The emergence of social structure and the question of naturalism. Dave Elder-Vass

Transcription

1 The emergence of social structure and the question of naturalism Dave Elder-Vass Dave Elder-Vass School of Politics and Sociology Birkbeck College University of London Malet Street London WC1E 7HX Paper for BSA annual conference, York, March 2005, Realism study group session

2 The emergence of social structure and the question of naturalism Abstract This paper discusses two classic problems in the philosophy of the social sciences. First, the question of social structure: can we justify treating social structures as having real causal powers of their own, and if so how? And, second, the question of naturalism: to what extent do, or should, or can, the social sciences resemble the natural sciences? Taking a realist stance, the paper will approach the question by examining the differing ontologies of the natural and social worlds. Although both, it will argue, are founded upon emergence, the entities that emerge at different levels may have radically different natures. After discussing the general theory of emergence, the paper will discuss how this can be applied to social institutions, and go on to examine some of the ways in which emergent social structures may differ from natural structures as a consequence of how they emerge. The paper does not, however, attempt a complete account of the question of naturalism, as it does not address the important question of the meaningful nature of social action. Nevertheless, it concludes that while there is a unity of sorts between the natural and the social sciences, there are good reasons why social science is radically different and these will (naturally) shape its practice. Introduction This paper discusses two perennial problems in the philosophy of the social sciences 1. First, the question of social structure: can we justify treating social structures as having real causal powers of their own, and if so how? The foundation of my response is the principle that all entities with causal powers are emergent from their parts, and thus that their properties and potential powers arise from the nature of their structure. The paper will begin the construction of a social ontology based upon this principle, looking in particular at the emergent nature of organisations and other social institutions. This is an approach inspired by general systems theory, but a very different style of systems theory from the functionalist variety that has predominated (unsuccessfully) in sociology. This will lead us on to the question of naturalism: to what extent do, or should, or can, the social sciences resemble the natural sciences? Here, my founding principle is the belief that appropriate methods of study always depend on the nature of the object of study and what one wants to learn about it (Sayer, 1999, p. 19). To be more precise, there may be a variety of methods of study that can be fruitful in any field, depending upon the nature of the objects of study, and we may then select from this list the method(s) that are most appropriate to the particular purposes of our study. This paper will be concerned with the first part of this problem elucidating how the nature of the objects of study determines what methods of study are viable and applying it to the social. It is, I suggest, the particular ways in which their structures shape the 1 I would like to thank Jason Edwards, Filipo Artoni, and the participants in the UCL/Birkbeck politics research seminar for their valuable comments on an earlier draft of this paper. D. Elder-Vass 2

3 emergent properties of social entities, that determines what methods are viable for their study. The key question to be addressed in this second part of the paper, then, is how far can an emergentist approach to social objects illuminate the problem of naturalism? and not naturalism or anti-naturalism. Like others in the realist tradition, I will conclude that the answer to this latter question can only be neither and both. Bhaskar, in particular, has advocated a qualified anti-positivist naturalism (Bhaskar, 1998 p. 2), and both Archer and Benton have pointed out that his position could just as well be labelled a qualified anti-naturalism (Archer, 1998, p. 190; Benton, 1985, p. 176). Realism implies what we might call ontological naturalism that both the natural and the social sciences are concerned with different parts of the same stratified reality but not methodological naturalism, since the different structures of these different parts can have very different implications for their study. This, of course, is not a statement that applies uniquely to the difference between the social and the natural sciences. The natural sciences are by no means homogeneous in their methods, for just the same reason as has been identified above: the different natural sciences differ in their objects of study, and these objects differ in their structures and properties. There are as a result many criteria that differentiate between different kinds of science, and on some of these criteria some or all of the social sciences fall into the same group as at least some of the natural sciences. We might, for example, contrast the experimental and the historical sciences, in which case we would find geology and evolutionary biology in the latter group along with the social sciences (Benton, 1985, p. 188; Machlup, 1994, p. 6). There are other criteria that draw a line more neatly between the natural and the social sciences, but a thoroughgoing anti-naturalist would have to justify privileging such criteria over others that demonstrate the commonalities of some social and natural sciences. This argument alone is enough to demonstrate that we must find an intermediate position between naturalism and anti-naturalism. The significant question becomes, not are the social sciences different but rather how are the social sciences different and how are they similar? The paper will begin by expanding on my assertion that social objects are emergent from their parts, providing a brief overview of the theory of emergence and the sorts of emergent entities that can be considered social objects. Then it will go on to examine the relationship of emergence to organisations and institutions in particular, and conclude by showing how the particular forms of social structures contribute to a level of dynamism and complexity that makes it difficult or impossible to apply some of the methods of the natural sciences to them. The paper will not, however, provide a comprehensive treatment of the problem of naturalism; in particular, it does not address the concept dependence or inherently meaningful nature of social action. Like Bhaskar, for example, I accept that social action is indeed inherently meaningful, and the consequent need to adopt some of the methods of hermeneutics (Bhaskar, 1998, p. 38). Like Sayer, however, I do not believe that this prevents us from looking at social life in causal terms: the subjectobject interactions merely become more complicated, and the realist proposition of the intransitivity of social phenomena as objects of social research stands (Sayer, 1999, pp.33-4). Specifically, I assume that the meaningfulness of social action does not prevent us from identifying social actions clearly (although fallibly), and that it does not render causal explanations redundant or impossible. If this is accepted, then the omission of the meaningfulness issue from the rest of this paper poses no threat D. Elder-Vass 3

4 to its argument, although it remains incomplete as an account of the issue of naturalism. Emergence, cause, and causal analysis Emergence is the idea that a whole can have properties (or powers) that are not possessed by its parts or, to put it more rigorously, properties that would not be possessed by its parts if they were not organised as a group into the form of this particular kind of whole. Such properties are called emergent properties, and any entity that has one or more emergent properties is an emergent entity 2. Emergentism claims that everything in our world is, or is made up of, emergent entities, and also that all of these entities are part of a single unified hierarchy of wholes and parts. This section will briefly overview those parts of the theory of emergence that are essential to the treatment of naturalism in this paper, then outline the beginnings of an emergentist social ontology 3. Emergent properties, including the real causal powers of an entity 4, arise from the organisation of the entity s parts. This can be illustrated with the classic example of emergence: the case of water (an example which goes back at least as far as Mill s System of Logic (Mill, 1900, p. 243)). Many of the properties of water, such as being liquid at room temperature, or being able to put out fires, are clearly different from the properties of its constituent hydrogen and oxygen atoms (Mihata, 1997, p. 31). If these atoms were present but simply as atoms, or organised into molecules of other types than water, the resulting substance would not have the properties of water. It is the fact of being organised into the specific form of water molecules that gives this collection of hydrogen and oxygen atoms the particular properties of water. The same argument can be applied to any emergent entity, to demonstrate that its characteristic properties and powers depend both on the presence of its characteristic parts and on their being organised into the characteristic structure of the higher-level entity 5. The critical role of organisation as the source of emergent properties has been identified by authors in all the well-developed literatures on emergence, e.g., (Archer, 1982, p. 475; Bertalanffy, 1971, p. 54; Buckley, 1998, p. 36; Cilliers, 1998, p. 43; Cunningham, 2001, p. S68; Emmeche et al., 1997, p. 106; Holland, 1998, pp ; Lloyd Morgan, 1923, p. 64; Sawyer, 2001, p. 560; Smith, 1997, p. 5; Sperry, 1986, p. 266). The presence of the parts and the relations that structure them into the form of any given whole, however, cannot be taken for granted. The original and continuing 2 I use the words entity and system as interchangeable synonyms in this paper. Occasionally I will use the word structure in this sense too, sometimes for consistency when discussing other authors who use the word in this way. Usually, however, I use the word structure to mean the way the parts of an entity are organised by the relations between them. I rely on the context to make clear which usage is implied in each case. This seems more accessible than Collier s more rigorous suggestion that we use structuratum as a synonym of entity, and structure in the second sense discussed here (Collier, 1989, p. 85). 3 I have given a fuller overview of the theory of emergence in an earlier paper (Elder-Vass, 2004). 4 Real is used here in Bhaskar s sense, to indicate that these powers exist independently of whether they are actualised in particular events, since entities can possess powers that are unrealised (i.e. not triggered in particular situations) and powers that are exercised ineffectively (as a result of being overridden by countervailing causes) (Bhaskar, 1978, pp ). 5 I follow the usual convention here of regarding wholes as being at a higher level than their parts. This carries no normative implication, but is a useful metaphor in describing emergence. D. Elder-Vass 4

5 existences of any entity are always contingent; they depend upon the operation of casual factors in the actual world. Generically, we can label the causes that bring an entity into being from the collection of its parts, or that modify its structure without destroying it, as its morphogenetic causes. But having come into existence, there is no guarantee that an entity will continue to exist; if it does so, we can label the set of causes that sustains this existence as its morphostatic causes 6. At any moment, it is always possible that countervailing causes will overcome the morphostatic causes sustaining an entity, and dissolve it back into its component parts as, for example, when a biological organism dies and decomposes or reuse those parts in some other entity as when a chemical reaction destroys a water molecule and reuses its constituent atoms as parts of other molecules. One consequence of the role of organisation or structure in determining the powers of an entity is that emergent entities cannot be eliminated by any reductionist strategy from causal accounts that depend upon the exercise of their powers. It may often be possible (and indeed desirable) to provide an explanatory reduction of how the powers of a higher level entity result from the properties of its parts and the way they are organised, but this does not entail an eliminative reduction in which the higher level entity becomes redundant to the explanation (Elder-Vass, unpublished). Returning to the case of water, we can explain its ability to put out fires by demonstrating the way that its chemical bonds are altered in certain conditions so that it captures the oxygen required to sustain the fire and in the process is itself transformed chemically. This becomes an explanation in terms of the atomic constituents of water and the way that the bonds linking them together are affected in certain circumstances, thus it is an explanatory reduction of the case. However, it remains true that it is only because these atoms are organised into the form of water in the first place that they can behave in this way. Since the emergent entity is nothing more than the combination of its parts and their organisation, any explanation that depends upon the properties of its parts and on the characteristic way that they are related within this type of higher level entity is in effect an explanation in terms of the higher level entity. Now, since each emergent entity is made up of parts in this way, and each part is itself an entity, each entity has a multi-layered hierarchical structure. The water molecule is made up of hydrogen and oxygen atoms, but these in turn are made up of electrons, protons, and neutrons, and the protons and neutrons are in turn made up of quarks. Everything else in the universe is also made up in some way from a group of particles like electrons and quarks, and physicists seem to expect that all such particles will ultimately be explained in terms of a yet lower level type of entities. Thus we can see the whole of material reality as a single hierarchy of emergence. This is often described using the metaphor of levels, with a physical level at the bottom, a chemical level above that, a biological level above that, and so on another view with a long history (McLaughlin, 1992, p. 50). This is a little problematic, since, for example, there are non-biological higher-level entities that emerge from chemical parts. There are also some questionable assumptions involved in lumping together a wide variety of entities under generic labels like the physical. A preferable metaphor, then, is provided by the tree, with different branches representing the different types of entity that can be constructed from any given type of lower level entity. Although this metaphor is more helpful, it remains inadequate in some 6 I owe the terms morphostatic and morphogenetic to Buckley, via Archer (Buckley, 1967, p. 58) (Archer, 1982, p. 480, n8). D. Elder-Vass 5

6 respects, since higher-level entities often emerge from a combination of entities from a number of different lower-level branches. Nevertheless, it does express the essential unity of the hierarchy of emergent entities that is the basis of the ontological naturalism implicit in an emergentist view of the world. As we have seen, emergence provides an account of the real causal powers possessed by emergent entities. This complements Bhaskar s account of causation in A Realist Theory of Science (Bhaskar, 1978). As Bhaskar has argued in his account of multiple determination, the causation of actual events is the outcome of the interaction of the real causal powers of the various entities causally involved in the event concerned (Bhaskar, 1978, pp ). In the experimental sciences it is sometimes possible to create controlled experimental environments in which we can prevent the causal powers of many other entities interfering with the entities being studied (closed systems), and thus isolate the effects of individual causal powers. As a result, it is possible in experimental conditions to produce constant conjunctions of events that are interpreted as laws in the positivist tradition, although in fact such constant conjunctions depend on the experimenter s ability to exclude potential effects of countervailing causal influences. Outside experimental situations, however, causation occurs in open systems, where it is often difficult to disentangle the different impacts of a variety of interacting causal powers (Bhaskar, 1978, ch. 1; Sayer, 1999, p. 16). In such situations, the outcome is inherently open-ended and somewhat unpredictable, although there may sometimes be partial empirical regularities (or demi-regularities (Lawson, 1997, ch. 15)) that are potentially valuable in helping us to identify underlying causes. This is of course true for both natural and social systems, and hence differentiates the experimental from the historical sciences, rather than constituting an argument against naturalism as such. This brief account of emergence and cause implies that a complete causal analysis of the real powers of any emergent entity would include five elements: (a) a list of its characteristic parts; (b) an explanation of how these must be structured (related to each other) to form the whole; (c) a morphogenetic account of how this comes about; (d) a morphostatic account of how it is sustained; and (e) an explanatory reduction showing how the powers or properties of the whole are produced as a result of it having the parts it does, organised as they are. To provide a complete causal analysis of an event would then require that we explain which entities have a causal impact on it, and how the causal powers (analysed as above) of these various entities interact with each other to produce the outcome. In practice, such complete explanations are impossible, in that there may be a vast range of entities that interact to produce any given event, and identifying the causal powers of any one of them may be highly problematic, let alone investigating how they interact in a particular case or even a class of cases. Nevertheless it is useful to have a description of the ideal type of causal explanation that is implicit in an emergentist realism. In practice we can sometimes arrive at an explanation which is adequate for a given purpose, when the set of entities with a significant causal effect is sufficiently small, although such explanations will always be fallible. This ideal type is of course naturalistic, in the sense that it is an ideal that applies to the causal explanation of all types of powers and events. But the particular natures of the different types of entity and event studied by different sciences impose limitations (and indeed opportunities) on the degree to which particular methods of investigation can be applied. Thus the ontology of emergence founds both a naturalistic methodological principle and an anti-naturalistic one. D. Elder-Vass 6

7 The beginnings of a social ontology In addition to general questions of ontology such as those discussed above, realism recognises that each discipline has its own regional ontology (Benton and Craib, 2001, p. 5). These regional ontologies identify the types of entity that are the subjects of the discipline, and may clarify some of their more general characteristics. This section, then, suggests the very beginnings of a regional ontology for the social sciences which will be further developed, at least in some aspects, as the paper proceeds. Let me begin by enumerating some of the kinds of entity that are the subject matter of the social sciences (I make no claim for the completeness of this list). 1) human individuals people. Although there have been some interesting attempts to analyse people into various parts with different roles to play in their social functioning 7, this paper will assume that human beings are unitary entities, while recognising that for other purposes we might want to analyse their decomposition into, for example, minds and bodies, functional subsystems, or collections of cells. 2) organisations and/or institutions. These social structures are the main focus of the remainder of this paper, which will argue that these structures can have emergent properties and as a result can be causally effective. The differences between these two concepts are explored below. It is clear enough, however, that both organisations and institutions are made from, as well as by, human beings. Organisations and institutions arise from the power of human beings to co-operate to use each other. 3) human artefacts. Artefacts are made by human beings (who thus enter their causal history as morphogenetic and generally also morphostatic causes) from physical materials, and may have emergent powers of their own as a consequence of their (designed) structure aircraft, for example, may have the power to fly, while undercarriages, cockpits, tailfins, and food trolleys generally do not. Artefacts arise from the power of human beings to construct to use tools. 4) symbolic entities. Symbolic entities are created by human interaction and used by human beings to communicate meanings. Examples include words, stories, theories, and ideologies. Clearly such entities have a central role in human social behaviour, and I suggest that these too are emergent in their own right as argued by Archer under the heading of cultural emergent properties (Archer, 1995, pp ). However, a detailed discussion of them is beyond the scope of this paper. They arise from the power of human beings to communicate to use meanings. 5) There are also hybrid types of the above categories. Thus, for example, cities would seem to result from the combination of artefacts and organisations/institutions to create an entity that encompasses both and has properties beyond any of them. A book is perhaps a combination of symbols and artefact, with emergent properties not possessed by either. And cultures would seem to be a particularly important hybrid type, composed of a combination of symbolic constructs with social practices that are parts of institutions. Further categories may be required in a more developed social ontology, but it is perhaps worth identifying one category that has been deliberately omitted: societies. I do not suggest that there is no such thing, only that the coherence of any bounded 7 Notably (Archer, 2000). D. Elder-Vass 7

8 concept of society is extremely problematic, and it is hard to see how such a poorly defined entity could have real causal powers 8. Most of the powers that have been attributed to societies, I suggest, belong to somewhat lower-level social entities. The objects of study of the social sciences, then, comprise both human individuals and a variety of types of higher-level structures that arise as a result of human activity: as a result of humans acting upon physical things, as a result of humans acting upon each other, and as a result of humans acting upon ideas. I refer to these higher-level structures in this paper as social entities. Unfortunately or perhaps fortunately for social scientists, the interaction between these various entities is enormously complex (Byrne, 1998, p. 20). As has already been noted, this is also true of the entities studied by the natural sciences, but it will be argued below that there are some varieties of interaction that are unique to the social sciences, and that pose particular problems for a purely naturalistic account of them. Organisations A great many social scientists have denied that social structure can be causally effective. In the realist tradition, a prime example is provided by Varela and Harré, who regard any such belief as the fallacy of reifying a property of a group of social actors into an entity (Varela and Harre, 1996, p. 314). Archer, in particular, has already replied effectively to such arguments, for example in (Archer, 1982), but their persistence indicates that a more detailed analysis of the case for the causal effectiveness of social entities would still have value. This section will develop such an analysis, focussing on the case of organisations, perhaps the type of institution whose causal efficacy is most straightforward to explain. The next section will extend the argument to other sorts of social institution. Any organisation, I argue, is an emergent entity composed of a group of human individuals, structured by a set of relationships between them. It is common in sociology to call these relationships the roles of the people in the organisation. Such roles implicitly represent rules that define how the incumbent must relate to other members of the organisation, and also how they must relate to outsiders when acting on behalf of the organisation. Occupying a role means (a) to be recognised as occupying it by the other relevant role incumbents, and (b) to perform the behaviours that define the role. In the terms of an emergentist ontology, roles are not entities but relations hence they are not constituted by parts but instead are occupied by actual people. They can therefore only have causal influence in the sense that, and to the extent that, they are so occupied, or to the extent that the role incumbents adopt their characteristic behaviours which is of course another way of saying the same thing. Now, when a role incumbent adopts the behaviours defined by a role (e.g. answering the phone in the call centre if your role is call handling agent ), we have a case of downward causation, in the sense that the behaviour of the role incumbent is influenced by (their understanding of) the institution s expectations of (rules for) a holder of that role. 8 The concept of the state, of course, is a different matter. States are organisations, usually with well defined spatial boundaries, at least in some respects. But they do not map neatly onto societies ; there are many potentially cross-cutting social systems that follow different boundaries, or none at all; cf. (Walby, 2005). D. Elder-Vass 8

9 This, of course, is to claim that human behaviour is caused, which may make some social scientists uncomfortable, but it is not to claim that it is ever caused exclusively by a single factor. I make no claim here that the role incumbent s behaviour is determined by the social structure in this case by the rules of the organisation as this would be to claim that there is only one causal factor operating on the incumbent. Rather, I argue that the action of the role incumbent is co-determined (as in all cases of actual causation) by a variety of causal powers, including the causal power of the organisation as exerted through its rules, as well as the causal powers of the individual role incumbent cf. (Archer, 1995, p. 184). Thus, there is a causal mechanism operating from the institution to the role incumbent, although it can be defeated by other countervailing mechanisms. To the extent, however, that this causal mechanism is effective, the behaviour of the role incumbent in the role is part of the behaviour of the institution 9, and the causal effects of the organisation are the aggregate of the causal effects of its role incumbents when they do act in role. Now, a methodological individualist would argue that this reduces the behaviour of the organisation to that of the individuals and there is no need for the organisation at all in this explanation, e.g. (King, 1999, p. 271). However, the argument made earlier against eliminative reductions in general is perfectly applicable to this case. The role incumbents have the effects that they do when acting in these roles only because they are organised into this organisation through their performance of these roles. If there were no organisation there would be no such roles and the people would behave differently. Similarly, if there were no organisation, then those with whom the role incumbents interact would treat them differently. I would not hand over my money to a person in an electrical shop, for example, unless I believed they had, through their role incumbency, the right on behalf of the business they represent to give me the television I expect in return. Although I am served by a person I am served by them as a role incumbent and expect them to act in a certain way to represent the retail business that owns the shop and its stock as a result of this role incumbency 10. Nor would I pay any attention to an instruction by my manager to perform my role in a particular way unless she had, through her role incumbency, the right to make such a request. In this latter case, the organisation has a downward causal effect on my behaviour that has operated through a fellow role-incumbent; but it is nevertheless an effect of the organisation on me because she too is operating as a role incumbent and only has the authority to give an instruction because she operates as a representative of the organisation. Now we must, of course, distinguish between the behaviour of an individual in general and their behaviour in the role. Thus the chief executive s actions are part of the organisation s actions when she is seen as representing the institution in the terms of her role (e.g. in making an announcement at a corporate event). But they are not 9 Mouzelis seems to intend something similar when he talks of the type of action that results from the incumbency of authority positions as a case of macro action (Mouzelis, 1991, p. 35). 10 This is the first of several occasions on which I will use companies as an illustration of the properties of organisations. This rests on the assumption that company is a sub-type of organisation. Clearly sub-types have specific features that distinguish them from the broader type, e.g. the legal personality of the contemporary limited liability company, and particular sub-types such as companies may represent historically specific forms of a more general type. They can still be used to illustrate the properties of the more general type, however, as long as those illustrations do not rest on features of the sub-type that are not shared by the more general type, and this is the strategy adopted in this paper. D. Elder-Vass 9

10 when she is acting outside the role in a private capacity (e.g. when going for a swim), or when she is acting in some other role (e.g. when speaking as a candidate for election under the banner of a political party). The behaviour of the organisation, then, is the aggregate of the behaviours of its role incumbents in the role. Although the relationship between these behaviours is additive, the organisation is nevertheless emergent, because it has a non-linear effect on these behaviours as a result of the fact that the role incumbents behave differently as role incumbents than they would have done in isolation (i.e. if they were not incumbents of these roles) 11. Thus far, I have given an analysis of organisations that addresses three of the five elements required for a full analysis of a case of emergence - it has identified the components of an organisation (people), the relations that constitute them into the organisation (roles), and how this gives the organisation emergent properties not possessed by its parts. A full analysis, then, still requires an account of the morphogenesis and morphostasis of organisations. Such an account is mostly beyond the scope of the present paper, but there is one aspect that is worth examining: the relationship between role incumbency and organisational morphostasis. The morphostasis of an organisation requires that (a) it has incumbents for all essential roles; and (b) those incumbents act within the expectations for their role. Now, there is, in the account so far, no necessity that these requirements will continue to be met. Being a role incumbent, for example, is purely contingent. A role incumbent may choose to leave the role (in most contemporary organisations) and if they do so then any downward influence of the organisation on the former role incumbent will cease. But most organisations provide for their morphostasis by replacing role incumbents who leave essential roles. Similarly, role incumbents may fail to perform according to the definition of the role, which could undermine the performance and ultimately even the continuing existence of the organisation. Again, this is generally dealt with quite simply either by removing and replacing the incumbent, or by managing their performance (e.g. by further training or by disciplinary threats) so that it does start to conform to the role s requirements. Any organisation that is unable to deal with either of these sorts of problem is likely to fail and dissolve although there are also of course, many other problems that could lead to such a result, and a successful organisation must have morphostatic processes in place to defeat these too. It is worth a brief digression to examine how this story compares to similar stories in the natural world. We might argue, for example, that a hydrogen atom H has a certain role as part of being part of a water molecule, that the water molecule only exists and has the powers of a water molecule as long as the hydrogen atom (or another equivalent one) performs this role, that countervailing causes may interfere with its continuing in the role in which case it may leave the molecule resulting in its dissolution, and so on. Other sorts of natural entities (e.g. diseased cells in a biological organism) may mis-perform their roles, with the result that the organism s performance degrades or ultimately it dies. And natural entities may have morphostatic mechanisms (e.g. an immune system) that defend them against such threats. In other words, the concept of role and the performance of roles are not unique to human organisations or dependent upon any specifically human property. 11 I thus see emergence in sociology in different terms than Sawyer, who sees emergence as depending on the multiple realizability of properties (Sawyer, 2001). D. Elder-Vass 10

11 On the other hand, of course, the way in which human beings perform roles is a product of the way that human beings act in a wider sense. Factors such as consciousness, reflexivity, and the concept dependence of human action mean that there are a whole variety of different factors involved in securing human role performance in particular, humans must be persuaded to perform a role. Unlike the parts of most lower-level natural entities, they cannot just be held in a particular spatial relationship as a result of which role performance becomes automatic. Furthermore, the ability of humans to perform multiple roles introduces a whole level of complexity into the social sciences that is unknown in the natural sciences. These are both questions that will be returned to below. Institutions The term institutions is used ambiguously in the social sciences (Jary and Jary, 2000, p. 302). For some authors particularly outside sociology institution is a synonym of organisation in the sense discussed in the previous section. But there is another sense, with which this section will engage: the idea of an institution as a characteristic set of social practices or relations that does not appear to form the basis of an organisation as such. Examples commonly used are marriage, monogamy, kinship, class, contract, and property. In this usage an institution is not a social entity with a set of causal powers (as an organisation is) but rather a type of social relation 12. Now, on the argument so far, it is possible that when such a relation obtains amongst a group of people, although it is not an entity itself, it constitutes them into a social entity. There do seem to be cases of institutions that conform to this pattern. Marriage, for example, is a relation between two people that constitutes them into a married couple. Once married, each person assumes the role of a wife or a husband (or simply a married person, if there are cultural contexts where the two roles are synonymous), and this role implies that they will observe certain (culturally variable) rules of behaviour in their relations to their partner and to other people outside the marriage. As a result, each married couple becomes an entity in its own right, composed of two human individuals as related by the roles that define them as a married couple. Clearly, the marriage may end, e.g. in divorce, so the relationship and hence the higher level entity are only contingently morphostatic; but in the meantime, there is a set of morphostatic causes maintaining the minimum role requirements. As long as the marriage survives, though, the couple is an entity with a causal impact on the world. It may have the tendency, for example, to prevent sexual relations between a partner in the marriage and an outsider, to lead to the couple living together, and to induce them to have children together (though any or all of these tendencies may be overcome by countervailing powers). It is therefore possible to see a married couple as a causally effective entity in its own right indeed, as an instance of an organisation with a downward causal effect on its role incumbents. This is not to say, however, that all of those things we call institutions can necessarily be seen as the structuring relations of organisations. In principle, it is possible that some institutions may be common practices or relations that do not lead to the formation of such an entity. These practices may not constitute entities because they represent only fleeting relationships such as the practice of thanking a passing stranger who does one a small favour. Or they may constitute more permanent entities 12 In this paper institution is used in the broad sense implicit in the literature; hence it will at some times refer to a social entity and at others to a social relation. D. Elder-Vass 11

12 that nevertheless have no emergent properties or powers perhaps this may be the case, for example, where two people regularly sit next to each other without establishing any more significant relationship (although one might argue that this is too tenuous a relation to constitute an institution or an entity in the first place). In such cases, the causal powers at work could only be those of the individual people who participate in these relations. Still other institutions may structure entities that are not organisations. Property is perhaps such a case. This is a relationship between (a) an owned entity (let me call this a chattel 13 ); and (b) an owning entity, usually a person or organisation. It can also be seen as a relationship with an excluded population, who are denied the rights with respect to the chattel that are reserved to the owner. Here, I suggest, if there is a higher-level emergent entity it is composed of the owner, the chattel, and the ownership relation between them, where that relation is an instance of the institution of property. This, of course, would be a hybrid entity in the terms discussed earlier, being composed of both a human part and a physical (usually but not necessarily artefactual) part. We have no label for such an entity. Even if we invented one, it is not clear whether it would be a useful way to look at this situation. We might prefer to argue that there is no entity here and the property relation can be socially significant without constituting an emergent entity. I suggest that it would be entirely possible to construct an entitybased account of this situation, which in fact wouldn t look terribly different from the account of a monogamous marriage given in the previous paragraph. It would, however, seem even more unfamiliar than the idea that a married couple has causal powers that are ineliminably different from those of the partners. It remains to be seen whether such an entity would be a more useful tool in causal explanations of the social than the commoner practice of seeing property merely as a relation, in which only the parts have a causal role to play. The field of institutions, then, is one where further analysis is required, and where each type of institution may need to be considered separately. Emergentism has no need to postulate emergent entities where none exists, but there may be cases where the predominance of a relational view of institutions blinds us to the existence of emergent entities arising from them with real causal powers. Structure and agency One response to the claim that social structures have real causal powers is to see this as antithetical to human agency, and some sociological structuralisms have indeed seemed to deny a causal role to individual human beings, at least with regard to the behaviour of social structures themselves, e.g. (Althusser, 1967; Durkheim, 1964; Luhmann, 1990). Critical realists have strongly criticised such views. Both Bhaskar and Archer, in particular, have invoked emergence in order to argue that the social is a realm in which human individuals and social entities interact, each conditioning the other but each possessing its own causal powers (Archer, 1982; Bhaskar, 1998). This section will examine how emergentism relates to human agency, to the extent necessary to dismiss agency as an argument against the causal effectiveness of social 13 The dictionary definition of chattel restricts its usage to certain sorts of property; I intend no such restriction in my usage of the term. D. Elder-Vass 12

13 structure, without developing a full theoretical account of agency itself. In particular, it will not address the relationship of agency to ideas of freedom of choice. Human individuals, as noted earlier, are themselves emergent systems, with properties and powers that arise from their components and the way these are organised to form a human being. For the sake of this argument I shall assume that the components of human beings are cells, although in a fuller account of agency we may want to identify some intermediate levels of structure 14. The group of cells composing any given human being is structured by the set of relationships described by physiology into the form of a human being. Although scientists know a great deal about how these relationships lead to many aspects of human functioning, we do not yet have a full explanation, particularly in terms of the functioning of our brains, of how they produce human behaviour. Nevertheless, emergentism (and indeed any scientific approach to human behaviour) rests on the belief that human powers are the consequence of the way our cells are organised into the characteristic structures of a human being. And it leads to the belief that we may some day be able to provide an explanation of human behaviour that shows how it is emergent from these parts and their organisation. Even a complete explanation of human functioning in terms of our parts, however, would not mean that human agency had been eliminated; just as explanatory reduction does not entail eliminative reduction in the natural sciences, nor does it in the human sciences. We have seen that the unique causal powers of any given social organisation, for example, follow from the fact that its members would not behave in the way that is characteristic of such an organisation (and thus have the effects that are characteristic of such an organisation) unless they were constituted into it by its characteristic relations (or roles). In a precisely parallel argument, the unique causal powers of human individuals follow from the fact that their cells would not behave in the way that is characteristic of a human individual (and thus have the effects characteristic of a human individual) if they were not constituted into one by a particular set of physiological relationships. This still allows us to construct explanations of human behaviour in terms of intentions, reasons, social learning, shared meanings, norms, values, institutions, and the like, although clearly the explanation of how all of these factors combine to produce a theory of human behaviour is beyond the scope of this paper. And it allows for the possibility that the causal explanation in these terms of human behaviour is extremely complex, with the consequence that it may in some respects seem unpredictable. The point that is essential here is that human beings are causally effective entities, whose behaviour is itself in turn caused by a variety of interacting factors, including the influence of social structures, but it is not determined monolithically by any single set of such factors, including social structure or indeed the biological structures of their bodies. The causal effectiveness of social structure, then, in no way undermines the argument for the causal effectiveness of human beings. Indeed, it rests upon it, since the emergence of social structure depends upon its parts being causally effective in their own right. And it rests upon precisely the same logic of emergence as does the causal effectiveness of human beings. If this logic could not in principle sustain the causal effectiveness of social structures, it could not in principle sustain the causal effectiveness of human beings or indeed of any other entity. As Buckley argues, if 14 As Archer, for example, attempts to do (Archer, 2000). D. Elder-Vass 13

14 social groups are not real entities then neither are individual organisms, cells, molecules, or atoms, since they are all nothing but the constituents of which they are made. But this nothing but hides the central key to modern systems thinking the fact of organization of components into systemic relationships (Buckley, 1998, p. 36). Those who deny that social organisation can be causally effective because they are nothing but the people who compose them must explain how people can be causally effective when they are nothing but the cells that compose them. On the basis of such an argument, King has suggested that the concept of emergence, in fact, involves a relapse into sociological reification where society comes to exist independently of individuals, although this relapse into reification is concealed by the continual emphasis on individual practices and beliefs (King, 1999, p. 270). I hope that this paper has shown that emergence, on the contrary, insists upon the dependence of society upon individuals. The possession by social organisations, for example, of causal powers of their own, is a consequence of their dependence upon the individuals that compose them. Although some writers may occasionally be tempted to render causal powers of their own as independent causal powers, this is not a usage of independent that constitutes reification in any sense that can reasonably be objected to, since it does not imply any denial of the role of the parts in the explanation of those causal powers. The argument in the previous sections, being intended as an explanation of the causal powers of social structures, has naturally focussed on only one part of the social dynamic described in Bhaskar s Transformational Model of Social Activity : the influence of structure on human individuals, which he labels socialisation (Bhaskar, 1998, pp. 34-7). For the avoidance of doubt, I must also make clear that, although it is not the primary focus of this paper, the emergentist account places equal weight on the second part of this dynamic: the influence of human individuals on social structure, which he labels reproduction/transformation. As I have already made clear, the continued existence of a social organisation (its morphostasis) depends upon the continuing role performance of its members, although also on other causal factors which may be external to the organisation. Likewise, changes in the structure and workings of an organisation (its morphogenesis) also depend upon the performances of its members; indeed it is often part of certain roles to help change an organisation in response to the challenges it faces. The link between role and change is less obvious in more relational forms of institution; changes in these are perhaps more often the unintended consequences of the role performances involved in the institution. Nevertheless, individuals play a crucial part in the transformation of these too. Indeed, this view of the world also allows a role to the otherwise problematic category of mega-actors, introduced by Mouzelis. These are individuals whose economic, political or culturally based social power makes the consequences of their decisions widely felt (Mouzelis, 1991, p. 107). Such actors can be influential in one of two distinct ways. First, they may have substantial influence in their capacity as private individuals. Thus, for example, a wealthy and prominent private art collector who patronises a particular style of art may have a significant effect on social tastes, and through this, for example, on the art-buying behaviours of both other private individuals and of organisations like public galleries and businesses corporations. Second, individuals may be immensely influential by virtue of the way in which they perform their roles in organisations. Such roles can be performed well or badly, because role specifications do not completely describe how they are to be performed. D. Elder-Vass 14

15 Role specifications constrain acceptable behaviour in a role, and they may provide criteria for standards of performance, but they also provide resources that are available to the role, and leave open many alternative ways of performing the role. This is one of the crucial ways in which social roles differ from role-equivalents in the structure of natural objects: they provide the opportunity for flexible behaviour within the role by its incumbents, and such flexibility enhances the possibilities for the individual role incumbent to have an exceptional impact, whether in the form of spectacular success or dangerous failure. Hence, for example, an exceptionally capable or lucky chief executive may contribute to the establishment of a dominant business corporation with a major impact on society. And, of course, mega-actors need not become so as a consequence of their own exceptional abilities. It is enough to be in a role that gives one personal influence over a powerful organisation. The President of the United States, for example, will inevitably be a mega-actor simply because of the combination of their personal discretion in performance of their role with the immense power of the US Government. Their actions in this role, unlike those of the wealthy private art collector, will be part of the actions of the organisation to which the role belongs, yet their role enables them to affect the behaviour of this organisation in a potentially influential way. Applying Bhaskar s account of actual causation, we can say that the causal powers of the agents who occupy roles within an organisation co-determine, along with the causal powers of other entities involved, the particular behaviours of the organisation. Yet the impact of the organisation s behaviour arises from the causal powers of the organisation itself, since these individuals could not have acted in this way, with the corresponding impact, had they not been parts of the organisation in the relation to it defined by their roles. To summarise this section: the possession of agency means that human beings can have a causal impact on the world in their own right, but this does not mean that theirs are the only causal powers that can influence the social world; and it does not mean that the actions of human beings are not themselves caused. Social events are the outcome of complex interactions between the causal powers of individuals, organisations, institutions, natural objects, human artefacts, and symbolic structures. The dynamism of social systems In discussing the contribution of individuals to the transformation of social structure, this paper has already acknowledged that social systems are not fixed in form. The role of morphostasis should perhaps be clarified here: morphostatic forces are those that ensure the continuing existence of a entity, but there is no necessity for this continuing existence to be predicated on a permanent fixity of form. Human beings, for example, move around constantly, and their staying alive depends upon it; for a human being morphostasis is not only consistent with change but absolutely requires it. Not only do we move around, but we also develop in our earlier years we grow, in our middle years our brains develop more and more useful connections and hence knowledge, and in our older years we develop signs of aging. Not only do we change, but our bodies also constantly rebuild and repair themselves, so that morphostasis of the biological organism implies not only a constant change in the arrangement of the parts, but also a constant change in the parts themselves, continually replacing one cell of a certain type with another. D. Elder-Vass 15