Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-12T17:54:34.209Z Has data issue: false hasContentIssue false
  • English
  • Français

The Ontological Fallacy: a rejoinder on the status of scientific realism in international relations

Published online by Cambridge University Press:  29 April 2009

FRED CHERNOFF
Rights & Permissions [Opens in a new window]

Abstract

This article argues that scientific and critical realism have embraced several mistaken claims, among them that social science enquiry cannot proceed unless the theoretical objects of study are specified in advance. The article argues, rather, that although pre-scientific, observable objects and events must be specified from the outset, theoretical objects come to our attention only in the course of formulating theories. The article advances an alternative to scientific realist and critical realist foundations, namely, causal conventionalism, which is an adaptation to the social sciences of several elements of Pierre Duhem's conventionalist account of physical science. The article argues that major goals of theorising that scientific realism and critical realism seek to fulfill are better satisfied by the conventionalist alternative. In an effort to clarify some important issues, the article identifies and responds to a series of related criticisms of my views offered by Colin Wight in his recent article ‘A Manifesto for Scientific Realism in IR: Assuming the Can-Opener Won't Work!’ in ‘Millennium’, and in his book, Agents, Structures and International Relations: Politics as Ontology.1

Information

Type
Research Article
Information
Review of International Studies , Volume 35 , Issue 2 , April 2009, pp. 371 - 395
Copyright
Copyright © © British International Studies Association 2009

Introduction

Scientific realists and critical realists in international relations have recently emphasised that studies of meta-theory must place questions about what exists (ontology) ahead of the traditional questions about what it is possible for us to know (epistemology). Prominent authors in IR, such as Colin Wight, Heikki Patomäki, and Alexander Wendt,Footnote 2 have made this claim, adding that the theoretical object of study must be specified before enquiry may proceed. Wight says that the ‘often heated epistemological debates are really “phony wars” and … the real differences between theories concern competing ontologies’.Footnote 3 Wight, Patomäki and others frequently cite Roy Bhaskar's point that the objects we know through observation and inference do not exhaust all that there is. Those who disagree with their views, and who hold that all that exists is what is observable, are said to run afoul of what Bhaskar calls the ‘epistemic fallacy’.Footnote 4 These authors emphasise the importance of focusing on ontological questions rather than epistemological questions. Wight says, for example, that there is no scientific method as such that can be generalised across all the sciences. Unless, of course, we want to construe the rule ‘start with ontology first’ as a tentative methodological proposal to all science.Footnote 5 This group claims that ‘social science is the study of social objects’, justifies the conclusion that ‘those objects must be identified before philosophical positions are taken regarding how we may come to make knowledge claims about them’. This article argues that such an inference is unwarranted, and refers to this inference as ‘the ontological fallacy’.

The article sketches how one might reasonably develop a set of ontological commitments (section I). It describes key elements of the view I have developed elsewhere, ‘causal conventionalism’ (section II).Footnote 6 The article next looks at a number of criticisms of that view published in the past year by Colin Wight (section III),Footnote 7 with particular emphasis on Wight's criticism of its ontological doctrine, namely, constructive empiricism (section IV). The article also tries to clarify some confusing terminology that critical realists have used (section V). The article notes that critical realists oppose a view of the social science that allows for ‘predictiveness’, and observes the difficulties for any application to policy-making that this engenders. It then focuses on the specific problem of the ontological fallacy (section VII). It closes with conclusions about pluralism and the ontology of IR theories (section VIII).

1. Whence an ontology?

Scholars who argue for a primary focus on ontology say that IR theorists have, in the past, mistakenly placed a priority on matters of epistemology and methodology. If these critics are right, then there must be a scientifically and/or philosophically justifiable method for choosing what objects go into our ontology, which can be articulated prior to a discussion of epistemology and methodology. But the latter claim is dubious at best. A scientific ontology (in the natural or social sciences) must be developed in a particular way, which does not allow for the development of a full, theoretical ontology prior to settling various questions in the methodology and epistemology of science.

In particular, as scientific investigators, we make observations about the world and the behavior of the things that we observe; those ‘things’ are what constitute our pre-scientific ontology. Such observations prompt us to ask various questions, especially about how we might explain the observed behavior. We try to answer the questions by formulating explanations and theories. The theories that make use of these explanatory answers postulate further entities. But these latter entities are knowable only by reference to the explanations and theories of which they form a part. There is no other way of judging which, if any, exist. For this reason the entities cannot be specified prior to theorising. Let us outline this procedure in a bit more detail.

1. We first note that ordinary discourse embodies a pre-theoretical ontology, which includes social, physical and other objects. This pre-theoretical language uses terms that refer to the objects that virtually all of us, as ‘common sense realists’, accept as real, such as tables, chairs, people, dogs, and cats. The doctrine of common sense realism regards the objects of ordinary experience in the way they are understood by the typical, normal-functioning adults in the Western world; it avoids some of the more counter-intuitive philosophical positions developed over the centuries that treat them as something substantially different. There are non-physical entities included in this view, references to which are part of our pre-theoretical discourse, such as political offices, money, states, and wars. Thus some familiar pre-theoretical objects are, strictly speaking, non-observable. The claim that these objects are identified without any need for a scientific theory requires no commitment to any of the competing philosophical accounts of how we come to understand those objects, such as empiricism, rationalism, post-structuralism, etc.Footnote 8 Whatever account one accepts, it is quite obvious that people had ideas about stones, birds, states, and promises before there was any developed theorising in what are today understood as natural.Footnote 9

2. Human beings raise questions about pre-theoretical entities and the events in which they participate. We ask questions about these pre-scientific entities and the connections between observable events that involve these entities. (For example, ‘Why did the boulder fall onto the road?’ or ‘Why did the two autocratic states sign a particular trade pact’?)

3. We then propose answers to those questions. These answers will often involve unobservable entities that have causal efficacy (such as ‘gravitational forces and/or someone pushing the boulder caused it to fall’ or ‘international regimes of certain sorts promote increased information flow and increased expectation of compliance with agreements’.)

4. We devise principles that make use of answers that combine with boundary conditions to form theories. (Examples would include Newton's theory of universal gravitation and Keohane's functional theory of regimes.)Footnote 10 That is, regardless of whether we take them as ‘real’ or ‘as if they were real’, we make reference to the theoretical entities as we formulate new questions.

We pose new questions concerning the behavior or properties of the unobservable entities, that is, those entities which we acknowledge as a result of their having been postulated by well-established theories. (‘Do gravitational forces bend light?’ Or ‘What causes regimes to break down’?)

6. Finally, we select or develop methods to test theoretical explanations of the observed phenomena against alternative theoretical explanations. We have a set of pre-theoretical or pre-scientific objects that the methods must conform to, and the competing theories have sets of entities that they postulate. The methods we choose must be appropriate, in the sense that they must have the capacity to test aspects of those particular entities. Thus, the methods will reflect the ontologies of the competing theories.

If this outline is at least roughly accurate, then it is possible to see that our ontology does not come from our methodology, as is sometimes argued. Rather, the proper methodology should reflect the character of the theories we are evaluating. And those theories are what entail particular sets of entities.

2. Causal conventionalism as a meta-theory

The position I have defended incorporates 1. the Duhemian principle of the conventionality of all science, 2. methodological pluralism, 3. a pragmatic theory of knowledge, and 4. constructive empiricist view of theoretical terms.Footnote 11

The first principle, the conventionality of all science, reflects the conventionalist discovery that all empirical science progresses only when the community of scholars accepts certain measure-related propositions that cannot be derived from any combination of objective observations of the world and rules of formal logic. To cite the most significant convention in physics, physicists conventionally stipulate that our measuring instruments are not subject to expansion and contraction as they are moved through space. Observation and formal logic cannot prove either that they do or do not systematically change length when moved. These conventionally agreed-upon propositions are, however, not ‘arbitrary.’Footnote 12 Philosophers of science are responsible for debating possible conventional claims and arriving at the most philosophically defensible conclusions. It is not merely a matter of agreement, but a matter of careful and rigorous meta-theoretical argumentation that produces the most effective conventions. Virtually all accounts of the natural sciences include an explanation of past ‘rational progress’ of science. The extension of conventionalism to the social sciences and IR includes the view that, at least in some areas of inquiry, progress can occur – and even that it has occurred.Footnote 13 While many ideas from the natural sciences have been imported and applied to IR and social sciences, some more successfully than others, Duhem's powerful arguments about the nature of theory and knowledge in physics have not. In my recent writings I have sought to develop an account of IR as a social science by making use of these fundamental insights of Duhem.

With regard to the second element, methodological pluralism, my view is that different sorts of questions may be raised in IR that are equally legitimate, and different sorts of methods would be appropriate to answer them. This methodological pluralism observes that some questions are moral/evaluative and require the methods of moral argument to answer them. Others are clearly interpretive and require hermeneutic and interpretivist techniques to produce the most compelling answers to them. Still others are primarily causal and are most directly answered by methods that are closer to those used in the natural sciences. The view of methodological pluralism I endorse does not imply that different meta-theory claims are all equally valid. Philosophical inquiry is generally able to distinguish superior from inferior methods.Footnote 14 Third, the pragmatic theory of knowledge and truth, advanced initially by Charles Sanders Peirce, holds that the ultimate arbiter of what is acceptable to believe is the ‘cash value’ of our beliefs, which in turn is based on what makes experience and action possible and successful.Footnote 15

Finally, the constructive empiricist view of theoretical terms, advanced by Bas Van Fraassen, acknowledges that there may be entities that underlay the observable empirical connections and processes. However, the history of science is one of continual theoretical change and improvement: even the most widely-accepted theories of the past have been shown to be erroneous. This ‘meta-induction on the history of science’ leads to the conclusion that even those theories that today appear to be the most certain will some day be discarded and replaced. As a result, we must resist concluding that the theoretical (and unobservable) entities postulated in any currently-accepted theory will ultimately prove to be true when measured against future observations and as-yet-to-be-proposed theories.Footnote 16 Thus, it is a source of unnecessary error to infer that the theoretical entities postulated by currently-accepted theories should be added to our ontology of ‘existing’ objects, which otherwise consists of observable, pre-scientific objects like tables and chairs and dogs and cats.

3. Wight's critique of conventionalism in science

A. Scientific progress

Wight asserts that positions that reject SR [deleted word], inhibit scientific progress and encourage dogmatism. According to Wight, this is particularly the case with instrumentalism, namely, the view that concepts and theories are merely useful instruments whose value is purely a function of how effectively they explain and predict phenomena and not whether they are true or false (or correctly depict reality). I have addressed this criticism briefly in a recent forum on scientific and critical realism in Millennium.Footnote 17 But more should be said about Wight's specific examples – Galileo's new interpretation of astronomical data and Bohr's account of the status of quantum theory.

i. Galileo and the Catholic Church

Wight criticises my rejection of scientific realism by citing the example of Galileo's treatment by the Catholic Church. Because ‘[t]he Church did not object to the claim that Copernicus's theory provided a useful, instrumental model of the solar system, or that it supplied accurate predictions’, Wight concludes that instrumentalism ‘is effectively taking sides with the Inquisition against science’.Footnote 18 Many errors are compacted into this brief example.

Because the earth and sun are referred to in the astronomical theories, there are two ways in which they might be treated – either as pre-theoretical objects or as theoretical objects. There are flaws in Wight's position either way. First of all, if they are pre-theoretical objects, then the argument simply does not apply at all. Instrumentalism treats theoretical terms as referring to theoretical objects ‘as if’ they existed, rather than asserting that they are real, like tables and chairs. Instrumentalism's ‘as if’ approach does not treat the existence of pre-theoretical objects such as the Earth and Sun, if they are so classified, differently from SR or other doctrines.

If Wight thinks they should be treated as theoretical objects, then it would still be clear that there is no basis for the claim that instrumentalism ‘sides with’ the Church against Galileo. First of all, it is obvious that if instrumentalism had been adopted in Europe as the best way to understand the nature of science, then there would have been no controversy to begin with, since the ‘as if’ account of the solar system would not have been taken as asserting that Galileo presents a timeless literal truth about the movements of these bodies (which in fact he did not). If instrumentalism had been accepted, Galileo would not have suffered the attacks of the Church, and the cause of scientific progress would have been further advanced by additional work by Galileo and other Copernican-inspired scientists. An instrumentalist view of science, had it been adopted by all parties, would have led to the greater progress of science.

There would not have been a controversy if the Church had viewed all science along instrumentalist lines; it initially interpreted Galileo's theory as making claims in the way that scientific realists would. Obviously, instrumentalists hold that instrumentalism should be adopted. Thus, Wight is blaming instrumentalists for something that took place because their doctrine was rejected by the side he regards as having been in the wrong.Footnote 19

Second, since Wight does not spell out what it means to say one side is ‘siding with’ the Church, we should ask, ‘What, in the most precise terms, would an instrumentalist say about this controversy?’ Since all parties at the time accepted that Galileo's theory was more useful for all scientific and practical purposes, the instrumentalist would, based on her view of the nature of science and the accepted criteria of theory choice (parsimony, empirical adequacy, comprehensiveness, etc.), conclude, ‘Galileo's theory of the solar system is superior, and Ptolemy's is inferior’. That is all the pure instrumentalist would say. It is extremely difficult to see how this could constitute to taking sides with the side the instrumentalist says has the weaker theory.

Wight might insist instrumentalists must go farther and say that Galileo's theory has to be taken as a literally correct and true description of the movement of the planets; he might add that if they do not, they are siding with the Church. The instrumentalist would resist this demand, but would do so on the (correct) grounds that Galileo's theory might not prove better than all other possible theories. On the basis of the evidence then available, Galileo's was superior to its rivals. But to say that it is a ‘true theory’ is to reject the possibility that a still better theory might come along. And of course, a still better theory did come along. Thus whether the earth and sun are taken as theoretical or pre-theoretical objects, Wight's charge that instrumentalism sides with the Church fails.

ii. The Quantum picture of the atom

Wight's second example should make it completely clear that instrumentalism does not hold back scientific thinking and progress. He cites Bohr and Heisenberg as anti-scientific realists who exhibit the same sort of obstinacy and anti-scientific spirit as Galileo's Church censors. It is true that Bohr and Heisenberg were vigorous anti-scientific realists. But they can hardly be portrayed as hidebound, conservative-minded, anti-scientific thinkers who are responsible for holding back scientific progress, since they were among the most creative and revolutionary thinkers who advanced scientific progress in the past century. They did, however, believe that their model of the atom was correct and would stand the test of time. But this is similar to scientific realists who have believed that various theories provided scientists with the final answers on various questions. Their insistence on the completeness of their theories are a small element in the overall picture that we must consider to gain an understanding of the impediments (or lack thereof) that instrumentalism places in the way of scientific progress. Wight concludes ‘that scientists treat their theoretical posits as mere instruments, and not real, is absurd.’ Wight may have arguments against instrumentalism and may claim that there are hidden flaws in it. But we should note: 1) Several of the most brilliant and creative scientists of their generation held the instrumentalist view. 2) Instrumentalism may or may not be correct. (To assert that their endorsement of it shows its correctness would be to argue from authority.) But the fact that many great scientists did hold it makes it impossible to dismiss that view as ‘absurd’. Thus, 3) any argument against the view requires substance that goes beyond dismissing it as absurd.

History is full of scientists who believed that their theories were stepping-stones to greater discoveries, and of scientists who believed that they had attained the final truth of the matter in their respective fields. Both groups include instrumentalists and scientific realists. The fact that Wight can cite several instrumentalists who believed that their theories were finished products would only sustain his claim that SR lends itself to scientific progress more than instrumentalism does if the scientific realist record included no scientific realist scientists who treated newly-developed theories as final products. While it is true that Bohr and Heisenberg believed their theory was complete, innumerable scientific realists, including even Newton, believed that their theories were ultimately correct and would not be superseded. Hence Wight's argument that instrumentalists hold science back in a way that scientific realists do not is not supported by his examples.

iii. Progress and falsifiability

Wight seems to say that SR accounts of science see science as falsifiable and anti-scientific realist accounts cannot account for falsifiability. ‘In order for one's theory to be open to refutation, it must be the case that the theory is attempting to provide an account of something; it is only this that allows the possibility for others to refute one's claims and hence disprove them’.Footnote 20 This is clearly mistaken. How, according to scientific realists, do we find out whether or not a theory has been falsified? The answer, according to Wight, is by examining the observable effects of the causal mechanisms. But of course, this is exactly the answer one would give whether one subscribes to an SR account of science or a constructive empiricist (anti-scientific realist) account.

When we look at Wight's own description of the process of falisification, it is clear that it does not involve any recourse to unobservable theoretical terms. The process is the same whether or not we believe in the ‘reality’ of the unobservable, theoretical, causal mechanisms. The different ways in which SR and anti-SR accounts treat theoretical terms have nothing to do with falsifiability.

The natural sciences have developed a set of methodological strictures that are observed by scientists, regardless of their position (if any) on SR. On the basis of these methodological and practical strictures, and on the basis of the criteria accepted by the community at any given time on theory choice, scientists appraise available theories as better or worse than their rivals. Among other things, whether they accept SR or not, they require ‘empirical adequacy’. But in the natural sciences, what counts as a superior theory for scientists is framed in terms of criteria of evidential support, good practice, and theory choice; it has nothing to do with whether or not one accepts SR.

Two more points on the subject of progress should be noted. One is that Wight, who endorses SR, has a tendency to identify principles he accepts, for example fallibilism and revisability, and then to attribute them to SR, with an implication that those who do not accept SR would be forced to reject them. But some scientific realists believed that they were developing final truths of science; and, virtually all non-scientific realist authors accept revisability as much as scientific realists do.

Finally, Wight argues against my defence of instrumentalism by claiming that SR is more likely to advance scientific progress than conventionalism, on the grounds that the belief that one is seeking the truth about reality is what ‘drives science on’. He says, ‘Indeed a conventional account of science would tend toward conservatism’.Footnote 21 The argument here is highly speculative and the conclusion is historically dubious. First, this appears inconsistent with Wight's worry that only scientific realists will be focused on positing ‘realistic’ theoretical entities. Whatever that may mean, it certainly appears to place more constraints on scientists who accept SR and allows more latitude for creativity for those who do not. But claims on either side are mere speculation. Second, Wight's argument is about what ontological prejudices are best for scientists to hold to aid their work rather than what a philosophical enquiry would reveal as true. As a psychological claim, the issue then becomes very much an empirical question to be settled by examining the relationship between actual scientists’ beliefs and those individuals’ contributions to scientific progress. Such an enquiry is beyond the scope of this article. But there are prima facie reasons to doubt that the study would result in the way Wight expects, considering that the seminal argument that no theory can be the final truth, unrivaled by others, given any finite store of evidence – the “Duhem thesis” – was developed by a thorough-going conventionalist. A final point regarding the charge of ‘conservatism,’ is that both Poincaré and Duhem argued for the possibility of curvilinear geometry of space before Einstein. In light of these well-known facts about the history of physics it is rather surprising that someone would claim that conventionalism would tend toward scientific conservatism.

B. Theory and practice

In this connection we note that Wight asks the question, ‘How can we put our theory into practice if our theoretical posits do not exist?’ This question does not pose any problem whatsoever for instrumentalism, since what that doctrine is about, above all else, is treating theoretical principles ‘as if’ they were literally true and theoretical entities ‘as if’ they existed (which Wight clearly acknowledges elsewhere). For this reason nothing should be more straightforward than putting an instrumentalist's system of theories into practice, which is to treat the theories (again, we should emphasise, for practical purposes only) as if they were literally true.

In both of his recent works Wight claims that scientific realists follow research methods that are different from others. In his book he asks, ‘What differences, then, will the adoption of a scientific realist philosophy, both in general and in the social sciences in particular, make to our research practice?’Footnote 22 However, in the three remaining pages of the chapter, after he first poses this question, he discusses a variety of issues (like the importance of Foucault's power-knowledge relationship, the fact that inquirers’ values affect their theorising, the fact that seeking theories with enduring, epistemic access issues connected to the open nature of social systems, and so forth) without ever saying anything about how anti-scientific realists are constrained from doing anything that scientific realists can do – or vice versa. That is, there is little that is peculiarly scientific realist in what follows. But more importantly there is nothing that has any practical research implications that would differentiate what Wight believes from any instrumentalist or empiricist. Every sentence of the two final concluding paragraphs of the section and chapter could easily be accepted by empiricists or instrumentalists. There is a similar gap in Wight's recent article, ‘A Manifesto’, when he raises the question of how a student who wanted to study democratic peace would proceed differently if she adopted SR. Wight answers the question by saying that such a student could not proceed without ‘ontological clarification’. But the clarifications he specifies are, as I have noted, exactly what an empiricist would demand.Footnote 23

C. Getting it right

Wight says that scientific realists, unlike anti-scientific realists, care about ‘getting it right’. He says, ‘If our assumptions are not intended to be realistic, then it does not matter how we describe things. So there is no point in trying to get positivism right … I disagree. Getting things right is … imperative’.Footnote 24 I fully agree with Wight that it is important to get right whatever it is possible for us to get right. But it is equally important to point out areas where there are inherent limits to what can be ‘gotten’ at all, like the measure-stipulation in physics that conventionalists identified. There are many other examples. If someone said that she did not have a system of first-order logic that could be shown to be internally consistent and at the same time could provide a basis for elementary number theory, a critic might demand a better system of logic. But there can be no such system of logic. If, after having Gödel's proof available to them, scholars had continued to demand such a system of logic, and reject as inadequate any system that was not demonstrably consistent and complete, they would have been holding back the advancement of knowledge. Similarly, if a physicist were to tell us that her theory does not allow us to identify our exact location in the universe in absolute terms, a critic might demand that physicists work toward a better theory. But modern physics tells us that position is completely relational. Hence, scientists who might have continued to demand, contra Einstein, absolute positions of space and time would, similarly, have been holding back science. Thus there are powerful intellectual grounds for accepting that there are certain, specifiable limits to what can be known in various fields of inquiry that contravene any naïve demands we might make upon them. Contrary to Wight's claim, the physical sciences have such limitations which we must recognise, which in turn must limit the desire to overcome conventional stipulations.

D. Conventionalism as a mere ‘moment’ in science

Wight acknowledges that conventional agreement plays a role in the development of scientific theories in physics and elsewhere.Footnote 25 But he sees that role as limited and as one that is overcome when the development advances beyond an early stage. This is not a plausible account of science.

In order for physicists to develop any theory, they must measure physical objects and processes. This involves use of instruments, the simplest of which would be a rod for measuring length. A century ago Poincaré and Duhem asked how we know whether a measuring rod changes or retains its length when it is moved from one position to another. They asked, how can we tell if there are forces acting on all objects, including the measuring instrument, that systematically change the length of each as it moves from position a to point b – and then changes its length back when it is returned to position a? They correctly determined (since the force would be operating uniformly on the objects to be measured and the measuring instrument itself) that there is no way, based on pure logic and observation, to answer this question. But some answer must be given, since calculations based on the assumption that the rod changes length when moved would lead to a different set of physical laws than would calculations based on the assumption that it does not change length. Physicists accept as a matter of convention that measuring rods have constant length.

Wight says, ‘So yes, as Fred Chernoff has argued, there is a conventional element to all scientific research’.Footnote 26 But Wight treats the conventional element as transitory; in his view it operates only in the short-term to facilitate further research because in the long run science overcomes any conventional element. Wight says, ‘this instrumentalist/ conventionalist moment is unable to explain how science deals with non-observables in the long term’.Footnote 27 This claim is not correct. Contrary to Wight's claim, the conventional element never ‘drops out’ as scientific progress advances because there is, even in principle, no way to replace it with some sort of objective, empirically-grounded principle that some superior future theory will incorporate. Physical scientists in the twenty-first century have to maintain the proposition that measuring rods maintain constant length when moved as a convention, just as they did a century ago.

E. Criteria of theory choice: simplicity and predictive capacity

Wight is critical of comments I make about the importance of simplicity and predictive capacity in making theory choices. In a footnote Wight explains says, ‘Chernoff puts too much stress on simplicity as a criterion of theory choice’.Footnote 28 But simplicity is a central feature of philosophical and scientific reasoning. For example, in the conclusion of his article, while discussing another matter, Wight says, ‘A fundamental hypothesis of science is that appearances are deceptive and that there is a way of looking at, interpreting, and/or organising the evidence that will reveal superficially disconnected and diverse phenomena to be manifestations of a more fundamental structure’.Footnote 29 Wight is quite plainly invoking the criterion of ‘simplicity’ in this passage. Why do we believe that the principles that unify many apparently ‘diverse and disconnected phenomena’ should be preferred over theories that invoke many different explanatory propositions for the phenomena and postulate a plethora of causal processes and theoretical entities? It is precisely because of scholars’ – including Wight's – acceptance of the fundamental principle of simplicity. Thus, Wight himself finds it difficult to formulate an argument about how scientific theories work without recourse to a belief in the power of ‘simplicity’.

Theory choice in the natural sciences is based on an array of criteria that may sometimes conflict with one another. Most deal with empirical evidence, including the accuracy of prediction, consistency with past observations, and the like. I have made this claim many times. Wight points out one passage where I abbreviate in a way that could be misleading. Wight says, ‘In suggesting that prediction is the determining factor in theory choice, Chernoff seems to contradict his later claim that “consistency, coherence, simplicity, and explanatory power” are part of the justificatory framework ([Chernoff] 2002: 195)’.Footnote 30 At several points in the article I list a variety of criteria.Footnote 31 In the passage Wight cites I was referring to two or more theories that may be subjected to experimental tests (and thus which have been judged roughly equal on the basis of the other criteria of theory choice). Typically, when experiments (or social science quasi-experiments) are conducted, the competing theories will yield different predictions for the outcome of the experiment. What then leads scientists to accept one over the others is the accuracy of the predictions – the (mis)match between the theories’ expected outcomes of the experiments and the actual outcomes.Footnote 32 The subsequent statements on the subject in that article make my position clear. But Wight is right that, as phrased, the passage should have noted that this determination is applicable when the theories have fared similarly well on the basis of the other criteria.

Wight goes on to say that my view cannot account for ‘explanation’ at all because to do so would be to reject an important part of Duhem's account of science. Indeed, I do not agree with Duhem's limitation on causal explanation, especially for the social sciences, and I do so often and prominently, in almost all of my recent writings. Indeed, I name my meta-theoretical account ‘causal conventionalism,’ which would seem to lay out clearly where I stand on causal explanation in the social sciences. In the article Wight cites I explicitly say that the Duhemian view I defend can account for ‘the social sciences’ greater need for some form of causal explanation’.Footnote 33 That is about as plain as it can be. And in my recent book, The Power of International Theory, I elaborate on that view and specifically note the fact that my account endorses explanation, while Duhem's does not. I acknowledge the contrast and defend my approach. I argue there that the principle of the conventionality of all science does not entail any sort of rejection of ‘explanation’ and is not incompatible with an account of scientific explanation.Footnote 34 In The Power of International Theory there are numerous discussions of the role of ‘causal explanation’ in the social sciences. Even a number of the section titles, by themselves, make clear that they defend a causal explanation in IR.Footnote 35

4. Constructive empiricism, theoretical terms, and theoretical entities

Constructive Empiricism. Wight chooses to attack Van Fraassen's doctrine of constructive empiricism, which I have endorsed, essentially by dismissing it.Footnote 36 Constructive empiricism agrees neither with the traditional empiricist view of theoretical constructs nor with the scientific realist view that talk of theoretical entities is either meaningless, or meant to stand for something that we cannot acknowledge as ‘existing’ because of the lack of empirical methods of verification or corroboration. Van Fraasen's view is that there are powerful reasons not to infer the real existence of the theoretical entities of the best theories we accept. He argues that there is no justifiable inference to an unseen world of theoretical entities because i) the evidence for such a realm is always by inference; ii) there are always alternative accounts of the observed phenomena that yield conclusions about different theoretical entities; and iii) the history of science shows that those entities probably do not exist. The most philosophically defensible conclusion about theoretical entities is that they may exist and they may not, and that the likelihood they may exist precisely as our currently favoured theory describes them is very low. Constructive empiricism leaves open the possibility of an underlying causal structure without affirming it or specifying what objects it contains. I have described this view, which I support, as a form of agnosticism with regard to theoretical entities.

Wight's critique of my position is, as noted above, essentially to dismiss it. He says:

I am aware that I have not dealt with Chernoff's agnostic position … because I do not think it is a coherent one to adopt. Much like in the theological domain, an agnostic that goes to church (tests positive for their reality) is probably a believer, one that does not probably an atheist (does not bother to test).Footnote 37

Wight admits that he offers no argument here and merely asserts that scientists and presumably others cannot suspend judgment – on questions like whether the terms of theories that appear to refer to theoretical, and in principle unobservable entities, do indeed refer to real things.

There are very persuasive grounds for rejecting this unsupported assertion. First, in ordinary life, as well as in philosophical and scientific enquiry, people suspend judgment on many questions. On some of these questions scientists may have conjectures, but they do not generally accept any set of claims as the true answer. This would include scientists. They often suspend judgment on questions like whether there is an omnipotent Creator and/or about whether their great-grandchildren will become members of the Royal Geographic Society. Second, scientists have the capacity to suspend judgment on scientific topics a) about which there is inadequate evidence and b) on matters that current theories tell us we cannot answer. Examples of the former include how we might explain the proportion of large animals to small animals on earth and why there is a particular distribution of galaxies in the universe. Examples of the latter would include fixing the precise position of an electron, given its momentum, and what is inside a black hole (since information cannot come out of a black hole, due to the ‘information loss problem’ of black holes).Footnote 38 (And a third sort of evidence is the testimony of some scientists who say that they suspend judgment on the reality of theoretical entities).

In IR many scholars suspend judgment about whether Operation Barbarossa might have succeeded had it been launched a month or two earlier, and/or on whether there will be a fundamental transformation of the anarchical nature of world politics by the end of the century. Some scholars have conjectures on the former and predictions on the latter, and some do not. There is simply no reason to accept the assertion that people cannot suspend judgment on important and unimportant questions, and there are good reasons to believe that they are indeed capable of doing so.

Scientific realism in the social sciences

Wight addresses my reservations about extending SR to the social sciences. He asks rhetorically, ‘Are there, for example, rules and norms that state leaders follow, break, or bend, in their interactions with other state leaders, despite the fact no one has ever observed such a rule? Equally, is there really no distribution of capabilities within the international system? Does Chernoff really deny that these things exist?’ Wight adds, ‘The fact that something is socially constructed and dependent on concepts, beliefs, language use, and so on, does not mean it is not real’.Footnote 39

First of all, constructive empiricism is very explicit in avoiding any denial of existence of such entities; it denies only that we can know them with confidence. Let us, in any event, consider in more detail how best to answer Wight's rhetorical question. The question clearly suggests that Wight believes that these things do exist. We must then ask why Wight insists that they do, given that we do not experience them directly and that they are not referred to in our common sense or pre-theoretical discourse. The only reason that seems feasible is that they are used by theories that explain the world better than theories that do not invoke those entities. That is, they serve a function in theories that do the scientific, explanatory, and practical work that we need our theories to do. The best reason we have to insist that a particular theoretical object referred to by theory T1 exists is that T1 is true. But we have noted that over the ages the physical sciences have a history of eventually refuting even the most widely accepted and seemingly certain theories. There are no good reasons at this time to argue that IR is farther advanced than contemporary physical theories and thus that our current social science theories will stand up to the scrutiny of decades or centuries better than physical theories have.

We must also note that constructive empiricism, contrary to Wight's claim, does not require that we ‘deny that these things exist’. Rather, it draws on the history of error in the natural sciences, which is at least as great in the social sciences, to conclude that we must be cautious in accepting the claim that theoretical principles in science harness the final truth on a particular question. And if the theoretical principles are not ultimately true, then we are probably speaking falsely when we assert that the entities postulated in those principles exist.Footnote 40 So, constructive empiricism does not require that we insist that there are no such entities. Rather it insists that we are likely to be wrong about the particular empirical theory we currently use to explain the observable world, because any conclusions about which particular theoretical entities exist are likely to be mistaken. We only use a set of entities, M, because, compared to the rival theories that postulate different entities, the theories that postulate set M provide a better explanation of the observable entities (cats and dogs and people) and events about which we pose questions. The purported entities’ explanatory power is the only reason we have for talking about them. So however attached Wight is to those theoretical entities, if a theory comes along that better explains what they were proposed to explain, then they can no longer be claimed to exist.

Theoretical terms

One of two central features of scientific realism is its treatment of theoretical terms. Critical realists endorse this aspect of scientific realism. This issue separates their view from empiricism and from the position that I have advanced, which includes the pragmatically-based theory choice criteria of Duhemian conventionalism along with a version of Van Fraassen's constructive empiricism. Scientific realists claim that if we accept a theory as the best available, then we must accept the entities postulated by it into our ontology. We include them alongside pre-theoretical entities like people and tables and dogs and regard them as equivalently ‘real’. But the meta-induction on the history of science, described above, does not warrant this conclusion. The form of conventionalism I have advocated takes a constructive empiricist view of theoretical terms, according to which the entities postulated by the theory may or may not exist. We can never be certain that our best theory, T1, is fully true and will continue to hold up over the decades and centuries, against many as-yet-undiscovered theories.

The connection between SR and the ontological fallacy

Why does the philosophical debate over SR focus on the status of theoretical terms? What justifies scientific realists and critical realists in accepting the existence of the referents of theoretical terms? The answer is in the doctrine known as ‘inference to the best explanation’ (IBE). The most prominent defenders of SR in IR emphasise the priority of ontological questions. There is, however, no inherent connection between SR and their claim about the priority of ontology. The most influential philosophers who have advocated SR, Popper, Maxwell, Putman, and Boyd, have identified IBE as one of the core elements of it. Anyone who insists on inference to the best explanation and the claim that theoretical entities should be admitted to our ontology must recognise that our access to (knowledge of) theoretical entities comes by way of inference. Thus it is not direct in the way that our access to (knowledge of) table, chairs, cats, and dogs is direct. Since traditional scientific realists in the philosophy literature identify IBE as the central justification for SR, it is clear that beliefs about theoretical entities are conclusions, ‘inferred’.

We might add here that, as noted above, Wight claims that one only general scientific methodological claim is ‘start with ontology first’.Footnote 41 In rejecting the idea that method is what holds scientific enterprises together and demarcates enterprises as sciences, Wight says that what makes a type of enquiry scientific is that it offers ‘depth explanations’. ‘What marks scientific knowledge out from other forms of knowledge is that it attempts to go beyond appearances and provide explanations at a deeper level of understanding’.Footnote 42 Although that is precisely what mythology and religion generally do, Wight would not wish to count them as science.

5. Tightening terminology

One of the persistent problems in the debate over scientific realism and critical realism, especially in the IR literature, is the confusing use of terminology. IR theorists such as Wight, Patomäki, Kurki, and others are all scientific realists but they endorse some version or other of critical realism, which is much more to the point. Some insist on using the term ‘scientific realism’. But critical realism is far more descriptive of their position, in that it is a much more specific doctrine. Only a small portion of the principles they defend are scientific realist principles. However, they tend to ascribe many of their critical realist views that most scientific realists would disavow, as part of scientific realism. Critical realists accept many contested claims that other sorts of scientific realists do not but tend to refer to the whole collection of claims as part of scientific realism.

Because these authors are invoking well-known philosophical terms, especially ‘scientific realism’, it is incumbent upon them either to use them in line with general usage in the discipline or to acknowledge that they have created a new meaning for a term.

It is very difficult for IR theorists to avoid some simplification as they discuss and apply debates in the philosophy of science and social science. But there must be an attempt to represent doctrines accurately. I have tried to cite quotations to support claims about what the core tenets of SR are and to cite them from the authors who have been most influential in defining that doctrine in the philosophy of science. I have also attempted to offer quotations from the authors whose views I endorse, for example, by citing versions of conventionalism (including the well-known tenets of Poincaré with which I disagree). Wight offers a wide range of principles that he attributes to SR but does not acknowledge them as not-widely accepted by scientific realists, and he does not acknowledge how fragmented supporters of SR have become.

At its core SR is a doctrine about the truth of scientific theories and the reality of the entities those theories postulate. The statement of the doctrine by its most influential supporters, who defined the doctrine, such as Hilary PutnamFootnote 43 and Richard Boyd,Footnote 44 make this very clear. However, over the years the earlier versions of scientific realism (like those of older versions of empiricism) encountered severe difficulties. Their defenders have ceded considerable ground by moderating and/or circumscribing them, or by using of a variety of devices in the philosophy of language to reconceptualise the ordinary notions of ‘truth’ and ‘reality’ to deal with the objections (with which Wight seems unsympathetic).Footnote 45 Even by the early 1980s a number of commentators began noting that key aspects of SR had become ‘watered down’ so far that, for most versions of SR, it was no longer clear what was ‘realist’ about SR.Footnote 46 And we should remember that it was the (now decreasing) straightforward belief in the reality of the world science studies that was responsible for so much appeal of SR.

Wight uses the term ‘scientific realism’ in a way that does not comport with the philosophical usage, often, as noted, conflating it with critical realism. And one clear indicator of Wight's penchant for conflating the two doctrines is to notice that in the section in chapter 1 of ‘Agents, Structures and International Relations’, in which he presents ‘scientific realism’, over a quarter of the 85 footnotes in that section are to one specific author, namely Bhaskar, the founder of ‘critical realism.’ Several other footnotes refer to Bhaskar's critical realist disciples, like Andrew Collier. And most of the other footnotes are included because they are explicitly opponents of SR (myself included).Footnote 47 Thus while Wight often attributes a wide range of philosophical views to SR, when he does rely on specific works to explicate the doctrine, they are the works of the most committed critical realists.

It is important that we exercise care in using these terms – which is not always done. Consider the comment by Wight that ‘all approaches, whether scientific realist or not, rely on implicit (sometimes explicit) realist metaphysic’.Footnote 48 To this he attaches the following note:

Chernoff (2002), for example, attempts to refute the role of this commitment to philosophical realism, but his critique only makes sense if we conclude that the explanations he does provide tell us something that we did not previously know. Hence, his own explanation already implies a depth ontology.Footnote 49

This vivid example of terminological carelessness, which can be spotted in several different ways.

First, the conclusion is a non sequitur because the claim ‘an author is telling us something that we did not previously know’ in no way is connected to any requirement of philosophical realism. For example, a logician may prove a theorem that was not previously known to be true without a commitment to any sort of philosophical realism whatsoever. One might object that this example is unfair because Wight is only talking about ‘explanations’. We might then imagine that we see a boulder on a road, which has recently fallen from a hillside spot that it had occupied for several years. We then ask why the boulder fell. A friend tells us that she hiked up in the middle of the night and pushed it down the hill. In this context the answer would be regarded as reasonable and sufficient. And the answer both explains the event and tells us something we did not previously know. But obviously, it does not require any ‘depth ontology’. Thus the conclusion that philosophical realism is connected to ‘telling us something new’ in the Wight's claims is a non sequitur.

Secondly, and perhaps more instructively, Wight's comment is a perfect example of the confusion that I wrote ‘Scientific Realism as a Meta-theory’ to warn against. Wight states that I oppose philosophical realism; but in my published work I have never questioned that doctrine. In fact, in the article Wight cites, the first reference I make to ‘philosophical realism’ is to distinguish it clearly from the scientific realist position that I am opposing. The passage reads as follows: ‘There are many forms of philosophical realism that should be carefully distinguished from SR (scientific realism) that are, typically, compatible with anti-SR’.Footnote 50 Thus it is very clear in that article that I am emphasising the importance of distinguishing different forms of philosophical realism in order to avoid the precise sort of mistake Wight makes in interpreting my critique of SR. Wight says that I am attempting ‘to refute’ the ‘commitment to philosophical realism’; whereas the passage shows that I am explicitly asserting that my opposition to SR does not require a general rejection of philosophical realism.

Finally, Wight and other scientific realists often abbreviate the term for anti-scientific realists to ‘anti-realists’. This adds to already-persistent confusion, since most anti-scientific realists are realists in many other senses, especially in that they accept common sense realism. Anti-scientific realists sometimes even endorse ‘motivational realism’, a view that describes scientists’ attitude toward the unseen world many of them seek to describe and explain. As I argue elsewhere, motivational realism neither entails nor is entailed by either scientific realist or anti-scientific realist principles.Footnote 51

6. Policy-making and the rejection of prediction in IR

One of the chief advantages science confers on humanity is the ability to do things that we could not do without it. The aid in doing things stems entirely from the way in which applications of natural science theories make use of the power that those theories, along with knowledge of initial conditions, have to produce accurate (or, sometimes, nearly accurate) predictions. Building bridges, sending rockets to Mars, and finding cures for diseases (though still lagging behind on jet-pack travel) all are possible only because the theories tell us what will (likely) happen under certain specific conditions. Similarly, in the social sciences all policy making is directed toward the future, by aiming to choose a current action that will produce desired consequences. If increasing manufacturing is a good thing for the treasury, it is because of the prediction that more goods made and sold next year will produce more jobs next year, which will, in turn, produce more tax revenue. While past observations and explanations for them may be the basis of certain principles, all of the inferences on which policy decisions rest require predictions. That is, all policy-making requires predictions (which may be probabilistic, contingent, etc.) along with a set of values or goals.

Scientific realists in the philosophy of the natural sciences, like virtually all philosophers of science, regard prediction as one of the chief hallmarks of and benefits of astronomy, chemistry, physics, etc. In IR theory, the use of the term ‘scientific realism’ would seem to connote a ‘scientific’ approach to IR meta-theory by its proponents. But despite sounding ‘pro-science’ (as Fine puts it), scientific realism does not confer this main advantage on IR theory. Critical realists and scientific realists in IR and the social sciences would like to be able to apply IR theory to policy-making.Footnote 52 But they thoroughly undercut that goal by stating very explicitly that IR and other social sciences are incapable of prediction.Footnote 53 Wendt, Wight, and Patomäki, quite clearly oppose the idea that IR theory can predict.Footnote 54

The implications are significant. Critical realists and scientific realists in IR (unlike other post-positivists) accept the almost-certainly false consequences of ontological claims that will later be rejected, while at the same time (and unlike most other naturalist approaches to IR meta-theory) missing out on the principle benefit of a scientific approach to IR and social sciences, namely, the ability to aid policy-making. Moreover, Wight, like other critical realists, and like the critical theorists who have influenced them, support ‘human emancipation’ as a goal of social enquiry.Footnote 55 While more needs to be said on this point, I will observe here only that the denial of predictiveness in all social sciences is inconsistent with the social goal of human emancipation, if the latter is meant to entail any successful attempts to change social conditions.Footnote 56

7. The ontological fallacy

Critical realists and some scientific realists in IR argue for the priority of ontology. They maintain that ontological questions must be treated before (and are somehow more important than) epistemological or methodological questions. They say that we must settle on which objects – both pre-theoretical and theoretical – belong in our ontology before we can inquire into what we know about them. These claims are mistaken and commit the ontological fallacy. The only way we can justify any existence claim for a theoretical object is by the superiority of a theory which postulates it, and the only way we can show the superiority of a theory is by having answered important questions of (natural or social) scientific methodology.

In the natural sciences there is a pattern of ‘approach to consensus’ that can be historically identified. When the scientific community seeks to explain a range of observable phenomena, the typical pattern is that, at some point, the community will come to accept a single theoretical answer, at least for a time – until another supersedes it. According to SR, the scientific community concludes that the theoretical postulates of that consensus theory are ‘real’. However, in the social sciences most questions about observable phenomena are explained by different groups of scholars in different ways; that is, competing theories invoke substantially different theoretical principles. And typically none of the theories will, over time, ‘win out’ in an approach-to-consensus pattern. The competing social science theories postulate very different theoretical entities. Since scientific realists are not ontological relativists (that is, since they do not endorse the view that ‘objects A and B are “real for you” and ‘unreal for me”’), they face a serious problem when they try to apply scientific realism to disciplines where the central theoretical questions are debated decade after decade without community-wide acceptance of a single theory.Footnote 57

Theories are accepted over rivals on the basis of a set of criteria. No philosophers of science claim that there is a single, specific concept or criterion that guides all scientific theory choice. Some of the criteria noted above are parsimony, empirical adequacy, and comprehensiveness. Many others have been proposed. Very few are ontological. The one that is most clearly ontological is an application of ‘simplicity’ – which may be applied epistemologically as theoretical parsimony, and ontologically, in the form of Occam's Razor. Interestingly, Wight tries to downplay ‘simplicity’, which undercuts the most widely accepted ontological criterion. But it should be emphasised that ontological simplicity is not ‘first among criteria’ and it is not capable of over-riding other criteria. A theory with a simple ontology that does not account for past observations well and produces inaccurate predictions would not be accepted by scientists over an ontologically less simple theory that explains and has predictive success.

Wight and other scientific realists and critical realists tried to invoke a notion of ‘realistic-ness’ of entities proposed. They have argued that empiricism and instrumentalism are deficient because they allow for the postulation of ‘unrealistic’ entities. The empiricist could answer that, as long as the new theory meets the philosophical criteria of theory choice better than its rivals, the ‘seeming-reality or -unreality’ of the entities it postulates is completely irrelevant. They point out that, from time to time there have been moments in history when scientists have proposed new theories that fit with current observations about ordinary objects. These new theories explained the behaviour of those objects more effectively than older theories and thereby led to the rejection of the older theories and the entities they postulated. Science has judged the entities they posit, of whatever sort they may be, as ‘acceptable’, if they provided a better explanation of the observable world. There are no a priori barriers to what a ‘realistic’ entity is. Instrumentalists simply apply the best philosophical criteria that have survived long debate, just as scientific realists do. A person unfamiliar with science would likely find the things that physical science has postulated as entities in the past century – antimatter, electrons, electrons, black holes, the universe as a string, etc. – as anything but ‘realistic’. The charge that instrumentalists lack a requirement of ‘realistic-ness’ for their theoretical posits is without force.

With regard to natural science, Wight does not defend the claim that all entities ever posited by good scientific theories throughout history existed then or now. He correctly points out that scientific realism is not committed to this claim.Footnote 58 Rather, it is committed only to the claim that the theoretical posits of the best current theory in each field must be said to exist. How do we apply this in the field of IR? Critical realists and scientific realists in IR must deal with the fact that there is, at least thus far, persistent disagreement on the vast majority of IR questions, which leaves multiple theories in play. There are various sets of observed phenomena that are explained by different authors by means of liberal, realist, Marxist, psychological, bureaucratic politics, and other sorts of theories. Are all of the entities postulated by all of the diverse current theories that are still in contention ‘real’? How does Wight's dictum ‘start with ontology first’ help? What can we do to select our scientific or theoretical ontology before we have tested theories? Can we eliminate some of the explanations before the theories are fully formulated and tested? That is, can we eliminate some of the competing theories on the basis of something problematic about their theoretical ontologies? There appear to be few, if any, guidelines for doing so.

The problem can be seen clearly by way of an example. Suppose we invent a theory (badly) designed to explain preventive war that postulates a social entity, let us call it ‘sogiston’, which is ‘an invisible humour that infests all governments that decide to launch preventative wars’.Footnote 59 Must we take this entity to be real just because it is postulated by an explanatory theory? If not, then presumably when we compare other sets of theories and find that some yield explanations inferior to their competitors, the inferior theories must be viewed as postulating entities that do not ‘really’ exist (and thus should not be entered into our ontology). The discussions of critical realism and scientific realism in IR thus far provide no grounds for differentiating those entities that can be said to exist, and which thus must be added to our ontology, from entities that should not be so entered.

In sum, many criteria are used to choose the best theory from a set of possibilities. Some, though not many, of the criteria typically relate to the theory's ontology. But those criteria cannot be applied to choose a theory in the absence of methodological and epistemological considerations that form the grounding for most of the other criteria that a wide array of non-ontological criteria (internal consistency, parsimony, range of phenomena explained, predictive accuracy with respect to observable objects and processes) that philosophers of science agree are central to the process of scientific theory choice.

8. Conclusion

Critical realists claim that ontological questions regarding observable and non-observable entities must be treated before epistemological or methodological questions are posed. That is, they say that we must settle on the objects of inquiry, both pre-scientific and scientific, that we accept into our ontology before we may pose theories or study methodology. These claims are erroneous and are examples of what I have called ‘the ontological fallacy’. I have argued that the only way we can provide evidence for or against any supposed theoretical entity is by means of the superiority of a theory of which it is a part, which results from its comparative performance in satisfying the criteria accepted at that time by the scientific community.

This article in particular has tried to provide some grounds for the claim that the goals of scientific theorising are better satisfied by alternatives, especially by causal conventionalism, than by scientific realism and critical realism. (Related arguments presented elsewhere have been indicated.). Causal conventionalism can accomplish what those doctrines seek to do without some of the problems and inevitable error that they bring with them. The goals of scientific enquiry include i) expanding the store of systematic empirical knowledge, ii) avoiding the acceptance of false beliefs and iii) aiding human beings to do things successfully in ways that would be impossible without such knowledge. I have argued that the causal conventional alternative does a much better job than SR and CR of fulfilling i)–iii); the SR and CR approaches do not allow us to do anything that we could not do treating the theoretical objects ‘as if’ they existed, and it almost certainly adds to the store of false beliefs in our epistemic corpus. Causal conventionalism provides an account for both the natural and social sciences that does not bring with it a store of almost-certainly false beliefs.

Scientific realism is debated in the philosophy of the natural sciences. There are at least good prima facie arguments that support it as a plausible account of the natural sciences. But there are also very powerful counter-arguments that they must be weighed against. And when SR is applied to the social sciences, virtually all of the natural science-oriented arguments lose force and the counter-arguments apply even more powerfully, as I have argued.Footnote 60 For example, one of the prima facie arguments in favour of SR in the natural sciences draws on the need to explain the approach-to-consensus history of most natural science debates; but in the social sciences where there is long-standing absence of consensus on most questions, there is no need to explain how consensus always develops. The pattern in the social sciences renders highly dubious any attempt to posit theoretical objects as ‘real’.

Critical realists and scientific realists purport to allow a ‘scientific’ view of IR. But they reject prediction, which is one of the major fruits of scientific knowledge, and it is all scientific knowledge does to help human beings navigate the world better. Causal conventionalism shows how prediction, at least in very limited and qualified ways, is justifiable in the social sciences.Footnote 61 Critical realism introduces the error that scientific realism introduces in the natural sciences – acceptance into our ontology of entities that almost certainly do not exist. But at the same time it does not offer the compensatory benefits that other naturalist accounts of the social sciences offer (that is, those that seek to parallel the naturalist sciences, like causal conventionalism), which include grounds for the application of theory to policy-making. (This is the case because the link to policy-making requires a justification of some form of prediction). In this sense, critical realism, or Wight's version of scientific realism in IR, offers the defects of SR without the compensating benefits. In this article and elsewhere I have, of course, not claimed to settle the philosophical debate between supporters and opponents of SR. Rather, I have tried to show that there are the defects in SR noted, that they are particularly acute and damaging for SR’s potential application to the social sciences, that they undercut other goals of applicability of social theory, and that the do not aid social scientists in doing anything that various non-SR views justify.

Critical realists in IR typically argue that epistemological debates have been given priority over ontological ones, which has impeded intellectually fruitful debate in IR.Footnote 62 This article has sought to show the general errors of going from the claim that ‘the objects we know of do not exhaust all of the objects in the universe’ to the claim that ‘major ontological questions must precede epistemological’ or to the claim that ‘a primary focus on methodology will lead into unnecessary controversy and error.’ The article has shown some of the flaws in the arguments for this extreme view of ontological priority and has adumbrated a general approach to developing a social scientific ontology that does not commit the ontological fallacy.

References

1 Colin Wight, ‘A Manifesto for Scientific Realism in IR: Assuming the Can-Opener Won't Work!’ Millennium: Journal of International Studies, 35 (2007), pp. 379–98, and Colin Wight, Agents, Structures and International Relations: Politics as Ontology (Cambridge: Cambridge University Press, 2006).

2 Wight, Agents, Structures; Heikki Patomäki, After International Relations: Critical Realism and the (Re)construction of World Politics. (London: Routledge, 2002); and Alexander Wendt, Social Theory of International Politics, (Cambridge: Cambridge University Press, 1999).

3 Wight, ‘A Manifesto’, p. 388.

4 According to the epistemic fallacy, ‘what is known is what can be … observed and what “is” is what can be known’, Heikki Patomäki and Colin Wight, ‘After Postpositivism: The Promises of Critical Realism’, International Studies Quarterly, 44 (2000), p. 217.

5 Wight, ‘A Manifesto’, p. 388.

6 Fred Chernoff, ‘Scientific Realism’ and The Power of International Theory: Re-forging the Link to Policy Making through Scientific Enquiry (London: Routledge, 2005).

7 Wight, ‘A Manifesto’ and Agents, Structures.

8 The objects may be construed as sense-data, concreta, social constructs, etc. We may even have a ‘socially constructed’ conceptual framework to pick out ordinary objects. (Though it is not clear why humans need such, if orang-utans do not.) But these are not scientific theories.

9 It is hard to imagine a philosophical view that does not acknowledge that human beings, and/or perhaps even other primates, were able to recognise what we have specified as ‘ordinary objects’ prior to the rise of science or the acceptance of any set of unobservable, theoretical entities. There are philosophical difficulties distinguishing precisely observables from non-observables. But the account of the development of ontologies here requires only that humans be able to pick out objects in their environment that do not require the use of sometime-contested scientific theories.

This article draws a standard distinction between the ontologies of ordinary experience, which include many objects we observe without mediation and inference, and the theoretical ontologies of sciences. One might push the distinction farther and separate a philosophical ontology from both of the above. That position is entirely consistent with the position taken in this article and the way in which scientific realists and critical realists in IR have argued for the priority of ontology. Philosophers who support a particular set of entia would use the generally accepted philosophical criteria for choosing a superior philosophical account of the domain specified. The arguments advanced in this article pertaining to the choosing a scientific-theoretical ontology would still mutatis mutandis apply. Some would, though, not separate the problem into three parts but would focus on the distinction between pre-theoretical and theoretical objects. The general principles of ontological commitment sketched in this article apply whether we treat a binary or tripartite ontology.

10 There are, of course, huge difference between Newton's mechanics and Keohane's theory of institutions, which point up some of the general differences between the natural and the social sciences. However, these theories share the fundamental features of theories noted.

11 Since it is not possible to defend all of those principles here, the reader may refer to Chernoff, ‘Scientific Realism as a Meta-Theory of International Relations’, International Studies Quarterly, 46 (2002), pp. 189–207, and ‘The Study of Democratic Peace and Progress in International Relations’, International Studies Review, 6 (2004), pp. 49–77; and The Power of International Theory.

12 This is explored further in the next section. Other conventionalists, following Poincaré, would say that they are arbitrary.

13 See James Lee Ray, ‘A Lakatosian View of the Democratic Peace Research Program’, in Colin Elman and Miriam Fendius Elman (eds), Progress in International Relations Theory: Appraising the Field (Cambridge, MA: MIT Press 2003), pp. 205–44.

14 The admission that that all meta-theories are equally valid, warrants equal acceptance of a meta-theory that rejects methodological pluralism. See also Chernoff, ‘Methodological Pluralism and The Limits of Naturalism in the Study of Politics’, in Richard Ned Lebow and Mark Lichbach (eds), Theory and Evidence in Comparative Politics and International Relations (New York: Palgrave-Macmillan, 2007).

15 For a fuller defence of these see Charles Sanders Peirce, The Collected Papers of Charles Sanders Peirce, 2, Elements of Logic, Charles Hartshorne and Paul Weiss (eds), (Cambridge, MA: Harvard University Press, 1932) and Christopher Hookway, Pierce (Boston: Routledge & Kegan Paul, 1985).

16 Examples abound. Newton's theory was taken as certain beyond doubt from the late 17th century onward. In the 20th century, the prestigious Leconte Prize was awarded to René Blondlot in 1904 for his ‘discovery’ of n-waves. But later theories rejected the n-wave. Arthur Fine argues that the emphasis on scientific success sometimes obscures the many failures; he puts it by saying that each breakthrough in scientific progress is the peak of a mountain otherwise composed of failure; see Arthur Fine, ‘Unnatural Attitudes: Realist and Instrumentalist Attachments to Science’, Mind (1986) 96, pp. 149–79.

17 See Fred Chernoff, ‘Critical Realism, Scientific Realism and International Relations Theory’, Millennium: Journal of International Studies, 35 (2007), pp. 401–9.

18 Wight, ‘A Manifesto’, p. 396.

19 Suppose we were to learn that a child, who said that ‘2+2=5’, was criticised by Mussolini, who insisted, rather, that the sum is four. The logic of Wight's criticism is exactly parallel to accusing someone who upheld the claim that 2+2=4 of ‘siding with fascism’.

20 Wight, Agents, Structures, p. 38.

21 Agents, Structures p. 29, note 63.

22 Wight, Agents, Structures, p. 59.

23 See Wight, ‘A Manifesto’, p. 385, note 23 and Chernoff ,‘Critical Realism’, p. 407. There is certainly no consensus among philosophers of science as to any difference that adoption of SR makes to practice. Authors like Fine and Kukla have argued that it does not make any practical difference whether one is or is not a scientific realist; see Fine, ‘Unnatural Attitudes’ and André Kukla, Studies in Scientific Realism (New York: Oxford University Press, 1998). It makes a difference for philosophers whether some version of scientific realism is right or whether the anti-scientific realist position is right – because they want the right answer. But it is a completely different question as to whether it makes any difference for practice. I am not asserting that one cannot make the case that adoption of scientific realism affects practice. Rather, the conclusion here is that it is difficult to see how; neither is the difference self-evident nor does Wight (or others) provide clear examples of how it could – despite claiming to have done so.

24 Wight, ‘A Manifesto’, p. 381.

25 Ibid., p. 383.

26 Ibid., p. 383, note 16.

27 Ibid., p. 383, emphasis added.

28 Ibid., p. 395, note 63.

29 Ibid., p. 398.

30 Ibid., p. 395.

31 For example, Chernoff, ‘Scientific Realism as a Meta-Theory’, pp. 201–2 and 204.

32 The quote Wight offers is, ‘the question of which theory should be adopted is precisely a matter of determining which theory is predicatively the best theory’; Chernoff, ‘Scientific Realism as a Meta-theory’, p. 195, cited by Wight, ‘A Manifesto’, p. 395.

33 Chernoff, ‘Scientific Realism as a Meta-Theory’, p. 203.

34 While Duhem (1954) himself did not endorse ‘causality’, there is no inconsistency between his principle of conventionality of science and causal explanation. In The Power of International Theory, p. 112–3, I note that the addition of a causal component is parallel to the effort by Patomäki (a co-author of Wight) to add a causal component to British Institutionalist theories that were developed without such a component; see Patomäki, After International Relations: Critical realism and the (re)construction of world politics (London: Routledge, 2002).

35 Chernoff, The Power of International Theory, pp. 111–3. These defences of social causation aremost explicit in the section in chapter 4 entitled ‘convention and causation’, and the lastthree sections of chapter 4, where I defend my view from charges that social causation is impossible.

36 See Bas C. van Fraassen, The Scientific Image (New York: Oxford University Press, 1980).

37 Wight, ‘A Manifesto’, p. 394, note 57.

38 See Sumit R. Das, ‘Hawking Radiation in String Theory’, Journal of Astronomical Physics 20 (1999), pp. 131–48. Scientists similarly generally suspend belief regarding why the principle of least action holds. I thank Lee Arnold for suggestions on this point.

39 Wight may be conflating ‘unobservable objects’ with ‘theoretical objects’. As noted in section I, there are non-theoretical objects that are unobservable, such as ideas, love, anger, and for some philosophical views, universals (apart from their instantiations).

40 It is possible though unlikely that the superior theory that is adopted in the future will postulate the same theoretical entities.

41 Wight, ‘A Manifesto’, p. 388. We should note that an emphasis on the importance of ontology does not require the fallacious inference that this article explicates. Patrick Jackson has made a strong case for inquiry into a philosophical ontology, and argues that such an inquiry will lead to a different sort of social enquiry that we otherwise would embrace. See Patrick Thaddeus Jackson, ‘Foregrounding Ontology: Dualism, Monism, and IR Theory’, Review of International Studies (forthcoming).

42 Wight, Agents, Structures, p. 18.

43 Putnam says, ‘A realist (with respect to a given theory or discourse) holds that (1) the sentences of that theory or discourse are true or false; and that (2) what makes them true or false is something external – that is to say, not (in general) our sense data, actual or potential, or the structure of our minds, or our language, etc.’;

Hilary Putnam, ‘What Is Mathematical Truth?’ Mathematics, Matter and Method: Philosophical Papers,1, pp. 60–78. (London: Cambridge University Press), pp. 69–70.

44 Richard Boyd's characterisation of SR is, ‘the doctrine that the sort of evidence that ordinarily counts in favour of the acceptance of a scientific law or theory is, ordinarily, evidence for the (at least approximate) truth of the law or theory as an account of the causal relations obtaining between the entities quantified over in the law or theory in question’; Richard N. Boyd, ‘Realism, Underdetermination and a Causal Theory of Evidence’, Noûs 7 (1973), p. 1.

45 Agents, Structure ch. 1.

46 One of the first to note this was John Worrall, ‘Scientific Realism and Scientific Theory Change’, Philosophical Review, 32(1982), for example, pp. 225–9. Another important scientific realist, Jerrett Leplin soon thereafter observed that scientific realism is a majority position whose advocates are so divided as to appear a minority; see Jarett Leplin, ‘Introduction’, in Scientific realism, Jarett Leplin (ed.), (Berkeley and Los Angeles: University of California Press, 1984), p. 1.

47 In the next and final section in the chapter, where Wight applies his version of realism to the social sciences, half of all the footnotes (22 out of 44 that have references to publications) cite Bhaskar and another six cite Collier. Put another way, there are 42 references to Bhaskar (and six more to his disciple and interpreter, Collier) in the 39 pp. of the sections on scientific realism and social scientific realism (Wight, Agents, Structures, pp. 23–61). This is but one indication of the heavy reliance on Bhaskar.

48 Wight, Agents, Structures, p. 18.

49 Ibid., note 23.

50 Chernoff, ‘Scientific Realism as a Meta-theory’, p. 191. For a concise and insightful description of the genesis of critical realism in IR, see Chris Brown, ‘Situating Critical Realism’, Millennum: Journal of International Studies, 35 (2007), pp. 409–16.

51 This is precisely the argument of Karen Darling in ‘Motivational Realism: The Natural Classification for Pierre Duhem’, Philosophy of Science, 70 (2003), pp. 1125–3.

52 Wight, Agents, Structures, pp. 51–2. See Arthur Fine, ‘Unnatural Attitudes’, p. 156.

53 Some who oppose traditional naturalism in IR have argued that some of the problems, especially with inductive reasoning, can be circumvented by substituting a notion of ‘alternative future scenarios’ from that of ‘prediction’; see Steven Bernstein, Richard Ned Lebow, Janice Gross Stein and Steven Weber, ‘God Gave Physics the Easy Problems: adapting social science to an unpredictable world’, European Journal of International Relations, 6 (2000), pp. 43–76. This is an illuminating and helpful approach that offers insights into how theory may be used by policy-makers. But I have argued elsewhere (Chernoff, 2005, pp. 172–6) that it does not avoid the philosophical problems its authors believe it does.

54 Patomäki says, ‘Qualitative changes and emergence are possible, but predictions are not’, After International Relations, p. 157; cf. also p. 191. Wight says, ‘Criteria for rational confirmation and/or rejection of theories in the social sciences cannot be predictive […] ’ Agents, Structures, p. 52. And although Wendt, in Social Theory of International Relations, offers perhaps the most extensive treatment of meta-theories in the IR literature, ‘prediction’ is nowhere endorsed and is, in fact, barely even mentioned.

55 Wight, Agents, Structures, p. 50–51.

56 That is, the inconsistency arises if ‘human emancipation’ is intended as anything beyond a purely individual, intellectual, or emotional emancipation of the sort that people could enjoy whether they are living under liberal, authoritarian or fascist regime that permits them to read what they want.

57 I have argued elsewhere that community-wide agreement is possible in the social sciences (Chernoff ‘The Study of Democratic Peace’) and there may be theoretical debate where the contending theories postulate the same theoretical entities. But the vast majority of questions debated in IR fit neither category of exception.

58 Wight does acknowledge that current theories may turn out to be wrong (‘A Manifesto’, p. 384.). But he uses two misleading cases in drawing conclusions. The first is his critique of the ‘pessimistic meta-induction on the history of science’, where he defends SR from the charge that it is committed to the claim that all entities ever postulated by scientific theories must exist. No philosopher of science to my knowledge has advanced this claim (though in IR it has been raised). The standard philosophical argument is far stronger, viz., that scientific realists are forced to acknowledge that they have been wrong in drawing ontological conclusions from all previously accepted and now rejected past theories; so the probability is high that they will be wrong about our current theories. The second is that Wight believes that the existence of atoms, postulated by current theories, is very unlikely to be proved not to exist in current theories. But other entia, noted above, which are much less familiar to the reader and which have endured over far shorter time spans are postulated in current theories just as atoms are. So the postulation of atoms is not the most enlightening example in the context of a discussion of the confidence a reader might have in the long-term survival of many of the entities of current scientific theories into the distant future.

59 The ‘sogiston theory’ is specifically offered as an example of a bad theory in order to dramatise some aspects of the problems for SR and CR in IR; but the other aspects of the problem may be dramatised by strengthening the theory to make it less outrageous and more of a serious, but still-losing, competitor vis-ò-vis existing When we look at Wight's own description of the process of falsification, it is clear that he does not involve any recourse to the unobservable theoretical terms. The process is the same whether or not we believe in the ‘reality’ of the unobservable, theoretical, causal mechanisms. The different ways in which SR and anti-SR accounts treat theoretical terms have nothing to do with falsifiability theories of preventative war.

60 Chernoff, ‘Scientific Realism as a Meta-Theory.’

61 Chernoff, ‘The Power of International Theory’, ch. 5.

62 Wight (‘A Manifesto’, p. 394) says, ‘ … we need to move beyond sterile and confused epistemological debate and concentrate on what really divides the various theoretical positions; their competing claims about what the world is like, how it should be, and what the most important causal factors are. Ontology is at the heart of all our disagreements and that is what we should be concentrating our attention on.’ Wight continues, ‘There are important epistemological and methodological issues to be decided, but what is the point of debating these in advance of an examination of what it is we want to study’.