic properties are dependent on microscopic ones, but, according to the strong thesis, this dependence is explanatory: high-level convergence can only be explained by identifying patterns of common behavior at the individual level.

8In both cases—weak and strong—aggregation is of paramount importance in the understanding of social phenomena. However, the strong thesis adds the claim that aggregation alone is unable to explain high-level order. Simply aggregating actions that have nothing in common, we will be unable to account for high-level convergence. For aggregative machines to deliver macroscopic order (the word “order” is crucial here), they should be fed with the right kind of inputs. According to the strong thesis, these inputs are individual properties that hold across agents.

9Let me explain this more precisely using two examples. The first example concerns the market of blood. In The Gift Relationship, Richard Titmuss attempted to explain some interesting features of this market (Titmuss, 1971). He has observed a surprising fact: the policy of giving economic retribution to blood donors was predated by unintended and undesirable macroscopic effects. The policy problem started because blood is always in short supply. There are basically two reasons for this: first, the difficulty of storing blood during long periods of time; second, the reduced number of donors. In order to increase the supply of blood, many countries—notably the United States—have decided to pay donors for their blood. According to Titmuss, the social consequences of this policy were catastrophic. The main upshot of this policy was that the quality of blood deteriorated drastically, increasing the impact of blood-transmitted infections. How can we explain this macroscopic fact? Why those countries implementing a monetary market for blood have poorer rates of non-infected blood samples than those countries which refused to do so?

10Titmuss identified two individual mechanisms that were capable of explaining this macroscopic feature. The first mechanism that seems to be at work here is that the fact of giving the donors money for their contribution launched a demotivational effect in those individuals who were guided by pro-social behavior. Most of the individuals motivated by altruistic behavior leave the market, increasing the relative number of donors motivated by purely economic gains. The second mechanism simply states that those individuals motivated by purely economic gains have strong incentives to keep hidden any possible transmissible sickness.

11It is very important to clarify how this interesting explanation is organized. These two mechanisms identify common reactions at the individual level that help to explain the observed high-level patterns. It is important to keep in mind that it is not just aggregation that is making the explanatory work here. Only when the aggregation process takes as input common patterns of individual behavior, are we able to explain high-level order. Given a specific external shock—payment for blood—individuals reacted in common ways, and by aggregating these individual actions we obtain the high-level properties that we want to account for. Aggregation alone would be explanatorily superfluous if not fed with the right common individual inputs.

12The second example that I would like to introduce in order to illustrate this feature of individualist explanations is Merton’s account of the emergence of science in Seventeenth-Century England (Merton, 1938). Following very closely in this aspect the pioneering work of Max Weber, Robert Merton explains the emergence of science as a result of a radical change in the “foci of interests” of the English intellectual élite. Puritanism played a powerful role in concentrating talent, that otherwise would have been dissipated, into scientific activity. The concentration process is achieved by providing individuals with a complex set of incentives:

“Natural philosophy was instrumental first, in establishing practical proofs of the scientist’s state of grace; second, in enlarging control of nature and third, in glorifying God. Science was enlisted in the service of individual, society and deity” (Merton, 2001, 85).

13Science was not only legitimized as a respectable activity, but also its practice was rewarded as a useful way to honor God’s handiwork.

14Observe that in this example the explanation depends not only in aggregating individual behavior. The explanation also makes the assumption that given a set of external incentives a sizeable number of talented individuals behaved in the same way: they invested time and effort in scientific activities. The result is the same as that in the case from Titmuss. This is, I think, a remarkable and elegant way to explain social phenomena. As we will see later, the main point that I am putting forward in this paper is that, although remarkable and interesting, this way of explaining social phenomena is not the only currently available.

15Once this characterization of methodological individualism is accepted, many issues hotly debated in mainstream social sciences become understandable. The central challenge for methodological individualism is how to make these individual commonalities visible. There are basically two ways to achieve this objective that have received special attention in the social sciences. The first one is to suppose that individuals have a common psychology. Having a common psychology will make individuals prone to react in the same way when confronted with the same set of external stimuli. Most discussions about the role of motivational assumptions in the social sciences, and especially in economics, are related to this problem; for instance, in the importance given to notions like rationality, self-interest, altruism, etc., in current discussion concerning the theoretical foundations of mainstream social sciences. All these motivational assumptions attempt to constrain dispersion at the individual level, making patterns of individual order more accessible to inspection. By describing individuals as similarly endowed, we make their behavior less variable. Creatures having common cognitive capacities are creatures that will behave less unpredictably, allowing patterns of individual order to both emerge and to be detected easier.

16Obviously this does not mean that methodological individualism is committed either to “hommo oeconomicus” or to any other specific motivational assumption. According to methodological individualism, we need underlying individual commonalities to have a successful explanation, but these underlying commonalities need not be grounded on specific motivational assumptions. Whatever motivational assumption contributes to putting order at the individual level will be welcomed. Methodological individualism is neutral between competing theories of motivation (Van Parijs, 1985). Pro-social behavior as well as anti-social or selfish behavior could be part of a successful individualist explanation. Motivations like envy, frustration or the sense of justice can be admitted into social explanations. Relative deprivation theory—just to mention a prominent example—has provided interesting examples where a richer set of motivational assumptions have been mobilized in order to provide microfoundations (Frank, 1985; Runciman, 1966; Frey, 1997). Recently, Brennan and Pettit have explored another interesting motivational factor—esteem—unjustly forgotten in mainstream social science; they have put it to work within a strongly individualistic framework (Brennan and Pettit, 2003; Rios, 2006).

17Assuming a common psychology however is not the only way to cause individual commonalities to emerge. There is an alternative way to achieve this objective: to reduce the margin of choice. Context is here the main tool at hand. Contextual constraints contribute to generate commonality of behavior at the individual level. The more constraining the context is, the more individuals will react to external shocks in regular, homogeneous ways. Methodological individualism does not rule out, as it is usually argued, contextual constraints. On the contrary, contextual constrains contribute to making individualistic explanations more predictive.

18In a recent splendid paper, Satz and Ferejohn have argued along similar lines. According to them, even neoclassical economic theory works better in a situation in which individual rational choice is severely limited by the context (Satz and Ferejohn, 1994). It is precisely the lack of strong contextual pressures that explains why the theory of voter behavior is weak in comparison with the theory of party behavior in electoral competitions. Parties are more limited in their available choices than voters (Satz and Ferejohn, 1994).

19Homogeneous individual reactions can then be obtained both through the assumption of a common psychology and through the introduction of a highly constrained choice leaving small room for decision making. Obviously both strategies need not be antithetic; they can rather be combined in very different fashions in empirical research.

20To summarize: methodological individualism developed an astringent conception of what a sociological explanation is. Some explanations will not be able to fit the bill. High level order should be explained by identifying low-level patterns of common reactions. Were these individual commonalities absent, high-level order would be impossible to explain. I will argue that this claim is unwarranted. There are important scientific explanations that do not require specifying the underlying individual commonalities in order to account for high-level order. These are not, however, black box explanations because they provide a mechanism—although not an individualist one.

II

21In the first part of the paper I have attempted to provide a charitable interpretation of what the main tenets of methodological individualism are. I have said that methodological individualism is committed to two ideas—aggregation and commonality of individual properties. In this chapter I will argue that the individualist project cannot be considered the only methodological strategy for the social sciences. Stated briefly—there are interesting explanations that could be given, without describing any underlying individual commonalities.

22But before proceeding I would like to be more precise in my criticism. I am not criticizing methodological individualism across the board. I am just arguing against its monopoly of the theoretical market: what I am claiming is that it has conceptualized as a mandatory requirement what is just a possibility. In a nutshell, whether a successful explanation of macroscopic order should be couched in terms of individual commonalities, is a contingent fact that depends on the specific properties of the problem that is at stake.

23Individualistic explanations are not just applied to the social sciences. In many other scientific disciplines, structures or states of affairs can be explained by identifying low-level commonalities. Take for instance an organism: the state of the organism is dependent on the internal behavior of its composing parts. Accounting for a disease amounts to finding out a common pattern of disarray at a low level of organization. Describing this common pattern is what science is looking for. The general idea underlying this position is that the inner structure of objects governs their high-level properties: how an object behaves is accounted for by its internal composition. In a way, this type of explanation is structural, because it is the internal organization of a given object which accounts for its superficial properties. Superficial properties are dependent on the low-level organization of reality: the macroscopic order is commanded—sometimes unintentionally commanded—by the organizational patterns located at the microscopic level.

24The interesting point that I would like to make is that selectional explanations do not follow this style of explanation. Selectional explanations do not explain by identifying low-level commonalities. The idea is that different low-level configurations can give rise to the same systematic high-level effects. Macroscopic order can be reached even in the absence of microscopic order. This is precisely where selectional explanations challenge methodological individualism: they provide a mechanism capable of accounting for macroscopic regularities without identifying low-level individual commonalities.

25Selectional explanations are structurally organized on two levels. First there is a variability requirement: this variability is a necessary condition because it is upon this that the selection process will operate. A completely homogeneous population will not be able to account for differential survival because all the elements being filtered have identical survival potential. Second there is a selectional mechanism. This mechanism could be discriminative in different degrees, ranging from very discriminative to less discriminative. These two conditions—variability and selection—are necessary conditions for a selectional process to take place. While variability allows for novelty, the selectional process sets the rules for conservation. Each selectional process will change the environment and hence will impose new pressures on the next generation. It is exactly this ongoing process that accounts for the progressive, though blind, modification of the population.

26Take as an example the struggle between gazelles and tigers. The starting point is a heterogeneous population of both gazelles and tigers. Individual members within this population have different abilities or skills to tackle an environmental problem—how to avoid predators (in the case of gazelles), and how to hunt prey (in the case of tigers). Some gazelles are faster than others, and some tigers are more skilful than others at catching fleeing gazelles. The problem for gazelles is how to escape from predators, and the problem for tigers is how to catch gazelles. Only those members that cope most efficiently with this problem will be represented in the next generation. As a consequence, the more efficient forms of escape behavior will win out over the less efficient ones. The same can be said concerning the hunting skills of tigers. In fact, this is an interesting case of co-evolution: changes in the skills of tigers put pressure on the gazelles; and changes in the ability of the gazelles to escape from tigers put pressure on the entire population of tigers.

27But observe that what is really relevant here is the production of an outcome—avoiding predators—rather than behavioral properties by which this outcome is achieved. As a matter of fact, it could happen that radically different behavioral tactics are possible to escape from predators—crying, flying, jumping, etc. None of these strategies is structurally fit in itself. Their fitness is dependent on the behavior of the predators. These escape strategies might have nothing in common, except the fact that they are successfully contributing to the gazelles not being eaten by the tigers. The only thing that matters is the survival differential of those members of the population that use one—rather than another—specific escape tactic. The explanatory work is being done by the commonalities of effects, not by the similarity of the underlying tactics for escaping from predators. Only outcomes count. This is the basic moral of selectional explanations—and this is precisely what puts them in stark contrast with the cannons of methodological individualism. Sameness of effect need not be accounted for in terms of sameness of structural properties, because, as we have shown, the same outcome could be produced by different—sometimes radically different—causal pathways. Whenever a selectional process is in place, objects having nothing in common can still be systematically retained, for regularly producing a given effect.

28This, I think, is a striking difference with individualistic explanations. Selectional explanations only take into account outcomes. They do not look for the causes of these outcomes, because there could be an indefinite number of causes producing the same given effect. In this sense selectional explanations are forward-looking explanations. Individualistic explanations, on the other hand, are looking for commonalities of structure: they explain high-level convergence as a consequence of these common structural low-level properties.

29Let me take an example from Sober to further clarify this point (Sober, 1984). Suppose that there is a multiple-choice examination to pass in order to gain access to a well known university. There are several candidates: some of them are successful and they are therefore admitted to the university; other candidates fail. Why is the university population made up of individuals who have passed their examination? One possible explanation is a selectional explanation: only those that have successfully passed the examination will be represented among the university population. But the important point that I want to stress is that those selectional explanations favor the production of certain outcomes, not the process by which they are obtained. The details of the individual process by which each one of the candidates obtained this result is simply irrelevant. Maria could have passed the examination thanks to her extraordinary intellectual ability; Peter, thanks to the patient help of his mother; and Paul just by chance. The case might be that there is not an underlying common individual mechanism that accounts for their successful performance. It might even be that the only property that these individuals share is that they all successfully passed the examination!

30In a nutshell: selectional explanations do not describe the construction of an outcome as the product of a common property of several individuals. In fact, in selectional explanations individual properties are irrelevant: this means that, whatever the individual structural properties, only specific types of outcomes will be selected for. Because many different sets of individual actions could produce the same outcome, the identification of them is explanatorily irrelevant.

31However it is important to stress that lacking an individual mechanism does not mean that there is no mechanism at all: it could be simply that the mechanism is not an individual mechanism but rather a selectional one. The important point is that selectional explanations provide a mechanism but this mechanism is not couched in terms of the behavior of the individual. It rather makes reference to the relational properties of the whole system. If I am on the right track, the unavailability of law-like regularities at the individual level does not mean that there are no law-like regularities at ecological levels. Individuals can arrive at the same end-point by absolutely different paths.

32Advocates of methodological individualism could counter this theory by arguing that selectional mechanisms are not really “true” mechanisms. I am skeptical about this argument. What does it really mean to have a “true” mechanism? If you stipulate that only “individual” mechanisms count as real mechanisms, then you have won, but your victory is a Pyrrhic victory. Having a mechanism is having, at least, a sketch of an elementary algorithm capable of systematically producing regular effects from a given set of inputs. This is, at least, the notion of mechanism that seems to be implicit in the writings of prominent advocates of methodological individualism like John Elster (Elster, 1982).

33An interesting example is the case of the theory of the firm in economics. Most orthodox analysis of the behavior of firms relies on profit maximization. Alchian has argued that the assumption of optimization on the part of firms is an unnecessary assumption (Alchian, 1950). Even if firms behave randomly rather than guided by the maximization of profit it will be possible to account for the decisions that we actually observe. Why? Because only those firms that behave in the right way will survive in the market.

34It is important to fully understand the meaning of this claim. The notion of “behaving in the right way” cannot be given in advance; it is a context-dependent notion. According to some advocates of the evolutionary theory of the firm, what is decisive is viability (Nelson and Winter, 1982). The selectional process takes place by rewarding whatever is a viable decision. But the viability of any decision depends on what other firms in the market are doing, not in general “a priori” assumptions about profit maximization. Wasteful, non-profit oriented behavior could be contingently selected for. Take the example of those complex protective devices that managers built up in order to avoid take-overs. These devices could have strong “survival” advantages, even if they are not profit-oriented. I am not claiming that this is certainly so; my claim is that it is an open empirical issue that could only be accounted for with fresh empirical data.

35It is interesting to observe that once a set of firms have been driven out of the market by this selectional process, new constraints on viability arise. This does not mean that in the long run only those firms that display profit-maximization behavior will be selected for. The key point is that by its very nature any selection process has an historical character (Simon, 1996). This means that the state of the system now is dependent on the contingent state of the system in the past. As the selectional process has no foresight, it operates mechanically over actual opportunities, rather than on global maxima (Jacob, 1977). This implies that the system could, but need not, attain global equilibrium. Whether or not profit maximization is the selected strategy will depend on the structural history of the system.

36This model has opened a fruitful research agenda in the sociology of organizations. Why are there so many (or so few) kinds of organizations? Traditional organizational theory sought to answer this question by individualizing the right strategy developed by existing firms for dealing with an environmental problem. An alternative explanation is to conceive the overrepresentation of certain kinds of organizations as the outcome of a selectional process (Hannan and Freeman, 1989). Hannan and Freeman sought to account for these general features of the population by identifying the selectional mechanism that, given a certain set of ecological constraints, increases the viability of certain kinds of organizations (Hannan and Freeman, 1989).

37This analysis is far from being an exoticism of organizational theory and the economic theory of the firm. There are certainly other examples that could fit the bill (Runciman, 1998). Cultural transmission, for instance, has been modeled by way of selectional processes (Boyd and Richerson, 1985). The theory of memes championed by Richard Dawkins has become a true industry in mainstream evolutionary social science (Dawkins, 1976; Dennett, 1995; Blackmore, 1999). The same could be said of the development of science (Popper, 1973). Scientific theories compete with each other in a struggle for existence: better theories spread through the population, while inferior ones disappear from the scientific marketplace. Different models have been explored in order to develop this insight (Popper, 1973; Hull, 1988).

38It is extremely important to understand clearly what I am claiming here. I am not saying that either the evolutionary theory of the firm or the ecological theory of organizations is a good theory. Maybe they are wrong. Whether the evolutionary theory of the firm or the neoclassical theory of the firm is the right one is an empirical issue that should be solved by empirical testing, rather than by philosophical argument. My argument only requires their possibility, not their truth.

39I am just pointing to an interesting conceptual issue: selectional explanations do not require a description of the underlying individual commonalities of a given sociological fact because they select whatever outcome is viable. As the same outcome could be implemented in different underlying supports, then selectional explanations are an alternative way of explaining social phenomena that does not require the identification of systematic commonalities at the individual level. Selectional explanations describe a mechanism—selection—that is not an individual mechanism, but an ecological one. The main insight that selectional explanations provide is that there is not a single path to success. Many different structural properties can be selected for by the same selectional mechanism. Viability is not tied to structural properties definable in advance (Hannan and Freeman, 1989).

III

40Before finishing this paper I would like to discuss some possible sceptic objections to the main line of the argument that has been defended here. Discussing these possible objections will help clarify more precisely the exact scope of the thesis that I have been advocating in this paper.

Conclusion

52In this paper I have analyzed the explanatory status of methodological individualism. In the first part I attempted to isolate two different thesis methodological individualism is committed to—the weak and the strong. According to weak thesis, social outcomes are made out of individual agents. According to the strong thesis, high-level convergence should by accounted for by identifying low-level individual commonalities. The strong thesis claims that aggregation alone will not deliver high-level convergence, unless it is fed by the right kind of input: common individual behavior. In this paper, I have not disputed the weak thesis of methodological individualism. The main focus has been on the strong thesis.

53In the second part of the paper I have attempted to make sense of the idea that selectional explanations could explain macroscopic order without relying on the existence of microscopic patterns of behavior. Selectional explanations do not rely on structural individual properties, but on outcomes: whatever behavior contributes to achieving a given outcome, will be selected for, even if it has nothing in common with the other items being filtered in the population. Individual commonalities can, but need not, be available in a selectional process.

54In the last part of the paper, I attempted to discuss some putative skeptic objections to the main claims of the paper. I have argued that although selectional explanations could be sometimes compatible with methodological individualism, it need not be so. The main line of my argument relied on the distinction between the weak and the strong thesis of methodological individualism. While selectional explanations require the validity of the weak thesis, they do not require the validity of the strong one. This means that the existence of low-level individual commonalities is not a necessary condition for selectional explanations to operate. A filtering process can operate validly even when populations are composed out of agents sharing no microscopically explanatory common properties.

55If the argument that I have developed is on the right track, some important theoretical and philosophical consequences could be drawn. The first one is that methodological individualism cannot claim to have the monopoly of sociological explanations; selectional explanations can also provide interesting and illuminating explanations. The second consequence is that the notion of mechanism is independent of methodological individualism: we may have mechanisms that are not couched in terms of common behavior of individuals. Selectional mechanisms, for instance, are a good example of non-individualistic mechanisms. This means that although it is true that having a mechanism (individual or not) is indispensable for opening the black box, it is false to think that these mechanisms should describe individual commonalities.