Realist Social Theory and Its Losing Battle with Concepts

1. Introduction

Social theoretical reason has traditionally been preoccupied with a number of (what are considered to be) hard problems, for example, of reconciling structure and agency or of specifying the proper form of explanation for social phenomena. Whatever we might think about the extent to which progress has been made on these fronts since the early days (and whatever we might think constitutes progress), it is clear that the problems are still with us, and it certainly appears that they will continue to be with us in the future. To the reflective mind, this observation directs attention away from what the problems aim at and toward the ways in which social theorizing is conducted, raising the question whether some central theoretical procedures instead of constituting indispensable tools might actually be obstructions, which prevent forms of progress.

In this article, I shall attempt to evidence this way of thinking. I will confine my attention to realist social theory and detail a number of conceptual mistakes that occur in the early realist work of Harré and Madden (1975) and Bhaskar ([1975] 2008), as well as in more recent theorizing, chiefly Dave Elder-Vass’s (2007a, 2007b, 2010), which attempts to offer an emergentist solution to the problem of structure and agency.

The conceptual mistakes I propose to dissect are embodied in the following (sets of) procedures that form a usually taken-for-granted part of theoretical reason:

I will proceed in the following sections to examine these procedures in the above order. Not dispensing with them, it will be argued, not only stands in the way of adequately dealing with chosen theoretical problems but also tends to make debates between theoretical schemes irresolvable, for theorists can always (a) insist on the a priori propriety of a model of explanation and (b) select different areas of the causal conceptual terrain as models of what counts as causal. Furthermore, such debates are bound to remain irresolvable as long as (c) the attachment to a theoretical scheme does not translate into a contentious factual claim that is necessarily connected to one of the competing schemes, for if the same facts can be recognized by anybody, they cannot be invoked to make a difference, and (d) the application of the theoretical scheme and model of explanation does not make a demonstrable difference in comparison to commonly available understanding of what it is applied to. Thus, to demonstrate the superiority of their proposed theoretical scheme, theorists need to discharge the burden of proof concerning (c) and (d). ¹

Although, as mentioned, in what follows, I will be discussing (critical) realist theorizing, my main purpose is not to refute Elder-Vass, Bhaskar, or Harré and Madden but rather to call into question their procedures for handling concepts. The following discussion, it is hoped, will provide the means by which a case as to the widespread use of (at least some of) these problematic procedures can be made (see Tsilipakos 2012, forthcoming), and thus the reader is encouraged to remain sensitive to the general social theoretical relevance of the discussion.

Let us first, then, scrutinize the search for a rigid model of concepts and the fixing of explanatory form, as evinced in the realist understanding of the concept of causal power and its role in explanations.

2. Explanatory Form and the Concept of Causal Power

A central place within the critical realist paradigm (Bhaskar ([1975] 2008), [1979] 1998) is reserved for the concept of causal power (deriving from Harré and Madden 1975) as is evident in the social theoretical premiss that structures and individuals possess different kinds of causal powers, one irreducible to the other (Archer 2003, 15). This fact coupled with the critical realist commitment to a model of scientific explanation in terms of underlying mechanisms requires that social theory demonstrate that structural and individual powers are connected as part of a causal mechanism, itself part of an open system (this, for example, is what Archer attempts to accomplish via her account of the mediating role of “internal conversation” and, more generally, reflexivity; see Sharrock and Tsilipakos 2013), and that it also illuminate the perhaps smaller scale mechanism that gives rise to such powers. In a similar vein, one of the latest versions of critical realism, one propounded by Dave Elder-Vass (2010), is designed to account for the way exclusive structural causal powers are produced by showing the process to be consistent with a general emergentist ontology captured in the core dictum that the relations between the parts of an entity give rise to emergent causal powers that the individual parts may not possess. ²

Nevertheless, the conception of explanation presupposed by the realist account of causality in terms of causal powers, which originates in Harré and Madden’s work, and finds its way into Bhaskar and Elder-Vass’s critical realism, rests on a mistake. The mistake lies in attempting to fix explanatory form by stipulating that citing the powers of particulars has explanatory force without any reference to an explanandum in question. To witness this conception in its origins, we need to turn, first, to Harré and Madden’s book.

The book comprises both an assault on the Humean theory of causality and the simultaneous effort to avoid the latter’s shortcomings by constructing “a conceptual system capable of accommodating the actual intellectual practices of science . . . [and] common sense” (Harré and Madden 1975, 1-2). The stated aim of the book is, however, vitiated by the fact that the authors focus on only one of the centers of variation (Hacker 2007, chap. 3) of our causal concepts. In the following passage, this fact is acknowledged by Harré and Madden, who nevertheless try to circumvent the difficulty by promoting their chosen employment as the ground of all other cases of causal talk. ³ This maneuver, of course, undermines their aim of doing justice to our actual practice of causal concept use and exposes their project as one of reconstruction/revision rather than elucidation:

Our most central idea is that of a powerful particular. We show how in terms of that notion the citation of the presence of a particular has explanatory force . . . Our conception of causality is deliberately in keeping with one of the commoner ways in which this concept is employed. In this sense, causation always involves a material particular which produces or generates something. But what may be singled out as the cause may be an event, a state of affairs, or even in certain contexts, a material substance . . . [nevertheless] in any specific application of the notion of causality, the crucial element, we argue, the presence of which makes the action causal, is a powerful particular. (Harré and Madden 1975, 5, emphasis added)

Instructive in the above for our present purposes is how Harré and Madden distinguish between the powers a particular has to generate and produce something, and the citation of the presence of a particular or singling out of a cause in an explanatory context. We might present this distinction as one between two different (but obviously related in the context of science) sets of activities where causal concepts are employed to different ends: the investigation of natures and powers of stuffs, and the identification of a particular cause (or a number of causes for that matter) in the context of an investigation or explanation of a particular phenomenon. The second type of activity, as Harré and Madden acknowledge, is not limited to picking “powerful particulars” or “agents with powers” as causes. For example, a large nail has the power to pierce a car tire and can thus be cited as the cause of a puncture, but it is the unexpected presence of a nail, the poor state of the tires or the wrong type of tires, the lack of a roadside barrier, excessive speed, bad road conditions, the level of alcohol in the driver’s blood or poor driving skills, which, although not classed as “powerful particulars” or “agents with powers,” can on occasion be disjunctively—or in some cases conjunctively—cited as the cause(s) of the accident. ⁴ In other words, being a powerful particular is neither a necessary nor a sufficient condition for something to be cited as a cause. It thus makes no sense to be constrained in our explanations by the search for powerful particulars, which is not to say that some of our explanations may not cite precisely such particulars as the cause(s); it all depends on what we are trying to explain.

Harré and Madden, however, still want to guarantee that powerful particulars serve as explanatory causes, and this is exactly the service that their concept of causal power attempts to provide by running together these two logically distinct activities. Thus, the concept is meant to fix the transition from having a power to do something to having the power to be identified as (one of) the cause(s) in the context of an explanation, in other words, the power to have explanatory force. As noted, however, the former power is neither necessary nor sufficient for the latter “power,” which is not a power of the particular in question but a function of the kind of explanandum and the circumstances in which an explanation is required. ⁵ It is not that citing a causal power can never have explanatory force but rather that it is not possible to guarantee that it will independently of what it is we wish to explain. Adequate specification of the explanandum needs to precede any such guarantee, which, of course, defeats the purpose of having a general and explanandum-independent guarantee.

Having sketched the rationale behind the introduction of the concept of a causal power, we can witness the selfsame conception repeating itself in Elder-Vass’s (2010) latest version of critical realism, centered on the thesis that social structures may have exclusive causal powers. By establishing this thesis, Elder-Vass’s purpose is to add such powers to the catalog of causal powers operative in the social world, which are to be invoked as (part of) the only admissible way of explaining particular events. But this is exactly the mistaken entailment that is being obscured by the fusion of two logically distinct activities in the concept of causal power as it originates in Harré and Madden’s account of the powers of the natural world.

Having examined the procedures of fixing concepts and what counts as an explanation, as manifested in the realist deployment of the concept of causal power, I will next consider problems that arise in the investigation of causal powers in the social world, especially when such investigation is conducted in general categorial terms and/or concerns concepts that are not social scientific inventions. The following section will attempt to show that conceiving of conceptual matters as subject to empirical verification/falsification and, more generally, the lack of care in distinguishing between the conceptual and the empirical lead to a host of further confusions.

3. Conflating Conceptual and Empirical Matters

As we have already seen with reference to the work of Archer and Elder-Vass, critical realists are interested in inquiring into the causal powers social structures may or can have. It should be noted that the modal expressions “may”/“can” introduce a distinct logic that excludes from consideration the particular causal powers of a particular social structure. The question, for example, Elder-Vass (2010) is asking is not “does this organization have the causal power to V?”, for instance “does BP have the capacity to purge the Gulf of Mexico of the adverse effects of an oil spill?” Instead, the question operates at the level of kinds and, pivoting on the modal “may” (or “can”) as it does, is concerned with the causal powers that organizations can (exclusively) possess. In this section, it will be contended that there is a conflation in realist social theory of the empirical with the conceptual. We will witness this in Elder-Vass’s failure to appreciate the differences between establishing that a particular thing has a causal power and that a kind of thing can have a causal power as they are both unhappily glossed as fallible knowledge claims. We will also trace a similar type of confusion in the Bhaskarian conception of the unobservable. Matters are made worse by misunderstanding the character of most concepts in the context of the social sciences.

Let us begin by considering how Elder-Vass responds to a critic, Charles Varela, who contends that the means critical realists use for ascribing causal powers are generally unprincipled as they do not follow Harré and Madden’s closely enough. Bhaskar and Archer’s causal criterion is found too open-ended, and Elder-Vass’s own relational emergence angle is thought to “violate the cardinal principle of causal powers theory” (Varela 2007, 208) in locating causal power in social structure (which under Varela’s conception does not constitute a powerful particular) thus adding this power to the particular (individuals) rather than predicating it of a particular. To repeat, Varela supports the “power of a particular” scheme deriving from Harré and Madden’s causal powers theory, which he considers as the only means of “principled ascription of causal powers to entities of the right sort” (Varela 2007, 202), and as such, like Harré and Madden, he too is committed to using a singular legitimate mold for causal concepts (which he thinks, however, does not necessarily apply to human affairs).

Elder-Vass’s (2007b) response to the challenge posed by Varela consists in denying the coherence of Harréan criteria and, ultimately, appealing to his own method for social ontology as what enables him to make systematic ascriptions. What he says in the process evinces his conception of the form of inquiry and of the logic of ascriptions of causal powers. He claims,

when we ascribe a causal power to an entity we are making a knowledge claim, and as all good realists know, all such claims are fallible since the thing we purport to know exists (or does not) independently of what we purport to know about it. Ascription of causal powers is therefore distinct from their existence, and so we need criteria by which to assess such ascriptions. (Elder-Vass 2007b, 468, emphasis added)

Now, the assertion that there are, indeed, criteria by which we may assess ascriptions does not turn on any ontological claim (as will be seen below) and is rather uncontroversial. Where controversy can and does arise, however, is when we consider what sort of criteria one may use and whether the need for criteria implies that ascriptions of causal powers to kinds are correctly represented as fallible knowledge claims. It will be argued that when we say that a kind of thing can have a causal power, we are not making a knowledge claim whose substantiation can lie beyond our current knowledge frontiers and which can depend on currently inaccessible information. Far from making an empirical claim, we are, in fact, issuing a claim about knowing our concepts. Some examples will help us appreciate a number of different kinds of knowledge claims, the backing required, and how the question of fallibility arises or not in each case.

First, let us consider the specific claim that an organization, a petrochemical corporation, has the power to purge a gulf of an oil slick. Knowing this hinges on one having a rough idea of the extent of the problem and the resources available to the organization. It makes sense to say that one may be mistaken in that estimation, if, for example, the extent of the damage caused has not been assessed properly or one has miscalculated the executive and financial capabilities of the organization. All in all, what is needed to back the claim is a sound understanding of the particulars of the case.

Second, consider the general claim that organizations have the power to make their employees redundant. What kind of backing does this claim require? One way we can understand this is as an assertion about the legal status of organizations, the law governing the relations with their employees, and, for example, their compensatory rights in case of redundancy. Accordingly, the claim can be backed by an appeal to statutory law. The application of the law in particular cases where disputes arise can be anything but straightforward, but as a question of what the law generally says, the claim can be easily settled (cf. Hart [1961] 1997, chap. 7). Note that once this is done, there can be no further question of fallibility.

Finally, consider the assertion that organizations have the power to mobilize their members in pursuit of some goal. Is this a fallible knowledge claim? Not if saying that organizations having the power to mobilize their members to accomplish an objective is a way of talking about the concept of organization: insofar as we understand the concept, no further question of fallibility arises.

It is instructive at this point to compare a scientific concept (that of acid) with the concept of organization. As discussed by Harré and Madden (1975, 8), the relation between acids and the power to turn longwood solution red is an internal one. Turning longwood solution red is a criterion for being an acid, and we may use it as an inference license to pass from the observation that a substance does not turn longwood solution red to the conclusion that it is not an acid. Accordingly, to entertain the thought that something might be an acid and not turn longwood solution red, the latter effect needs to be dropped from its criterial role. But that would entail a change in the concept of acid (call the new concept acid*) so that the claim that it turns out that acid* does not turn longwood solution red is not the negation of “acid turns longwood solution red.” ⁶ We can express this as a distinction between physical and logical possibility. In this case, the difference between physically possible powers and logically possible powers is decided by the way that acidity and the power to turn longwood solution red fit into an investigation: are they internally related and used as an inference license—that is to say, are they criterial, or do we use a concept of acid that does not depend on such a criterion and is accordingly an intelligible empirical matter whether it has the power in question? The “can” or “may” that befits the investigation of criteria is one of logical, not physical possibility. The crucial point, then, is that if one is interested in what causal powers acids can have, then one is interested in exactly those causal powers that are used as criteria for something to count as an acid.

The chemical concept of an acid is a theoretical concept, by which I simply mean that it is part of a theory and stands or falls with it. The concept is designed to classify and account for a number of phenomena and is based on a set of criteria that crystallize and systematize experimental work. However, the examination of the concept of acid is not an empirical investigation nor are statements laying down the criteria that constitute concepts fallible, unless we are prepared to change the concept and modify or even jettison the theory that it plays a part in. The criteria for knowledge are the criteria for understanding the concept, and those criteria are neither true nor false although we can be wrong or right about them and, naturally, what we say using the concept of acid can be true or false.

On the contrary, if the concept in question is not tied to any scientific theory but is rather part of ordinary English, then knowledge is not knowledge of a theory but knowledge of the language. This crucial observation removes further grounds that might serve to make talk of fallibility plausible. Concepts that are part of scientific theories allow for Harré and Madden’s idea that necessary relations between concepts reflect natural necessity, as well as for a sense in which one can speak of the concept’s fallibility as attached to the fallibility of a theory with regard to some kind of ontological beyond, misleading though such a way of phrasing matters might be. But when it comes to the concept of organization, there is no equivalent beyond or sense of fallibility of this kind. The concept of an organization is neither part of any scientific theory nor is it formed after experimentation to serve theoretical explanation (cf. Bittner 1965). Thus, there is no room for conceptual relations to reflect anything. The concept is what it is; it has its lay and legal uses, and if the investigation is to be concerned with that concept (for there is no case to be made that Elder-Vass is concerned with a different concept), then it is not up to one to change it or to portray it as if it were a concept of the same type as acid is. ⁷

Thus, saying that the claim attributing a causal power to a kind is a fallible knowledge claim when that attribution is the expression of a rule constitutive of the concept in question, especially when one is engaged in an investigation focusing on precisely these constitutive rules, evinces a conflation of the empirical and the conceptual. For what is being sought after are not some contingent powers that a particular organization might have but the conceptually related powers of organizations that enter into the concept of organization. This is manifestly the case in the concern with the powers kinds and not particulars may or may not, can or cannot have, steering interest toward what makes organizations organizations and what is logically possible for them to do. In these cases, the issue is of knowing a concept and portraying conceptual relations accurately; there is no ontological remainder, which might go beyond what we know.

Critical realists are, nevertheless, programmatically committed to such an ontological remainder. Elder-Vass’s epistemological discussion (see quotation above) has an air of ontological gravitas whereby—through running together “knowing X” (in the sense of being acquainted with) and “knowing that p”—knowledge claims are spoken of as (involving) existence claims: to know some thing or some one presupposes that that same thing or one exists, in the way that knowing John presupposes that John exists. But such knowledge is mostly irrelevant to scientific investigation where most of “the things we purport to know,” for example, the mass of Pluto (or what Pluto’s mass is), the length of the gestation period in mice, and that Greek sovereign debt exploded after 2009 cannot be said to presuppose that what we know exists (nor, of course, that it does not exist). None of these things we know are entities nor can claims to know them be spoken of as existence claims like the paradigmatic postulation in physics of, say, gravitons and dark matter. The point of the ontological seriousness produced by this sleight of hand is to maintain the impression of unknown entities beyond our frontiers of knowledge, which, in turn, lends some plausibility to critical realism’s insistence on a stratified ontology.

Before concluding this section, it will pay off to inquire into the matter in more detail. Bhaskar’s stratified ontology (empirical, actual, real) ⁸ extending beyond what we know and can observe is (at least in part) motivated by (a) the identification of what is real with what has real effects and (b) the realization that a causal power cannot be identified with its exercise.

Starting with (a), at the most general level, it entails that we have to take into account things that can be said to cause even though they might not be observable. This constitutes, then, a reason for holding that the real is more extensive than what we can perceive. The improvement of this conception is presumably most pronounced when contrasted with an empiricist ontology, which is limited to (what are considered) observables. In advocating this position, nevertheless, Bhaskar (and others) commit the conceptual error of turning a distinction between observables and non-observables into a dichotomy.

Dichotomies are, in effect, a way of stipulating the unrestricted application of a pair of terms, in this case, the terms observable and non-observable. But there are concepts in the language that allow for the connection with observability or non-observability and concepts where that connection is not provided for. There is what is observable, non-observable, and where the notions of observability or non-observability have no application (or none yet). ⁹ For example, the question of the observability or non-observability of society (or the social) is a case where the connection is unclear: thus, it is better to avoid giving a “yes or no” answer to the question, for example, whether the social is observable or non-observable. Critical realists such as William Outhwaite (1998, 284), however, wish to argue that “Society is not observable” because “There is no such thing as observing society as such.” But if the latter is true, then there is no such thing as not observing society either, and we might as well say that society is not unobservable. Consider, further, the “non-observables” Margaret Archer casually lists:

Some of the non-observable social factors which concern us may be open to humanistic hermeneutics (my motives in writing this book or yours in reading it), but frequently we just are dealing with non-observables which remain that way (i.e. they are distinct from a psycho-analyst’s attempts to help clients to awareness of something unconscious). In this category would go international finance markets, institutional contradictions, ecological imbalance, third dimensional power or ideological mystification. (2000, 48-49)

In reading through this passage, we might suspect that Archer adheres to a very restricted idea of what can be observed, one that derives from some version of empiricism, and which relegates everything that does not conform to the picture of a directly given to the category of un-observables. Moreover, it is (perhaps for this reason) hardly clear what it means to say that the above items are non-observable. Why, for example, are institutional contradictions non-observable, are they in principle or only contingently so, and what would it mean to observe them? Is there a clear sense in which we are, indeed, restricted in our inquiry due to (an intelligible) something we cannot do? And, finally, is it not possible to specify ways of observing some of the above items? We might take stock of these probing thoughts for now and move on to consider (b).

It is noteworthy that Varela and Harré (1996, 318) seem to think that (b) constitutes the only reason Bhaskar has for being committed to a stratified ontology featuring non-observables, that is, the fact that causal powers are not exhausted by their manifestation. While this thought is correct in itself (see Hacker 2007, chap. 4), it cannot warrant cataloging powers under non-observables. It is plain that Smith’s power to drive a car is something one can observe when one sees her driving. This is what observing her power to drive amounts to. The fact that she does not lose the power when she is not driving does not mean that the power is present somewhere but impossible to observe, least of all in some transcendental realm. Nor does being unable to tell whether someone can drive just by looking at them render such a power non-observable. The mistake here lies in overlooking that what it means to observe something depends on the something. Observing powers or abilities might not be the same as observing colors or shapes, but this does not render the former non-observables.

Bhaskar’s ([1975] 2008) programmatic idea that, in open systems, powers (or what he sometimes ¹⁰ interchangeably calls “tendencies”) “may be possessed unexercised, exercised unrealised, and realised unperceived (or undetected) by men” (175) also fails to support their non-observable status because it is deeply incoherent. If we look at the examples he invokes, ¹¹ it is clear that none of them warrant this conception: First, although the power to lose one’s temper can be possessed without being “exercised,” the tendency to do so cannot be possessed without ever having been “exercised.” Moreover, we would rather say that one manifests that tendency or power (because it is not clear that it makes sense to say that one exercises it, otherwise that person would be faking losing their temper), for example, when in a particular occasion they respond angrily to, say, a mild provocation. This, if anything, is also to realize that tendency or power. To say that one “exercises” the tendency or power to lose his temper without realizing it is unintelligible for reasons we will shortly come to see. Second, to say that someone has the tendency to win (as opposed to the power to win) is not to report what he will do given certain circumstances (once, say, besides enabling conditions, stimulus conditions are also fulfilled) but what the usual outcome of him competing is: the tendency to win is a statistical notion. For this reason, having the tendency to win cannot be possessed unexercised, that is, unless someone has frequently won in the past. Finally, and most importantly, powers and tendencies are individuated by what they are tendencies to do. Winning is not the realization of the tendency to run fast—running fast is. When someone is running fast, they are not exercising their tendency to win, which remains unfulfilled if in the end they lose; they are exercising their tendency to run fast. This removes any space for non-observables opened up by the supposed gap between a power or tendency’s exercise and realization.

To bring the discussion of “non-observables” to a close, it is important to emphasize that the pervasive analogy in critical realism with the postulated entities of natural science (e.g., the Higgs boson) is completely out of place, as the difficulty for a social theory is the conceptual one of specifying what we mean by saying, as Archer does above, for example, that society, institutional contradictions, financial systems, and the rest are non-observable and not (before we do that) to find a technical means of detection (which is arguably what is lacking in the postulation of particles in physics).

In this section, it has been argued that the programmatic critical realist commitment to non-observables, like Elder-Vass’s response to Varela with regard to the logic of ascriptions of causal powers to kinds, confuses empirical and conceptual matters and results in the falsification of investigation that is concerned with concepts as being one about hidden entities (or processes) the existence of which we cannot conclusively prove. Moreover, thinking of (statements about) concepts as fallible in light of scientific progress has been shown to be mistaken, even more so when the concepts in question are not technical or scientific ones.

We now move on to examine the third and final set of problematic procedures, those having to do with the theoretical use of examples, again drawing materials from contemporary critical realism and the work of Dave Elder-Vass.

4. Problems in the Use of Examples

This section scrutinizes some of the ways in which examples are employed to support a theoretical scheme. This issue is discussed with detailed reference to Elder-Vass’s attempt to justify his thesis that social structures may possess exclusive causal powers (or “causal powers in their own right” [2010, 26]), for example, that organizations can have certain causal powers their parts (i.e., their members such as administrators, managers, and other employees taken as individuals) may not possess. While engaging with this thesis at some analytical depth, we will witness how the examples that are conscripted as supporting Elder-Vass’s theoretical scheme are, at worst, misdescribed by the imposition of the scheme onto them and, at best, when they are accurately described, are merely illustrative and cannot (exclusively) support the scheme since they are uncontroversial. Let us start by examining how Elder-Vass uses examples to introduce the general emergentist ideas, which he will then apply to social structures:

[T]ake the example of a human finger pressing a key on a computer keyboard. . . . First, is it the finger or the person of which it is a part that has the causal power to press the key? And secondly, is it the finger or the person that actually does press it?

The causal power question is easily answered by considering the counterfactual question of whether the finger would be able to press the keyboard if it were not part of the whole human. . . . Hence the causal power to press the key is a power of the whole human and not of the finger.

The second question is more open. The most plausible answer, it seems to me, is that both the person and the finger press the key. The finger does so very directly and the person does so through the finger, which is one of its parts. Thus when the finger presses the key it acts both as a finger and as a part of a larger whole. The finger, I would want to say, is the part of the person that implements the person’s causal power to press the key. In such cases, the two elements cannot be empirically but only analytically disentangled: there is no event “finger pressing key” that can be empirically distinguished from the event “person pressing key” and yet using the counterfactual method we can distinguish between the two corresponding causal claims.

I will suggest that this argument can be generalised to people and social structures: that sometimes, when a person acts, they do so both as an individual and as a part of a structure. (Elder-Vass 2010, 27-28)

There are a number of dubious moves in the above illustration that are consequential as regards the claim to apply the argument to social structures. As far as the “who has the causal power” question is concerned, it is objectionable whether the counterfactual test is sufficient to establish the exclusive possession of powers; one may contend that, by the same token, were one’s finger cut off, they would not have the power to press the key. ¹² The question cannot be settled by the counterfactual test that appears to work by removing a part of the functioning causal mechanism because the question is rather a different one. Far from being concerned with physical possibility, it is, in fact, the logical question whether it makes sense to say that a human and a finger can do the same thing, a consideration applying before any mechanism removal test can gain a foothold. What actually makes the difference in this case is not the counterfactual test but the conceptual point that fingers do not press anything of their own accord, people do.

The response to the second question, that is, “who is it that actually presses the key?” also features a number of conceptual difficulties. The seeming interchangeability of “person pressing key” and “finger pressing key” is only possible due to the lack of an adequate specification of the situation, which will allow us to decide which description to apply. Although when I am pressing a key, my finger is pressing against the key, “human finger pressing key” and “person pressing key” are not interchangeable descriptions because the former can in some cases imply, either that the finger is a severed one or that I pressed the key by accident, inadvertently, and so forth. Furthermore, people do not press buttons through their fingers, they do it with their fingers. It is nonsensical to describe the situation as if I am at a distance and need to use some device, that is, my finger in this case, to help me reach the key, or as if I am somehow encaged in my body so that I can only touch the key indirectly. I can be said to press a key directly when I do so without using some kind of tool; I cannot be said to be pressing the key indirectly when I am using my finger, especially not in contrast to my finger itself, which is doing the job directly. Moreover, the assertion that the two descriptions are empirically indistinguishable operates on the very impoverished, empiricist sense of what is observable we encountered in the previous section. Insofar as we use these two different forms of words to describe different situations, it follows that there is an observable difference between such situations, even though such a difference need not be one that can be captured by means of still photography (which seems to be the constraint Elder-Vass has imposed on the case). But even if one were to allow that these two descriptions are interchangeable, it would still not follow that there are two empirically indistinguishable events going on simultaneously. Instead, the case would rather be of one event admitting two or more descriptions.

The above conceptual moves on Elder-Vass’s part are meant to lend support to the scheme of emergence, but they fail because they misdescribe the chosen example. It will now be argued that the attempt to generalize the scheme to individuals and social structures makes no less problematic a use of the examples it employs.

Elder-Vass (2007a, 33) illustrates his argument with the example of a person buying a television (TV), which he introduces in the following way: “I walk into an electrical shop and purchase a TV from a sales assistant, who arranges for it to be delivered to my home in a few days time.” This familiar situation serves as the scene for a number of no-less-familiar and uncontroversial observations. First, Elder-Vass notes that I ask the person at the shop in their capacity as a sales assistant (and not, say, as squash player), a role the person assumes as part of the organization. Accordingly, it is as part of the organization that she accepts the payment and can arrange for the TV to be delivered to my home address. Finally, he notes that the sales assistant does not own the TV and does not sell it on behalf of herself. To these observations, then, Elder-Vass (2007a) forces the questions that were singled out in the finger example: (my formulation) “who is it that sells me the TV?” and “who has the power to sell me the TV, the sales assistant or the organization?” He responds, thus,

It is the organisation that sells me the TV, though it does so through the sales assistant, who is one of its parts . . . if a dog bites me, it does so through its teeth, yet we understand perfectly well that it is the dog that is primarily causally responsible for the bite, and not the teeth. This is perhaps more difficult to accept when the parts have a mind of their own, but the principle is similar. Certainly we must accept that the individual agency of the sales assistant co-determines the outcome, but it only co-determines it. Just as we accept that human beings are causally responsible for the behaviour of their parts when it is directed by their decisions, so we must accept that organisations are causally responsible’ for the behaviour of their members or employees when that behaviour is motivated by organisation policy. (Elder-Vass 2007a, 33, emphasis added)

In its discussion of organisations, this paper has already stated the positive case for seeing people plus relations as constituting a higher level entity with causal powers. The key points in this argument were (a) that if the people concerned were not organised into such an entity, those powers would not exist; and (b) that the people plus the relations are the higher level entity, so to say that the people plus the relations have a power is the same thing as to say that the higher level entity has the power. Thus, for example, in an organisation that practices the division of labour (e.g., Smith’s pin factory) if we say that the productive capacity of the organisation depends on both the workers and the relations between them that exist when they are organised as they are in this organisation, this is necessarily equivalent to saying that the productive capacity is a causal power of the organisation and not of the workers. (Elder-Vass 2007a, 38, emphasis added)

There are three pivotal conceptions displayed in the above passage: (a) the similarity of principle between cases, (b) the co-determination of the outcome, and the correlated idea that (c) there is a higher-level entity, which is the proper bearer of the causal power and the main determinant of the outcome.

First, let us take up the contention that there is a similarity of principle involved between the dog-teeth and the organization-sales assistant, something considered as displaying the virtue of consistency for the principles of emergence. But, we might ask, does this uniformity derive from the examples or is it rather the outcome of an imposition of the emergentist scheme and its requirement of consistent application?

In fact, if we carefully look at the examples, we can see that the comparison between the dog and the organization (or the division of labor) is misleading because in the case of the dog, it is not conceptually open to us (and not because we are “causally myopic” as Elder-Vass [2010, 199] thinks) to suggest that the teeth of the dog bite, although the dog bites with its teeth in the same way that it sees with its eyes, nor, in another variation of the canine example, that the lungs of the dog bark, unless somehow the dog barked without meaning to (if we can get that to stick as a description of its behavior), nor would we say that the lungs, windpipe, and mouth of the dog barked. Similarly, one can intelligibly speak of the behavior of one’s parts when these are not under his control, and of directing the behavior of a person by their decisions, but to say that one is responsible for the behavior of his parts when directed by his decisions is to force a way of describing organizational life onto humans and their limbs and not, as Elder-Vass’s project requires, to derive the former from the latter. There is, thus, nothing to be judged as similar (or dissimilar for that matter) by the emergentist part-whole relations. The importance of preserving sense militates against the idea of a systematic application of the scheme, which can, after all, only be achieved through conceptual gerrymandering.

The “similar principle” idea is also used as ammunition against the individualist opposition (Elder-Vass has Anthony King in mind who nevertheless disavows the label [King 2007]), which is deemed inconsistent and seemingly reduced to absurdity because it is held to entail that it is not the dog that barks but its lungs, windpipe, mouth, and so forth. Opposing this violation of reason, Elder-Vass (2007a, 37) sees himself as a non-reductionist who—guided by his scheme—can always preserve the higher-level entity: “Emergentism is therefore recursively consistent: it ascribes causal powers at each level on the basis of the same ontological argument.” However, if using the “same argument” is tantamount to riding roughshod over conceptual distinctions, then it does not follow that rejecting what Elder-Vass says concerning one case commits one, on pain of inconsistency, to denying what he is saying in another case. One can deny that there is a higher-level entity in the organization-sales assistant example, while also denying that the teeth or the lungs of the dog can be said to bark.

It remains for us to examine what Elder-Vass’s claim for the higher-level entity might amount to and what justification he offers for accepting it. This, after all, is the nub of the disagreement between Elder-Vass and King and the central aspect of the former’s theoretical scheme. The argument turns, according to Elder-Vass, on the counterfactual test and the “redescription principle” that “people plus relations equals higher-level entity.” Attention will be primarily directed to the thesis entailed by the “redescription principle,” that is, that the causal power of people plus relations is the causal power of the organization. This will be considered together with the idea of co-determination.

First, it must be made clear that even if it were the case that the causal power is properly only the organization’s causal power, it would not follow that there is an additional entity involved. The reason is that entities are not individuated by their exclusive possession of causal powers. A boat engine may have the causal power of propelling a vessel, but it is not an entity additional to the engine parts. Rather, one is dealing with descriptions of different logical categories. It may be objected, and rightly so, that Elder-Vass does not speak of an additional entity but rather of a higher-level entity. This may be granted. The point, however, is that speaking of a higher-level entity enables him to argue, at least on first inspection, that there is an additional cause to be taken into account. This is what the idea of co-determination is meant to describe. Elder-Vass claims that the “individual agency of the sales assistant co-determines the outcome” but the main responsibility lies with the organization. Indeed, it seems that, viewed through the lens of co-determination, they are both determinants but the main one is the organization.

The idea of co-determination is a crucial element in Elder-Vass’s scheme as it attempts to reconcile structure and agency. In attempting to make sense of it, one option is to read it as an expression of the critical realist understanding of multiple determination where the idea is that a multitude of causes determine the event taking place. This conception would seem to suggest that the outcome (presumably described as “what took place,” for example, that a TV was sold) is some kind of resultant event, taking place once component causes have been computed, much like in mechanics, the forces acting on a body can yield a component force whose vector represents the direction the body will move toward. We will see that the understanding of (the individual agency of the) “sales assistant” and the “organization” as simultaneously acting separate causes is problematic.

It is also questionable to portray the selling of a TV as in part the organization’s doing, in part the sales assistant’s doing, in the same sense that eating a cake was partly Dick’s doing who ate eight slices and in part the cat’s doing who ate the remaining two but, therefore, mainly Dick’s doing. Although this picture seems to work well with the co-determination idea, it cannot apply to the example in question: it is not the case that the sales assistant sells one-fifth of the TV, while the organization the remaining four-fifths, although the former may be paid 20% of the value of each sale.

Alternatively, Elder-Vass may have in mind something of the following sort: salespersons are quite different to each other. Despite the common sales policy of an organization, they are unique individuals each bringing something of their own to the job. Thus, the specification of the job description will not tell us what it is that every person does unless we look at how the individual implements this. Moreover, some organizations are more liberal than others in the procedures they lay down regarding employee behavior, attire, and so forth. Co-determination construed, thus, would more likely come to bear on the way a particular sales assistant makes a sale. Thus, the individual qualities of the sales assistant may enter as explanatory factors when some feature of the making of a sale cannot be accounted for by organizational procedures, but it is hard to see how, apart from any such specifics, “the outcome” is a result of the organization and the sales assistant as individual being related as major and minor co-determinants.

In elaborating the idea of co-determination, Elder-Vass appeals to the counterfactual test that, he hopes, will yield causal co-determinants; unfortunately, it fails in that role too. It seems acceptable to say that were it not for someone interested in buying the TV and for a sales assistant who was capable at her job and part of an organization having the power to sell TVs, “the outcome” would not have taken place. However, by the same token, if it were not for an operational banking system and the earth continuing its rotation, “the outcome” would not have occurred either. It is patently absurd to say that all these things are involved as co-determinants. Moreover, it is certainly ill-advised to speak of “the outcome” in such an underspecified manner that it becomes difficult to decide which factors are indeed relevant. The counterfactual test also fails if it is meant to establish the equivalence between “people and relations” and “higher-level entity” as a further determinant, for to say that people are organized in a certain way does not imply that they are organized into any entity, in the sense of an additional cause.

What accounts for the ideas of co-determination and higher-level entity is the fact that action takes place within an organizational setting. While it is obviously true that the organizational background needs to be taken into account if we are to understand what took place, it need not do so as an additional cause. Describing the situation in the way we normally would, which—let it be noted—is also the way Elder-Vass introduces it, that is, talking about purchasing a TV from a sales assistant is what settles such a background. The understanding of acting within the organizational setting is typically involved in the description “sales assistant.” If Elder-Vass wants to separate the individual’s causal powers from the organization’s causal powers, then using the description “sales assistant” is a choice that misleads. ¹³

Although the emergentist account Elder-Vass (2010) puts forth is one of relational (or weak) emergence and, therefore, states that wholes have exclusive powers when compared to their unorganized parts, it seems that this is not always respected, and Elder-Vass appears to run together organized-part and unorganized-part senses of being a part of something. Unlike “individual,” “sales assistant” is obviously an organized-part descriptor. Thus, when we read that “[i]t is the organisation that sells me the TV though it does so through the sales assistant who is one of its parts,” it is more than a little puzzling. Sales assistants, precisely because they are “organized parts,” do have the power to sell TVs: that is what their job is after all. Hence, to claim that the organization has exclusive powers of selling compared to the sales assistant is either to support something different to weak emergence or, by understanding “sales assistant” to mean the person divested of their capacity as sales assistant, to offer a claim logically unconnected to the higher-entity thesis: the fact that Tom, Dick, and Harry as individuals do not have any such power is not an observation in favor of bestowing the power upon the organization in contrast to the sales assistant.

Speaking of “the sales assistant” and of a “higher-level entity” also invites confusion because it makes it appear as if besides distinguishing between acting qua individual and acting qua sales assistant, there is a further distinction to be had between the sales assistant selling the TV qua sales assistant and the organization selling it. Crucially, not only is there no reason to choose in this particular case between “sales assistant” and “organization” but it also seems that it is hard to find an intelligible contrast between the two. Both interrogatives “who sells me the TV?” and “who has the power to do so?” are not determinate enough questions in Elder-Vass’s discussion. When adequately specified, they can be answered by “the sales assistant” or by “the organization,” but one cannot serve as a (even minimal) contrast to the other. We are not correcting someone who said “the sales assistant sold me the TV” if we respond by asking whether it was mainly or also “TVs’R’US” who did so but rather misunderstanding what they said. The reason is that “the organization” and “the sales assistant” can intelligibly be said to sell the TV in different senses and, thus, do not constitute competing descriptions. All in all, Elder-Vass’s idea, first, that there is a higher-level entity, namely, the organization, which mainly sells me the TV although not of itself but rather through the sales assistant who nevertheless does not really sell the TV but rather implements the organizational power, and, second, that there is also the sales assistant as an individual who minimally sells it to me and, thus, is a co-determinant, ends up making significantly less rather than more sense of the example.

We have witnessed in this section, then, a significant part of Elder-Vass’s attempt to present the emergentist theoretical scheme under a plausible light. We saw that the examples are misdescribed and subordinated to the legislative use of the scheme itself, thus constituting illustrated versions of it, not independent evidence for it. It is important to note, nevertheless, that even in instances where the examples are accurately represented, they still do not warrant the unrestricted application of the emergentist scheme. This is not because there is a problem with induction as a general procedure but because there are no grounds to justify induction in this case. The examples could potentially recommend the scheme for those chosen instances (and thus warrant an extension or generalization of the scheme) but that depends on the scheme constituting an improvement over the commonly available understanding of those cases. In the absence of any such improvement, there is not much to recommend widening the scheme’s scope.

5. Conclusion

By way of conclusion, I would like to draw out some implications from the above discussion that have broader relevance for what I suspect are widespread but flawed practices across social theory. Let us first soberly reflect on Elder-Vass’s thesis that social structures may possess exclusive causal powers and consider what is at stake in the contention and what would somebody who denied it be denying. Elder-Vass’s (2007a, 37-38) description of his disagreement with King with regard to the example of the division of labor (the same could be said about the TV example) takes the following form: “he [King] recognises the same facts of the case as Archer and myself . . . where we differ is on the question of whether these facts entail that the group as such has causal powers in its own right.” But the facts concerning the examples used can be commonly acknowledged by both theorists (and by anyone else for that matter since, as we saw, they are utterly banal) precisely because they do not necessitate Elder-Vass’s thesis. In fact, what is at stake is not an additional factual issue but rather a notational one, which means that the “denial of structure” Elder-Vass attributes to opponents is really the denial of the critical realist notation, that is, of the stipulation that we should talk about a higher-level entity with exclusive causal powers. We can, indeed, decide that from now on, we will talk in the manner favored by Elder-Vass’s scheme. But there is no reason why we must talk like this, for there is no factual question at stake that hinges on adopting the new way of speaking. In other words, there is not an explanation missing from the described cases that is provided by the notation; on the contrary, the cases are commonly understood, and adopting the proposed scheme does not increase our understanding of them. Moreover, the notation is dependent for its sense on the commonly available, recognizable ways of describing the situation (that both Elder-Vass and his critics resort to) and the understanding those ways afford. Thus, Elder-Vass (2007b, 475)—like, it should be stressed, many other social theorists—is placing the cart before the horse when he says, “[o]nce we have understood the various entities and powers at work, we can then begin to understand social events by examining how these different powers of different entities interact to produce them.” In fact, however, there is no actual gain from the long and arduous process of redescribing examples we already understand and possess the means to talk about in the restricted terms of the relational emergentist ontology. The proposed emergentist notation is language idling. It changes nothing, it adds nothing to our understanding of each case, and, thus, it explains nothing.

If all this is true, some might worry at this point, what is to be made of realist social theory as well as, broadening the scope of the argument somewhat, of more generally used social theoretical procedures of redescription? Are we to believe that theoretical debates have been in vain? Even worse, is the above meant to show that it is impossible for social theory to constitute an improvement over commonly available ordinary understanding? And what about the issue of exclusive causal powers, does it not make a difference, for example, with regard to corporate responsibility? And is that not important? I wish to argue for none of these conclusions and, in fact, none of them follow from the arguments provided in this article. The latter’s point has not been to deny that what social theorists and philosophers hold dear is important but to avoid pronouncements about the importance of theorizing and to show us instead in detail how we practice theory, what could be wrong with it, and, thus, what stands in the way of making progress.

Returning to the central theoretical procedures that embody the conceptual mistakes that constitute the focus of this article, we may conclude by pointing out what procedures may replace them and thus what kind of social theoretical practice might be considered more defensible:

The above, I submit, despite their innocuous ring, are far more difficult to implement than might be thought and far more radical than social theory is currently prepared to handle. They can be used to form stepping stones to a substantially different social theoretical practice, one that is lucid, intellectually satisfying, and perhaps inevitably more modest.