Historical sociologists on critical realism

Critical realism took its origin within the philosophy discipline, arising at the time that there was profound debate over the adequacy of logical positivism as a basis for the philosophy of science. Carl Hempel represented the fruition of positivist philosophy of science, with his hypothetico-deductive model of confirmation, his deductive-nomological model of explanation, and his covering-law model of historical explanation. These all amount to the same idea, of course: that scientific knowledge takes the form of a set of general theoretical principles or laws, a set of empirical statements about existing conditions, and a set of deductions from the laws and statements of consequences for the observable phenomena. There was a strong reaction in the 1960s to the orthodoxies of logical positivism and Hempelian philosophy of science by philosophers such as Norwood Hanson, Paul Feyerabend, and Thomas Kuhn. Compelling criticisms were offered of the strict distinction between observation and theory, concerns were raised about the putative coincidence of explanation and derivation from general laws, and more nuanced theories of scientific rationality than the hypothetico-deductive method were offered.

A particular sticking point within the positivist theory of science was its common adherence to a Humean theory of the meaning of causation as constant conjunction. Hume derided the idea of “causal necessity” and sought to replace this notion with the idea of conformance to a strong regularity. Rom Harré and Edward Madden undertook a strong critique of this assumption in Causal Powers: Theory of Natural Necessity, also in the mid-1960s. And this anti-positivist strand of thinking about causation was more important to the emergence of critical realism than any other influence.

Several gifted sociologists joined this debate in the 1990s. Especially astute were contributions by George Steinmetz and Margaret Somers, both colleagues at the University of Michigan, and Philip Gorski at Yale. In a review article in Society for Comparative Study of Society and History in 1998 Steinmetz provides a careful review of the intellectual background and the central ideas that Roy Bhaskar introduces in his writings on naturalism and realism (link).  Steinmetz reviews the mainstream assumptions that defined positivist philosophy of social science through the 1960s and the echo of these assumptions in mainstream sociology; and he provides a fairly detailed description of Bhaskar’s alternative. He emphasizes several central ideas:

• the transcendental nature of Bhaskar’s reasoning: “discovering what must be true about the world for science to be possible” (176)
• the distinctions among the real, the actual, and the empirical
• the scientific importance of “open systems” — systems lacking causal closure and displaying contingency
• a specific idea about emergence — "Emergence is defined as the relationship between two levels such that one arises diachronically (or perhaps synchronically) out of the other but is capable of reacting back on the lower level and is causally irreducible to it (Bhaskar 1993:73) [178]

Following Peggy Somers in "We're No Angels" (1998; link), Steinmetz distinguishes between "theoretical realism" and critical realism; essentially the former concept refers to the standard version of scientific realism found within post-positivist philosophy of science (Hilary Putnam, Richard Boyd). The key attribute of scientific realism, according to Steinmetz, is its continuing adherence to the idea of scientific laws; "theoretical realism is strongly deductivist" (173). So Steinmetz believes that "theoretical realism" is more closely aligned with positivism than is critical realism, and that critical realism fits the practice of historically minded social scientists better.

I will argue that most historical researchers, whatever their self-description, are critical realists rather than theoretical realists, positivists, or neo-Kantian idealists, and that this stance is the most defensible one for the social sciences in general on ontological and epistemological grounds. (171)

Steinmetz believes that the philosophy of science articulated within critical realism accords very well with the practice of historically minded social scientists like himself. He closes his article with these words:

Critical realism is especially “liberating” for historical sociology. It provides a rebuttal to the positivist and theoretical realist insistence on the dogmas of empirical invariance, prediction, and parsimony (see Bhaskar 1989:184). Critical realism guards against any slide into empiricism by showing why theoretical mechanisms are central to all explanation. At the same time, critical realism suggests that contingent, conjunctural causality is the norm in open systems like society. Yet critical realism’s epistemological relativism allows it to accept the results of much of the recent history and sociology of science in a relaxed way without giving in to judgmental relativism. Historical social researchers are reassured of the acceptability of their scientific practice, even if it does not match what the mainstream misconstrues as science. Critical realism allows us to safely steer between the Scylla of constricting definitions of science and the Charybdis of solipsistic relativism. (184)

The methodology that Steinmetz commends is one that highlights social contingency and conjuncture, while at the same time discovering explanatory relations among circumstances based on the causal mechanisms we can identify that connect them. These are all important aspects of sociological research, and we should indeed seek out philosophies of social science that make room for them.

That said, I am not persuaded by the unfavorable distinction that Steinmetz and Somers draw between scientific realism and Bhaskar-style critical realism. I am inclined to think that the tradition of scientific realism has less baggage (from logical positivism) and critical realism has more (from Bhaskar's sometimes arcane philosophical arguments and distinctions deriving from transcendental philosophy). Here is Steinmetz on the deficiencies of theoretical realism:

Theoretical realism disparages explanations which invoke unique, nonrepeatable constellations of causal mechanisms in accounting for specific historical conjunctures. (174)

But this doesn't really seem to be an accurate portrayal of a wide range of scientific realists, including Richard Boyd. In fact, we can better look at the tradition of scientific realism as being closer to another tradition that Steinmetz admires, that of American pragmatism. (For a long time Harvard's department of philosophy was the home of scientific realism, and it was also the intellectual heir of James and Peirce.) Scientific realism, when considered as a meta-theory of the work of social sciences, simply extends to the social sciences the ontological elbow room that the natural sciences have long enjoyed: when we postulate unobservable entities, causes, and processes, we are sometimes justified in believing that these entities actually exist -- provided that our hypotheses are appropriately linked to observation and inference emanating from a dense field of scientific inquiry.

Take a sociological construct from Bourdieu that Steinmetz finds to be very useful, the idea of an intellectual field (link). This construct plainly invokes an extended and intangible social structure or entity -- an interconnected system of individuals, values, and institutions that steer the progress of persons and ideas through their careers. The concept has proven to be a plausible and contentful way of conceptualizing sociological phenomena across a broad range of contexts (intellectual and cultural history, imperialism, scientific research, political ideology), and Steinmetz and other sociologists are justified in attributing real existence to this construct. But this realist interpretation of the construct does not require esoteric philosophical reasoning; we can look at it as a very ordinary and pragmatist inference from the orderliness of a specific range of social phenomena to the best explanation -- that there is an underlying field of interrelations that generates this orderliness. And it seems to me that mainstream scientific realists like Boyd and Putnam would be very satisfied with this line of reasoning.

So I would look at these comments as a minor corrective to Steinmetz's argument here: social scientists are indeed well advised to be anti-positivist; they are well advised to be realist in their theorizing; but there is nothing in the case that suggests that Bhaskarian realism is the particular variant of realism they should assume. A more pragmatic and pluralistic version of scientific realism seems more suitable to research in sociology. (Here is a brief discussion of a more pluralistic and eclectic version of scientific realism for the social sciences; link.)

Naturalizing causal powers

Several earlier posts have considered Tuukka Kaidesoja's very interesting recent book, Naturalizing Critical Realist Social Ontology (NCR). The book is an important contribution to the evolving literature on next steps for critical realism, and TK is an exceptionally clear and perceptive philosopher. Here I will focus on Tuuka's contribution to the causal powers literature.

The topic of causal powers is important for current debates within the philosophy of social science. This is especially true when it comes to the question of the causal role that supra-individual social entities play. Like Dave Elder-Vass in The Causal Power of Social Structures: Emergence, Structure and Agency, I want to support the idea that social structures (for example, organizations) have causal powers and properties, and a social structure is supra-individual entity. E-V presents this notion in terms of the idea of emergence, whereas I propose to understand it in terms of the notion of relative explanatory autonomy (link, link). But in each case, we hold that it is legitimate to attribute a causal power to a composed social entity, and that there is no compulsion to “reduce” that power to the individual powers of the persons who compose the entity. What is it about the social structure that gives rise to the causal power?

There are two important points to consider here. First, we need to ask what the terms of the causal relation are thought to be. Is it the abstract structure of the organization (shared with other organizations of the same type) that exerts causal power; or is it the concrete particular, this particular instantiated organization, that is the causal agent? I want to maintain that it is the particular social structure, not the abstract structure, that bears the causal role and exerts the causal power.

Second, the traditional account from critical realism and Bhaskar would hold that the powers of a social structure derive from its “essential” properties. But following Kaidesoja, it is both reasonable and justified to drop the essentialism associated with this line of thought. Instead, we can say that the powers of the structure derive from its contingent but current features of organization and functioning. In the case of a social organization, this comes down to the particular set of rules and practices that drive the organization at a point in time. As long as these rules and practices persist, the organization will continue to have the powers that we attribute to it. When those rules and practices undergo change and innovation, it is an open question what changes will result for the causal powers of the organization.

Kaidesoja approaches a view very similar to this in his treatment of Harré and Secord’s analysis of individual and collective powers:

I suggest that these views [advanced by Harré and Secord] presuppose that rules and institutions possess causal powers that are ontologically irreducible to those of individuals. (115)

So what about the assumption of essentialism that is often part of the definition of a causal power? TK takes up the issue of essentialism and natural kinds within causal-powers theory, and argues that we need to "naturalize" this issue as well. Whether there are natural kinds in a particular domain is a question for the sciences to answer, not the philosophers. TK notes that modern biology does not support the notion that biological things (including species) fall into natural kinds defined by distinctive essential natures.

Biological variation between and within species (or populations) is thus a normal state of affairs in nature and there is no a priori limit for such variation…. This means that it is no longer plausible to conceive biological species as natural kinds in Harré and Madden’s (1975) sense. (111-112)

So natural-kind essentialism does not fit the entities and processes of the biological realm.

Whether or not the essentialist notion of causal power an be applied to a certain collection of objects studied in a specific discipline should be decided by means of empirical analysis of the scientific research practices, theories and models that are developed in this discipline. (112)

But TK does not believe that this invalidates the idea that biological entities have causal powers; and this entails that there is a separation between essentialism and the attribution of causal powers.

I have argued at many points here that this feature of heterogeneity and change in some of the core characteristics of entities is fundamental to the social world as well (link). So TK's central insight here is important for the philosophy of social science as well as for biology: causal powers should not be defined in terms of the essential properties of an entity; causal-power theory should not be constructed in such a way as to presuppose essentialism.

One thing I especially appreciate in TK's treatment of causal powers is the light he sheds on the difference between logical or conceptual necessity, on the one hand, and natural necessity, on the other (106). This is relevant to the earlier discussion here about whether causes necessitate their effects (link). There I argued against the views of Mumford and Anjum, who reject necessity, on the grounds that their argument turns on features of logical necessity that do not attach to causal necessity. Kaidesoja's discussion here reinforces my conviction that it is reasonable to assert causal "necessitating" even when we acknowledge that causes are sometimes not followed by their effects. Discussing Harré and Madden TK writes:

The concept of natural necessity is thus carefully distinguished from the concepts of logical, transcendental and conceptual necessity (ibid., 19–21). (107)

Kaidesoja emphasizes the similarity of views that exists between Harré and Bhaskar concerning the specification of a causal power. Here is one typical statement from Bhaskar's A Realist Theory of Science, among many that TK quotes:

To say that a thing has a power to do something is […] to say that it possesses a structure or is of such kind that it would do it, if appropriate conditions obtained. (RTS, p. 88) [118]

The parallel with Harré’s formulations is evident. TK finds that Bhaskar’s main innovation on this point is his attempt to make a transcendental argument for the necessity of attributing real causal powers to entities, and this is a move that he rejects. TK finds that Harré and Madden’s account is more convincing exactly because it locates causal powers in the realm of “concrete powerful particulars”, not in the transcendental realm (121, 122).

Due to the aforementioned problems in the transcendental realist account of the concept of causal power, I prefer Harré and Madden’s Aristotelian conceptualization of causal powers which interprets them as efficient causes and ties them inseparably to the concrete powerful particulars. (122).

And this in turn provides an additional reason to reject the essentialism associated with Bhaskar’s broader conception of causal powers (that the causal power of a thing derives from its essential nature). This becomes the heart of TK’s concept of a “naturalized version of the concept of causal power” (136), and it seems to be a very plausible position.

Quantitative and qualitative social science

The social world is one reality, but the methodologies associated with quantitative and qualitative research are quite different. Quantitative research allows the researcher to discover patterns, associations, correlations, and other features of a population based on analysis of large numbers of measurements of individuals. Qualitative research usually involves studies of single individuals, based on interviews and observations, with the goal of identifying their internal psychological and behavioral characteristics. Quantitative research is directed at identifying population characteristics, patterns, and associations. Qualitative research is directed at teasing out the mental frameworks and experiences of individuals within specific social and cultural settings. Qualitative researchers are generally not interested in discovering generalizations or regularities, and are more interested in identifying particular features of consciousness, culture, and behavior.

What kinds of interface or bridging are possible between these two levels of social research?

Take the example of race studies. Both qualitative and quantitative research studies have been conducted in this field, with the goal of shedding light on the phenomenon of race in American society. Quantitative research has often been concerned to identify the features of inequality which are associated with race within American populations, including income, wealth, education, health, employment, and other important features. For example, the National Survey of Black Americans provides voluminous data on a range of characteristics of African American individuals, with surveys extending from 1979 to 1992 (link). Here is a list of the variables included in these studies (link). Several hundred research studies and reports have been completed making use of these data sets; here is a representative study by James Jackson making use of data sets like these to probe health disparities by race (link). These quantitative studies permit the researcher to use advanced statistical tools to measure and evaluation the strength of associations among characteristics and to evaluate causal hypotheses about the linkages that exist among characteristics.

Qualitative research on race takes several forms. There are ethnographic studies, through which the researcher attempts to identify the phenomenology and lived experience of race. Here I would include several research efforts that have been discussed here previously -- Al Young's study of young inner city Chicago men (The Minds of Marginalized Black Men: Making Sense of Mobility, Opportunity, and Future Life Chances) and Loïc Wacquant's ethnographic study of a boxing club on Chicago's south side (Body and Soul: Notebooks of an Apprentice Boxer). There are theoretical studies, which explore possible structures or mechanisms which produces racial and racialized behavior and disparities. Here is a good example from Elizabeth Cole on the construct of intersectionality as a way of theorizing about racial and gender identities (link). And there are studies of social psychology designed to identify the ways in which racial attitudes, presuppositions, and ideas contribute to behavior in American society. Here is a nice example of such an analysis by Lawrence Bobo and Cybelle Fox (link).

It is clear that studies based on all of these methodologies are insightful and valuable. We will arrive at a better understanding of the meaning and causal importance of "race" through all these approaches. The question raised above remains an important one, however: how should we think about the relations among these bodies of inquiry and knowledge?

One way is to think in terms of levels of analysis (link): we might say that quantitative studies examine facts about race at a more macro level (large populations), whereas qualitative studies are more meso- or micro-level studies. This isn't a very satisfactory view, however, because each of these approaches is concerned about individual-level facts; what differs is the level of aggregation of those facts that is chosen.

Another approach seems more promising: to consider the suite of qualitative studies of race as being a tool box for identifying the social mechanisms through which the patterns and associations that are discovered at the large population level come about. Qualitative studies (studies aimed at discovering or theorizing the mentalities and behaviors through which race is constructed and carried out) permit us to understand racialized behavior in groups that in turn allow us to understand the population outcomes that quantitative studies identify.

A third possibility is that these different methodological approaches do not admit of "bridging" at all. Here the idea would be that these are fundamentally different forms of knowledge, and they belong in different parts of the toolbox. Sometimes this approach is taken by advocates of one methodology or the other in dismissing the scientific credentials of the other approach -- quantitative researchers who dismiss qualitative research as anecdotal and qualitative researchers who dismiss quantitative research as positivist. This approach seems fundamentally wrong. We should look at the various ways of studying important aspects of social life as being complementary and fundamentally consistent.

My own predilection is to think of the qualitative approaches as providing insight into how various social processes work; how it is that socially constructed actors bring about the patterns of behavior and outcome we observe at various levels of aggregation. A quantitative study of racial attitudes might suggest that cities with effective public transportation have higher (or lower) levels of racial mistrust across groups. We would want to be able to form some hypotheses about what the underlying behaviors and attitudes are that bring about this effect. What are the mechanisms through which access to public transportation influences racial trust? And for this kind of inquiry to be possible, we need to have some good empirical theories about racial identities and mental frameworks.

So it does in fact seem both possible and desirable to try to integrate the findings of both quantitative and qualitative studies of racial attitudes; and this finding seems equally valid in almost all areas of the social sciences.

(Thanks to Mosi Ifatunji for his stimulating seminar at the University of Michigan on new approaches to the study of black ethnic disparities, which caused me to think about this topic a little further. Here is Mosi's webpage with some links to his work; link.)