Critical realism

Critical realism is neither a method nor a theory (Mutch, Delbridge & Ventresca 2006). It is a philosophy of science (Fairclough 2005, Sayer 1992) which acts as philosophical under-labourer of the social sciences (Mutch, Delbridge & Ventresca 2006). Therefore, critical realism is not a cookery book for research design, but an analytic model which assists and strengthens reflection on the fit between data and theory. It is a way “to combine an account of process on the one hand with one of the enabling context on the other, without collapsing the one into the other” (Mutch, Delbridge & Ventresca 2006 p. 622). The term critical has a number of meanings in relation to critical realism. It signifies a critique of

 the taboo on ontology which characterized economic schools at the time when Roy Bhaskar introduced critical realism (Buch-Hansen 2005)

 the artificially closed systems in experiments which are applied in order to get access to and isolate the operation of one causal law, but excluding all others. Such experiments assume that the causal law under study will operate in the same way in the open systems of nature and society outside the experimental setup (ibid.)

 the restrictive view of scientism which centers around the search for regularities and hypothesis testing (Sayer 1992)

But it is also a critique of the reduction of social science to the interpretation of meaning (Sayer 1992), and of the reduction of discourse to a representational role, with no generative properties. This reduction eliminates the capacity of language to reshape human agency, and the structural patterns it generates, reproduces, and elaborates (Reed 2000). Critical realists agree that entities are conceptualized through discourse, but they are not made through discourse (Fleetwood 2005). There is a social world outside our heads which is socially defined through language based distinctions (Tsoukas 2000).

A final aspect of the term critical refers to the potentiality of critical realism to offer critique by questioning and re-conceptualizing the understanding of reality. One of the possible roles for the social sciences is to show why a false belief of a subject matter is generated, and this endows social science with an emancipatory role (Buch-Hansen 2005).

If it is possible to expose false beliefs, it follows that it may be possible to expose what generates the false belief. This being so, change is possible.

Figure 2.1: The temporal aspect of how structure and interaction influence each other (Archer 1998 p. 376)

Potential change is linked to the ability of human agency either to reproduce or to transform existing structures and practices through action (Fairclough 2005). But the study of potential change makes it necessary to break into the cycle of action and impose an analytical starting point at a certain point in time and space. In this arbitrary point of analysis, structures will pre-exist action and interaction, which may reproduce or transform the pre-existing structure as illustrated above in figure 2.1. And it is construction of a narrative of this situation or event which is the core of a critical realist analysis (Mutch, Delbridge & Ventresca 2006).

2.1.1 Ontology

Critical realists state that knowledge claims (epistemology) cannot exist in a vacuum: the validity of such claims depends on the assumptions of what the world has to be like (ontology) if claims are to be valid. Moreover, the adequacy of knowledge claims must be ascertained in the empirical (Tsoukas 1989). This diversification and stratification of the reality into the domain of the real, the actual and the empirical are core tenets of critical realist ontology. In addition to the observable empirical domain of experience, there is another domain of actual reality in terms of events. But the operation of these two levels is under influence of the powers or mechanisms of entities and structures in the domain of the real.

Table 2.I illustrates the classification of the domains. Reading the table from right to left it illustrates that experience in the empirical domain presupposes the occurrence of events in the actual domain. Events are independent of a researcher’s taking notice of these events. But events presuppose the existence of mechanisms in the real domain, which are responsible for the generation of events (Tsoukas 1989). Or reading the table from left to right, the real is the domain of structures and their associated ‘causal powers’, the actual is the domain of processes and events, and the empirical is what is experienced by social actors (Fairclough 2005).

Domain of real Domain of actual Domain of empirical

Mechanisms X

Events X X

Experiences X X X

Tabel 2.I: Classification of the real, the actual and empirical domains (Bhaskar 1978a p. 13)

The analysis rests on the distinction between internally linked or necessary relations on the one hand, and contingent or externally linked relations on the other hand. Internally linked objects, entities or individuals depend in their being on the relationship with the rest of the components in the structure: e.g. slaves, masters and the institutions supporting slavery.

Structure is defined as necessarily related objects or practices and the powers that they possess (Sayer 1992). These mechanisms or powers are “something that makes something else happen” (Buch-Hansen 2005 p. 61). They operate as tendencies, but their activation as well as the effect of their activation are not given, but contingent (Tsoukas 1989).

Contingencies consist partly of the powers of other entities in the structure, and partly of the contingent conditions in terms of externally linked relations; i.e. a relation between elements which may influence the two elements involved, but which is not necessary for the two elements to exist. Consequently, mechanisms are irregular. They do not display an automatic or general pattern of acting; their workings depend on the specific conditions in terms of the co-acting of other mechanisms or contingent relations as illustrated below in figure 2.2. And “while the elements of structures are necessarily related, it is contingent whether any structure as a unit, exists” (Sayer 1992 p. 97).

Fig 2.2: The structures of causal explanation (Sayer 1992 p. 109)

Research designed for establishing and generalizing causality in the empirical only touches upon the surface of phenomena and cannot catch this irregularity of forces which are not observable in the empirical. Therefore, the analysis of causality is substituted by causation in a critical realist analysis. The purpose of causation is to analyze how the powers necessarily possessed by objects influence events under the specific conditions of the externally linked (contingent) relations.

However, the domain of the real is not readily observable or deducible. It has to be inferred in a movement of analogy or metaphor from the conception of a phenomenon (the transitive dimension) to the conception of something quite different in terms of a structure or mechanism (the intransitive dimension) (Lawson 1997). This is why Bhaskar’s (1978b) explication of realism is also called transcendental.

Transcendental realism implies that neither deduction nor induction is satisfactory or sufficient modes of inference for this type of theory-building. Whereas deduction is the mode of inference which enables a move from a general law-like claim to a particular instance, induction is the mode of inference applied for a move from particular

observations to a general claim. None of them are applicable for a move from a number of data to a theory of a mechanism which predisposes the observed phenomena in a specific direction. For that purpose, retroduction is a more appropriate mode of inference allowing the move from surface phenomena to deeper causal structures (Lawson 1997) by

inference from theories as well as singular phenomenon. This allows the move from an understanding and explication of what is going on, to an attempt to explain why. This process is achieved through causation and abstraction (see below).

2.1.2 Epistemology

Sayer (1992) defines epistemology in general terms as the theory of knowledge. He focuses on a number of concepts of epistemological importance in a critical realist perspective.

One of the fundamental issues is the fallibility of knowledge. In a realist perspective this supports the tenet that the world exists independently of our knowledge of it. Sometimes reality strikes back, and we have to realize that our understanding of the world does not reflect reality. This is only possible because there is a difference between the world and our knowledge of it. If knowledge were purely conceptual this could not happen, and truth would be relative to our conceptual schemes.

Nevertheless, our perception depends on concepts in order to rend observations and experiences intelligible in daily life, as well as in science. The world is understood in terms of, if not determined, by concepts. In that sense our conceptualization is pre-structured

like the material world. Therefore, observation is neither theory-neutral nor theory determined, but theory-laden. In spite of the fact that our conceptual world is pre-structured, and that new concepts are built on old ones, concepts and practices are not static. They move and change in a dynamic process of learning and knowing, of practicing and conceptualizing. In this process the role of theory is to negotiate and modify the relationship between what we sense and how we understand and express this understanding (Sayer 1992), or in other words to re-conceptualize our scientific understanding.

This re-conceptualization is a matter of sorting out the elements of the structure in necessary and contingent relations (abstraction), in order to create the foundation for a causal analysis (causation) of process and change. The difference between the analysis of causality and the analysis of causation is that the former is concerned with the relationship between discrete events, whereas the latter is concerned with the continuous processes;

the way in which relations and objects act (Sayer 1992). In other words, causation is a way to make sense of the events in the actual by retroduction; inference from theories as well as singular phenomena.

Similar events Different events

Mechanisms

(nessecary relations) Same mechanism Different Mechanisms

Similar set of Mechanisms

activated by

Conditions

(contingent relations) Similar conditions Similar conditions Dissimilar conditions

Table 2.II The workings of mechanisms and conditions in producing similar and dissimilar events

Whereas causation is concerned with the analysis of process, abstraction is concerned with the analysis of structure. It is a matter of dissolving the concrete whole into its constituent aspects and properties in terms of relations. This identification and description support the conceptualization and identification of the processes by sorting out the necessary

relations, and separating them from the contingent ones. Similarities between events occur due to the influence from the same generative mechanism and similar type of contingencies activating the mechanism. Differences may result either from different mechanisms or result from dissimilar contingencies activating similar sets of mechanisms in different ways (Tsoukas 1989), see table 2.II above.

Figure 2.3: The links between strata and analysis in a critical realist epistemology The definition of structure as the internally related objects and practices directs the attention towards the necessary relations (Sayer 1992), which can give an answer to the question why. But causation rests on abstraction, because the working of mechanisms depends on the conditions, therefore

“valid explanatory knowledge in this realist epistemology requires the researcher to identify the contingent causal powers that are operating in the particular situations under research and the ways in which they combine and interact in order to create the particular events in the empirical domain”

(Easton 1998 p. 79)

Or in other words; abstraction is a matter of sorting out and describing what mechanisms and how they act, given the contingent conditions of when, where and whom of the study, whereas causation is a matter of the why– the theorizing of explanation. The links between the stratified reality and analysis are illustrated in figure 2.3 above.

However, the answer to the question ‘why?’ is a temporary one, because mechanisms are not characterized by regularity. They only have effect when triggered by contingencies, and the effect differs according to the conditions. This being so, the criteria for judging the validity of a knowledge claim cannot refer to a universal truth, but has to be judged in terms of practical adequacy which means its explanatory quality (Buch-Hansen 2005).

“To be practically adequate, knowledge must generate expectations about the world and about the results of our actions which are actually realized” (Sayer 1992 p. 69)

Practical adequacy is a tentative approach to the issue of truth, a way to replace or modify the concept of truth and to include the importance of practice (Sayer 1992). However, the concept of a stratified reality has lead critics to conclude that when critical realists are looking for a causality residing in underlying mechanisms, they have to believe in the existence of a universal truth (Dreyer Hansen 2004). This is not the case.Critical realism does not postulate any type of privileged access to truth, but attempts to find ways to formulate understanding (Mutch, Delbridge & Ventresca 2006). Not exclusively of singular phenomena, but of the workings of forces and contingencies. So in spite of the modest critical realist claim on truth, and the difficulties posed by the irregularity of mechanisms, the ambitions are not modest. Sayer (1992) states this ambition as a matter of finding inputs to robust theories which are practically adequate; i.e.

 work in other contexts

 are consistent with other knowledge and practices

 explain the situation under study

 give an account of what produced the situation

These criteria of practical adequacy are the ones to apply when judging the validity of theory in a critical realist approach.

One of the limitations of a critical realist approach is its incapacity for prediction. The irregularity of mechanisms implies that they only have effect when triggered by contingencies, and that the effect differs according to the conditions. Consequently,

“The result of studies of all but the more durable social structures are therefore likely to be theoretically-informed and informative narratives rather than formal analysis of apparently timeless mechanisms” (Sayer 1992 p. 145)

Such narratives can inform us of what an object or mechanism is like, and what it can do.

And they can describe the conditions under which the activation of the mechanism was observed. They cannot predict the reoccurrence of the specific combination of a situation in which the mechanism and the triggering conditions are present. But they increase the

“chances of either removing or changing the mechanism, preventing its activation or suppressing the damaging effect of its exercise” (Sayer 1992 p. 135)

Moreover, a critical realist approach only offers a weak foundation for studies which apply quantitative statistical methods. This is so in spite of the claim that critical realism enables the application of a variety of methods and theories (Tsoukas 1989). The preoccupation with causal explanation of mechanisms and their activation under specific contingent conditions demands a qualitative analysis of relations; abstraction and causation.

Therefore, studies of cause-effect relations between discrete events, or co-variance between variables based on quantitative statistical analysis are not easy to sustain within critical realism. Critical realist favours the use of qualitative data. They create a better foundation for abstraction and causation than quantitative statistical analysis of co-variance and cause-effect relations. In consequence a critical realist approach cannot offer generalization from samples to populations or the type of normative prescriptions which an analysis of quantitative data can produce.