18 Ontology and Epistemology in Political Science David Marsh and Paul Furlong 19 Ontology and epistemology Ontological and epistemological positions are related, but need to be separated. To put it crudely, one's ontological position affects, but far from determines, one's epistemological position. Ontology Ontological questions are prior because they deal with the very nature of 'being'; literally, an ontology is a theory of 'being' (the word derives from the Greek for 'existence'). This sounds difficult, but really it is not. The key question is whether there is a 'real' world 'out there' that is independent of our knowledge of it. For example, are there essential differences between genders, classes or races that exist in all contexts and at all times? A simple illustration easily makes the point. Over the last ten years John Gray's book Men are from Mars and Women are from Venus (1992) has sold seven million copies in the USA and millions more in forty countries worldwide. He argues that men and women are very different and that men and women can only understand and deal with one another better if they recognise this fact of life. This book takes a clear ontological position; there are fundamental differences between men and women that are features of their very existence. These differences persist over time and are common across cultures. This is an essentialist or a foundationalist ontological position. So, its proponents argue that there are essential differences of 'being' that provide the foundations upon which social life is built. Of course, this is a contentious position; one which is strongly attacked by many, if not most, feminists. They believe that the differences between men and women are socially constructed. As such, they are not essential differences but are particular to a given culture and time. They are the product of patriarchy, in which male dominance shapes the culture and values of society, affects patterns of socialisation and perpetuates gender inequality. This argument reflects a different ontological position that is anti-foundationalist and emphasises the social construction of social phenomena. Epistemology If an ontological position reflects the researcher's view about the nature of the world, their epistemological position reflects their view of what we can know about the world and how we can know it; literally an epistemology is a theory of knowledge. Again, this sounds difficult, but the basic concerns are not too difficult. There are two key questions. Can an observer identify 'real' or 'objective' relations between social phenomena? If so, how? The first question itself subsumes two issues. Initially, it takes us back to ontology; if one is an anti-foundationalist, then one argues that there is not a 'real' world, which exists independently of the meaning which actors attach to their action, to discover. At the same time, such an anti-foundationalist would also suggest that no observer can be 'objective' because they live in the social world and are affected by the social constructions of 'reality'. This is sometimes called the double hermeneutic; the world is interpreted by the actors (one hermeneutic level) and their interpretation is interpreted by the observer (a second hermeneutic level). The second question raises another important, and clearly related, issue, j, To the extent that we can establish 'real' relationships between social g phenomena, can we do this simply through direct observation, or are there % some relationships which 'exist' but are not directly observable? The ^ answers one gives to these questions shapes one's epistemological position. Of course, there are different ways of classifying epistemological §, positions and there is no agreement as to the best way. Probably the most f, common classification distinguishes between scientific (sometimes positi- % vist) and hermeneutic (or interpretist) positions. We shall begin with a J brief review of that distinction, before proposing an alternative, which !J; distinguishes between positivist, realist and interpretist positions. fe Scientific versus hermeneutic approaches % Social science was influenced by the ideas of science as the nomenclature P clearly indicates. In particular, the empiricist tradition played a crucial role p in the development of social science. David Hume argued that knowledge starts from our senses. On the basis of such direct experience we could "" develop generalisations about the relationships between physical phenom- ena. The aim was to develop causal statements which specified that, under a given set of conditions, there would be regular and predictable outcomes \ (on this see Hollis and Smith 1990: ch. 3). The adherents of the scientific tradition saw social science as analogous to science. In ontological terms •5 they were foundationalists; they thought there was a real world 'out there' which was external to agents. Their focus was upon identifying the causes of social behaviour. The emphasis was upon explanation and many felt •k that the use of rigorous 'scientific' methods would allow social scientists to & develop laws, similar in status to scientific laws, which would hold across time and space. 20 Ontology and Epistemology in Political Science In methodological terms, the scientific tradition was greatly influenced by logical positivism that posited a very straightforward characterisation of the form of scientific investigation. As Hollis and Smith put it (1990: 50): To detect the regularities in nature, propose a generalisation, deduce what it implies for the next case and observe whether the prediction succeeds. If it does, no consequent action is needed; if it does not, then either discard the generalisation or amend it and [test the] fresh [predictions]. In contrast, there is an alternative hermeneutic (the word derives from the Greek for 'to interpret') or interpretist tradition. The adherents of this position are anti-foundationalists, believing that the world is socially constructed. They focus upon the meaning of behaviour. The emphasis is upon understanding, rather than explanation. As such, in the interpretist tradition it is not possible to establish causal relationships between phenomena that hold across time and space. Positivist, realist and interpretist positions We prefer this classification because the scientific tradition identified by Hollis and Smith conflates two distinct positions, positivism and realism. Positivists adhere to a foundationalist ontology and are concerned to establish causal relationships between social phenomema, thus developing explanatory, and indeed predictive, models. The realist is also foundationalist in ontological terms. However, realists, unlike positivists, do not privilege direct observation. The realist believes that there are deep structural relationships between social phenomena which cannot be directly observed, but which are crucial for any explanation of behaviour. So, as an example, a realist might argue that patriarchy as a structure cannot be directly observed, although we can see many of the consequences of it; we return to this example later. The distinction between positivist, realist and interpretist approaches is examined in much more depth in the next section. However, the key point here is that any classification that we adopt would annoy some social scientists. We use this particular distinction because we are realists and, as such, do not like the conflation between positivism and realism involved in the first distinction. However, many other authors would question our distinction. In particular, many, like Bevir and Rhodes (see below) would want to make further distinctions within the tradition of interpretive theory. The point is that any way of classifying epistemological positions can be contested; we choose one, but are aware of the criticism of it. In addition, we shall deal with many of those criticisms when we look at the variants within the three positions we identify. David Marsh and Paul Furlong 21 Why are such distinctions important? In our view, ontological and epistemological concerns cannot, and should not, be ignored or downgraded. Three points are important here: 1. 2. First, these concerns should not be put in what the Australians, with typical directness, call the 'too hard basket'. Certainly, the issues involved are not easy, but neither are they difficult, if they are explained simply and with appropriate examples. Second, ontological and epistemological positions should not be treated like a sweater that can be 'put on' when we are addressing such philosophical issues and 'taken off when we are doing research. In our view, the dominance of a fairly crude positivist epistemology throughout much of the postwar period encouraged many social scientists to dismiss ontological questions and regard epistemological issues as more or less resolved, with only the details left to be decided by those interested in such matters. Such social scientists have tended to acknowledge the importance of epistemology without considering it necessary to deal with it in detail; positivism has been regarded as a comforting sweater that can be put on where necessary. In contrast, we would argue that epistemology, to say nothing of ontology, is far from being a closed debate. Third, researchers cannot adopt one position at one time for one project and another on another occasion for a different project. These positions are not interchangeable because they reflect fundamental different approaches to what social science is and how we do it. This is the key point. As we pointed out in the introduction, a researcher's epistemological position is reflected in what is studied, how it is studied and the status the researcher gives to their findings. So, a positivist looks for causal relationships, tends to prefer quantitative analysis (for a more detailed discussion of the relationship between ontology, epistemology and methodology, see Chapter 11) and wants to produce 'objective' and generalisable findings. A researcher from within the interpretist tradition is concerned with understanding, not explanation, focuses on the meaning that actions have for agents, tends to use qualitative evidence and offers their results as one interpretation of the relationship between the social phenomena studied. Realism is less easy to classify in this way. The realists are looking for causal relationships, but think that many important relationships between social phenomena cannot be observed. This means they may use quantitative and qualitative data. The quantitative data will only be appropriate for those relationships that are directly observable. In contrast, the unobservable relationships can only be established 22 Ontology and Epistemology in Political Science David Marsh and Paul Furlong 23 indirectly; we can observe other relationships which, our theory tells us, are the result of those unobservable prelationships. We return to these issues in the next section. Different approaches to ontology and epistemology Here we outline the positivist, the interpretist and the realist positions in more detail. We shall focus on: the major criticisms of the positions; the variations within these positions; and the way the positions have changed over time. At the outset, however, it is important to emphasise that the distinctions between the positions, and more specifically that between interpretism and realism, are not clear-cut. Positivism The core of positivism is fairly straightforward, although of course there are variants within it: • Positivism is based upon a foundationalist ontology. So, to the positivist, like the realist, but unlike those from the interpretist position, the world exists independently of our knowledge of it. • To the positivist, natural science and social science are broadly analogous. We can establish regular relationships between social phenomena, using theory to generate hypotheses which can be tested by direct observation. In this view, and in clear contrast to the realist, there are no deep structures that cannot be observed. Traditionally, positivism contended that there is no appearance/reality dichotomy and that the world is real and not socially constructed. So, direct observation can serve as an independent test of the validity of a theory. Crucially, an observer can be objective in the way they undertake such observations. Researchers from the interpretist tradition rarely accept any notion of objectivity. Realists accept that all observation is mediated by theory (to the realist, theory plays the crucial role in allowing the researcher to distinguish between those social phenomena which are directly observable and those which are not). • To positivists the aim of social science is to make causal statements; in their view it is possible to, and we should attempt to, establish causal relationships between social phenomena. They share this aim with realists, while interpretists deny the possibility of such statements. • Positivists also argue that it is possible to separate empirical questions -that is, questions about what is - from normative questions - that is, questions about what should be. Traditionally, positivists thought that the goal of social science was to pursue empirical questions, while philosophy, metaphysics or religion pursued the normative questions. Because we can separate empirical and normative questions, it is possible for social science to be objective and value-free. Realists and, especially, those from within the interpretist tradition would reject that proposition. Many social scientists are positivists, although much of the positivism is implicit rather than explicit. The behavioural revolution in the social sciences in the 1960s, dealt with by David Sanders in Chapter 2, was an attempt to introduce scientific method into the study of society. It was an explicit reaction to political theory, which it saw as concerned with normative questions, and institutionalism, which it saw as lacking theoretical and methodological rigour. In contrast, it was based upon a foundationalist ontology and, most often, a quantitative methodology (but see below and Chapter 10). The view was that a social 'science' was possible if we followed the scientific method; deriving hypotheses from theory and then testing them in an attempt to falsify them. We needed 'objective' measures of our social phenomena, our variables; as such, we would focus upon 'hard' data - from government statistics, election results and so on - rather than soft data - from interviews or participant observation. So, for example, if a positivist was studying political participation, they would be interested in measuring the level of voting, party or pressure group membership, direct action or whatever, and relating it to demographic variables such as class, gender, race and education. The aim would be to establish the precise nature of the relationship between these variables and participation in order to produce causal models. We shall return to this example later. The key point here is that, as always, the ontological and epistemological position adopted had clear methodological implications. The criticism of positivism takes two broad forms. The first line of criticism broadly argues that, in following the methods of science, positivists misinterpret how science really proceeds. Two lines of argument have been particularly important here. First, there is the pragmatist position of Quine (1961) who develops two crucial critiques of positivism (for a fuller exposition see Hollis and Smith 1990: 55-7; they deal with a third, less important, criticism): (i) Quine argues that any knowledge we derive from the five senses is mediated by the concepts we use to analyse it, so there is no way of classifying, or even describing, experience without interpreting it. (ii) This means that theory and experiment are not simply separable, rather theory affects both the facts we focus on and how we interpret them. This, in turn, may affect the conclusions we draw if the facts 24 Ontology and Epistemology in Political Science David Marsh and Paul Furlong 25 appear to falsify the theory. As such, if we observe 'facts' which are inconsistent with the theory, we might decide that the facts are wrong rather than that the theory is wrong. Of course, this undermines the notion that observation alone can serve to falsify a theory. Second, there is Kuhn's view (1970) that, at any given time, science tends to be dominated by a particular paradigm that is unquestioned and which affects the questions scientists ask and how they interpret what they observe (for a fuller discussion, see Hollis and Smith 1990: 57-61). Consequently, scientific investigation is not 'open', as positivism implies, rather certain conclusions are almost unthinkable. There is a 'paradigm shift' when a lot of empirical observation leads certain, brave, scientists to question the dominant paradigm, but until that time, and for the most part, scientists discard observations which do not fit (obviously this fits well with the second of Quine's criticisms above) and embrace the results which confirm the paradigm. The second main line of criticism of positivism is more particular to social science. It argues that there are obvious differences between social and physical or natural phenomena that make social 'science' impossible. Three differences are particularly important. First, social structures, unlike natural structures, do not exist independently of the activities they shape. So, for example, marriage is a social institution or structure, but it is also a lived experience, particularly, although not exclusively, for those who are married. This lived experience affects agents' understanding of the institution and also helps change it. Second, and related, social structures, unlike natural structures, do not exist independently of agents' views of what they are doing in the activity. People are reflexive; they reflect on what they are doing and often change their actions in the light of that reflection. This leads us to the third difference. Social structures, unlike natural structures, change as a result of the actions of agents; in most senses the social world varies across time and space. Some positivist social scientists minimise these differences, but, to the extent they are accepted, they point towards a more interpretist epistemological position. Many positivists avoid these critiques which are put in the 'too hard basket'. However, the more sophisticated positivists are aware of these criticisms and the position has changed significantly as a result. Fortunately, this volume boasts two sophisticated behaviouralists who are positivists, Sanders and John. It is particularly worth examining David Sanders' view in a little more detail because it represents an excellent example of the modern, more sophisticated, positivist position. Sanders (Chapter 2) accepts he has been strongly influenced by the positivist position, but acknowledges the 'ferocious philosophical criticism' to which it was subjected. He argues that 'post-behaviouralists', who might also be called 'post-positivists': acknowledge the interdependence of theory and observation; recognise that normative questions are important and not always easy to separate from empirical questions; and accept that other traditions have a key role to play in political and social analysis. As such, this post-positivism has moved a significant way from more traditional positivism, largely as a result of the type of criticisms outlined here. However, the ontological and epistemological problems have not gone away, rather they have been elided. Two quotes from Sanders illustrate the point. First, he argues (see Chapter 2: 51): Modern behaviouralists — 'post-behaviouralists' — simply prefer to subject their own theoretical claims to empirical tests. They also suspect that scholars working in non-empirical traditions are never able to provide a satisfactory answer to the crucial question: 'How would you know if you were wrong?' Later he continues (Chapter 2: 54): For modern behaviouralists, the ultimate test of a good theory is still whether or not it is consistent with observation - with the available empirical evidence. Modern behaviouralists are perfectly prepared to accept that different theoretical positions are likely to produce different observations. They insist however, that, whatever 'observations' are implied by a particular theoretical perspective, those observations must be used in order to conduct a systematic empirical test of the theory that is being posited. This is a sophisticated statement of a positivist epistemological position, but it is still essentially positivist. The aim is to use observation (of whatever type) to test hypothesised relationships between the social phenomena studied. Research from within other traditions must still be judged against the positivists' criteria: 'observation must be used in order to conduct a systematic empirical test of the theory that is being posited'. Yet, that is not a standard most researchers from within an interpretist tradition could accept (even Bevir and Rhodes 1999 could only do so with major qualifications), because they do not believe that direct observation can be objective and used as a test of 'reality'. Most realists would also have a problem with Sanders' position because they would see many of the key relationships as unobservable. One other aspect of Sanders' position is important here. He accepts that interpretation and meaning are important, which might suggest that the differences between positivist and interpretist traditions are beginning to dissolve. So, Sanders argues (Chapter 2: 53), in criticising prior studies of voting behaviour: 'There are other areas — relating to the way in which individuals reflect, to a greater or lesser degree, upon themselves — where 26 Ontology and Epistemology in Political Science behavioural research has simply not dared to tread.' He recognises that such factors might, or might not, be important, but emphasises that they would be difficult to study empirically. However, the crucial point is that Sanders wants to treat interpretation and meaning as intervening variables. In this view, how a voter understands the parties and their own position may affect their voting behaviour. At best, this acknowledges only one aspect of the double hermeneutic; the interpretist tradition would argue that we also need to acknowledge the subjectivity of the observer. So, positivism has changed in response to criticism. Post-positivism is much less assertive that there is only one way of doing social science. However, it still emphasises explanation, rather than understanding, and the primacy of direct observation. In our terms, it is still foundationalist and firmly located in the scientific tradition. The interpretist position The interpretist tradition is the obvious 'other' of positivism. However, it is a much broader church than positivism and much of this subsection will deal with its variants. Nevertheless, it is useful to begin with an outline of the core of the position. • In the interpretist tradition researchers reject the notion that the world exists independently of our knowledge of it. Rather, they contend that the world is socially or discursively constructed. This view is diametrically opposed to positivism, but shares certain features with some modern variants of realism. In ontological terms, then, this position is anti-foundationalist. • This means that for researchers working within this tradition social phenomena do not exist independently of our interpretation of them; rather it is this interpretation/understanding of social phenomena which affects outcomes. As such, it is the interpretations/meanings of social phenomena that are crucial; interpretations/meanings which can only be established and understood within discourses or traditions. Consequently, we should focus on identifying those discourses or traditions and establishing the interpretations and meanings they attach to social phenomena. • However, we must also acknowledge that 'objective' analysis is impossible. Social 'scientists' (of course interpretists would not use this term) are not privileged, but themselves operate within discourses or traditions. Consequently, knowledge is theoretically or discursively laden. As such, this position acknowledges the double hermeneutic. This position has clear methodological implications. It argues that there is no objective truth, that the world is socially constructed and that the role If David Marsh and Paul Furlong 27 of social 'scientists' [sic] is to study those social constructions. Quantitative methods can be blunt instruments and may produce misleading data. In contrast, we need to utilise qualitative methods - interviews, focus groups, vignettes and so on — to help us establish how people understand their world. So, for example, someone operating from within this tradition studying political participation would start by trying to establish how people understand 'the political' and 'political' participation The major criticism of the interpretist tradition comes, unsurprisingly, from positivists, though some realists would agree with elements of that critique. To positivists, the interpretist tradition merely offers opinions or subjective judgements about the world. As such, there is no basis on which to judge the validity of their knowledge claims. One person's view of the world, and of the relationship between social phenomena within it, is as good as another's view. To many positivists, this means that such research '■ is akin to history, or even fiction, whereas they aspire to a science of society. It is difficult for someone in the interpretist tradition to answer \ this accusation, because it is based on a totally different ontological view *; and reflects a different epistemology and, thus, a different view of what |, social science is about. However, as we shall see, most researchers do \ believe that it is possible to generalise, if only in a limited sense. Perhaps "ij more interestingly, Bevir and Rhodes (forthcoming) attempt to defend 4 their approach against this positivist critique by establishing a basis on tr which they can make knowledge claims; on which they can claim that one J interpretation, or narrative, is superior to another. We shall return to their / argument below. Bevir and Rhodes (forthcoming: ch. 2) distinguish between the hermeneutic and postmodern, or post-structuralist, strands in the interpretist position. In essence, the hermeneutic tradition is idealist; it argues that we need to understand the meanings people attach to social behaviour. So, hermeneutics is concerned with the interpretation of texts and actions. This involves the use of ethnographic techniques (participant observation, transcribing texts, keeping diaries and so on) to produce what Geertz (1973) calls 'thick description'. As Bevir and Rhodes put it, quoting Geertz, the aim is to establish 'our own constructions of other people's constructions of what they and their compatriots are up to'. However, ethnographers do generalise. They develop a narrative about the past based upon the meanings which the actions had for social actors. Then, on the basis of this 'thick description', they offer an interpretation of what this tells us about the society. The point is that these interpretations are always partial, y in both senses of the world, and provisional; they are not 'true'. J More recently, as Bevir and Rhodes (forthcoming) emphasise, post- 11 structuralism and postmodernism have provided a powerful challenge to foundationalism in both philosophy and social science. Yet, as Bevir and 28 Ontology and Epistemology in Political Science David Marsh and Paul Furlong 29 Rhodes also point out, this variant of the interpretist tradition is itself so diverse that it is difficult, if not impossible, to characterise. They overcome this problem by focusing on the work of Michel Foucault, who is perhaps the best-known writer in this broad tradition. He, like most post-structuralists and postmodernists, is a strong opponent of foundationalism and the modernisation project associated with the Enlightenment. This project argues that: the basis of human knowledge is direct experience; as such, it is possible to develop an 'objective' view of the 'real' world (thus, it denies both elements of the double hermeneutic); language is transparent or neutral; and that human history is inevitably progressive, with present knowledge building on past knowledge to improve our information about the world and our ability to control it. In contrast, Foucault argues that experience is acquired within a prior discourse. As such, language is crucial because institutions and actions only acquire a meaning through language. Thus, as Bevir and Rhodes (forthcoming) argue, for Foucault: 'to understand an object or action, political scientists have to interpret it in the wider discourse of which it is part'. This means that, as Bevir and Rhodes stress, it is the social discourse, rather than the beliefs of individuals, which are crucial to Foucault's version of the interpretist position. The identification of that discourse, and the role it plays in structuring meanings, is thus the key concern of those adopting this approach (for an example of this broad approach in use see Howarth (1995). Bevir and Rhodes (forthcoming) develop their own take on the interpretist tradition. It is particularly interesting because it directly addresses the key issue raised in the positivist critique of this tradition. They argue that social science is about the development of narratives, not theories. As such, they stress the importance of understanding and the impossibility of absolute knowledge claims, but they want to explain and they defend a limited notion of objectivity. Broadly, Bevir and Rhodes are within the hermeneutic, rather than the postmodern, or post-structuralist, stream of the interpretist tradition. As such, they follow Geertz and others in arguing that it is possible to produce explanations within the interpretist tradition. However, their understanding of explanation is very different from that of a positivist. In their view, the researcher can produce an explanation of an event or of the relationships between social phenomena. But, this explanation is built upon their interpretation of the meanings the actors involved gave to their actions. What is produced is a narrative which is particular, to that time and space, and partial, being based on a subjective interpretation of the views of, most likely, only some of the actors involved. Consequently, any such narrative must be provisional; there are no absolute truth claims. However, Bevir and Rhodes do wish to make some, more limited, knowledge claims. They argue (forthcoming): 'Although we do not have access to pure facts that we can use to declare particular interpretations to be true or false, we can still hang on to the idea of objectivity.' They suggest that a field of study 'is a co-operative intellectual practice, with a tradition of historically produced norms, rules, conventions and standards of excellence that remain subject to critical debate, and with a narrative content that gives meaning to it' (original emphasis). They continue: [Practice, tradition and narrative provide] for a negotiated and dynamic set of standards through which rational debate and argumentation between proponents of rival perspectives or approaches is possible [where] these standards are historically embedded within social practices, traditions and narratives which provide 'embedded reasons' ... for judging an argument true or false or an action right or wrong. Such criteria are not universal or objective; rather, they are 'shared criteria for assessing ... knowledge claims'. To Bevir and Rhodes, postmodernism errs in failing to acknowledge 'significant, grounded rationality' that is to be found in these practices and traditions. In Bevir and Rhodes' view, such knowledge claims are not self-referential because they can be 'reconfirmed' at three distinct points: The first is when we translate our concepts for fieldwork: that is, are they meaningful to practitioners and users and if not, why not? The second is when we reconstruct narratives from the conversations: that is, is the story logical and consistent with the data? And the third is when we redefine and translate our concepts because of the academic community's judgement on the narratives: that is, does the story meet the agreed knowledge criteria? Overall, they argue: To overcome this difficulty, we should conceive of objective knowledge, less as what our community happens to agree on, and more as a normative standard embedded in a practice of criticising and comparing rival accounts of 'agreed facts'. The anti-foundational nature of this practice lies in its appeal, not to given facts, but to those agreed in a particular community or conversation. In addition, and of key importance, the normative, critical bite of our approach lies in conducting the comparison by the rules of intellectual honesty. These rules originate in anti-foundationalism and not in a straightforward acceptance of the norms of the relevant community or conversation. 30 Ontology and Epistemology in Political Science David Marsh and Paul Furlong 31 As we can see then, there are a number of variants within the interpretist tradition. However, they are all anti-foundationalist and critical of positivism. These approaches have become much more common in political science since the 1970s for a number of reasons. First, increasingly philosophical critiques have led to the questioning of positivism. Second, the postmodern turn in social science has had an affect on political science, although much less so than in sociology. Third, normative political theory has changed fundamentally. Historically, it was foundationalist; the aim was to establish some absolute notion of the good or of justice. As Buckler argues in Chapter 8, that is no longer the case. Some normative political theorists have been influenced by postmodernism, again variously defined, and more by the work of Quine and others. Now, most political theorists are anti-foundationalists or, at the very least, have a very limited conception of any universal foundations. Fourth, as Randall shows in Chapter 5, much, but by no means all, feminist thought has been strongly influenced by postmodernism; it is anti-foundationalist and operates within the interpretist tradition. As such, we can see the influence of this interpretist tradition very broadly across political science. Realism Realism shares an ontological position with positivism, but, in epistemo-logical terms, modern realism has more in common with relativism. The core views of classical realism are again fairly clear and owe much to Marx's work: • To realists, the world exists independently of our knowledge of it. In ontological terms they, like positivists, are foundationalists. • Again like positivists, realists contend that social phenomena/structures do have causal powers, so we can make causal statements. • However, unlike positivists, realists contend that not all social phenomena, and the relationships between them, are directly observable. There are deep structures that cannot be observed and what can be observed may offer a false picture of those phenomena/ structures and their effects (for an excellent exposition of this position see Smith, in Hollis and Smith 1990: 205-8). But, as Smith puts it, although we cannot observe those structures, 'positing their existence gives us the best explanation of social action. To use a phrase familiar to the philosophy of science, we are involved in "inference to the best explanation" (Hollis and Smith 1990: 207). As such, to a realist there is often a dichotomy between reality and appearance. This is a very important issue because it has clear methodological implications. It means that realists do not accept that what appears to be so, or, perhaps more significantly, what actors say is so, is necessarily so. As an example, classical Marxism, and Marxism is the archetypal classical realism, argued that there was a difference between 'real' interests, which reflect material reality, and perceived interests, which might be manipulated by the powerful forces in society. Given this view, we cannot just ask people what their interests are, because we would merely be identifying their manipulated interests, not their 'real' interests. The criticisms of classical realism were of two sorts, which reflect different epistemological positions. The positivists denied the existence of unobservable structures. More importantly, they argued that positing them makes the knowledge claims of realism untestable and, thus, unfalsifiable. As such, realist claims that rely on the effect of unobservable structures have the same status to positivists as the claims of scholars from within the interpretist tradition. In contrast, authors from the interpretist tradition criticise the foundational claims of realism. In their view, there are no structures that are independent of social action and no 'objective' basis on which to observe the actions or infer the deep structures. So, the realist claim that structures cause social action are rejected on ontological and epistemological grounds. In our view, contemporary realism has been significantly influenced by the interpretist critique. In particular, this modern critical realism acknowledges two points. First, while social phenomena exist independently of our interpretation of them, our interpretation/understanding of them affects outcomes. So, structures do not determine; rather they constrain and facilitate. Social science involves the study of reflexive agents who interpret and change structures. Second, our knowledge of the world is fallible; it is theory-laden. We need to identify and understand both the external 'reality' and the social construction of that 'reality' if we are to explain the relationships between social phenomena. Realism also has clear methodological implications. It suggests that there is a real world 'out there', but emphasises that outcomes are shaped by the way in which that world is socially constructed. As such, it would acknowledge the utility of both quantitative and qualitative data. So, for example, realists might use quantitative methods to identify the extent to which financial markets are 'globalised'. However, they would also want to analyse qualitatively how globalisation is perceived, or discursively constructed, by governments, because the realist argument would be that both the 'reality' and the discursive construction affect what government does in response to global pressures. We shall return to this example later. Modern realism then attempts to acknowledge much of the interpretist critique, while retaining a commitment to causal explanation. The key 32 Ontology and Epistemology in Political Science problem here of course is that it is not easy, indeed many would see it as impossible, to combine scientific and interpretist positions because they have such fundamentally different ontological and epistemological underpinnings, one focusing on explanation and the other on understanding (on this point, see Hollis and Smith 1990: 212). One of the main criticisms of realists has been that they often treat concepts as if they related to some fixed, or at least more or less given, 'essences' or cores. It should be noted first that this is not a necessary tenet for realists; it reflects rather the philosophical traditions from which they derive. Nevertheless, the question of what a concept is for is an important one, and it affects ontology directly. If a concept cannot be tied firmly to an underlying reality, as traditional philosophy seems to imply, the concept of 'being' itself may be detached from the real world of experience. This is one of the reasons why modern philosophy has considerable difficulty even recognising that there may be a subject of ontology. It should also be noted that this is one of those issues on which positivists and interpretists can find themselves temporarily in agreement, even though, as we have seen, they have fundamentally different views about knowledge and being. Any apparent agreement between them, however, has limited scope, as they have different origins and are heading for different destinations. Having considered how these categories relate to some important issues in the social sciences, we can now move on to apply the arguments to particular cases so as to illustrate their use and their limits. Ontology and epistemology in political science: two cases The aim in this section is to examine how a researcher's ontological and epistemological position affects the way they approach empirical questions in political science. We shall focus on two areas: globalisation and multilevel governance. These areas have been chosen because they reflect a broad spread of the concerns, but, in our view, similar arguments could be made in relation to other substantive areas. Case J: Globalisation The literature on globalisation mushroomed in the 1990s. It has been common to distinguish between processes or aspects of globalisation: so many authors have distinguished between economic, political and cultural processes, while acknowledging that they are interrelated. In this vein, many have argued that economic globalisation has grown apace and that this process has significantly restricted the autonomy of the nation state. r David Marsh and Paul Furlong 33 Indeed, Ohmae (1990) goes as far as to argue that only two economic forces, global financial markets and transnational corporations, will play any role in the politics of the future. In his view, the future role of states will be analogous to the current role of local councils. At the same time, other authors have focused on cultural globalisation, suggesting that world culture is becoming increasingly homogeneous: in the view of most, reflecting a growing US hegemony. Certainly, there is little doubt that the issue of globalisation in a crucial one for those interested in questions of contemporary political economy and governance. Political scientists have probably been most concerned with economic globalisation and the way in which it restricts the autonomy of the state, and have utilised a foundationalist ontology and a positivist epistemology, although, as we shall see below, some more recent work is realist. In contrast, sociologists, particularly those who focus on cultural studies, concentrate upon cultural globalisation, operating from an anti-founda-tionalist and interpretist position. The main debate about economic globalisation has concerned the extent to which it has increased. There are two main positions. Some authors, like Ohmae (1990), who are christened hyperglobalists by Held et al. (1999) and seen as first-wave theorists by Hay and Marsh (2000), argue that there has been a massive increase in various indicators of economic globalisation: direct foreign investment; international bank lending; transnational production; international trade and so on. In contrast, authors such as Hirst and Thompson (1999), christened sceptics by Held et al. (1999) and seen by Hay and Marsh (2000) as second-wave theorists, argue that the process is more limited. More specifically, they suggest that: globalisation is not a new phenomenon; regionalisation, rather than globalisation, is a better description of the changes that have occurred; and the only area in which there has been significant globalisation is in relation to financial markets. We are not concerned here with the detail of this argument. Our point is that both sets of authors agree about what constitutes evidence of globalisation and how we can go about studying that evidence. Globalisation is an economic process that can be measured quantitatively, indeed there is large agreement as to the appropriate measures, and which, to the extent that it exists, has an effect on patterns of governance. More recently, other authors have been, in most cases implicitly rather than explicitly, critical of this ontological and epistemological approach. The point is easily made if we return to two ways of classifying the literature on globalisation to which we have already referred. Held et al. (1999) contrast hyperglobalist and sceptical approaches to globalisation with a third approach to which they adhere: the transformationalist thesis. In contrast, Hay and Marsh (2000) identify a third wave of the globalisation literature that builds upon a critique of the first two waves. These two 34 Ontology and Epistemology in Political Science 'third ways' share something in common, but do differ significantly in a manner that reflects ontological and epistemological debates. The transformationalists differ significantly from the sceptics in that they share: a conviction that, at the dawn of a new millennium, globalisation is a central driving force behind the rapid social, political and economic changes that are reshaping modern societies and world order ... In this respect, globalisation is conceived as a powerful transformative force which is responsible for a massive shake out of societies, economies, institutions of governance and world order. (Held et al. 1999: 7) Held et al. also emphasise the major way in which the transformationalist account parts company with both the other two positions (1999: 7): The transformationalists make no claims about the future trajectory of globalisation ... Rather [they] emphasise globalisation as a long-term historical process which is inscribed with contradictions and which is significantly shaped by conjunctural factors. So, they argue that: there are 'real' social, political and economic changes occurring in the world; globalisation is a cause of these changes, a transformative force; but there is no inevitable process of globalisation which, as social scientists, we can identify. This last point is especially important here. The putative development of globalisation is dependent on the actions of agents, whether individuals, companies, institutions or states; as such it is a socially constructed process. It seems clear then that the transformative position is a realist one. This position has methodological consequences. It points strongly to comparative analysis, because the emphasis is upon how different countries, and indeed different companies and markets, are affected by, and respond to, this process of globalisation in different ways. If globalisation is not an inevitable or universal process, then we need to focus on how it is differently experienced in different contexts. This point is even clearer if we turn to what Hay and Marsh call the third-wave literature on globalisation. Hay and Marsh (2000: 6) follow Held et al. in arguing that we: 'shouldn't make essentialising and reifying assumptions about the effects, consequences, or even the very existence, of globalisation'. Rather, globalisation is a series of contradictory and contingent processes. More specifically, they suggest that, for many authors, especially the hyperglobalists, globalisation is a process without a subject. In contrast, they argue that it is agents who construct globalisation and, as such, the researcher should identify the actors involved and how they perceive and discursively construct globalising tendencies. r David Marsh and Paul Furlong 35 However, Hay and Marsh go further to contend that these discursive constructions have significant effects on outcomes. So, they suggest that it is the discursive construction of globalisation that affects government economic policies, rather than the 'real' processes of globalisation. As such, and taking the UK as an example, their argument would run along the following lines: • While there has been a significant increase in regionalism in patterns of trading and a globalisation of financial markets, there is limited evidence that Britain is locked into a globalised political economy which determines the economic policy which the British Government can adopt. 1» However, British governments, and especially the Blair Government, have argued that it is constrained in that way. It suggests that the extent of globalisation is such that the pursuit of neo-liberal policies is inevitable: there is no alternative. • As such, it is not the 'reality' of globalisation that is shaping British economic policy, but the dominant discursive construction of that reality. We are not concerned here about the validity or otherwise of this argument. The crucial point for us is that this view clearly marks a break with the positivism that underpins most work on globalisation. To Hay and Marsh, there may be 'real' processes at work, but the way they affect outcomes is mediated by the discursive construction(s) of these processes. This argument has both realist and interpretist elements. There is an appeal to a real world, but the emphasis is on the discursive construction of that world. This position illustrates how realist and interpretist positions interface. In our view, this position is a realist one if it recognises |* that there is an interactive or dialectical relationship between the 'real' | world and the discourses. A realist would acknowledge not only that * discourses have real effects, in this case that the dominant discourse of globalisation shapes economic policy, but also that the 'real' processes of globalisation constrain the resonance of different discourses. So, if the dominant discourse is at odds with the 'reality', alternative discourses can appeal to that 'reality deficit'. However, if it is merely the discourses that have the causal power, then, in our view, it is an interpretist position. There are other approaches to globalisation which are clearly located in an interpretist tradition. As we emphasised above, most of these ap-proaches stress cultural globalisation. Of course, as Held et al. point out | (1999: 328), the concept of culture has a long and complex history but | 'normally refers to the social construction, articulation and reception of meaning'. This definition immediately suggests an anti-foundationalist K ontology and, most often, an interpretist epistemology. 36 Ontology and Epistemology in Political Science David Marsh and Paul Furlong 37 It is possible to approach the issue of cultural globalisation utilising a positivist epistemology. So, one could focus empirically on the extent to which certain cultural icons, such as Coca-Cola, McDonald's or Madonna, have become universal, or whether colonialism was associated with a similar global culture. However, the focus of a cultural studies approach to globalisation is much more likely to be on difference. Two points are important here. First, the argument would be that there are various discourses about globalisation, none of which is 'true', although at any time one discourse may be dominant. Second, while one discourse may dominate, it can be, and will always be, resisted: different agents - citizens and researchers - will offer different narrations of globalisation and its effects. In this way, this alternative 'cultural studies' approach reflects an anti-foundationalist and an interpretist position. Case 2: Multilevel governance Multilevel governance and intergovernmentalism: realism versus positivism The term 'multilevel governance' covers a variety of familiar phenomena that are normally located in the areas of regional policy and European integration. Though 'multilevel governance' (MLG) is rapidly acquiring the status of a fashionable mainstream concept, it is not as established as 'globalisation' in the vocabulary of politicians and commentators. Here again, the contemporary debates in large part reflect different ontological and epistemological positions. In this case study, we concentrate not on the different uses of the term, although these can be significant, but rather on the contrast between MLG and its main opponent, which is liberal intergovernmentalism. A useful definition of MLG is provided by Hunt (1999): '[According to multilevel governance theories] the policy process involves the interaction between a constellation of public and private actors located at the supranational, national and sub-national level.' This interaction is usually understood as non-hierarchical and as lacking a central, predominant authority, and similar usages can be found in Marks et al. (1996) and Armstrong and Bulmer (1998). These theorists argue against the view of the EU as an international organisation whose decision-making is based predominantly on national interests determined by member states, a view known as intergovernmentalism. The intergovernmentalist perspective is closely associated with international lawyers, but an influential political analysis is provided by Andrew Moravscik (1993), who argues that the European policy process can be understood as a nested game played out both in the domestic politics of member states and in the international arena of the EU's institutions. While the MLG theorists derive their frameworks from institutionalist perspectives (see Chapter 4), arguing that 'institutions matter' in shaping interaction, analysts such as Moravscik generally utilise rational choice perspectives (see Chapter 3 below). Both approaches would claim to be empirically grounded, but the nature of the empirical grounding differs. Most MLG theorists are realists in epistemological terms, emphasising how the continuity of rules, norms and operating procedures, and sometimes of 'deep, non-observable structures', can, and does, determine the outcomes of decision-making in the long term. As such, their logic is inductive rather than deductive. Overall, MLG is not so much concerned with the debate between neo-federalists and intergovernmentalists as with the consequences of different possible forms of integration for normative issues such as democratic participation, effective government and distributive justice. In contrast, liberal intergovernmentalists seek to identify the preferences and the parameters of the individual actors (usually member states) and show how, after the event, the outcomes can be understood as the result of rational calculated behaviour. Their logic is therefore deductive: they argue from the general to the particular. Liberal intergovernmentalism is foundationalist in ontological terms and operates with a positivist epistemology. In its treatment of European integration it is clearly unsympathetic to neo-federalism and to supranationalism. The normative underpinnings of multilevel governance and intergovermentalism MLG theorists argue that, rather than conceptualising regional policy as a national issue in which the lead role is taken by the national state institutions, it should be identified as an arena in which the European Union plays an integral role in policy-making, together with the separate regional authorities and the central national institutions. In this sense, theories of multilevel governance make a distinction between 'government' and 'governance'. 'Government', it is argued, is too narrowly concerned with the formal structures of state authority, and with the associated processes and issues, whereas 'governance' is concerned with much wider notions of politics, encompassing the production, accumulation and regulation of collective goods at all levels including the international. Power relations in multilevel governance are structured by reciprocal interdependence on each other's resources, rather than on conflict over either scarcity or fundamental values. Typically, these theories argue that relations of decision-making between the various levels should be seen as 38 Ontology and Epistemology in Political Science loosely interconnected rather than as tightly nested; that is, characterised by multilateral links, and non-hierarchical in form, rather than by a hierarchical chain of bilateral links in which the national state authority has a predominant role, as is the case with intergovernmental approaches. There is thus a strong normative element in multilevel governance. From describing the increased evidence of the multiplicity of decision-making forms and levels in European integration, proponents move to arguments about the value of multilevel governance in enhancing democratic legitimacy and effective decision-making under conditions of globalised political economy. In comparison with state-centred accounts, multilevel governance is said to be 'closer to the people', and therefore more acceptable, and more flexible and adaptable, so better able to respond to the rapidly changing economic climate (Marks et al. 1996). The arguments against multilevel governance, if it is regarded as a policy prescription as well as empirical analysis, concentrate on two main issues (Moravscik 1993; Scharpf 1988). The first is what is known as the 'joint decision trap'. This focuses upon the danger of deadlock in decisionmaking where there are many participants, interdependent arenas and a variety of possible combinations of policy-making processes. Though multilevel governance may offer the prospect of policy-making close to the people and greater legitimacy, it risks sacrificing efficiency in decisionmaking if there is no authoritative procedure for resolving disagreements among equal participants. A second criticism denies even that multilevel governance provides greater legitimacy and argues that, when the smaller units and more local levels of decision-making are included, the greater complexity of procedures results in opacity of decision-making and, therefore, in less accountability. In practice, multilevel governance can mean obscure elite-led agreements and public incomprehension. Neo-liberal arguments try to resolve these problems by emphasising how the member states in the EU remain both the focus of popular legitimacy, albeit with some rebalancing towards regional authorities, and the main guarantors of effective governmental decision-making. In response to this, Marks et al. (1996) have three main criticisms of the intergovernmental approach. Underlying these disagreements is a fundamental dispute about the nature of social reality. First, positivist explanations of societal phenomena neglect the structural constraints within which individuals operate. These are varied in kind, but the most important are generally the impact of differential allocation of resources, the culturally-given nature of the value framework within which individuals choose and the unpredictability of external factors, such as the international economic and security climate. Second, the realist perspective emphasises how the institutional frameworks have a primary effect in shaping decision-making through their David Marsh and Paul Furlong 39 formal rules, their informal procedures, their value structures and their effect on office-holding and internal role-oriented behaviour. In one sense, the institutions are no more than the sum of countless individual choices, but merely to state this does not get us very far. Realists seek to find ways of characterising different institutional frameworks so as to move beyond this and to introduce other levels of analysis and explanation which recognise the weight of the long-term structural and institutional context. Third, it is argued that intergovernmentalists are insufficiently critical about what time-frames are relevant and why. The term 'path dependency' used in this context (Pierson 1996) does not just refer to the given nature of resource allocations at 'point zero' which the researcher takes as the start. It also directs our attention to the impact of decisions prior to point zero, and of the ways in which the institutional frameworks lock actors into particular sets of choices. This implicitly asks positivists to justify why they adopt diachronic modes of explanation, which imply social understanding as a set of discrete operations in fixed points in time, as opposed to synchronic explanations, which emphasise a more continuous and context-led understanding of the social nature of time. Despite these epistemological and methodological differences, writers such as Moravscik and Scharpf (Scharpf 1988, 1997) appear to be able to integrate some of the concerns of multilevel governance into their own perspective, so that, despite the methodological differences between the positivists and realists, we can identify these as distinct strands in the study of the European policy process, marked by an attentiveness to similar policy problems and with some of the same language. The constructivist approach This is not true of the social constructivist approach (see, for example, Jorgensen 1998; Weldes 1996; Wendt 1994). This rejects the language of causality, with which positivists and realists are content in their different ways, and in contrast, is based upon an interpretist epistemological position. Constructivists argue that, if there is a problem of increasing complexity of decision-making associated with the decline of the nation state, this complexity must be understood as an intentional social construct on the part of decision-makers, part of a set of political projects associated with responses to perceptions of external and internal constraints. The questions which arise are concerned with political decision-making as a series of attempts to resolve conflicts over meaning and identity, understood in the broadest sense. Constructivists take issue with the positivist understanding of the nature of political choice. They argue against the acceptance of individual preference as a given and instead interrogate specifically why and how preferences come to be formed and 40 Ontology and Epistemology in Political Science David Marsh and Paul Furlong 41 how these preferences and choices relate to the strategic aims of powerful interests in society. Multilevel governance then would not be seen as a set of objectively perceived phenomena, but as a normative framework which is itself part of the political conflict between the interests associated with neo-liberal economic restructuring and those seeking a more social democratic accommodation with technological change. This locates the arguments about multilevel governance within the discussion about the nature of globalisation, which we dealt with above, and in which one of the main disputes is about the underlying realities of technological economic and social change and their relationship with the discursively constructed political uses made of them within specific political projects. Conclusion The point here is not to attempt to resolve these disputes. Rather, what we have sought to do is to show how the different approaches in different issues relate to epistemological and methodological assumptions, and to one another. The terms introduced here can be used as signposts, suggesting how we can come to terms with the deeper implications of the theories and groups of concerns which are the focus of the individual chapters which follow. One of the temptations in so doing is to attempt to find a synthesis of all the available positions, in the hope that, at some level of analysis, agreement is possible over fundamental issues. Unfortunately, experience and logic combine to warn against this temptation. These debates have been part of the intellectual and moral climate of Western thought for centuries and continue because they reflect disagreements not just about logic or technicalities but also about the proper scope of human action in society. In other words, they are questions which relate to deep-rooted moral positions that may be internally coherent, but are incompatible with one another, except in so far as they all include some appeal to intellectual and ethical tolerance of diversity. In the face of these difficulties, another strategy, alluring at least to risk-averse researchers, is to avoid the issue. Far from being safe, this position is actually rather unsafe, since it does not enable one to distinguish between good and bad research and between good and bad arguments. The least one can say about these issues is that they are of sufficient importance to warrant a genuine commitment to coming to terms with them. Coming to terms with the issues requires one to think through the different arguments separately, to compare them and to evaluate them. As we argued at the beginning of this chapter, this means identifying, as far as possible, what are the epistemological and ontological underpinnings and what these imply in terms of argumentation, practical research method, explanatory logic and research construction. The purpose of this chapter has been to encourage this and to attempt to provide an introduction to some of the main ideas and methods involved. Further reading • The best introductions to the philosophy of science and social science are Chalmers (1985, 1990) and Winch (1958). • For an accessible overview of ontology and epistemology, see Hav (2002). • On the positivist approach, see Kuhn (1970), Hempel (1965, 1966) or Halfpenny (1982). • On the interpretive approach, see Bevir and Rhodes (1999). • On realism, see Sayer (1992).