A prominent faultline that continuously surfaces in debates over modes of inquiry in political science is the one between positivist or scientific approaches on the one hand and postpositivist or interpretive approaches on the other.Footnote 1 Alas, these debates are often conducted with very little reflection on the philosophical underpinnings of the relevant approaches.Footnote 2 In these debates, concepts like “positivist” and “postpositivist” are often associated with methodological choices—quantitative or qualitative—at least as much as philosophical commitments—naturalism or anti-naturalism. This lack of philosophical reflection can result in a skewed understanding of the issues at stake in the debates over an adequate political science.Footnote 3 For example, when methods are judged solely in pragmatic terms (i.e., in terms of their substantive utility for certain lines of inquiry), it might seem possible to reconcile methods that are in fact irreconcilable from a philosophical standpoint. We will argue, more particularly, that when political scientists lump all qualitative approaches together, they neglect the philosophical chasm that separates naturalist and anti-naturalist uses of qualitative and interpretive methods.
The distortion that arises from a neglect of philosophical issues is perhaps nowhere as evident as in the division between qualitative and quantitative approaches. In methodological debates, this division often gets mapped onto that between positivism and postpositivism. But, in philosophical terms, the qualitative methods camp is in fact split between some who share the philosophical naturalism of so much positivism, and others who seek to distance themselves from just such naturalism.Footnote 4
A split between naturalists and anti-naturalists haunts even the recently formed Organized Section on Qualitative Methods within the American Political Science Association. On the one hand, many qualitative scholars neglect philosophy while making implicit naturalist assumptions and trying to build bridges between qualitative and quantitative methods.Footnote 5 Thus the first issue of the Section's Newsletter defined its scope in methodological terms: “case study methods, small N analysis, comparative methods, concept analysis, the logic of inquiry, comparative historical methods, the ethnographic tradition of field research, constructivist methods, interpretive methods.”Footnote 6 On the other, some qualitative scholars raise philosophical concerns, challenge attempts to build bridges to the positivism they associate with quantitative approaches, and insist on the interpretive nature of political science. Indeed, there has been an upsurge of avowedly constructivist or interpretive approaches in most subfields of the discipline including policy analysis, international relations, and feminist research.Footnote 7 Thus, in a recent issue of the Section's Newsletter, Bernhard Kittel complained that “the qualitative response to the quantitative template seems to have embraced too many [of the latter's] assumptions” in a way that “blatantly disregards important developments both in the natural sciences and in the philosophy of science.”Footnote 8 In an earlier issue of the Newsletter, Dvora Yanow went so far as to argue that an interpretive approach was not a subfield of qualitative methods since it does “not live under the same philosophical umbrella.”Footnote 9
In this article, we offer a philosophical critique of the naturalist tendencies within the qualitative methods group. Specifically, we defend an anti-naturalist philosophy, and then use it to challenge the naturalist tendencies within qualitative concept formation as developed most prominently by Giovanni Sartori and David Collier. Sartori and Collier have largely neglected the philosophical dimension of issues of concept formation, and adapted conceptual strategies dominated by implicit naturalist assumptions. But, in our view, these assumptions are inappropriate to the human sciences given the meaningful and contingent character of human action and the situatedness of the social scientist.
We begin our essay, in the first section, by articulating the philosophical distinction between naturalism and anti-naturalism. In the second section, we then show how naturalism and anti-naturalism inspire different strategies of concept formation. We argue that naturalism encourages the formation of concepts characterized by reification, essentialism, and an instrumentalist view of language. In the final section of the paper, we draw on the framework set up in the previous parts of the paper to subject qualitative concept formation to a philosophical critique. We focus on the two most prominent theorists and practitioners of qualitative concept formation, Giovanni Sartori and David Collier, and show how their conceptual strategies embody a reification, essentialism, and instrumentalist view of language associated with naturalism.
So, we seek to draw attention to the philosophical dimension of methodological disputes. We offer a challenge to qualitative methodologists to explain how they would defend the appropriateness of their tools and strategies in the face of an anti-naturalist critique. We also prepare the philosophical groundwork for an alternative, anti-naturalist perspective on the formation of social science concepts. We do so in the belief that a debate about philosophical issues will ultimately improve and enrich political science regardless of its particular outcomes. Indeed, we want above all to expose the shaky philosophical foundations of qualitative methods; we are less concerned to provide a detailed account of an alternative methodological edifice. Such critique constitutes, of course, a well-established intellectual practice in its own right within modern philosophy and social science. Critique serves to expose unfounded assumptions, demarcate new fields of debate, and stimulate scientific innovation. We hope that our present critical voyage will likewise lead us, as well as our readers, into new geographies of political inquiry.
Naturalism and Anti-Naturalism
One problem with the usual faultline between positivism and postpositivism is the unhelpful way in which it muddles philosophical and methodological concerns. The term “positivism” often fuses foundationalist empiricism and naturalism with quantitative methods in opposition to a “postpositivism” that fuses postfoundationalism and perhaps anti-naturalism with qualitative and interpretive methods.Footnote 10 But things are not that simple. Many proponents of qualitative methods are, for example, naturalists who have doubts about positivist epistemologies, while some proponents of qualitative methods appear to embrace a rather naïve empiricism in justifying their methods as ways of getting at facts that elude quantitative scholars. We want, therefore, deliberately to break with the usual fault line between positivism and postpositivism to focus on philosophical issues. We focus initially on the specific distinction between naturalism and anti-naturalism.
The dazzling achievements of the natural sciences have exerted an enormous pressure on the human sciences, including a powerful drive to model the latter on ontological and epistemological foundations associated with the former. Naturalism arises from the belief that similarities between the natural and social worlds are such that they should be studied in the same ways. Initially naturalists wanted mainly to preclude appeals to supernatural explanations: they argued that humans were part of nature and so amenable to empirical study, and they insisted upon a scientific method based on the rigorous collection and sifting of facts. Before long, however, naturalism became ensnared with the positivist conviction that the same logic of inquiry applies to both the natural and human sciences. Hence we can define naturalism as the idea that the human sciences should strive to develop predictive and causal explanations akin to those found in the natural sciences.Footnote 11 In the classic statements of this view, the human sciences study fixed objects of inquiry that possess observable and, at least to some extent, measurable properties, such that they are amenable to explanations in terms of general laws, even if these general laws sometimes involve assigning probabilities to various outcomes.
In the past few decades, however, philosophers of social science have typically come to favor anti-naturalism. The critique of naturalism has developed over the past half-century within a variety of philosophical traditions. Anti-naturalism has been most clearly and consistently articulated within the hermeneutic tradition, starting with the work of Wilhelm Dilthey at the turn of the twentieth century and developed more recently by Hans-Georg Gadamer, Paul Ricoeur, and others.Footnote 12 In the social sciences, Max Weber was one scholar who incorporated some hermeneutic themes. He insisted that causal explanation in social science relied in large part on verstehen (interpretive understanding).Footnote 13 Weber also insisted on the singularity of such causal explanation; it is a form of explanation that seeks the contextually specific causes of historical particulars.Footnote 14 Indeed, Weber developed his ideal-typical strategy of concept formation, in his famous essay on “Objectivity,” by way of an explicit critique of the naturalist tendency “to require the analysis of all events into generally valid ‘laws’.”Footnote 15 Nowadays anti-naturalism has also become dominant within analytic (or post-analytic) philosophy. Its dominance therein began in the latter half of the twentieth century following the leads provided by Ludwig Wittgenstein,Footnote 16 Alasdair MacIntyre,Footnote 17 and Charles Taylor.Footnote 18 Additional contributions to the rise of anti-naturalism have come from phenomenologyFootnote 19 and pragmatism,Footnote 20 and, within the social sciences, from ethnomethodologyFootnote 21 and cultural anthropology.Footnote 22
Anti-naturalists argue that constitutive features of human life set it apart from the rest of nature to such an extent that the social or human sciences cannot take the natural sciences as a model. The relevant features of human action are that it is meaningful and historically contingent. Let us explore them in turn before then emphasizing that they apply as much to social scientists as to those whom they study.
We will begin with the meaningfulness of social life. Some naturalists hold a positivist epistemology according to which causal explanations are validated by their fit with observations, and meanings are irrelevant because they are not observable. These positions informed, for instance, classical behaviorism as propounded by John B. Watson and B. F. Skinner.Footnote 23 However, because this positivist epistemology is rarely espoused nowadays, we will concentrate on naturalists who would agree that human actions have meanings for those who perform them. It is widely accepted today that agents act for reasons of their own, albeit that we sometimes take the reasons to be tacit, subconscious, or even unconscious, as opposed to explicit and conscious. What divides naturalists and anti-naturalists is the role they give to meanings in the explanation of actions and so in the explanation of practices and institutions arising out of actions.Footnote 24 Naturalists typically want meanings to drop out of these explanations. Philosophical exponents of naturalism argued, for instance, that to give the reasons for an action was merely to redescribe that action. If we want to explain an action, they added, we have to show how it—and so no doubt the reason for which the agent performed it—conforms to a general law couched in terms of social facts.Footnote 25
Anti-naturalists refuse to let meanings or beliefs drop out of explanations in the human sciences. They argue that meanings are constitutive of human action. Hence, as Clifford Geertz famously claimed, social science needs to be “not an experimental science in search of law but an interpretive one in search of meaning.”Footnote 26 Some naturalists attempt to rebuff anti-naturalism by equating social science with the study of systems or structures that cannot be understood as the intended consequence of a single action. Traffic jams are often evoked as examples of such structures.Footnote 27 But traffic jams and other such structures scarcely undermine anti-naturalism. Most of what we want to know about traffic jams comes down to intentional action. To explain why people are driving when and where they are, we want to know whether they intend (consciously or not) to go to work, to a sports game, shopping, visiting relatives, and so on. Even more generally, we might explore the wider webs of belief that constitute the social practices within which these intentions are embedded. Why do people believe that driving to work is better than using public transportation? Why don't they take political action to increase public investment in transportation infrastructure? All such questions are questions about meaningful intentionality. If an account of traffic jams or other such structures really did ignore intentionality, it would be a very thin and inadequate account. It could tell us only in purely physical terms that the traffic jam arose because a given number of people tried to drive cars along a stretch of road of given dimensions. It could tell us nothing about the actions that led to these physical consequences; it could not tell us why these people were driving their cars or why the road system is as it is.
Anti-naturalists uphold the centrality of meanings for social science on the grounds not only that actions are meaningful but also that these meanings are holistic. In this view, we can properly understand and explain people's beliefs only by locating them in a wider context of meanings. Meanings cannot be reduced to allegedly objective facts since their content depends on their relationship to other meanings. The human sciences require a contextualizing form of explanation that distinguishes them from the natural sciences. Anti-naturalism has increasingly drawn, then, on holistic theories of meaning. The very idea of a hermeneutic circle asserts such meaning holism. As Gadamer wrote—referring back to the work of Friedrich Schleiermacher at the end of the eighteenth century—“as the single word belongs in the total context of the sentence, so the single text belongs in the total context of a writer's work.”Footnote 28 Semiotics too treats signs as acquiring content or meaning from their place within a system of signs; it does so in the strand of semiotics that derives from Charles Peirce's pragmatism as well as in that which derives from Ferdinand de Saussure's structuralism.Footnote 29 Likewise, many analytic and post-analytic philosophers argue that concepts can refer, and propositions can have truth conditions, only in the context of a web of beliefs or a language game.Footnote 30 Hermeneutics, semiotics, and contemporary analytic philosophy thus point to the importance of elucidating and explaining meanings by reference to wider systems of meanings, rather than by reference to categories such as social class or institutional position, and rather than by construing ideas or meanings as “independent variables” within the framework of naturalist forms of explanation. All these forms of philosophy thus lend support to an anti-naturalism that is at odds with the leading ways of doing political science. It is worth adding perhaps that the dominance of holism in contemporary philosophy—as observed even by skepticsFootnote 31—suggests that naturalism might prove a difficult doctrine for political scientists to defend.
Let us turn now to the historically contingent nature of human action. When naturalists try to let meanings drop out of their explanations, they are usually hoping at least to point toward classifications, correlations, or other regularities that hold across various cases. Even when they renounce the ideal of a universal theory or law, they still regard historical contingency and contextual specificity as obstacles that need to be overcome in the search for cross-temporal and cross-cultural regularities. Greg Luebbert, for example, discusses a number of discrete national case studies but his ultimate aim is to find “a single set of variables and logically consistent causal connections that make sense of a broad range of national experiences.”Footnote 32 Naturalists characteristically search for causal connections that bestride time and space like colossi. They attempt to control for all kinds of variables and thereby arrive at parsimonious explanations. But they can do so only by “freezing history.”Footnote 33
In stark contrast, anti-naturalists argue that the role of meanings within social life precludes regularities standing as explanations. That said, we need to be careful how we phrase what is at issue here. Anti-naturalists have no reason to deny that we might be able to find or construct general statements that cover diverse cases. Rather, they typically object to two aspects of the naturalist view of generalizations. First, anti-naturalists deny that general statements constitute a uniquely appropriate or powerful form of social knowledge. To the contrary, they consider statements about the unique and contingent aspects of particular social phenomena to be at least as apposite and valuable as general statements. Generalizations, in the anti-naturalist view, often deprive our understanding of social phenomena of what is most distinctly and significantly human about them. Second, anti-naturalists reject the claim that general statements can provide explanations of features of the particular cases: just as we can say that X, Y, and Z are all red without explaining anything else about them, so we can say that X, Y, and Z are all democracies but that does not explain any other feature they might have in common. Anti-naturalists oppose explanations of human actions in terms of trans-historical generalities because they conceive of human action as being inherently contingent and particular. Human life is characterized by ineluctable contingency, temporal fluidity, and contextual specificity. Hence we cannot explain social phenomena adequately if we fail fully to take into account both their inherent flux and their concrete links to specific contexts.
Anti-naturalists argue, in other words, that the human sciences require a historical and contingent form of explanation that distinguishes them from the natural sciences. Wittgenstein's claim that the meaning of a word cannot be elucidated in abstraction from the specific context in which that word was used points toward just such a historically contingent mode of knowledge.Footnote 34 Indeed, this account of meaning leads Wittgenstein explicitly to conclude that no explanation is ever final since it is always limited to a specific context.Footnote 35 Likewise, analytic and post-analytic philosophers now often argue that the human sciences deploy languages that presuppose ideas of choice and contingency that are quite at odds with the forms of explanation found in the natural sciences. In this view, the meaningful nature of actions implies that to explain them we have to invoke the reasons of the actors, thereby implying that the actors could have reasoned and acted differently: actions are the products of contingent decisions, not the determined outcomes of law-like processes.Footnote 36 Much contemporary philosophy thus points to the importance of narrative explanations that work by unpacking the contingent and particular conditions of actions and events, rather than by searching for trans-historical models, classifications, or correlations.
So, anti-naturalists emphasise the meaningful and contingent nature of action. To conclude this section of the article, we want to point out that meaningfulness and contingency apply to social scientists as much as to their objects of inquiry. Social scientists too come to hold particular webs of belief against the background of contingent traditions. Naturalists usually treat the situatedness of the social scientist as an obstacle to be overcome in the creation of proper knowledge. They require social scientists to try to abstract themselves from their historical perspectives. They argue that social scientists can produce valid scientific knowledge only if they divest themselves of their prejudices. In contrast, anti-naturalists usually deny the very possibility of abstracting ourselves from our prior webs of beliefs. They suggest that social science always takes place from within particular linguistic, historical, and normative standpoints. The questions asked and the concepts formed by social scientists are always informed by their existing webs of belief and by their ordinary language, which thus play an active role in shaping their scientific work.
The combined recognition of, on the one hand, the situatedness of the social scientist and, on the other hand, the meaningfulness of social life introduces a dialogical dimension to social science. Naturalists typically construe explanation as the product of a unidirectional subject-object relationship. Their neglect of the constitutive role of meanings leads them to see the social scientist as the only agent involved in crafting explanations: the objects of social science are just that—passive objects to be studied. In contrast, anti-naturalists often conceive of explanation as the product of a kind of dialogue between social scientists and those they study. Social science generally involves a subject-subject interaction in which the scholar responds to the interpretations or meanings of the relevant social actors. It involves a “fusion of horizons,”Footnote 37 that is, a process of reaching some kind of shared interpretation in which the social scientists' own views are often transformed. An encounter with the beliefs or meanings of social actors always has the potential to send out ripples through a scholar's own beliefs, altering their understanding of, say, their research agendas, the traditions in which they work, or their normative commitments.
We have thus arrived at the point where the meaningful and contingent features of action combine with the situatedness of the scholar and the dialogical view of social science. For anti-naturalists, actions are meaningful, meanings are contingent and liable to change over time, and these facts apply to the actions and beliefs of social scientists as well as those whom they study in a way that points to dialogical forms of explanation.
Two Views of Concept Formation
We have described our anti-naturalist philosophical perspective as a response and an alternative to naturalist conceptions of the social sciences. We now move on to discuss how naturalism and anti-naturalism inspire different approaches to concept formation in the social sciences. On the one hand, naturalism encourages strategies of concept formation that involve aspects of reification, essentialism, and an instrumentalist view of language. On the other, anti-naturalism challenges reified concepts for eliding the place of meanings, essentialist concepts for eliding the place of contingency, and linguistic instrumentalism for eliding the situatedness of the scholar and the dialogical nature of social science. Let us examine each challenge in turn.
Reification is one dimension to debates about concept formation. Anti-naturalism implies that many—perhaps even all—social science concepts denote objects that are composed at least in part of meanings or intentional states. Reification occurs whenever these concepts are defined either in ways that neglect relevant meanings entirely or in ways that neglect the holistic character of meanings, thereby likening human action to meaning-less “things.” Naturalism encourages such reification in that it ignores the constitutive role of meanings in social explanations, or at least it tears meanings from their holistic and contingent contexts so as to embed them in mechanistic explanations. Naturalists usually rely on reified concepts to elide the place of meanings in social science. Because reified concepts neglect intentionality, they enable naturalists to treat their objects of inquiry as if they were no different from those of the natural sciences. Indeed, reification occurs whenever the attributes of a social science concept are regarded as reducible to causal laws, probabilities, or fixed norms. For example, the concept of “social class” is reified insofar as it is understood in terms of supposedly objective socio-economic criteria such as relation to the means of production or income level, without taking into account how the members of a given social class themselves construe and experience their social situation. William Sewell, in his study of the development of the French working class from the ancien régime to 1848, shows the importance of workers' experiences and consciousness for the conceptualization of social class.Footnote 38
Social science concepts can exhibit reification in two possible ways. One type of reification consists in removing meanings from any constitutive or defining position in the conceptualization of human beings and their actions. Reified concepts of this kind never have referents that include meanings, except perhaps as epiphenomena of an allegedly non-meaningful stratum of social existence such as the relations of production. The more mechanistic versions of Marxist theory exemplify this kind of reification.
The second type of reification is more complex. It arises when meanings are given a role in social life and social explanation but when there is no recognition of the holistic and contingent nature of meanings. Reified concepts of this kind have referents that include meanings, but the relevant meanings are torn from their holistic contexts in order to be cast as “independent variables” within naturalist explanations; the relevant meanings are atomized in a way that enables them to be contained within naturalist forms of explanation. Sheri Berman, for example, appeals to beliefs or attitudes as explanatory factors only then to conceive of them in naturalist terms, casting them as mechanistic “independent variables.”Footnote 39 Similarly, the sociologist Richard Biernacki attempted, in his The Fabrication of Labor, to demonstrate that culture, which he construed as an “independent variable”, “parsimoniously explains a wide range of phenomena.”Footnote 40
Essentialism is another dimension to debates about concept formation. Anti-naturalism implies that meanings and actions are historically specific; we can understand actions only by locating them in their particular contexts. Essentialism occurs whenever social science concepts are defined in ways that ignore the historical specificity of the various objects to which they refer. Naturalism promotes essentialism in that it neglects historical contingency in order to postulate cross-temporal and cross-cultural regularities. Naturalists usually rely on essentialist concepts to elide the particularity and contingency of the objects they study, and to suggest that their regularities and correlations constitute explanations as opposed to mere generalizations. In contrast, because anti-naturalists emphasize historical specificity and contingency, they often emphasize the diversity of cases to which an aggregate concept might refer. They would regard empirical diversity not as an obstacle to concept formation, but on the contrary, as a fundamental aspect of social reality toward which social science concepts must be orientated. Essentialist approaches to concept formation are, in other words, a corollary of the naturalist neglect of contingency and particularity.Footnote 41 The concept of social class can serve us to illustrate essentialism just as it did reification. The conceptualization of social class in terms of objective socio-economic criteria is essentialist insofar as it ignores the way in which that concept is construed differently within different cultural or historical settings, as Gareth Stedman Jones and Dror Wahrman have argued in their respective work on the working class and British middle class.Footnote 42
It is important to recognize that essentialism can appear in a strong or a weak form. Strong essentialism occurs whenever the intension of a social science concept is simply associated with one or more core attributes, where these core attributes are said to characterize all cases to which we might apply that concept, and even where the core attributes are supposed to explain other characteristics of the relevant cases. This strong essentialism exhibits a logic of commonality, according to which concepts ought to be defined by a set of fixed attributes to be found in all relevant cases. This logic construes commonality, not as a contingent empirical finding that then gets registered in the concept, but rather as a precondition of the validity of social science concepts. Sartori's conceptual strategy, as we shall discuss, offers an example of strong essentialism.
Weak essentialism emerges out of a limited acknowledgement that strong essentialism is in some cases too rigid in the face of the “messiness” of social life. It therefore opens up a space for diversity but it heavily circumscribes this space. The space arises from allowing for variations in the degree to which a concept's attributes are manifested in various empirical cases. Crucially, however, weak essentialism retains the logic of commonality as its normative horizon. That is to say, it still shares the strong essentialist assumption that, beyond a rudimentary level of flexibility, any slackening of the reins of commonality undermines the validity of social science concepts. Weak essentialism introduces a peripheral modification to strong essentialism while leaving the essentialist core intact. David Collier's strategies of concept formation, as discussed below, exemplify weak essentialism.
Linguistic instrumentalism is a final dimension to debates about concept formation. Anti-naturalism implies that whenever social scientists formulate concepts, the content of the concepts is ineluctably informed by their situation. It also implies that the concepts should be developed through a kind of dialogue with the social actors being studied. Linguistic instrumentalism occurs when social science concepts are defined in ways that tear them from their location in the ordinary language of the scholar in an attempt to fashion them either into neutral instruments or into pragmatic instruments that nonetheless place the scholar in a subject-object relationship to the social actors being studied. Naturalism encourages this linguistic instrumentalism in two ways. First, naturalism implies that the situatedness of the scholar threatens the validity and reliability of social science, much as unsterile laboratory tools threaten the work of the biologist. Second, naturalism conceives of the scholar as the only agent involved in crafting social explanations. Hence naturalists typically believe that it is possible to detach social scientists from their linguistic situation, and also to place concepts at their service as sterile instruments directed at mute social objects.
Linguistic instrumentalism can appear on the subject side (i.e., the relationship between scholar and concept), on the object side (i.e., the relationship between concept and social world), or on both sides simultaneously. Subject-side instrumentalism portrays social scientists as the anonymous wielders of sterilized linguistic instruments that are shielded from their own ordinary language. Hence it elides the situatedness of the scholar. Sartori is, as we will see, a typical example of such linguistic instrumentalism. Object-side instrumentalism portrays the social world as a neutral object on which the concept is set to work as an instrument of discovery, description, classification, and explanation, without recognizing that the actors within this world form concepts with which to understand it and to act within it. Hence even when object-side instrumentalism might allow for the situatedness of the scholar, it still elides the place of dialogue or a “fusion of horizons” within social science. Collier is, as we will see, a typical example of this kind of instrumentalism.
We have described in general terms the ways in which naturalist philosophical assumptions get registered in approaches to concept formation. Because naturalism neglects meanings, contingency, the situatedness of the scholar and the dialogical principle, it encourages strategies of concept formation characterized by reification, essentialism, and linguistic instrumentalism. In the rest of this article, and within the general philosophical framework established so far, we will examine the approaches to concept formation adopted by qualitative methodologists from Sartori up until Collier. We will show how reification, essentialism, and linguistic instrumentalism bedevil the conceptual strategies of these scholars and embody their continuing dependence on a largely discredited naturalist philosophy. We have chosen to focus on Sartori and Collier because they are, by a wide margin, the two scholars who have developed the most elaborate qualitative theories of concept formation in political science.Footnote 43
Before we turn to them, however, we wish to reiterate that our critique is a philosophical one. We attempt to unearth the philosophical assumptions in their methodology, showing them to be naturalistic and hence, given the foregoing arguments, inappropriate for political analysis. We seek thereby to shift the debate from the practical advantages of methodological strategies to their underlying philosophical assumptions. Given our philosophical agenda, there is no need for us to examine the soundness or quality of the substantive outcomes of Collier and Sartori's approaches to concept formation. Rather, our critical task will have been fulfilled once we manage to demonstrate that those scholars' methodologies are marked by a discredited naturalism.
Qualitative Concept Formation
Giovanni Sartori
Sartori's naturalism, and its relationship to his methodological principles, appears in an essay he published in 1991 on “Comparing and Miscomparing.” In this essay, he set out to “explain the disappointing performance of the field of comparative politics”Footnote 44 by moving from a certain understanding of explanation in social science, through an account of the role of comparison in such explanation, and the role of concept formation in the comparative method, back to the failures of concept formation which he thereby suggested were responsible for the malaise of comparative politics. Sartori adopted the naturalist vision of a social science generating “law-like generalizations endowed with explanatory power.”Footnote 45 He argued that the role of comparison is precisely to “control (verify or falsify) whether generalizations hold across the cases to which they apply:”Footnote 46 comparison is the “method”Footnote 47 by which a hypothetical explanation can be discovered to be true or false. Next he goes on to suggest that just as explanations in the social sciences depend on the comparative method, so the comparative method depends on proper classifications. He argues that classifications establish our grid of similarities and differences, so if we get them wrong, we undertake wrong comparisons.Footnote 48 Finally he is thus able to suggest that the malaise of comparative politics stems from failures of concept formation. He complains that comparative politics has bred “cat-dogs,” that is, misconceptualizations that have led political scientists astray by denoting phenomena that “[do] not exist.”Footnote 49
Given that Sartori's interest in concept formation arose out of a concern with the disappointing record of comparative politics in generating naturalist explanations, we might not be surprised to find that when he turned to concept formation, he fell foul of reification, essentialism, and instrumentalism. His account of concept formation can be found in his edited volume, Social Science Concepts—a book that did much to pioneer conceptual analysis in comparative politics and so the rather newer concern with qualitative methods.Footnote 50 Sartori himself wrote the book's first chapter on “Guidelines for Concept Analysis.” The other chapters contained case-studies of conceptual analysis that were written more or less in accord with Sartori's theoretical roadmap. Crucially, while this roadmap made an occasional gesture toward anti-naturalism—it evokes approvingly Charles Taylor's claim that “language is constitutive of the reality”Footnote 51—it ultimately remains defiantly naturalist in its instrumentalist, reified, and essentialist view of concepts, or so we will now argue.
Let us begin with Sartori's instrumentalism. At the very start of his essay, he writes, “if language is the sine qua non instrument of knowing, the knowledge-seeker had better be in control of the instrument.”Footnote 52 Sartori's appeal to a consciously and purposively crafted use of language is, of course, compatible with anti-naturalist philosophies: it is arguable, for example, that the whole field of speech-act theory, as pioneered by Austin and Searle,Footnote 53 is all about how agents intentionally deploy ordinary language in order to express certain beliefs. Indeed, Sartori is surely doing us a service when he draws attention to issues about how social scientists might better deploy language (perhaps in creative and innovative ways) to convey their ideas with lucidity and precision. However, we will show that Sartori's own discussion of these issues then exhibits a narrow subject-side instrumentalism.
Sartori's instrumentalism appears partly in what he does not say. He ignores the situatedness of the scholar. His essay concentrates, instead, on the internal economy of the concept as an analytical tool. It explores the relationships between term, meaning, and referent; between intension and extension; between declarative and denotative and precising, operational, and ostensive definitions; between “accompanying properties” and “defining properties”; and between homonymy and synonymy. Sartori's anatomy of social science concepts, in other words, is purely “internalist.” His exposition leaves no room for his readers to consider how the situation of the scholar in the world might affect the “internal” aspects of social science concepts, the strategies for their formation, and the ways in which they can be used. It thus seems that Sartori thinks of social scientists as occupying a space outside of any particular life-world. Going back to the philosophical terminology set up earlier, we can say that Sartori elides the situatedness of the scholar in a way that shows his tacit reliance on naturalism.
Equally, Sartori's instrumentalism sometimes appears in what he does say. He treats concepts as tools over which the social scientist should seek perfect control, and he suggests that such control comes from various technical operations that remove concepts from ordinary language. Indeed, the explicit purpose of his essay is to overcome the two major “defects” he associates with “natural language.” These defects are “ambiguity” in the relationship of a concept's meaning to the word that expresses it, and “vagueness” in the relationship of a concept's meaning to its referents.Footnote 54 These are clear attempts to steer social scientists as much as possible away from their situatedness in the world.
Let us turn next to reification. Sartori's reified vision of concepts appears most dramatically in his definition of a concept's referent as “whatever is out there before or beyond mental and linguistic apprehension.” As he writes, “referents are the real-world counterparts (if existent) of the world in our head,”Footnote 55 where the “world in our head” refers to the intensions of our concepts, and the “real-world counterparts” of this world are empirical social phenomena. Sartori's definition of a concept's referent thus assumes a sharp distinction between our concepts and social phenomena. It leaves no room for a consideration of the constitutive role of meanings in actions. It treats actions and social phenomena generally as if they are distinct from the meanings, concepts, and beliefs “in our head.” Sartori thus treats social objects as if they were akin to physical ones. His theory of concept formation does not allow for any consideration of differences between the referents of concepts in the social sciences and objects in the physical world. In terms of our philosophical framework, Sartori's approach embodies a naturalist elision of meaningfulness.
Some of the contributors to Sartori's Social Science Concepts discuss concepts that refer to meanings or beliefs. The naturalist perspective governing the book, however, permeates even these concepts in the form of the second type of reification presented above: that is to say, these concepts neglect the holistic nature of meanings in order to craft concepts that can be incorporated within naturalist explanations as, say, independent variables. Consider Glenda Patrick's chapter on political culture. On the one hand, Patrick defines political culture in a way that clearly includes meanings: it is “the set of fundamental beliefs, values and attitudes that characterize the nature of the political system and regulate the political interactions among its members.”Footnote 56 On the other hand, however, she moulds the concept of “political culture” to fit naturalist explanations. Indeed she explicitly endorses Carl Hempel's naturalist analysis of concepts: they should be designed to “permit the establishment of general laws or theories by means of which particular events may be explained and predicted and thus scientifically understood.”Footnote 57 She appeals to the need “to determine the extent to which political culture constitutes a ‘causal’ factor—an explanatory and predictive term—for the explanation and prediction of political phenomena.”Footnote 58 Although Patrick's chapter (like all the others) does not include an application of a concept to explain empirical cases, it seems safe to conclude that what she has in mind is something like a correlation between “political culture” as a rigorously demarcated “variable” and other phenomena such as “political stability”—indeed she scrutinizes Almond and Verba's usage of “political stability” as a “dependent variable.”Footnote 59 Patrick, in compliance with Sartori's “Rule 8,” systematically cordons off political culture from adjacent concepts in the same “semantic field”—particularly “national character,” “political style,” “public opinion,” and “ideology”—in order “to isolate the critical differentia of the concept.”Footnote 60 This atomization of concepts forecloses the possibility of holistic explanations that would open out on to the whole web of beliefs of social actors. Here too we thus find the naturalist elision of meaningfulness.
Finally let us turn to Sartori's essentialism, which is of the strong type. His essentialism appears most clearly in his “Rule 7,” according to which “the connotation [=intension] and the denotation [=extension] of a concept are inversely related.”Footnote 61 Rule 7 (also called the “ladder of abstraction”) implies that the greater the number of attributes that comprise a concept's intension, the smaller the number of empirical cases that comprise its extension, and vice versa. Hence Sartori implies that when a concept is applied to new cases, then, if those cases do not share the core features that are shared by previous cases, the validity of the concept diminishes. Sartori's Rule 7, in other words, expresses the logic of commonality characteristic of strong essentialism. It excludes the historical uniqueness and particularity of individual cases at the very moment when we form concepts. Perhaps Sartori would reply that we allow for diversity (if not particularity) by forming varied concepts at a “low level of abstraction.”Footnote 62 However, if we formed a number of varied concepts, the diversity would appear only in the spaces between the concepts. The concepts themselves would still be defined by essential properties or at least commonalities. By thus espousing essentialism, Sartori's approach concurs with naturalism.
The strong essentialist approach of Social Science Concepts is exemplified by the chapter on power.Footnote 63 The authors of that chapter identify ten definitions of power, which they label DF1, DF2, and so on. But, instead of remaining content with this diversity, they subordinate all of the definitions (except DF10) to a single schema guided by the principle of “genus proximum et differentia specifica.”Footnote 64 This schema involves an attempt to construct a conceptual hierarchy. The top of the hierarchy consists of DF1 which defines power in terms of causality: “X has power over Y with regard to Z only if there is a relation of causality between X and Y with regard to Z.”Footnote 65 DF1, as the top of the conceptual hierarchy, is then treated as an essential core, common to all but one of the other definitions of power: “no relations between whatever social units are involved are called ‘power’ unless there is a relation of causation between the power holder and the power subject.”Footnote 66 The chapter on power argues that DF2 through DF9 all include DF1 and its idea of causality within their intensions. The authors expand on the core definition (DF1) in various ways simply to cope with the various antinomies and insufficiencies that arise from the minimum nature of the core definition itself. It is true, of course, that the authors identify a DF10 that does not share the essential feature of the other definitions: DF10 is “X has power in situation S if X is pivotal or decisive in S,” which makes power a matter of non-relational decisions rather than relational causality. Nonetheless, the authors treat DF10 in a way that further illustrates their debt to essentialism: although they accept that both causation-based and decision-based definitions of power are equally legitimate, they nevertheless regard their inability to eliminate this diversity as a “semantic puzzle.”Footnote 67 They view the irreducible plurality of definitions as an anomaly rather than a normal state of affairs, and, in doing so, they reaffirm their adherence to the essentialist view of concepts embodied in Sartori's “Rule 7.”
David Collier
If Sartori's work reveals the presence of naturalism at the birth of conceptual analysis as a qualitative methodology, Collier's shows how it still lingers. Collier is, of course, a—probably the—leading political scientist working on conceptual analysis in relation to qualitative methods. Collier, like Sartori, approached questions about concept formation against the background of concerns about comparative methodology. He too wrote an essay in 1991 reflecting on the state of comparative inquiry—“The Comparative Method: Two Decades of Change.” Collier might appear to be less wedded to naturalism than Sartori: he identifies contextual interpretive inquiry as one of three sub-categories of small-N comparative analysis,Footnote 68 and he calls for sensitivity to contextual diversity within comparative analysis.Footnote 69 Nonetheless, Collier ultimately appears to define even interpretation in naturalist terms: he assimilates it to a general comparative method based on “a commitment to systematic qualitative comparison that often involves a number of nations and evaluates each national case over a number of time periods,”Footnote 70 and he unpacks such comparison as dependent upon “systematic measurement and hypothesis testing.”Footnote 71 Collier's vocabulary here points to a vague and implicit adherence to naturalism. One gets the general impression that Collier's implicit scientific imaginary or model is that of the natural sciences, not only because of the use of the term “measurement” which draws on the prestige of quantitative research, but also because that measurement is expected to be systematic and repeatable. Collier implicitly presupposes that “nations” are objects whose core properties remain essentially the same “over a number of time periods” rather than being subjected to the fluidity of historical contingency.
Collier's naturalism has become clearer in his recent work, especially in the book that he coedited with Henry Brady, Rethinking Social Inquiry. Footnote 72 This book recasts the qualitative-quantitative relationship as one of “diverse tools, shared standards” based on “essentially similar epistemologies.”Footnote 73 The common epistemological ground is, of course, naturalist. This naturalism appears, for example, in the contrasting definitions of “interpretation” and “explanation” in the glossary written by Collier and Jason Seawright.Footnote 74 Interpretation is defined as “a description … of the meaning of human behavior from the standpoint of the individuals whose behavior is being observed.”Footnote 75 And explanation is defined in explicit contrast to just such descriptions.Footnote 76 The clear implication is that proper explanations have to be kept apart from interpretive studies of the meanings social phenomena have for social actors. Indeed, the glossary unpacks explanation in terms of dependent and independent variables—a view that pervades the rest of the book with the possible exception of the chapter by Charles Ragin. There is no room for explanations that point to beliefs and meanings that have a constitutive relation to actions and other social phenomena. In these ways, then, meanings and particularity are brushed aside in the search for more general explanations.
So, Collier expresses a flexible view of comparative analysis that nonetheless typically assumes a naturalist philosophy. We would suggest that the combination of an “eclectic”Footnote 77 view of comparative analysis with naturalist premises explains the dual movement in his approach to social science concepts—a movement that appears to be away from reification, instrumentalism, and especially essentialism but actually leads him back to them. Collier regularly appears to take a step toward the anti-naturalism of interpretive political science. He moves toward recognition of the constitutive role of meanings when he explicitly warns us of the danger of reification.Footnote 78 He moves toward recognition of contingency when he acknowledges that the meanings of social science concepts change along with the historical flux of the social world.Footnote 79 And he moves toward recognition of the relevance of the situatedness of the scholar when he allows that concepts rightly change along with our research goals and traditions.Footnote 80 Yet, as we will argue, although Collier thus makes the occasional move toward an anti-naturalist and interpretive political science, the larger path is still set by an overarching naturalism. Ultimately he walks the same naturalist road as Sartori, and it leads him too to reification, linguistic instrumentalism, and essentialism.
Let us consider first the issue of reification. When Collier and his coauthor, Robert Adcock, warn us of the dangers of reification, they define reification as “the mistake of overstating the degree to which the attributes one seeks to conceptualize cohere as if they were like an object.”Footnote 81 This definition equates reification with an elision of contingency, that is, the danger of understanding social phenomena as fixed objects rather than as in historical flux. Now, while we have pointed to the problems of neglecting contingency, we have also sought to distinguish reification from these problems. Reification consists less of an elision of contingency than of elision of the meaningful or intentional nature of action.Footnote 82 Collier's entanglement with reification conceived as a neglect of meanings thus appears in large part in what he does not say: even when he is explicitly discussing reification, he does not leave any room for considering the way in which social science concepts characteristically have to refer to objects that are constituted in part by meanings or beliefs. The meaningfulness of social action is not mentioned—let alone integrated into strategies of explanation or concept formation—in any of Collier's essays. This elision of meaningfulness embodies, as we have already argued, a now discredited naturalism.
The persistence of reification within Collier's approach also appears in some of his more concrete essays on concept formation. It appears in his essay, coauthored with Steven Levitsky, on concepts of democracy in comparative research.Footnote 83 This essay discusses the suitability of various definitions of democracy for different historical and geopolitical contexts. Collier and Levitsky suggest, for example, that the “procedural minimum” definition of democracy is inadequate for several Latin American countries in which the civilian government, even if elected freely, lacks the effective power needed to rule. They argue that in such cases we need to add the attribute of effective power on top of the procedural minimum definition.Footnote 84 Although Collier and Levitsky exhibit here sensitivity to context, they completely ignore the meanings that actors themselves attach to what they are doing. As a result they offer us a series of reified concepts of democracy that elide the place of meanings in social life. Their conceptualization of democracy is, in short, naturalistic insofar as it is devoid of reference to meanings as constitutive of actions and as a crucial aspect of contextual diversity within social science. Their conceptualization leaves no room for the possibility that democracies differ from each other, for example, because they are constituted by different beliefs (and so actions) about, say, “voting,” “parties,” “power,” and “legitimacy.” For anti-naturalists, the recognition of contextual diversity thus consists, not in adding or dropping reified attributes (as done by Collier and Levitsky), but in registering the different beliefs or meanings with which political actors imbue such attributes.
Let us turn now to the way naturalism entangles Collier in a kind of linguistic instrumentalism. Unlike Sartori, Collier acknowledges the importance of the situatedness of the social scientist, and he thereby avoids what we called subject-side instrumentalism. Collier and Adcock propound a “pragmatic approach” based on two aspects of the situatedness of the scholar: they recognize, first, that the meanings of social science concepts can change from one research agenda or tradition to another, and, second, that normative considerations often inform choices of method.Footnote 85 Nonetheless, because Collier elides the constitutive relationship of meanings to actions, he neglects the fact that the objects of social inquiry have accounts of themselves, and so forecloses the possibility of developing a dialogical form of social inquiry. His approach to concept formation thus exhibits the naturalist tendency toward what we have called object-side instrumentalism.
The following example from Collier and Adcock's article demonstrates the coexistence of their pragmatic approach with object-side instrumentalism. According to Collier and Adcock, when social scientists choose between a dichotomous and a graded conceptualization of democracy with respect to non-democracy, they can sometimes justify their choice on the basis of normative concerns.Footnote 86 To demonstrate how such normative justification operates, Collier and Adcock explore the case of O'Donnell and Schmitter who, they tell us, adopted a dichotomous concept of democracy in order to capture “what they saw as appropriate targets (neither too low nor too high) at which political actors should aim in pursuing democratization.”Footnote 87 While this example testifies to Collier and Adcock's acknowledgement of the scholar's situatedness, it also reveals the way in which they block the road to a dialogical form of social science. Collier and Adcock do not even consider whether or not O'Donnell and Schmitter adopted their concept of democracy as a result of taking any account of the beliefs of political actors. Consequently, the normatively-informed process of concept formation ends up being presented as an entirely solipsistic exercise; the possibility of a dialogical dimension is not considered. Collier escapes the Scylla of subject-side instrumentalism, in other words, only to find himself with the Charybdis of object-side instrumentalism.
Let us turn, lastly, to Collier's essentialism. Whereas Collier's discussions of reification and linguistic instrumentalism are somewhat cursory, he has written at length about the problem of contextual specificity for concept formation, so his views here will command a greater proportion of our attention. Collier's main treatment of the kinds of issues covered by essentialism occurs in an essay that he coauthored with James Mahon.Footnote 88 This essay concerns the problem of how to adapt concepts “to fit new contexts” without depleting their explanatory and classificatory power, or, in other words, how to allow for “conceptual traveling (the application of concepts to new cases)” without suffering “conceptual stretching (the distortion that occurs when a concept does not fit the new cases).”Footnote 89
Collier and Mahon's starting point is Sartori's strategy for avoiding “conceptual stretching,” according to which the essentialist core of a concept should be preserved in its application to new contexts by ascending the “ladder of abstraction” (renamed by the authors as the “ladder of generality”).Footnote 90 They explicitly reaffirm the fundamental validity of this strategy, and, to that extent, they already share much of Sartori's essentialism.Footnote 91
Yet Collier and Mahon do not think that Sartori's strategy can be applied to all social science concepts. Rather, they describe as “classical categories” those concepts that are amenable to Sartori's strategy for avoiding conceptual stretching, and they then go beyond Sartori in arguing that not all concepts are “classical” ones. It is important, therefore, to consider whether or not Collier and Mahon avoid the pitfalls of essentialism with the other types of concept that they consider. The two other types of concepts that they consider are “radial categories” and “family resemblance categories.” “Radial categories” clearly echo the strong essentialist tropes of “classical categories”; they are even defined as categories that have a core attribute (or “central subcategory”) which constitutes a kind of prototype, albeit that this prototype applies to cases only in conjunction with other attributes (or “noncentral subcategories”) so that the relevant cases might share their defining features not with each other but only with the prototype.Footnote 92
It is to “family resemblance categories” that we must look, therefore, to consider whether or not Collier and Mahon offer a way out of the essentialist traps set by naturalist assumptions. When Collier and Mahon introduce family resemblance categories, they certainly appear to be moving away from essentialism. They begin: “Wittgenstein's idea of family resemblance entails a principle of category membership different from that of classical categories, in that there may be no single attribute that category members all share.”Footnote 93 And they continue: “the label for this type of category derives from the fact that we can recognize the members of a human genetic family by observing attributes that they share to varying degrees, as contrasted to nonfamily members who may share few of them,” and “the commonalities are quite evident, even though there may be no trait that all family members, as family members, have in common.”Footnote 94 Collier and Mahon's discussion of family resemblance concepts is a significant advance on Sartori. It suggests that they recognize that a principle of mutual fit between concept and cases is too rigid for the social sciences. As we will show, however, their analysis of family resemblance concepts ultimately draws them away from Wittgenstein's anti-essentialism and back into the essentialist snares of a naturalist social science, even if their essentialism is admittedly of the weak rather than the strong version.
We can reveal Collier and Mahon's departure from Wittgenstein by distinguishing between two ways of thinking of a “family.” At one pole, a family might be a cordoned-off and relatively cohesive formation, but, at the other pole, it might be a looser cluster of people with unstable relationships and multiple step-members.Footnote 95 Collier and Mahon use the “family” metaphor in the former sense: members of a family are easily “contrasted to nonfamily members,” and their own “commonalities are quite evident.” But Wittgenstein used the metaphor of “family” in a looser way. He was quite skeptical about the possibility of drawing a boundary around all members of a family.Footnote 96 He also would have rejected the idea of commonalities among family members being “quite evident.” Indeed he did not use the term “commonality;” he preferred notably looser words such as “resemblance,” “similarity,” and “relationship.” In short, Wittgenstein used the notion of a “family” to convey a sense of indeterminate plurality, not common membership.Footnote 97
These two different uses of the “family” metaphor signal very different stances toward essentialism. Wittgenstein's looser sense of a “family” is clearly anti-essentialist. Hence the problem of conceptual stretching does not even arise for family resemblance concepts as he describes them: because family resemblance concepts do not have any single, fixed definition to begin with, it follows that there is nothing to be stretched. To the contrary, Wittgenstein's account of family resemblance concepts suggests that they develop and thrive precisely in being applied to new empirical contexts, which are analyzed on their own terms so as to elucidate their specificity. In contrast, Collier and Mahon's account of family resemblances reintroduces the problems of weak essentialism insofar as they insist on clearly evident commonalities. Hence they continue to regard conceptual stretching as a problem even for family resemblance categories.
Collier and Mahon's essentialist insistence on evident commonalities also suffuses the solutions they offer for coping with conceptual stretching. They appear to allow some leeway for diversity in that they advocate strategies such as “emphasiz[ing] that the category is an analytic construct which the researcher should not expect to be a perfect description of each case,”Footnote 98 or “identifying attributes that are present to varying degrees in particular cases, rather than being simply present or absent.”Footnote 99 However, this diversity remains heavily circumscribed by their insistence that social scientists do not violate the requirement of evident commonalities. It is this requirement that restrains Collier and Mahon within the confines of a weak essentialism.
Why do Collier and Mahon shy away from Wittgenstein's account of vague concepts and contextual specificity? The answer appears to be that an implicit naturalism drives them to seek concepts that can function within general, causal explanations. Their naturalism gets in the way of a properly anti-essentialist view of concepts: after all, if the cases covered by a concept do not share a common attribute, social scientists will struggle to explain the cases by reference to some common cause, as opposed to explaining them by reference to their particular origins and then using family resemblance concepts to capture the overlaps among the cases. Their naturalism leads them to privilege concepts that refer to objects of which we can provide a common explanation, and so have evident commonalities demarcating them from the objects to which the concept does not refer.
Conclusion
It seems not unreasonable to maintain that social science needs to be congruent with philosophical premises appropriate to its subject matter. We began this article by showing that there has arisen a widespread agreement (among philosophers if not in the unreflective practice of many social scientists) that anti-naturalist premises are the most appropriate for social science, where anti-naturalism highlights the meaningful and contingent nature of social life, the situatedness of the scholar, and so the dialogical nature of social science. Thereafter we suggested that, in stark contrast, qualitative approaches to concept formation often embody a discredited naturalism apparent in their entanglement with problems of reification, essentialism, and linguistic instrumentalism. Although Collier's work on concept formation is much more flexible and nuanced than the earlier work of Sartori, it too remains attached to a discredited naturalism that still entangles him in these same problems.
Our philosophical critique of qualitative concept formation sheds a distinct light upon various issues in political science. First, it highlights the profound affinity existing between quantitative methods and certain prominent practitioners of qualitative methods, and the concomitantly profound disparity between that kind of qualitative work and interpretive approaches. These affinities and disparities do not remain at the philosophical level, but spill over into innumerable other aspects of political analysis, from concept formation and strategies of comparison to forms of explanation. Second, our critique suggests that far too much political science might have a problem of philosophical appropriateness. It appears that the practice of many political scientists rests upon, and alas probably perpetuates, highly dubious assumptions about the nature of the social world, the role of the scholar in political analysis, and the relationship between scholar and world.
We have tried to show how important instances of qualitative methods and analyses of concept formation are beset by problems that arise from a lack of systematic philosophical reflection. The absence of philosophical reflection created a vacuum that has been filled by an over-emphasis on—perhaps even a fetishization of—methodology. We hope that our essay will stimulate our colleagues to engage those philosophical questions that alone can enable them adequately to consider whether or not any particular method is or is not appropriate to any particular subject of inquiry. Whatever the ultimate outcomes of such philosophical reflection, we are confident that our discipline can only benefit from it.
