email icon Email this citation

CIAO DATE: 09/00

Scientific Explanations of Social Phenomena: Overcoming the Positivist-Interpretivist Divide

Rebecca Johnson

Dept. of Government
Georgetown University

International Studies Association
41st Annual Convention
Los Angeles, CA
March 14-18, 2000

This paper is attempts to construct an epistemological basis for studying social phenomena. The term ‘social phenomena’ includes identity (individual and collective), culture, symbols, ideas, norms, principles, narratives, and collectively held beliefs. In order to improve scientific explanations of social phenomena, it is important to take a step back and evaluate the way they have been approached in the field in recent decades. There has been significant movement in the field over the past decade to accept and incorporate notions such as identity, culture, and norms, 1 but scholars’ attempts to study them in the same way that they have studied material factors for the past four decades have left the explanatory potential of social phenomena largely unfulfilled. Scholars have failed largely to evaluate whether accepted explanatory approaches fit. Social phenomena differ in kind from natural phenomena (they are time-space specific, do not exist apart from their social context, and are “a function of belief and action”). 2 They are not captured sufficiently by directly causal explanations offered by mainstream positivist scholars. 3 While the most scholars have dealt with social phenomena by trying to incorporate them into the field’s dominant approach to theorizing (positivism), the obverse is needed: approaches to theorizing international relations need to be incorporated into a social framework. This paper undertakes that task.

Scholars who have been able to employ successfully a new conception of social factors belong primarily to the interpretivist school. 4 Their work is to be applauded for its conceptual ingenuity and generally rigorous methodological standards, but interpretivists lack the desire to (or even belief in the ability to) speak beyond the particular empirical event they are investigating. Interpretivists question the practice of generalizing findings across types in the form of causal explanations. Interpretivists provide historical narratives rather than causal explanations, focusing their attention on the construction of social meaning. How do certain practices combine to constitute meaningful entities? This sort of ‘constitutive theorizing’ comprises the bulk of intepretivist work.

Both positivist and interpretivist approaches are able to view a narrow portion of complex and integrated social issues only. There have been two important attempts over the past decade to overcome the limitations of positivism and interpretivism—one by Friedrich Kratochwil in 1989 with the publication of Rules, Norms, and Decisions , and the other a decade later by Alezander Wendt with Social Theory of International Relations . Each makes the argument that positivism and interpretivism need not stay locked into their epistemological frames; indeed, much is gained by combining certain aspects from each. For lack of a better title to give theorists undertaking this pursuit, I will refer to them as social scientists. Social, because their intent is to capture the inherently social nature of international politics; scientists because they wish to offer explanations for types of events and processes that occur in international politics. While these initial contributions to the attempt to develop a truly social science of international relations are to be applauded for their insight and theoretical scope, neither is able to provide an explanatory scheme for the construction, manifestation, or destruction of social phenomena better than positivists. Any approach that seeks to explain the seemingly endogenous processes and events that combine to produce social phenomena must contain an understanding of causal complexity that most positivists lack, and must overcome interpretivists’ dislike of generalizations of outcomes to focus on generalizations of processes. This paper looks to positivist and interpretivist disputes over what it means for scholars to conduct “scientific” investigations, the appropriate definition of causality (and how this forces scholars to conceptualize social phenomena), and the functional methods, operating parameters, and goals of investigation to construct such an explanatory framework. This attempt builds on the work done by Kratochwil and Wendt, yet seeks to overcome their conceptual limitations by breaking investigations down to the dynamics at play in any given social phenomena (rather than looking at structures as does Kratochwil, or ‘bracketing’ agents or structures as advocated by Wendt). In order to position that approach to social scientific explanation, it is useful to examine the various understandings of ‘science’ and ‘explanation’ in the field today.

What is Scientific Investigation?

A fundamental point of contention between positivists and interpretivists is the definition of science. In his new book Alexander Wendt tries to cross the epistemological Rubicon between social phenomena and scientific explanation by characterizing the division between positivists and interpretivists as “a majority who think science is an epistemically privileged discourse through which we can gain a progressively truer understanding of the world, and...a large minority who do not recognize a privileged epistemic status for science in explaining the world out there.” 5 Daniel Little offers a broad understanding, writing that “[s]cience aims at producing knowledge about natural and social phenomena. And this aim brings with it a concern for truth, a concern for rational standards of belief assessment, and a commitment to the notion that the standards of belief assessment are conducive to truth.” 6 He argues that science includes the epistemic criteria of “an empirical-testability criterion, a logical coherence criterion, and an institutional commitment to intersubjective processes of belief evaluation and criticism.” 7 Social scientists must construct theories that can stand up to empirical validation and are internally consistent, and they must be committed to the project of peer evaluation. This paper develops the idea that means exist to explain social phenomena in a way that complies with these criteria and does not suffer the same shortcomings as many positivist approaches.

Post-positivists strike back that history is inherently indeterminate and this indeterminacy undermines any scientific orientation toward uncovering an objective truth by invalidating the empirical testability criterion. In examining historical explanation Olav Njølstad notes that

in an historical explanation, the causes and reasons referred to are themselves inferred relationships between empirically established actions and happenings in the past. Or, to rephrase Michael Oakeshott, in history; the relation between events is always other events’, which implies that ‘the conception of cause is thus replaced by the exhibition of a world of events intrinsically related to one another in which no lacuna is tolerated’. (Oakeshott, 1966, p. 209). Consequently, the causes and reasons which are said to ‘explain’ a particular historical event are only one of many possible interpretations of the linkage between it and other events equally established by empirical evidence—and equally open for rival interpretation. 8

He cites Kennan’s long telegram as one example of a key piece of historical evidence that left itself open to many competing interpretations. Njølstad sees scholars as incapable of approaching empirical evidence and events objectively, arguing that “since they have to interpret performatives and make inferences from speech acts to beliefs; and since their final criteria of plausibility will always be a function of their own cultural experiences—there can be no definite, non-ambiguous explanation of complex historical events. Instead, we are left with alternative interpretations, of varying validity.” 9 Njølstad’s concerns speak directly to the issue of whether scholars can say with reasonable confidence that they are able to investigate social phenomena in such a way that they can expect that the explanations that they offer and the inferences that they draw are valid. There are two issues here—the inference of beliefs from speech and the influence of observers’ cultural experience on analysis, both of which corrode some scholars’ belief in the ability to validate theories empirically. While these issues have been characterized in the literature as epistemological, I will contend that they are largely methodological and need not undermine the scientific enterprise if sufficient attention is paid to Little’s other two criteria for “science” (internal consistency and peer review). While there is an epistemological divide between those who can stomach some mingling of interpretation and observer bias with valid empirical investigation (positivists) and those who argue that the presence of contending interpretations and observer bias render scientific investigation problematic (interpretivists) or untenable (post-modernists), attention to method can reduce the influence of both in social scientific research. Little agrees that there are biases and distortion in the process of empirical study. He argues, however, “that the empirical methods of the social sciences serve as a substantial check on these deficiencies, and that over time it is reasonable to expect improvement in the verisimilitude of our social science hypotheses and theories. 10 In fact, most historical ‘indeterminacy’ is nothing more than “the humdrum limits of practical research: limited availability of data or contested questions, imperfections of available data, limits on research resources, and the like.” 11 Post-positivists may be correct to note that research can produce unacceptably distorted findings given the limited supply of empirical data that can be brought to bear on most questions interesting to social science researchers (and that scholars are somewhat constrained in their ability to distinguish between findings that are legitimate or distorted), but this does not mean that scholars should give up the pursuit for scientific knowledge. Scholars possess the ability to look for means to uncover additional empirical data, and refine the level of methodological and conceptual error implicit in most research projects. While this position would be rejected by some post-positivists and all post-modernists, those working within the interpretivist school have embraced these very tasks.

Writing in the interpretivist school Jennifer Milliken offers an interesting and useful discussion of methodological rigor in discourse analysis that shows that even analyses of open discourses can reach acceptable levels of accuracy and certainty. 12 By paying keen attention to potential observer bias through theory comparison and second-reader verification analysts can minimize the extent to which they inject themselves into their observations. While multiple interpretations of texts (and social events) exist, their presence need not stymie empirical investigation. Milliken takes the same approach advocated by Van Evera, King, Keohane, and Verba, and Lakatos. 13

An analysis can be said to be complete (validated) when upon adding new texts and comparing their object spaces, the researcher finds consistently that the theoretical categories she has generated work for those texts. This is also a partial response to the issue of the reliability of discourse analyses, i.e. that the interpretation offered has been checked and reworked until it fits with and explains consistently texts that were not originally part of its empirical base. 14

Little is also instructive on this point, pointing out that objectivity differs from certainty. “To say that I have objective knowledge about the statement ‘the apple is in the refrigerator’ is not to say that I am certain of its truth. It is rather to say two things: first, that I have good empirical reasons to believe that the statement is true; and second (at the meta-level), the fact that I have this body of evidence makes it probable that the statement is true.” 15 But is conceiving of knowledge as existing and ‘knowable’ sufficient to define science? If so, interpretivists like Milliken are as scientific as hard-nosed positivists like J. David Singer. Positivists include an additional feature to their definition of ‘science’ many interpretivists would reject—a commitment to generalizibility.

Robert Klee argues that the metaphysical point of science is to reduce as many things as possible to the fewest numbers of brute states of affairs. Systematizing knowledge and determining the hierarchy of basic and derived concepts becomes essential. 16 King, Keohane, and Verba make a similar assumption, writing that “[g]ood social science attempts to go beyond...particulars to more general knowledge.” 17 They continue, “[w]ithin the constraints of guaranteeing that the theory will be falsifiable and that we maximize concreteness, the theory should be formulated so that it explains as much of the world as possible.” 18

This equation of science with systematized knowledge lies at the root of many interpretivists’ rejection of the scientific enterprise; the destruction of nuance and the promotion of generalized explanations strips empirical investigations of its potential meaning and content. 19

Building Causal Theories?

As noted in the previous discussion concerning the acceptable boundaries of science, scholars’ positions concerning whether speaking across categories of cases is possible or advisable serves as a dividing line between positivism and interpretivism. This distinction has been accepted in the discipline, along with the logical corollary that positivists offer causal explanations that can speak invariantly to categories of cases and interpretivists focus on understanding specific empirical events. 20 While this division of labor is accepted fairly universally by the field, it is wrong. It may be the case that causal explanations apply to categories of social phenomena, but there is no precise dividing line for how many cases must fit into a category to render an investigation a “causal explanation.” Additionally, while the distinction between ‘explanation’ and ‘understanding’ has been presented as one of perspective (with the investigator looking either ‘outside-in’—explanation, or ‘inside-out’—understanding), it is fundamentally one of intent. 21 Before delving too deeply into an argument for why interpretive research may be considered explanatory, an initial discussion of how causality is approached by positivists is in order.

King, Keohane, and Verba define causal theory as

designed to show the causes of a phenomenon or set of phenomena. Whether originally conceived as deductive or inductive, any theory includes an interrelated set of causal hypotheses. Each hypothesis specifies a posited relationship between variables that creates observable implications: if the specified explanatory variables take on certain values, other specified values are predicted for the dependent variables. 22

This understanding of causality (variation in a specified independent variable has a direct and invariant effect on a specified dependent variable) depends on what King, Keohane, and Verba call ‘conditional independence,’ which “is the assumption that values are assigned to explanatory variables independently of the values taken by the dependent variables.” 23 King, Keohane, and Verba present a second assumption implicit in positivists’ notion on causality—verification of causal explanations relies on manifest behavior. They state this explicitly, arguing that “the standard for explanation in any empirical science like ours must be empirical verification of falsification. Attempting to find empirical evidence of abstract, unmeasurable, and unobservable concepts will necessarily prove more difficult and less successful than for many imperfectly conceived specific and concrete concepts.” 24 While they do not deny the important role unobservable factors play in social research, King, Keohane, and Verba ultimately argue that only those factors possessing observable implications are appropriate for empirical study. Positivist conceptions of causality include a third component; they must withstand counterfactual reasoning (change in dependent variable X would not have obtained were it not for the intervention of independent variable Y). Finally, a causal explanation, in distinction from a precise historical narrative, “identifies a causal process that recurs across a family of cases....a case study that invokes or suggests no implications for other cases, falls short of being explanatory.” 25

To summarize, a causal explanation as defined by positivists possesses the following four criteria:

  1. The independent and dependent variables must be conditionally independent;
  2. The relationship between dependent and independent variables must possess observable implications;
  3. A causal explanation must be counterfactually valid;
  4. The explanation, once posited, must apply across cases of the same type.

This produces theories of ideational phenomena that reflect this understanding of bounded and mechanically functioning variables that react unreflectively to the intervention of the independent variable. 26 King, Keohane, and Verba argue that to

show that ideas are casually important, it must be demonstrated that a given set of ideas held by policymakers, or some aspect of them, affect policies pursued and do not simply reflect those policies of their prior material interests....the observed dependent variable (policies) and the explanatory variable (ideas held by individuals) must be compared with a precisely defined counterfactual situation in which the explanatory variable takes on a different value: the relevant individuals had different ideas. 27

Positivist Causality and Social Phenomena 20

Many theorists within the constructivist school in International Relations have accepted this approach exactly. 28 This leads them to conceptualize social phenomena in ways that prevent them from capturing the truly social nature of their various objects of study. First, they conceive ideational phenomena (ideas, identity, norms, etc.) as variables they set next to interests as competing explanations. 29 While it may be useful to separate two ideal-types for heuristic purposes, retaining this separation during empirical investigations depoliticizes those investigations and denies the necessary and intimate relationship between ‘ideas’ and ‘interests’. Second, constructivist literature cannot escape a dichotomous ‘agent-structure’ conception of social interaction that relies on the methodological practice of bracketing one facet in order to study variance on the other. 30 The burden of explanation relies on the assumption of the presence of uni-causal streams of agency and structural forces in any particular encounter. However, multiple streams between and among agents and structures exist in any given interaction and depending solely on what one can ‘bracket’ at any point in time ignores many simultaneously occurring streams of action and produces an incomplete and potentially incorrect explanation. I will address how we can deal with multiple streams in the section on methodology below. Finally, constructivists’ application of a positivist conception of causality forces them to focus solely on the behavioral manifestation of social phenomena. While some social phenomena can and do reach a point where they are reified and internalized to the extent that they employ an externally binding force on individuals’ behavior, the presence and importance of social phenomena in the workings of international politics exceeds this limited selection.

Ideas as Variables 31

Many theorists have tried to incorporate ideas into International Relations theory by adding ‘ideas’ as an additional competing, explanatory variable. Mainstream constructivism “begins with the same model of rationality and introduces additional variables in order to account for deviations from expected outcomes, specifically by identifying the ‘ideas’ on the basis of which the actors engage in substantively rational action. This has prompted some critics to describe the ‘ideas’ literature not as a challenge to the rationalist approach but rather as its completion.” 32 What are the ramifications of conceiving ideational phenomena in this manner? Certainly ‘ideas’ exist as such; they may be ‘relational’ and constantly in the process of redefinition, but they exist as conceptually dense entities for a duration sufficient to justify conceiving them as things. Even language does this: “an idea, multiple identities .” Why should they not be seen as variables that change over time and exert influence over other variables? There are two negative ramifications of conceiving social phenomena in this way: it shifts the scholars’ focus from social phenomena that encompass actors to actors operating in a world of things. 33 Objectifying actors’ worlds risks depoliticizing social phenomena and people’s actions by creating a false and conceptually unhelpful firewall between ideas and interests. People may be strategically trying to realize goals, but the contest is one of strategies, not social meaning. Conceiving ideas in distinction from interests “is to ignore the political context in which actors strategize and are potentially organized across a political space and toward a policy outcome.” 34

Some scholars focus on ideas shaping interests, 35 while others focus on how interests co-opt ideas for their instrumental purposes. 36 Of course scholars argue that there is a relationship between ideas and interests and that each influences the other, but the wording of this conceptualization alone underscores the shortcoming of this interpretation. While there are observable instances in which ideas can be seen to influence interests and vice versa, this relationship is merely the most observable and superficial one that exists. At any point in time in any given context ideas and interests are inseparable because it is the product of their co-mingling that is the environment in which actors engage each other and the world around them. 37 Conceiving a given context at a point in time in terms of separable tensions between ideas and interests shifts the contest to ideas or interests (an untenable dichotomy analytically and empirically), not how and why ideas and interests play out as they do.

If ideas are relatively stable for enduring periods in time in any given context, what is the disadvantage of conceptualizing them in terms of discretely bounded variables? Scholars are clear to note that these variables are constantly in flux internally, 38 but that their apparent expression is relatively fixed for prolonged periods of time, making them suitable for empirical study as if they were static. Contra some of the more “purist” social constructivists I think this is probably correct and that, once reified, ideational phenomena take on the characteristics of other ‘things’ in international politics and can be portrayed in theory as such. But this position includes two substantive caveats—most ideational phenomena are not sufficiently reified at any point in time in a given context to warrant this form of conceptualization (and focusing on this severely circumscribed selection of reified social entities to the exclusion of the bulk of ideational phenomena weakens our theoretical potential) and this shift in focus toward things interacting away from processes developing, competing, and connecting takes theoretical power from a potentially fertile explanatory field.

One particular area of research that often falls prey to this dichotomy is work in ethnic conflict and nationalism. Theorists have tried to explain the role of ideational phenomena in ethnic or nationalist conflicts by examining how certain political leaders exploit collective identities, cultural symbols, and social images to manipulate public sentiment and action. 39 While this practice of consciously manipulating symbols to induce a desired effect is prevalent in politics generally, the question becomes whether ideas can reasonably be limited to this conceptualization. Much is lost if they are. Even attempts that focus on the role of moral entrepreneurs in exploiting windows of opportunity to reconfigure normative standards focus less on how entrepreneurs are able to change prevailing belief systems or preference orderings than the fact that they were able to do so. That fact alone renders this vein of research descriptive rather than explanatory, but other concerns exist as well. Mathew Evangelista’s new book on the role of transnational movements in U.S.-Soviet military policy, an impressive empirical contribution to the field, takes up the task of showing the impact of transnational movements in arms control and reduction from the 1950s through the end of the cold war. He argues that transnational interests were able to serve as conduits that shared information and ideas, appealed to norms, and coordinated policy initiatives to achieve their ends. 40 The bulk of his explanatory power, however, resides in political leaders’ ability to pick up the catchphrases developed by transnational movements in justifying their already-defined interests. Moral entrepreneurs served less as normative missionaries as they did normative spin doctors. So long as social phenomena are portrayed as politically manipulative tools to be employed by well- or ill-intentioned elites, International Relations scholars would do better to borrow from marketing strategists than social theorists. But social phenomena are more than political tools; they can be political forces. In situations where leaders are able to employ certain collective identities or images to rally a population (as Hitler did, for example) questions remains: why was the population vulnerable to this manipulation? What is it about the relationship between a leader and his public and the resonance of the collective image that that creates the possibility for this manipulation?

In any given context, it is likely that multiple leaders are vying for popular loyalty: what is the process by which collective images compete for dominance? How does this influence the public—do they split into groups along belief lines, or do they attempt to reconcile the competing images in some manner? Is the contest for dominance reducible to leaders’ power or rhetorical skill, or do collectively held images exert a force independent of those promoting them? These questions rely on an understanding of the world as greater than political leaders’ opportunism. Causal theories of this process would necessarily proceed from the perspective of explaining the social space that allowed leaders to manipulate collectively held images and the contested process by which leaders conveyed these images and publics accepted or rejected them.

Agents and Structures?

Conceiving ideas as variables forces theorists into a false bifurcation between agents and structures that relies on a conceptually untenable methodological practice of bracketing. This technique was made popular by Alexander Wendt and has been adopted by scholars such as Martha Finnemore in her book on shifting international norms. This tendency to set up agents and structures as bounded entities, fixing one to measure variation on the other, works against theorists’ attempts to model the relations that supposedly exist between and among agents and structures—the generation, articulation, change, and destruction of social phenomena. By constructing mental pictures of ideas as bounded while in motion, theorists lose the essence and import of social process. Again, the problem is one of emphasis. By focusing on a social world that is composed of aggregations of agents and forms of social structure, theories emphasize the interaction that occurs between and among agents and structures. Interaction is an important phrase because the prefix ‘inter’ denotes ‘between two things’. There is this process that goes on between and among agents and structures that theorists cannot really get around, but cannot accurately depict or explain in their writing either. So they fix the images in their mind and in their articles and then set one in motion while holding the other one constant in the hopes of seeing the dynamic that exists between the two. Theorizing reads at times as if ideas are seen as two cars that have gotten into a fender-bender. A bit of the paint from one car rubs off on the other and both look a little different for the encounter, but this transformation is physical and superficial. Agents and structures bump up against each other, and the result may be agents and structures with strikingly different appearances, but no true transformation has taken place.

An interesting and theoretically useful characterization is one that sees social and material phenomena as overlapping webs/streams. 41 This conception captures the true recursive nature of social phenomena as well as their relation with material factors in the world. But this approach confounds positivist causal explanations.

The Interpretivist Response

While causal accounts by positivists to date deal with the social equivalent of syntactics, interpretivists have focused on semantic and pragmatic types of investigation. 42 Interpretivists differ from positivists along the following three dimensions:

  1. The assumption of conditional independence and the distinction between Subject and Object is artificial and false;
  2. The relationship among social and material phenomena and their outcomes may not be refuted by observable occurrences;
  3. Investigations must remain time-space specific.

Interpretivists see experience as both arbitrary (reality is only one of many feasible) realities) and non-arbitrary (one can look into the historical conditions that make the extant reality dominant). 43 Interpretation attempts to make sense of social practices by examining them on their own terms, in their “original language.” 44 Charles Taylor lays out three characteristics that define the object of investigation of an interpretivist enterprise, “it must have sense or coherence; this must be distinguishable from its expression; and this sense must be for a subject.” 45 Interpretation occurs at the level of the subject, and any attempt to pull back from this perspective for the purposes of offering an explanation that extends to other, comparable phenomena destroys the understanding gained. Interpretivists have a different goal in mind from positivists. Taylor notes that “a good explanation is one which makes sense of the behavior; but then to appreciate a good explanation, one has to agree on what makes good sense; what makes good sense is a function of one’s readings; and these in turn are based on the kind of sense one understands.” 46 What makes good sense to positivists is objective study to determine direct causal relations that can be abstracted across cases of similar types. What makes good sense to interpretivists is study to uncover the perspectives and meanings understood by certain individuals and groups in a given situation in time and space. These perspectives and meanings are unique to each particular situation and cannot be abstracted from. While an interpretivist approach can supply rich, insightful, and useful accounts of social phenomena, any theory of international relations that wishes to find similar patterns of how different processes and events combine must be willing to compare different accounts for the purposes of drawing generalizations. Interpretivists conduct their research at the proper level of detail and specificity, but social science must be able and willing to pull its gaze outward as well, searching for patterns of action and combinations of processes that transcend specific instances.

Attempts to Bridge the Divide: Kratochwil and Wendt

How can social scientists focus on intricate, complex social relations “on the ground” at the same time that they attempt to keep larger patterns of interaction in sight? There have been two attempts at building social scientific approaches that contextualize and explain. The first attempt, made by Friedrich Kratochwil made a stark break for the positivist approach to direct causality and relies heavily on linguistic theory to explain social constitution. Alexander Wendt has tried to remain true to positivism by arguing a place for both directly causal and constitutive causality by conceiving a two-step explanatory process that begins with constitutive analyses followed by directly causal explanations. While both scholars are to be applauded for their moving the field of international relations closer to explaining social phenomena socially, each falls short. Kratochwil runs into trouble by relying too heavily on language as a proxy for social structure, and Wendt does not go far enough to break from a purely positivist epistemology.

Friedrich Kratochwil argues that the world of observational facts differs from the social world and knowledge because it contains mental notions such as intention, wishing and wanting, as well as norm-governed interactions among social subjects. Causality must account for meaningful action. 47 In the natural world (what Kratochwil refers to as ‘physicalist’) “cause and effect have to be determined independently from each other through neutral measurements. But the same is not true in the case of motivational accounts, where ‘causal’ motives can be imputed by the observer after a goal is assumed to be controlling.” This raises problems of refutability (missing the train does not necessarily refute the person’s intention to catch it, to use Kratochwil’s example). He notes that “[s]ince rules and norms are valid even if they fail to guide action in one or several cases, the question of whether and how rules and norms influence decisions cannot be “tested” if one uses strict refutability as a condition sine qua non of appropriate explanation.” 48 Furthermore, social action situates in the world of intention of meaning because individuals’ conduct is meaningfully oriented toward others. Meaningful action is placing an action within an intersubjectively understood context. This means that explaining an account is not cause-effect; rather, it is seen in terms of “elaborating, justifying, or possibly excusing the action rather than simply ‘refuting’ the hypothesis.” 49

This approach relies heavily on linguistics theory as its basis for explanation. Language is the logical foil for investigating the role of ideational phenomena in social life because it is both the link to the social world and a rule-governed activity (as Kratochwil sees human action to be), 50 allowing theorists a purportedly deeper and more nuanced account of ideational phenomena. Wesley Salmon sets up the distinction well using linguistic theory as the analogy to the debate in social science. In his encompassing history of the debates that have captured the philosophy of science since the 1950s, Salmon notes the tension that arose in Britain between the Wittgensteinians (ordinary language philosophers) and the logical empiricists (who Salmon characterizes as the “artificial language” theorists). 51 Salmon notes that linguistics is composed of three branches, syntactics, semantics, and pragmatics.

This approach to explaining social phenomena produces three important conceptual limitations as well—it underplays the material realities of life, potentially depoliticizes social phenomena, and relies on an optimistic assumption about social structure. With respect to the harsh realities of material life, Wendt takes exception to theorists such as David Campbell who focus on the construction of meaning by looking as the process of discourse. 53 He argues that theories that focus on social phenomena such as narratives and images cannot “account for the resistance of the world to certain representations, and thus for representational failures or misrepresentations.” 54 He offers the useful and unfortunate example of Montezuma confronting the undeniable presence of the Spaniards. Understanding the Spaniards’ intentions and socially conveyed images (and how those imaged mapped onto Montezuma’s images of god-like creatures) may have helped elaborate or justify the account, but it could not alter the very real, material effect of their presence on the conquistadores.

Second, focusing on linguistics (and rule-following or norm compliance using linguistics) runs the risk marginalizing the political nature of social phenomena. If the world is seen as a compilation of social rules that elicit compliance, the focus of investigation shifts to dominant rules that compel actors to conform. 55 This emphasis on rule-following conceptualizes individuals with severely limited agency who unreflectively follow dominant rules. Explanations of how rules compete for, attain, and lose dominance rely on their relative legitimacy, underplaying the role of power in rule-making.

Third, linguistic explanations of structure falter because they will be successful “only if the structural features of those formal derivations happily match the structural aspects of the external phenomena allegedly explained. But what rational grounds could there be for expecting to find such a happy coincidence between the order of our linguistic representations and the structural order of nature?” 56 Interpretivists argue that all knowledge is mediated though language, but attempts by social scientists like Kratochwil to rely on linguistic structures to explain behavior and social action rely on the assumption that this mediation is likely accurate. This assumption is questionable at best.

Wendt has taken an entirely different track in his attempt to construct a viable social theory. Rather than move toward an interpretivist understanding of the limits of science and causal explanations, Wendt wants to locate social science theorizing within a positivist frame. Understanding the initial counter-intuitiveness of this task he comments in a somewhat confessional tone that “[g]iven my idealist ontological commitments, therefore, one might think that I should be firmly on the post-positivist side of this divide, talking about discourse and interpretation rather than hypothesis testing and objective reality. Yet, in fact, when it comes to the epistemology of social inquiry I am a strong believer in science--a pluralistic science to be sure, in which there is a significant role for ‘Understanding’, but science just the same. I am a ‘positivist’ 57 His work is true to this claim, and unfortunately exhibits the same conceptual shortcomings as other constructivists who have paid less explicit attention to the subtleties involved in theorizing social relations.

Alexander Wendt is forced to recreate the bifurcation between structures and agents and rely on the practice of bracketing. He writes in the introduction to his book that he is, “concerned with the fundamental assumptions of social inquiry: the nature of human agency and its relationship to social structures, the role of ideas and material forces in social life, the proper form of social explanations, and so on.” 58 The mental distinctions conjured by his characterization are agents in relation to structure (each bounded and existing distinct from the other) and ideas and material forces (though Wendt goes farther than many to reinforce the fact that each is inherently implicated in the other). He is on the right track in his theoretical conceptualization of the mutual constitution of agents and their structures, but his positivist commitments force him from a position where he can examine this dynamic in any useful way.

Wendt slips into a directly causal account of social phenomena as well. He notes that the thinness of international society means that most of a state’s identity construction occurs domestically, keeping in mind that identity’s are made possible and embedded within an international system. 59 As I have noted above, social phenomena take up a social space larger than reified symbols or objects and any theory that examines even these relatively dense and stable reified images of ideas must incorporate the larger, more diffuse body of social phenomena (remembering the inextricably political and ‘interested’ nature of these phenomena in a given context at any particular point in time). The lack of observable social phenomena may be the result of so many norms, identities, etc., competing that none gain uniform articulation. This does not disprove either the existence or importance of the contest. Competition among norms, identities, etc., is important in its own right for explaining the nature of a particular systems as well as in the process of building theories that explain how particular social phenomena come to gain dominance over other competing phenomena

Toward Causal Explanations of Social Phenomena

Is there a way to establish and investigate causal relationships among social and material factors that avoids the conceptual limitations of positivism and interpretivism? Daniel Little offers an understanding of causality I think can knock down the firewall between positivist and interpretivist approaches to international relations. He understands that “social causal relations are constituted by the causal powers of various social events, conditions, structures, and the like, and the singular causal mechanisms that lead from antecedent conditions to outcomes. 60 By unpacking the conceptual space between antecedent conditions and their outcomes theorists can break down the murky notions of ‘mutual constitution’, ‘construction of meaning’ and the like. Social phenomena may be constitutive of social reality in the ways described by scholars such as Wendt and Kratochwil, but not by some magical wave of the ‘meaning wand’.

Little is correct to point out that many positivists are moving in the direction of examining causal mechanisms in an attempt to see how causes induce their effects, but the practice has a long way to go. King, Keohane, and Verba define causal mechanisms as “linked series of causal hypotheses that indicate how connections among variables are made,” 61 or tautologically, “the mechanisms by which a cause has its effect” (King, Keohane, and Verba, 1994: 86). They equate causal mechanisms with “‘process tracing’...‘historical analysis’, and ‘detailed case studies’.” 62 While certain methods may be important means of identifying and examining causal mechanisms, merely locating them in an empirical investigation does little to explain how or why the mechanisms operate in the manner they do. As more attention is paid to how scholars can break down the causal chain that exists between social catalysts and various social outcomes, theorists become better able to study the process of social constitution. 63 What are the dynamics at play in the recursive processes that generate social meaning but causal mechanisms?

This helps analysts get past the problem of endogenaity by breaking down larger social phenomena into the series of overlapping social, political, material, and psychological processes and events that combine to produce the phenomena. For example, by examining the development of a collectively held identity as the product of various processes and events analysts can begin to get a picture of how construction occurs. What combinations of conditions and processes (in what magnitudes) prompt discrete individuals to coalesce around a given identity? The distinction of subject and object may indiscernible at an aggregate level (analysts know that individuals are interacting with, shaping, and being shaped by their social environment in some way but cannot draw neat causal arrows), but this distinction becomes visible as researchers break the constellations of processes down into their various combinations.

The difficulty is how to carry this out conceptually and methodologically. Conceptually it is essential for scholars to remember that uni-causal explanations are seldom valid and that any causal relationship is likely complex. Investigating causal mechanisms keeps researchers attentive to this fact, but entails researchers’ focusing on the dynamics of the mechanism that enable particular results (or, often as importantly, non-results). Patrick Jackson and Daniel Nexon have offered a potential means for this sort of investigation. They focus on an approach to social science they call ‘relationism’. This approach “treats various configurations of ties—recurrent sociocultural interaction—between social aggregates of various sorts and the component parts as the building blocks of social analysis.” 64 By looking at how processes and events ‘link up’ analysts can make better sense of how different configurations (groups of processes and events) constitute specific practices of international politics. 65

This sort of approach to explaining social phenomena relies on an understanding of causality that can sustain complexity. Direct causality often will not suffice. Complex causal relationships pose some methodological concerns as well. Little points to INUS and probabilistic causal explanations as ways to focus on the casual mechanisms that are at play in any given social process. INUS conditions were developed by John Mackie and refer to an “insufficient but necessary part of a condition which is itself unnecessary but sufficient for the result.” 66 Little explains, “suppose A&B cause F and A&C&D cause F, and no other set of conditions cause F. Here A is a necessary condition for F. A&B are jointly sufficient conditions for F. And neither conjunct is necessary for the occurrence of F.” 67 Think of each condition as a various social, political, material, or psychological process or event. INUS causality allows researchers to examine how various combinations of processes and events combine (in which combinations and in what magnitude) in creating social outcomes. While any one process lacks causal weight on its own, it may exert force in certain combinations of other processes or events.

Probabilistic causation is defined as “causes alter the probability of occurrence of their effects relative to the prior probability of the effect.” 68 This form of causation is still deterministic as defined by Little (“causes work through exceptionless regularities”). 69 This is because variation in phenomena carrying causal weight prompts uniform variations in outcomes (barring the introduction of intervening forces) while allowing for an explanation that contains more nuance and complexity.

Kratochwil takes exception with probabilistic interpretations of causality when applied to ideational phenomena. “Probabilistic statements about empirical regularities...are solely constituted by their reference-class, by the ranges of numerical values within which the measurements of the crucial variables have to fall, and by the ‘methodological rule’ which decides that reproducible deviations from those ranges refute the theory.” 70 He claims that this approach to causality fails to capture how explanations of action are shaped by moral discourse and pragmatic interest. He fails to illustrate, however, why this is necessarily the case. What prevents a probabilistic account of intentions? This account may be difficult to operationalize, but difficulty is hardly an acceptable justification for theoretical and methodological complacency. By developing causal explanations that attempt to reduce as much methodological vagueness as possible, social scientists will be able to better approach conceptually vague social phenomena. Developing theories that conceive social phenomena in terms of probability of occurrence rather than actuality of occurrence may offer a productive way to address positivist causality’s implicit behavioralism. This approach to causality enables causal explanations for events that did not occur (why the dog could not bark; or, why it did not bark when it could have). It also promotes multi-causal and indirectly causal explanatins by conceiving social phenomena as the combination of multiple social and material phenomena. By viewing social phenomena this way, researchers are in a position more hospitable to investigating how variations in the various combinations of social and material phenomena alter the likelihood of particular empirical outcomes.

Little places generalizibility as the distinction between a causal explanation and a ‘causal story’, arguing “as a methodological maxim that a causal assertion is explanatory only if it identifies a causal process that recurs across a family of cases.” 71 He continues by arguing that “good causal explanation requires at least two things. First, the explanation needs to provide a causal analysis that identifies some of the conditions to contributed to the occurrences of the explanandum....And second, the account must invoke causal relations that attach to the type, not the token.” 72 Bhaskar and Njølstad offer compelling arguments about the contexuality of social phenomena. This point highlights the importance of theories specifying their boundary conditions carefully. King, Keohane, and Verba note that this step of questioning where the boundaries of our theories lay and why they exist where they do helps to illustrate which factors need to be incorporated to expand the theory’s scope. 73 Regardless of where particular social scientists come down with respect to the need to build expansive theories, the heuristic and analytically useful project of specifying a theory’s scope and domain conditions cannot be overemphasized.

My approach to conducting causal explanations of social phenomena contains four criteria:

  1. Social and material phenomena and the outcomes their combinations produce vary in their conditional independence;
  2. The relationship among social and material phenomena and their outcomes may (but need not) possess observable implications;
  3. The explanation must be counter-factually valid;
  4. Investigations focus on phenomena located in time and space with the intent of discovering processes that occur across types of phenomena.

Some recent empirical work comes close to approximating this understanding of causality. Laffey and Weldes look at how conceiving ideas as constitutive of social interaction shapes empirical investigation. This form of “analysis would require an investigation of those ideas which made it possible to understand the situational features as particular...conditions requiring a particular sort” of response by actors. 74 Constitutive theories look at how actions, ideas, institutions, etc., become intelligible, meaningful, and legitimate. These theories focus on questions regarding the “possibility conditions for different courses of action.” 75 This requires a different form of causality. Laffey and Weldes suggest causality be “conceived of generatively and relationally, as something that exists in the world rather than only logically or in theory. Causal claims refer to the ‘causal powers’ of social agents, which are conferred on those agents by the social structures and relations that constitute them.” 76 Michael Barnett picks up on this in an interesting and useful article on Rabin’s ability to create the social space necessary to enable the Oslo Accords. He focuses on the role of ideational ‘frames’ as connecting identities to interests to particular courses of political action, writing that identities “provide a grammar of action and makes certain action legitimate and possible.” 77 Barnett notes that he was trying not to identify the causes of the Accords, but to explore “the cultural preconditions, preconditions that were made possible by political elites.” 78 Ultimately his work fails to provide a satisfactory explanation. He writes,

Rabin’s moves toward a new historical narrative were not occurring in a political or intellectual vacuum. There were broader current within society that made Rabin’s message resonate and politically conceivable. Rabin was following a solid decade of intellectual and cultural developments within Israel that were challenging and questioning many of the most closely cherished interpretations of the past, symbols and taboos of Israeli society and history. 79

Barnett is unable to offer an account for how these factors combined as they did allow Rabin the discretion to act. What intellectual and cultural developments occurred (and in what fashion) to create this shift in cultural space?


This paper has offered an alternative way of approaching social scientific research that avoids the conceptual limitations of positivism and interpretivism. It builds from the contributions made by social scientists such as Friedrich Kratochwil and Alexander Wendt in the hopes of continuing the discussion of how to best construct social scientific causal explanations of social phenomena that capture the complexity that exists in international politics while retaining an eye to generalizibility. Rather than generalize outcomes, however, this approach advocates generalizing the combinations of processes and events that enable them. This requires that researches bring their investigations to focus on the dynamics at play in social phenomena, not the outcomes of these phenomena. While outcomes of interaction (a given dominant identity, for example) may be the manifestation of the combination of forces that have produced it (or allowed its production), an explanation of that identity must come from a level deeper than its observable presence. Attention to causal mechanisms and configurational analysis and the use of INUS and probabilistic causality can give social scientists the tools they need to perform these tasks.


Archer, Margaret (1995) Realist Social Theory: The Morphogenetic Approach , Cambridge: Cambridge University Press.

Barnett, Michael (1999) “Culture, Strategy, and Foreign Policy Change: Israel’s Road to Oslo,” European Journal of Internationa lRelations , 5, 5-36.

Bennett, Andrew (1999) “Causal Inference in Case Studies: From Mills Methods to Causal Mechanisms,” presented at the 1999 APSA annual conference, Atlanta, GA.

Bennett, Andrew and Alexander George (forthcoming) “Case Study Methods and Political Science: Similar Strokes for Different Foci,” in Colin Elman and Miriam Fendius Elman, eds. Bridges and Boundaries: Historians, Political Scientists, and the Study of International Relations , MIT Press.

Bhaskar, Roy (1979) The Possibility of Naturalism , Atlantic Highlands, NJ: Humanities Press.

Campbell, David (1992) Writing Security , Minneapolis: University of Minnesota Press.

Dessler, David (1989) “What’s at stake in the agent-structure debate?” International Organization , 43, 441-473.

Dessler, David (1996)

Checkel, Jeffrey (1997) Ideas and International Political Change: Soviet/Russian Behavior and the End of the Cold War , New Haven: Yale University Press.

Checkel, Jeffrey (1999) “Norms, Institutions, and National Identity in Contemporary Europe,” International Studies Quarterly , 43, 83-114.

Evangelista, Mathew (1999) Unarmed Forces: The Transnational Movement to End the Cold War , Ithaca: Cornell University Press.

Finnemore, Martha (1996) National Interests in International Society , Ithaca: Cornell University Press.

George, Jim, and David Campbell (1990) “Patterns of Dissent and the Celebration of Difference: Critical Social Theory and International Relations,” International Studies Quarterly , 34, 269-293.

Giddens, Anthony (1984) The Constitution of Society , Berkeley: University of California Press.

Goldstein, Judith and Robert Keohane, eds. (1993) Ideas and Foreign Policy , Ithaca: Cornell University Press.

Hollis, Martin and Steve Smith (1990) Explaining and Understanding International Relations , Oxford: Oxford University Press.

Hollis, Martin and Steve Smith (1991) “Beware of gurus: structure and action in international relations,” Review of International Studies 17, 393-410.

Hollis, Martin and Steve Smith (1992) “Structure and action: further comment,” Review of International Studies 18, 187-188.

Jackson, Patrick T., and Daniel H. Nexon (1999), “Relations before States: Substance, Process, and the Study of World Politics,” European Journal of International Relations .

Katzenstein, Peter, ed. (1996) The Culture of National Security , New York: Columbia University Press.

King, Gary, Robert Keohane and Sidney (1994) Designing Social Inquiry , Princeton: Princeton University Press.

Klee, Robert (1997) Introduction to the Philosophy of Science: Cutting Nature at its Seams , Oxford: Oxford University Press.

Kratochwil, Friedrich (1989) Rules, Norms, and Decisions , Cambridge: Cambridge University Press.

Laffey, Mark and Jutta Weldes (1997) “Beyond belief: Ideas and symbolic technologies in the study of international relations,” European Journal of International Relations , 3, 193-237.

Lakatos, Imre (1970) “Falsification and the methodology of scientific research programmes,” in I. Lakatos and A. Musgrave, eds. Criticism and the Growth of Knowledge , Cambridge: Cambridge University Press, pp. 91-196.

Little, Daniel (1998) Microfoundations, Method, and Causation , New Brunswick: Transaction.

Mackie, J.L. (1974) The Cement of the Universe , Oxford: Oxford University Press.

Mearscheimer, John (1994/1995) “The false promise of international institutions,” International Security , 19, 5-49.

Milliken, Jennifer (1999) “The Study of Discourse in International Relations: A Critique of Research and Method,” European Journal of International Relations , 5, 225-254.

Neumann, Iver (1996) “Collective Identity Formation: Self and other in international relations,” European Journal of International Relatios , 2, 139-174.

Neumann, Iver and Jennifer Welsh (1991) “The Other in European slef-definition: An addendum to the literature on international society,” Review of International Studies 17, 327-348.

Njolstad, Olav (1990) “Learning from History? Case Studies and the Limits to Theory-Building,” in Nils Petter Gleditsch and Olav Njostad, eds., Arms Races: Technological and Political Dynamics , London: SAGE Publications.

Ruggie, John (1983) “Continuity and transformation in the world polity,” World Politics , 35, 261-285.

Salmon, Welsey, (1989) Four Decades of Scientific Explananation , Minneapolis: University of Minnesota.

Snyder, Jack (1991) Myths of Empire , Ithaca: Cornell University Press.

Van Evera, Stephen (1997) Guide to Methods for Students of Political Science , Ithaca: Cornell University Press.

Wendt, Alexander (1987) “The agent-structure problem in international relations theory,” International Organization , 41, 335-370.

Wendt, Alexander, (1991) “Bridging the theory/meta-theory gap in international relations,” Review of International Studies 17, 383-392.

Wendt, Alexander (1992) “Levels of analysis vs. agents and structures,” Review of International Studies , 18, 181-185.

Wendt, Alexander, (1999) Social Theory of International Relations , Cambridge: Cambridge University Press.


Note 1: Goldstein and Keohane (1993); Katzenstein, ed. (1996) are notable examples. Back.

Note 2: Wendt (1999: 71). Discussion of the differences between natural and social phenomena from Bhaskar (1979: 48-49). Back.

Note 3: Here I am using the term ‘positivist’ as it is employed in the field today, understanding that the definition of what it means to be a ‘positivist’ in International Relations in the year 2000 differs importantly from what it meant to be a positivist when the term generated. Back.

Note 4: Barnett (1999); Neumann (1996); Neumann and Welsh (1991). I am consciously referring to interpretivists specifically, and not other post-positivists. Interpretivists differ from critical scholars in that they conduct empirical investigations with the intent of giving an account of the empirical events that occurred (by uncovering the social and psychological phenomena at work in a particular empirical instance). Critical theorists undertake a similar empirical investigation, but with the intent of uncovering the power dynamics implicit in any particular empirical instance. See George and Campbell (1990); George (1995); Ashley and Walker (1990). Post-modern scholars adopt a relativist claim that renders empirical investigation untenable and so deny the relevance of epistemological or methodological development entirely. Back.

Note 5: Wendt (1999: 38-39). Back.

Note 6: Little (1998: 173, emphasis in original). Back.

Note 7: Little (1998: 178). Back.

Note 8: Njølstad (1990: 223). Back.

Note 9: Njølstad (1990: 225). Back.

Note 10: Little (1998: 174-175). Back.

Note 11: Little (1998: 193). Back.

Note 12: Milliken (1999: 231-236). Back.

Note 13: King, Keohane, and Verba (1994); Lakatos (1970); Van Evera (1997). Back.

Note 14: Milliken (1999: 234-235). Back.

Note 15: Little (1998: 177). Back.

Note 16: Klee (1997: 118). Back.

Note 17: King, Keohane, and Verba (1994: 35). Back.

Note 18: King, Keohane, and Verba (1994: 114) Back.

Note 19: Njølstad (1990: 221). Back.

Note 20: This divide is most obviously noted in Hollis and Smith (1190, 1991, 1992); Wendt (1991, 1992); Dessler (1989). Back.

Note 21: Hollis and Smith (1990, 1991, 1992); Wendt (1991, 1992). Back.

Note 22: King, Keohane, and Verba (1994: 99-100). Back.

Note 23: King, Keohane, and Verba (1994: 94). Back.

Note 24: King, Keohane, and Verba (1994: 110). Back.

Note 25: Little (1998: 207-208). Back.

Note 26: See Ruggie (1983) for this critique of Waltzian neo-realism. Back.

Note 27: King, Keohane, and Verba (1994: 191). Back.

Note 28: Checkel, (1997); Evangelista, (1999); Finnemore, (1996). Back.

Note 29: Sikkink (1998). Back.

Note 30: Wendt (1987), Finnemore (1996). See Archer (1995) for a critique of this approach. Back.

Note 31: Laffey and Weldes note the prevalence of conceiving “ideas as objects,” 1997: 194. Back.

Note 32: Laffey and Weldes (1997: 196). Back.

Note 33: Dessler (1991). Back.

Note 34: Barnett (1999: 16). Back.

Note 35: Checkel (1997). Back.

Note 36: Mearscheimer (1991); Snyder, (1991). Back.

Note 37: Dessler (1989); Giddens (1984). Back.

Note 38: Wendt (1999). Back.

Note 39: Snyder (1991). Back.

Note 40: Evangelista (1999: 12). Back.

Note 41: Jackson and Nexon (1999: 292-316). Back.

Note 42: Dessler (1996). Back.

Note 43: George and Campbell (1990: 287). Back.

Note 44: Taylor (1971). Back.

Note 45: Taylor (1971). Back.

[46]Taylor (1971). Back.

Note 47: Kratochwil (1989: 24). Back.

Note 48: Kratochwil (1989: 100). Back.

Note 49: Kratochwil (1989: 25). Back.

Note 50: Kratochwil (1989: 6). Back.

Note 51: Salmon (1989: 35). Back.

Note 52: Salmon (1989: 35). Back.

Note 53: Campbell (1992). Back.

Note 54: Wendt (1999: 56). Back.

Note 55: Checkel (1999). Back.

Note 56: Klee (1997: 126). Back.

Note 57: Wendt (1999: 39, emphasis added). Back.

Note 58: Wendt (1999: 5). Back.

Note 59: Wendt (1999: 21). Back.

Note 60: Little (1998: 198). Back.

Note 61: King, Keohane, and Verba (1994: 225). Back.

Note 62: King, Keohane, and Verba (1994: 86). Back.

Note 63: For work in this vein, see Andrew Bennett, “Causal Inference in Case Studies: From Mill’s Methods to Causal Mechanisms,” presented at the 1999 APSA annual conference, Atlanta; Andrew Bennett and Alexander George, “Case Study Methods and Political Science: Similar Strokes for Different Foci,” in Colin Elman and Miriam Fendius Elman, eds., Bridges and Boundaries: Historians, Political Scientists, and the Study of International Relations (MIT Press, forthcoming). Back.

Note 64: Jackson and Nexon (1999: 291). Back.

Note 65: Jackson and Nexon (1999: 295). Back.

Note 66: Mackie (1974: 62). Back.

Note 67: Little (1998: 199). Back.

Note 68: Little (1998: 201). Back.

Note 69: Little (1998: 201). Back.

Note 70: Kratochwil (1989: 100). Back.

Note 71: Little (1998: 207). Back.

Note 72: Little (1998: 208). Back.

Note 73: King, Keohane, and Verba (1994: 104). Back.

Note 74: Laffey and Weldes (1997: 200). Back.

Note 75: Laffey and Weldes (1997: 201). Back.

Note 76: Laffey and Weldes (1999: 204). Back.

Note 77: Barnett (1999:25). Back.

Note 78: Barnett (1999: 8). Back.

Note 79: Barnett (1999: 22). Back.