Reply to Sica: Epistemological and Ontological Debates in Historical Sociology

The debate over the role of general theory in historical sociology continues with Alan Sica’s (2005) response to my article (Mahoney 2005) and now this reply from me. Sica expresses skepticism about the ability of any general theory to contribute important insights to sociology. He specifically argues that the epistemology of this approach—rooted in certain positivistic assumptions—is deeply problematic. Among other things, he sees the formalization entailed in the use of this theory as unproductive and leading to fruitless results; the deductive logic of general theory as having been tried countless previous times only to yield failures; and the effort to make predictions about human behavior as a badly misconceived project and an unfortunate legacy of pseudo-scientific research dating back to Saint-Simon and Adolphe Quetelet. All told, his view is that general theory should be seen as part of “sociology’s heart breaking quest to become a science” (2005:4 of the draft).

In this reply, I want to try to accomplish two things. First, I want to show that several of Sica’s criticisms are directed at a mode of scientific analysis endorsed by Karl Popper, Carl Hempel, and other mid-twentieth century philosophers, but not the mode of general theory that I discussed in the article. In this spirit, I will spell out the differences between the “scientific realist” epistemology of general theory and the “positivist” epistemology advocated by an earlier generation of philosophers.[1]  Second, I want to explore the ontological assumptions that animate Sica’s critique of general theory. Here I argue against Sica’s skeptical remarks concerning the utility of unobservables, and then say a little more about my views on the relative merits of the five general theories that I discussed.

Epistemological Debates

Epistemological debates are notoriously difficult to resolve, and I do not hope to persuade Sica that the epistemology of general theory is the best way for sociologists to proceed. However, I do wish to make clearer the important differences between general theory and positivism as traditionally understood. In doing so, I want to show that Sica conflates the epistemology of general theory with positivism.

The Virtues and Vices of Formalization

As I discussed and defined it, general theory commits scholars to the formalization of their arguments. Formalization refers to the process of attempting to make explicit:  (1) the key assumptions of an argument; and (2) the logical procedures an argument uses to draw conclusions. I sought to formalize—for illustrative purposes only—the arguments of some leading historical sociologists. This effort became a target of criticism by Sica, who appears to support a mode of social science analysis in which arguments are not formalized. Here I briefly review Sica’s position, outline a radical alternative in which good social science research is fully equated with formalization, and then state my own view which falls between these two extremes.

Sica is skeptical of formalization in part because the process of explicitly stating underlying assumptions can lead one to trivialize more complex arguments. Thus, for example, he pokes fun at my effort to explicitly state the assumptions underlying Rueschemeyer, Stephens, and Stephens’ (1992) balance of class power model, noting that my formulation does not capture many of the nuances of the argument.[2]  More generally, however, Sica is skeptical about formalization because he views history as inherently intractable. Historical processes themselves do not follow any coherent and ascertainable logic, and thus we cannot formalize arguments in the social sciences.

The mirror image of Sica’s position is represented by outspoken advocates of formal modeling, especially in political science and economics. For example, Cameron and Morton (2002) argue that formalization yields three main virtues:  (1) Clarity—testable hypotheses are derived from explicit assumptions in a transparent way;  (2) Rigor—reasoning can be confirmed (or disconfirmed) by outside researchers; and  (3) Unity—contradictory assumptions are exposed. In response to those who raise concerns about formalization, Cameron and Morton have the following to say:  “many [researchers] seem to see the clarity, rigor, and unity of formal modeling as its principle disadvantage for empirical work. For some researchers, a fuzzy, possibly erroneous prediction from a nonformal model with unknown assumptions is somehow preferable. The unspoken belief seems to be that if you keep your assumptions implicit or hidden (perhaps even from yourself), you haven’t made any assumptions!  Of course, this position is utterly mistaken” (2002: 794).

The general theory approach that I discussed stands between these two positions. On the one hand, this approach certainly does encourage historical sociologists to explicitly state their key assumptions and lines of reasoning. In this sense, it accords with Cameron and Morton’s position that, ceteris paribus, formalized arguments are better than nonformalized arguments, because formalized arguments have greater clarity, rigor, and unity—all of which enhance the ability of scholars to determine if their arguments are valid. At the same time, however, the approach recognizes that complete formalization is not possible for the kinds of arguments that are developed in historical sociology. This is true because these arguments rest on an infinite range of tacit assumptions, and it would be pointless to try to uncover and expose all of these assumptions. Hence, scholars using general theory must focus on making explicit their key assumptions and main lines of argumentation, recognizing that this formalization will inevitably leave out certain aspects of more complex arguments. I believe this standard sets a reasonable and realistic set of goals for historical sociology.

Beyond these observations, I would add that I am sympathetic to Sica’s contention that sometimes even the central arguments of historical sociologists may not be subject to easy formalization. And I share his view that these arguments are not necessarily false:  an inability to explicitly state the assumptions of an argument does not make the argument wrong. However, unlike Sica, I do not believe that historical processes are inherently immune to formalization; rather, the problem rests with the scholarly community’s current inability or unwillingness to model certain kinds of assumptions. Furthermore, I believe that the failure to identify key underlying assumptions leaves an argument open to the legitimate criticism of lacking a high level of clarity, rigor, and unity.

The D-N Model versus General Theory

At the core of Sica’s critique is the argument that general theory is committed to a positivist model of explanation, and that this model has failed despite two hundred years of attempts to make it work. Here I concur with Sica that traditional positivist explanation rooted in the “deductive-nomological” (D-N) model is problematic. However, I argue that general theory (as I defined it in the article) is grounded in scientific realism and differs in major respects from this D-N model.

The D-N model associated with Karl Popper, Karl Hempel, and Ernest Nagel holds that an event (E) is explained when it can be logically deduced from the existence of one or more conditions (C1, C2, Ck) and one or more general laws (L1, L2, Lk).[3]  For example, the event of social revolution in Cuba might be explained by:  (1) the conditions of a strong guerrilla movement, a weak state, and a permissive international environment; and (2) the general law holding that the combination of a strong guerrilla movement, a weak state, and a permissive international environment is sufficient for a social revolution. Although this mode of explanation has some intuitive appeal, it embodies three critical flaws that make it inappropriate for social science analysis (see Bennett and George 2005). In turn, general theory (as I have constructed it) largely avoids these shortcomings.

The first problem concerns the deterministic character of “general laws” as originally formulated by Hempel and Oppenheim. Sica rightly notes that the sociological quest for “laws” that permit fully reliable prediction and control has been a hopeless venture (see also Fay 1975). All nontrivial social regularities are either probabilistic and/or bound by scope conditions that limit their applicability. Hence, the D-N model must be modified such that probabilistic regularities, including assertions about probabilistic necessary and probabilistic sufficient causes, replace general laws as the means of linking initial conditions with events.

The version of general theory that I offer is built around the use of these probabilistic postulates rather than law-like postulates. In this sense, general theory is more consistent with Hempel’s (1965:381–425) inductive-statistical (I-S) model, which assumes that laws are conditional probabilities, not universal regularities. To be sure, this move toward probabilistic postulates raises major problems for logically deducing propositions, which also must be probabilistic.[4]  However, it does mean that general theory stands in sharp contrast to models that use “laws” in the sense of deterministic propositions that apply broadly across time and space. As I suggested in the original article, general theories are “general” by virtue of their use of abstract mechanisms, not because they depend on a conception of the social world in which law-like regularities obtain.

A second problem with the D-N model concerns the discovery of the general laws (or the probabilistic regularities) that work in conjunction with conditions in the world. Hempel and others never resolved the issue of how to explain the laws themselves. In the field of historical sociology, by contrast, we know that regularities and causal associations are developed inductively in the course of case research. Inductive reasoning is thus fused with deductive reasoning in the following way:  (1) the overall relationship between postulates and propositions is deductive—i.e., if the postulates are true, then the propositions are also true; and (2) the individual postulates are often inductively derived—i.e., they are conclusions drawn from expert case analysis. Given this fusion, induction must be assigned a much more important role than implied by the D-N model.

A final limitation concerns the D-N model’s equation of explanation with prediction. For various reasons, the idea that prediction equals explanation is deeply unsatisfying. To cite only one famous example:  a barometer’s reading may predict weather patterns, but the barometer certainly does not causally explain these patterns (see Salmon 1990:46–47). As an alternative, scientific realists of various stripes maintain that to explain a phenomenon is to identify the potentially unobservable causal mechanism that generates the phenomenon (e.g., Bhaskar 1979; Harré 1970; McMullin 1984; Sayer 1992). This scientific realist model of explanation is what I discussed in the article.

The scientific realists’ emphasis on explanation through causal mechanisms is often praised by philosophically-oriented sociologists, but the concept of “causal mechanism” remains confusing (Mahoney 2001; Mayntz 2004). I therefore sought to clarify this concept and its role in explanation. I drew heavily on writings that suggest causal mechanisms can solve the black box problem of correlational research. I argued that causal mechanisms help scholars overcome this problem because they represent entities that transcend time and space. I suggested that only such abstract entities could effectively fill all temporal and spatial gaps standing between two or more associated variables. I further offered five specific examples of causal mechanisms and discussed instances in which these mechanisms are applied in real research. In this sense, I wanted to present a concrete illustration of the ways in which explanation through causal mechanism actually works in practice.

Positivistism versus Scientific Realism

I have argued here that Sica attacks a positivist epistemology that only partially corresponds with the epistemology of general theory as defined in my article. To tie this argument together, it is useful to review the ways in which the general theory program does and does not correspond with positivist epistemology.

First, consistent with positivism, advocates of general theory are committed to formalization, in the specific sense that they believe it is a good idea to explicitly state their important assumptions and lines of reasoning. Indeed, all other things being equal, an argument that is stated formally is better than one that is not stated formally. At the same time, however, advocates of general theory must recognize that all other things are not always equal. For instance, it hardly makes sense to try to formalize all of the millions of implicit assumptions embodied in any argument. Nor should one necessarily dismiss an argument because it cannot be formally stated in its entirety. Rather, general theory encourages scholars to pursue formalization within reasonable limits, recognizing that there are trade-offs in research. Depending on these trade-offs, less formalized arguments are sometimes more appropriate than more formalized arguments even within the framework of general theory.

Second, like traditional positivists, proponents of general theory follow a mode of analysis in which propositions are logically derived from postulates. Indeed, general theorists treat postulates as jointly sufficient for the truth of propositions. However, unlike some positivists, they advocate the use of probabilistic postulates that are inductively formulated, and they use these postulates to derive propositions that are also probabilistic. Accordingly, they reject both the notion that good science involves explanation through deterministic laws and the idea that good research is purely or primarily deductive in form. Moreover, supporters of general theory avoid the positivistic tendency of equating prediction with explanation. Instead, they follow scientific realism in explaining social phenomenon in terms of specific but unobservable causal mechanisms.

I therefore prefer to use the label scientific realism to describe the epistemology of general theory, and in my view this epistemology stands in sharp opposition to the positivist epistemology associated with both the Vienna Circle of the early twentieth century and the D-N model that took hold in the mid-twentieth century. Sica has good reason to be hostile toward these positivist epistemologies, but he unfortunately has aimed his sights at the wrong target in attacking the version of general theory that I described. General theory represents a major epistemological break with traditional forms of positivism.

Ontological Debates

Scientific realism also has important ontological commitments that differentiate it from alternative philosophies. A discussion of these commitments provides a gateway to other concerns raised in Sica’s reply and allows me to weigh in on the relative merits of alternative general theories.

Sica is skeptical of my arguments about the utility of unobservable mechanisms in social science explanation. Indeed, he seems to favor a “nominalist” position in which social science concepts (e.g., state, gender, class) are deemed to be nothing more than mental constructs, lacking a true existence in reality. My realist position, by contrast, allows one to potentially grant the status of existence to social science concepts, including when they refer to unobservables.

I follow the realist literature in believing that one can legitimately treat unobservables as real entities insofar as one can show that these entities have real effects in the world (Hacking 1983). If, for example, one can show that the positing of unobservable “resources” facilitates the explanation of important outcomes and associations, then one has reason to believe that these resources are a real entity despite their unobservable status. The burden of proof for the existence of an unobservable entity is thus grounded in explanatory criteria; unobservables that repeatedly generate successful explanations can persuasively be treated as real.

Sica’s comments more fundamentally raise the question of whether social entities such as classes, states, and cultural groupings can be regarded as having an existence independent of human ideas about them. Sica’s contention seems to be that social entities lack this mind-independent existence, which in turn motivates a post-positivist argument that sociologists should pursue a kind of interpretive or hermeneutical approach in which actors’ subjective understandings are the central focus.

To be sure, social entities like states and cultural communities are not the same kinds of things as natural entities like trees and water. In particular, social entities cannot exist in the absence of actor practices and beliefs, whereas natural entities can (Bhaskar 1979:48–49). Nevertheless, there are very good reasons to treat social kinds as real and independently existing things. In the first place, social entities do not depend on the beliefs or practices of any single individual for their existence; rather, they exist independently of particular individuals (Wendt 1999:75). As Bhaskar (1986:52) puts it, “the objects of scientific enquiry exist and act, for the most part, quite independently of scientists and their activity.”  In consequence, the individuals who seek to analyze social kinds can treat them as objective features of the world that are constituted by outside collectivities.[5]  Equally important, insofar as one subscribes to a material explanation of human consciousness (e.g., Dennett 1991), the collective ideas that help constitute social entities can themselves be reduced to ontologically prior material conditions. On this view, to which I am sympathetic, both social kinds and human ideas about social kinds are ultimately the product of material conditions.

In short, the positing and use of unobservables is fully consistent with good scientific practice, and unobservables that are shown to exert effects in the world should be regarded as real entities. Furthermore, many social kinds are objective entities that exist by virtue of conditions other than specific researchers’ ideas about them.

The five general theories that I discussed assume that different causal agents and causal mechanisms produce social events and empirical regularities. Accordingly, even scholars who agree on the epistemology of general theory may still disagree on the question of which ontology should guide social research.

Sica’s remarks related to this issue mostly revolve around a critique of instrumental rationality. He especially draws on Diesing’s (1950) early work, which calls into question the idea that all or most action is driven by economic utility maximization and other cost-benefit forms of calculation. Sica further argues that rational choice theory as I described it ’smacks of the politically reactionary, the culturally imperious, and what might overall be termed ‘the Popperian worldview’” (2005:12 of  text).

I do not wish to engage here the normative implications of rational choice theory. As I suggested in the article, all general theories have both empirical and normative entailments. And while both kinds of entailments are important, it is useful to distinguish between the two, and my focus is on the empirical side. In this empirical regard, I have no special interest in defending rational choice theory. In fact, I have criticized rational choice theory elsewhere, in part because its scope of application is inherently limited (Mahoney and Rueschemeyer 2003). Nevertheless, I believe it is important to judge individual rational choice works on a case-by-case basis. From this perspective, some works in this area—such as Alexander’s The Sources of Democratic Consolidation (2002)—are major successes; others are not. In the case of Brustein’s (1996) book, I believe he presented a coherent argument with interesting data to support the idea that German voters were partially motivated by economic considerations in their decisions to vote for or join the Nazis. I disagree with Sica that the sources of support for Hitler and the Nazis are so obvious that one need only read the Versailles Treaty. Likewise, if I may make one normative remark, I do not believe that the effort to formalize the causes of support for the Nazis “verges on the morally repugnant” (Sica 2005:15). I find it normatively helpful to test the hypothesis that “decent people” may be motivated by economic considerations to vote for even the most “indecent leaders.”

In the article, I was more interested in proposing power theory as a rival to rational choice theory. Among other things, I sought to show that many leading historical sociologists are theoretically unified in the sense that they (implicitly) treat resources as foundational causes in the social world. I further sought to show that rational choice theorists hardly have a monopoly over general theory. Indeed, if we are willing to count implicit applications, power theory is probably the most commonly employed general theory in the field of historical sociology today.

Sica argues that my short discussion of functionalism oversimplifies more complex arguments, doing them great injustice in the process. Yet my goal was never to systematically evaluate these arguments. Rather, I started with the recognition that most contemporary sociologists already reject functionalism. To explain why, I attempted to show how scholars have criticized functionalist arguments in ways that are consistent with the epistemology of general theory. Beyond this, I did not seek to make any novel claims about functionalist theory.

Sica is correct that my short presentations on neo-Darwinian theory and cultural theory leave some unanswered questions. However, I still believe that the discussions are useful for illustrating the range of general theories that are available to historical sociologists. Regarding neo-Darwinian theory, I hope that more sociologists will become familiar with evolutionary psychology, for the approach merits serious attention within the discipline. Regarding cultural theory, Sica concludes by applauding my argument that cultural theorists sometimes reject causal analysis as the central goal of research. I want to emphasize, however, that cultural theorists are in fact divided over causal analysis. Some, like Sica, feel liberated by embracing an interpretive mode of cultural analysis that rejects causality as the central goal of research. Others see cultural theory as primarily a mode of causal analysis. Still others may be uncertain regarding the relationship between cultural theory and causal analysis. My article sought to offer an epistemology for those cultural theorists who believe it is profitable to treat semiotic practices as a foundational cause of associations and outcomes in the world.

Conclusion

Notwithstanding Sica’s critical reply, one of my goals in writing about general theory was to build a bridge across diverse scholarly groups. This bridge-building effort involved identifying an epistemological common ground that might connect scholars representing different theoretical traditions. It specifically took the form of the argument that general theory is potentially consistent with the research of many practicing historical sociologists, including those associated with functionalist theory, rational choice theory, power theory, neo-Darwinian theory, and cultural theory. Despite their different ontologies, these traditions have more in common with one another and with scientific theory than Sica and other skeptics may care to admit.

References

Alexander, Gerard. 2002. The Sources of Democratic Consolidation. Cornell University Press.

Bennett, Andrew, and Alexander L. George. 2005. Case Studies and Theory Development. MIT Press.

Berger, Peter L., and Thomas Luckman. 1966. The Social Construction of Reality: A Treatise in the Sociology of Knowledge. Doubleday.

Bhaskar, Roy. 1979. The Possibility of Naturalism: A Philosophical Critique of the Contemporary Human Sciences. Humanities Press.

———. 1986. Scientific Realism and Human Emancipation. Verso.

Brustein, William. 1996. The Logic of Evil: The Social Origins of the Nazi Party, 1925–1933. Yale University Press.

Cameron, Charles R., and Rebecca Morton. 2002. “Formal Theory Meets Data.”  Pp. 784–804 in Political Science: State of the Discipline, edited by Ira Katznelson and Helen V. Milner. W.W. Norton.

Costner, Herbert L., and Robert K. Keik. 1964. “Deductions from ‘Axiomatic Theory.’”  American Sociological Review 29:819–35.

Dennett, Daniel Clement. 1991. Consciousness Explained. Little, Brown, and Company.

Dessler, David. 2003. “Explanation and Scientific Progress.” Pp. 381–404 in Progress in International Relations Theory, edited by Colin Elman and Miriam Fendius Elman. MIT Press.

Diesing, Paul. 1950. “The Nature and Limitations of Economic Rationality.” Ethics 61:12–26.

Fay, Brian. 1975. Social Theory and Political Practice. George Allen and Unwin.

Hacking, Ian. 1983. Representing and Intervening. Cambridge University Press.

Harré, Rom. 1970. The Principles of Scientific Thinking. University Chicago Press.

Hempel, Carl G. 1965. Aspects of Scientific Explanation and Other Essays in the Philosophy of Science. Free Press.

Kiser, Edgar, and Michael Hechter. 1991. “The Role of General Theory in Comparative-Historical Sociology.” American Journal of Sociology 97:1–30.

Mahoney, James. 1999. “Nominal, Ordinal, and Narrative Appraisal in Macrocausal Analysis.” American Journal of Sociology 104:1154–96.

———. 2001. “Beyond Correlational Analysis: Recent Innovations in Theory and Method.” Sociological Forum 16:575–593.

———. 2004. “Revisiting General Theory in Historical Sociology.” Social Forces 83:459–89.

Mahoney, James, and Dietrich Rueschemeyer. 2003. “Comparative Historical Analysis: Achievements and Agendas.” Pp. 3–38 in Comparative Historical Analysis in the Social Sciences, edited by James Mahoney and Dietrich Rueschemeyer. Cambridge University Press.

Mayntz, Renate. 2004. “Mechanisms in the Analysis of Social Macro-Phenomena.” Philosophy in the Social Sciences 34:237–54.

McMullin, Ernan. 1984. “Two Ideals of Explanation in Natural Science.” Midwest Studies in Philosophy 9:205–20.

Ragin, Charles C. 1987. The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. University of California Press.

———. 2000. Fuzzy-Set Social Science. University of Chicago Press.

Rueschemeyer, Dietrich, Evelyne Huber Stephens, and John D. Stephens. 1992. Capitalist Development and Democracy. University of Chicago Press.

Salmon, Wesley. 1984. Scientific Explanation and the Causal Structure of the World. Princeton University Press.

———. 1990. Four Decades of Scientific Explanation. University of Minnesota Press.

Sayer, Andrew. 1992. Method in Social Science: A Realist Approach, 2nd ed. Routledge.

Sica, Alan. 2004. “Why ‘Unobservables’ Cannot Save General Theory: A Response.”Social Forces 83:491–501.

Wendt, Alexander. 1999. Social Theory of International Politics. Cambridge University Press.