1

This article has been published in 2005 in Journal of Critical 4(1), 28-61.

The Trouble with Trancendental Arguments: Towards a Naturalization of ’s Early Realist

Tuukka Kaidesoja

Abstract. This article analyzes and criticizes the transcendental arguments Roy Bhaskar uses to justify his transcendental realist ontology. They are compared to Kant’s in the Critique of Pure Reason and a detailed reconstruction of those formulated in A Realist of is presented. It is argued that Bhaskar’s formulations contain certain ambiguities and are beset with other, more serious, problems. First, his descriptions of scientific practices (which form the premise of his transcendental arguments) are far more controversial than is presupposed in his arguments. Second, Bhaskar uses the Kantian of transcendental necessity in his transcendental arguments which inevitably connects them to certain Kantian doctrines that are inconsistent with his transcendental realist ontology. Some qualifications of his formulations Bhaskar made in his later writings are also considered and another possible interpretation of his transcendental arguments is presented from the point of view of these qualifications. On this interpretation, Bhaskar seeks to naturalize Kantian transcendental arguments by combining a posteriori premises with a priori philosophical reasoning. It is argued that this kind of naturalized version of transcendental arguments is also problematic. Therefore, it is concluded that Bhaskar’s transcendental arguments fail to justify his transcendental realist ontology. Nevertheless, it might be possible to justify at least some of Bhaskar’s ontological claims from the perspective of meta-philosophical . It is maintained that this requires naturalization not just of transcendental arguments but of the whole transcendental realist ontology. One possible form of naturalistic argument that might replace Bhaskar’s problematic transcendental arguments is sketched. It is, however, admitted that, desirable though the naturalization of Bhaskar’s early ontology may be, more work needs to be done to achieve this goal.

Keywords: Bhaskar, meta-, naturalism, ontology, transcendental argument

Introduction

Bhaskar justifies his transcendental realist ontology in RTS1 by using transcendental arguments. He also deploys them in PN to derive his specific social ontology which is compatible with transcendental realist ontology. These ontological are important for Bhaskar, because his realist theories of the natural and human presuppose the . Bhaskar’s transcendental arguments have been very influential in the critical realist tradition, where they are taken to provide a quite strong warrant for his ontologies.2 In this paper I argue that Bhaskar’s transcendental arguments are beset with problems in of which they fail to justify the ontologies. I also point out that it might be possible to a naturalistic

1 I refer to Bhaskar’s books by the following abbreviations: PN (The Possibility of Naturalism, Sussex: The Harvester Press, 1979); RR (Reclaiming , and New York: Verso); RTS (A Realist Theory of Science, Sussex: The Harvester Press, 2nd edition, 1979); SRHE (Scientific Realism & Human Emancipation, London and New York: Verso, 1989. 2 Bhaskar’s transcendental arguments are cited approvingly, for example, in Andrew Collier, Critical Realism: An Introduction to Roy Bhaskar’s Philosophy, London and New York: Verso, 1994; B. Danermark, M. Ekström, L. Jakobsen, & J.C. Karlsson, Explaining Society: Critical Realism in the Social Sciences, Routledge: London and New York, 2002; Jeffrey C. Isaac, ‘Realism and reality: some realistic reconsiderations’, Journal for the Theory of Social Behaviour, vol. 20, no. 1, 1990, pp. 1-31; , Economics and Reality, London: Routledge, 1997. 2 argument to warrant at least some of the most important ontological claims Bhaskar formulates. I argue that this requires that the whole transcendental realist ontology be reinterpreted from the perspective of meta- philosophical naturalism.

In order to show why Bhaskar’s transcendental arguments are problematic, it is necessary to relate them to ’s transcendental philosophy, which is the source of the doctrine of transcendental arguments. I accordingly analyze Kant’s use of transcendental arguments in the Critique of Pure Reason. There they are tied tightly to the doctrine of transcendental , hence can adequately be understood only in the framework of that doctrine. I also briefly compare Kant’s transcendental arguments to Bhaskar’s, then analyze the latter more thoroughly. I point out that Bhaskar’s manner of articulating transcendental arguments incorporates some ambiguities and argue that there are in addition serious problems in their conceptualization in RTS. First, he does not clearly distinguish scientific practices from their descriptions, hence seems to assume that his descriptions, which form the premises of his transcendental arguments, are unproblematic and that he is able to establish the of these descriptions without recourse to empirical analysis of the practices. I argue that, to the contrary, his descriptions of certain scientific practices are fallible and can be justified only by empirical analysis of such practices. I illustrate this point further by showing that his description of experimentation in science is controversial. Second, I point out that he assumes that he is able to justify a priori3 the claim that it is a necessary condition of the possibility of certain scientific practices (e.g. experimentation) that the ontological structure of the world contain certain features (e.g. structures and mechanisms which are ontologically independent of any of events). I argue that Bhaskar’s transcendental argument requires that the concept of necessity in this claim should be interpreted in a Kantian way. I also provide textual evidence for this interpretation by citing his postscript to RTS and PN, where he explicitly uses the Kantian term ‘transcendental necessity’ in this context. I argue that this Kantian concept of necessity connects Bhaskar’s transcendental arguments to certain Kantian doctrines that are inconsistent with his transcendental realist ontology. I also point out that without the Kantian concept of transcendental necessity Bhaskar’s transcendental arguments break down. Therefore, I conclude that Bhaskar’s transcendental arguments in RTS are problematic and fail to justify his transcendental realist ontology.

Bhaskar, however, offers some qualifications of his transcendental arguments in his later texts (e.g. in the postscript to RTS, and in PN and SRHE). I point out that there are serious terminological ambiguities in these qualifications, because he adopts Kantian terminology without clearly specifying what he means by it. I also present a modified interpretation of his transcendental arguments from the point of view of these qualifications. On this interpretation, he is trying to naturalize Kantian transcendental arguments by combining a posteriori premises with a priori philosophical reasoning. I argue that this kind of ‘naturalization of transcendental arguments’ is not tenable, because the concept of necessity is problematic in them too. I also point out that scientific practices are in always compatible with two or more incompatible ontological interpretations and it is not possible to demonstrate a priori that one of these interpretations is true and others false. Furthermore, I show that the attempt to naturalize transcendental arguments can be criticized from both the Kantian and the naturalistic perspective. My conclusion is that, notwithstanding the qualifications, Bhaskar’s transcendental arguments remain problematic. I then try to show that at least some of the doctrines of Bhaskar’s realist ontology can nonetheless be saved by naturalizing the whole transcendental realist ontology. By this I mean that it is possible to reinterpret and justify such an ontology from the perspective of meta-philosophical naturalism. I present some general claims of the programme of

3 By the term ‘a priori’ I mean (1) claims or propositions of which truth can be known independently of any and (2) arguments or justifications whose validity does not depend on experience. By the term ‘a posteriori’ I mean (1) knowledge claims or propositions of which truth cannot be known without recourse to experience and (2) arguments or justifications which validity depend on experience. 3 meta-philosophical naturalism. These include, for example, a denial of the possibility of a priori philosophical theorizing and foundationalist philosophical doctrines and an emphasis on the relationship between science and philosophy as continuous and on philosophical doctrines as fallible, just like scientific theories. I mention some critiques that have been directed against some forms of philosophical naturalism. Despite the critiques, I maintain that the programme of meta-philosophical naturalism can be applied fruitfully at least in the context of philosophical ontology (or ). Then I sketch the form of one possible naturalistic argument that can be developed to justify naturalized critical realist ontology. I try to show that these kinds of naturalistic arguments are more promising than Bhaskar’s problematic transcendental arguments. However, much more work remains to be done to naturalize critical realist ontology.

Bhaskar’s Transcendental Realism and Kant’s Transcendental Idealism

Bhaskar’s aim in RTS is to present a realist theory of science which is grounded in a transcendental realist philosophical ontology. His transcendental realist ontology should be distinguished from both external realism and scientific realism, although he defends both of these doctrines too. According to John Searle, external realism is a view that ‘the world (or alternatively, reality or universe) exists independently of our representations of it’.4 This is quite a minimal formulation of ontological realism, because it does not say anything about the entities that populate the world or about the possibility of acquiring knowledge about these entities. It only says that reality is not dependent on our ways of representing it. This is also a very general formulation, because the concept of representation includes, for example, , , beliefs, , descriptions, pictures, maps, languages and knowledge. The central claims of scientific realism are that theoretical entities (e.g. quarks, , electromagnetic fields, mental states, social structures) are real (exist) and that it is possible to acquire fallible theoretical knowledge about them.5 In addition to these philosophical doctrines, Bhaskar’s transcendental realist ontology includes certain specific claims about the ontological structure of the world. It says, for example, that structures and generative mechanisms exist independently of any patterns of events and experience. Bhaskar also defends a view that natural necessity, natural kinds and the causal powers of things are real features of the world; their is not dependent on our experience and knowledge of them. These are stronger and more specific claims about the ontological structure of reality than external and scientific realists usually defend.

Bhaskar uses transcendental arguments to justify his transcendental realist ontology. Transcendental arguments in RTS concern the necessary conditions of possibility of certain historically located natural scientific practices, and their conclusions refer to the ontological structure of the world. Bhaskar writes that ‘Philosophical ontology asks what the world must be like for science to be possible; and its premises are generally recognized scientific activities. Its method is transcendental; its premise science; its conclusion the of our present investigation.’6 He uses inter alia scientific experimentation, scientific criticism, practical use of scientific knowledge, scientific change, scientific classification and scientific education as premises of his transcendental arguments. He maintains that his transcendental realist philosophical ontology does not presuppose the existence of any particular entities that are postulated by substantive scientific theories.7 It concerns rather the general categorical structure of the world that is interpreted in a realistic way.

4 4 John R. Searle, The Construction of Social Reality, London: Penguin, 1995, p. 150. 5 There are, however, many different kinds of scientific realism developed and defended in the philosophical literature. Recent formulations include, for example, Ilkka Niiniluoto, Critical Scientific Realism, Oxford: Oxford University Press, 1999, and , ‘Scientific realism’, Stanford Encyclopedia of Philosophy (http.//plato.standford.edu/entries/scientificrealism/, 2003). 6 Bhaskar, RTS, p. 36. 7 Ibid., pp. 29-30. 4

It is important to notice that Bhaskar’s way of using transcendental arguments in RTS is quite different from Kant’s transcendental argumentation. Kant’s famous transcendental arguments8 in the first Critique concern the necessary conditions of possibility of cognitive experience. Kant maintains that it is possible to establish by their means that certain forms of sensible intuition ( and space) and certain categories of understanding (e.g. , substance, possibility, reality) are necessary conditions of any possible experience. As such, they are fixed and (i.e. they do not change and they are the same for every human ). Kantian categories may best be understood as kinds of formal rules which impart form to our experience. Kant argues that, in addition to categories of understanding, experience also requires sensible intuitions which give content to experience. He points out that all of our sensible intuitions are always spatial and temporal, but he denies that time and space exist independently of us. Consequently, for Kant time and space are the features of our sensibility that determine the form in which sensible intuitions are given to us.9

Kant emphasizes that these necessary conditions of the possibility of experience do not concern the ‘fundamental’ ontological structure of reality which supposedly exists independently of our sensibility and understanding. They refer rather to the forms of our sensible intuition and the organizing of our understanding that partly constitute the objects of our experience. In other words, Kant thinks that the objects of our cognitive experience (and our empirical knowledge) have to conform to our ways of cognizing (or knowing) them, not vice versa. Nevertheless, he admits that there are also things-in-themselves (Dinge an sich) which are not constituted by our sensibility and understanding. It is, however, controversial how Kant’s claims about things-in-themselves should be understood. There are at least two incompatible interpretations concerning their epistemological and the ontological status, namely the so called ‘two worlds interpretation’ and the ‘two perspectives (or logical) interpretation’.10

According to the two worlds interpretation, things-in-themselves exist independently of the objects of experience. Thus, there is a noumenal world of things-in-themselves behind a phenomenal world of objects of experience. We cannot know anything about things-in-themselves except that they exist and have some kind of effect on the objects of our experience. From this point of view, the of the effect of things-in- themselves on the objects of experience and, more generally, the relationship between these two supposedly different worlds in Kant’s philosophy becomes problematic. It is, for example, problematic to say that things- in-themselves cause the objects of our experience, because the concept of causality is one of the categories of understanding and, thus, cannot be applied to things-in-themselves which transcend the domain of possible experience. The two perspectives interpretation denies that things-in-themselves exist totally independently of the objects of experience. It says that things-in-themselves are rather the objects of experience considered independently of the conditions of cognizing them. So, according to this interpretation, there is only one world of things which is analyzed from two different perspectives. The attractiveness of this interpretation is that it avoids many of the intractable problems of the two worlds interpretation. However, I not go into details of this debate here, because this problem of interpretation—as important as it is in understanding Kant’s philosophy—is not relevant to the ensuing arguments. What should be emphasized here is that Kant forcefully rejects application of the categories of

8 In Kant does not use the term ‘transcendental argument’ at all. He talks instead of ‘transcendental proofs’ and ‘transcendental deductions’. However, for the sake of simplicity, I will use the term ‘transcendental argument’ to refer to these both. 9 Immanuel Kant, Critique of Pure Reason, trs. by Werner S. Pluhar, Indianapolis/ Cambridge: Hackett, 1996. 10 For more about these interpretations see, for example, M. Leppäkoski, The Transcendental How: Kant’s Deduction of Objective Cognition, Stockholm Studies in Philosophy 13, Almqvist & Wiksell International: Stockholm, 1993, pp. 159, 170-72. 5 understanding to all metaphysical speculations about transcendent (i.e. super sensible) things, because he maintains that these categories are applicable only to the possible objects of experience.11

Now, from the point of view of the arguments I go on to develop, there are two crucial relevant differences between Kant and Bhaskar. First, Kant thinks that the categories are some kind of formal rules or organizing principles of our understanding. He argues that there is no world existing independently of our understanding that corresponds to the categories of understanding, because these categories partly constitute the objects of our experience that are the only possible objects of cognition (and empirical knowledge) for us (i.e. human ). In this sense categories for Kant are subject-sided.1212 Bhaskar thinks by contrast that a real categorical structure of the world exists independently of our experiences and historical conceptualizations of that world. Thus, categories for Bhaskar are object-sided. In other words, they are real features of the world, which exist independently of our experience and knowledge. Bhaskar, however, grants that historical conceptualizations of these categories are subject-sided (or transitive). Second, Kant maintains that the domain of scientific knowledge in empirical sciences is limited to the possible objects of experience. Bhaskar thinks by contrast that the objects of scientific research are unobservable structures and mechanisms that generate events or phenomena which may become the objects of our experiences. In other words, according to his transcendental realism, scientific knowledge about unobservable (or transcendent) structures is possible. Therefore, he denies that there exist any things-in-themselves which cannot in principle be the objects of scientific knowledge. Nevertheless, he acknowledges that it is possible that some entities will always remain unknown to us, because our cognitive capacities and material resources for scientific investigations are limited.

It is important to notice that Kant thinks it possible to justify a priori propositions about the objects of our experiences precisely because our understanding partly constitutes the objects of our experience. It may be said that these synthetic a priori propositions refer both to the structures of our understanding and to the objects of our experience at the same time. This feature of Kantian transcendental arguments ties them tightly to the doctrine of transcendental idealism. Transcendental idealism says that the objects of our cognitive experience (and knowledge a posteriori) are always appearances (or constituted objects), not things-in-themselves.13Kant maintains that it is possible to justify certain propositions concerning the objects of our experience without recourse to experience, because these propositions concern such features of the objects of experience that our understanding has imposed on them. Therefore, the framework of transcendental idealism has to be presupposed if one tries to justify a particular synthetic a priori proposition by using transcendental argument.

Kant also makes it very clear that it is impossible to combine empirical claims and transcendental philosophy. He maintains transcendental philosophy always proceeds a priori; it’s premises, justifications and conclusions are not based in experience. Transcendental philosophy consists rather in the a priori operations of pure reason. These operations are a kind of intellectual probing of the forms of our sensible intuition and the formal structures of our understanding that make our cognitive experience possible.

Accordingly, Kant distinguishes the empirical subject from the transcendental subject. While the empirical subject is the object of empirical psychology, the transcendental subject is the object of transcendental philosophy. Therefore, he asserts that transcendental philosophy has to be carefully distinguished from the empirical sciences: all empirical claims (including claims of empirical psychology) are only hypothetical,

11 E.g. Kant, Critique A, pp. vii-viii (‘A’ refers the first edition of Critique of Pure Reason and ‘B’ to the second.) 12 I think Kantian categories cannot be adequately characterized as subjective, because they are universal. For this reason I use the terms ‘subject-sided’ and ‘object-sided’ here. These terms are borrowed from Paul Hoyningen-Huene, Reconstructing Scientific Revolutions: Thomas S. Kuhn’s , Chicago: The University of Chicago Press, 1993. 13 E.g. Kant, Critique A, pp. A369-A373. 6 whereas there is no room for hypothesis in transcendental philosophy. Propositions of transcendental philosophy for Kant are apodeictic necessities, not hypothetical claims.14

Even though Bhaskar seems to be aware of these features of Kant’s transcendental arguments, he still thinks that it is possible to justify his transcendental realist ontology by using Kantian transcendental arguments. His adoption of Kantian transcendental arguments is rather surprising, because transcendental realist ontology is incompatible with transcendental idealism and transcendental idealism is, as mentioned before, the framework that is presupposed in Kant’s transcendental arguments. Transcendental realism is also one of the doctrines that Kant forcefully criticizes.

Bhaskar’s transcendental arguments in RTS

The premises of Bhaskar’s transcendental arguments in RTS are generally recognized natural scientific practices and the conclusions of these arguments refer to supposedly real features of the world. The is to show that certain features of the world are the necessary conditions of the possibility of certain scientific practices. Bhaskar does not say that the existence of scientific practice is a necessity. He rather states that it is contingent, but since certain scientific practices do in fact exist, it is possible to argue that there have to be certain features in the world that make these practices possible. His aim is also to criticize competing of science. He tries to show that certain scientific practices would be impossible or unintelligible if the world really were like these philosophies of science (explicitly or implicitly) claim it is.15 So I think his arguments in RTS can be reconstructed roughly as follows:

1. X is generally recognized natural scientific practice. 2. It is a necessary condition of the possibility of X that the world is P1, . . .,Pn. 3. X is possible because it is real. 4. If the world were Q1, . . .,Qn, as is presupposed in competing philosophies of science, then X would be impossible or unintelligible. 5. Therefore, it is conditionally (i.e. given that X exists) necessary that the world is P1, . . .,Pn. The transcendental arguments that Bhaskar presents in RTS do not exactly follow this reconstruction. He does not, for example, clearly distinguish between the conditions of intelligibility, the conditions of and the conditions of possibility of scientific practices. Nevertheless, I think his arguments should be reconstructed as concerning the conditions of possibility of scientific practices, for two reasons. First, the conclusion (5) is plausible only if we know that practice X is impossible without P1, . . .,Pn. The intelligibility and rationality of the practice X concern our conceptions and judgements concerning it rather than the features of the world that make it possible. Second, Kant’s transcendental arguments concern the necessary conditions of possibility (not intelligibility nor rationality) of cognitive experience. It seems to me that the intelligibility and rationality of the claim that ‘experience is the source of all objective cognition (and knowledge a posteriori)’ is rather presupposed than justified in Kantian transcendental arguments. Therefore, the aim of Kant’s arguments seems to be to demonstrate how cognitive experience is possible by establishing it’s necessary conditions of possibility. Accordingly, in his transcendental arguments in RTS Bhaskar seems to rather presuppose than justify that scientific practice X is intelligible and rational. Therefore, the aim of his arguments seems to be to demonstrate how certain intelligible and rational scientific practices are possible by establishing their necessary conditions of possibility.

14 E.g. Kant, Critique A, pp. xi-xxii. See also preface to I. Kant, Critique of , trs. by Werner S. Pluhar, Indianapolis/Cambridge: Hackett 2002; and Leppäkoski, The Transcendental How, pp. 21, 62-92. 15 Bhaskar, RTS, pp. 23-4, 29-30, 36, 106-7, 116-17. 7

Bhaskar’s transcendental arguments in RTS display some additional ambiguities. First, he does not clearly distinguish the description (or the concept) of the practice X from the practice X itself. On the contrary, he seems to think, at least in some contexts, that the concept of the practice It should be noted here that Bhaskar’s way of articulating this argument in RTS is ambiguous, because he does not clearly distinguish statements X and the practice X are the same thing.16 Second, he also does not make a clear distinction between conceptually mediated ontological presuppositions (which describe the ontological features of the world) and the ontological features of the world. Sometimes he seems to claim, for example, that it is possible to produce knowledge about reality solely by analyzing .17 Third, he does not clearly enough specify in what way the conclusions of his arguments are necessary. He writes, for example, that philosophical ontology does not provide ‘a set of necessary about a mysterious underlying [meta-?] physical realm, but [. . .] a set of conditionally necessary truths about our ordinary world that is investigated by science’.18 It is not at all clear how the of ‘conditionally necessary truth about the world’ should be understood and how it differs from ‘fallible claim about the world’. I have incorporated these ambiguities into my reconstruction of Bhaskar’s transcendental arguments, because, as I shall point out, they are closely linked to certain problems in them.

Bhaskar argues, for example, that scientific experimentation is possible (or intelligible) only if generative mechanisms (or causal laws) are ontologically independent from patterns of events. He also states that the ontological basis of causal laws are the causal powers of things which they possess necessarily due to their essential intrinsic structures. Furthermore, he argues that experimentation is not possible (or intelligible) unless the world consists of open systems, where constant conjunctions of events (or empirical regularities) do not occur, and unless scientists are causal agents participating in the production of constant conjunctions of events in experiments. He also criticizes and neo-Kantian transcendental idealism inter alia for their commitment to the doctrine of empirical realism which denies the existence of unobservable generative mechanisms and structures. This doctrine includes the following ontological presuppositions: (i) constant conjunctions of events are necessary for the existence of causal laws; and (ii) the world consists of closed systems where constant conjunctions of events prevail. Bhaskar argues that because empirical realists deny the existence of unobservable structures and mechanisms, they cannot establish the possibility (or intelligibility) of scientific experiments.19

According to my reconstruction, the practice X in this argument is scientific experimentation; P1 claims that events are ontologically distinct from unobservable generative mechanisms; P2 claims that things possess causal powers; P3 claims that world consists of open systems; and P4 claims that scientists are causal agents. Competing philosophical positions in RTS are positivism and neo-Kantian transcendental idealism. From the point of view of experimentation, the common ontological presupposition (Q1) of positivism and transcendental idealism is, according to Bhaskar, that they are both committed to the doctrine of empirical realism.

If we accept my reconstruction of Bhaskar’s transcendental arguments in RTS, then it is possible to show that these arguments are problematic. As mentioned earlier, Bhaskar does not clearly distinguish descriptions of scientific practices from the practices themselves. It seems to me that this leads him to overlook problems concerning his descriptions. It is quite obvious that the truth- of descriptions of scientific practices cannot be known a priori, because scientific practices are not operations of our understanding but activities

16 See, for example, ibid., pp. 23-4. 17 See, for example, ibid., pp. 23-4, 257-8; PN, pp. 8-10. 18 Bhaskar, RTS, p. 52. 19 19 Ibid., pp. 33-6, 45-62. It should be noted here that Bhaskar’s way of articulating this argument in RTS is ambiguous, because he does not clearly distinguish statements about causal laws from causal laws themselves. I think, however, that his argument can be reconstructed to incorporate this distinction quite easily. 8 of real people in world. It is not enough to refer solely to the conceptions of other philosophers of science to establish these descriptions either, because their conceptions of scientific practices differ and their descriptions of these practices can be misleading or false. For example, advocates of and advocates of scientific realism describe the function of scientific experiments quite differently. The same can be said about scientists’ conceptions of scientific practices. Conceptions of what they are doing in some particular scientific practice might be quite different. Moreover, it is possible for such conceptions to be misleading or even false, as Bhaskar is the first to admit.20 Scientists influenced by Popperian falsificationism might, for example, describe their practices along the lines of Popper’s philosophy of science, while their practices in reality might be quite different from the descriptions they give. Therefore, the truth of the description of a certain scientific practice X can be known only a posteriori. In other words, descriptions concerning scientific practices are always hypothetical and hence fallible. It follows that the description of a certain scientific practice X can be justified only by analyzing this practice empirically. It seems to me that Bhaskar is quite reluctant to admit these , because they surely weaken the warrant of the conclusions of his arguments.

To push this point further, it can be argued that Bhaskar’s description of scientific experiments is one-sided, because it deals only with experiments in physics and chemistry.21 It totally ignores experiments in, for example, the life sciences, medicine and psychology. In these sciences it is not usually possible to isolate the mechanism under study from the other active mechanisms and build closed systems (in Bhaskar’s sense) where constant conjunctions of events occur. Therefore, experiments in these sciences are usually quite different from those in physics and chemistry. In their most basic form, experiments in the life sciences, medicine and psychology are conducted by using a test and a control group. Members of the groups are chosen by the experimenter so that they posses in relevant respects similar causal histories before the experiment. Then members of the test groups are somehow manipulated by the experimenter, while members of the control group are not. Other relevant conditions and factors are held similar in both groups during the experiment. After the experiment, the changes in members of the control group are compared with the changes in members of the test group. These kinds of experiments help scientists to create and test theoretical hypothesis about the different kinds of causal mechanisms that might have produced the observed differences.2222 Bhaskar says nothing about these kinds of experiments, which are common in many sciences. Moreover, it can be argued that his descriptions of the experimental practices in physics and chemistry are oversimplified. Experimentation has been the focus of many studies in recent philosophy of science and science studies23. I think that these studies show that Bhaskar’s descriptions of scientific experiments are in need of certain specifications and revisions.

In addition to the difficulties concerning the description of the practice X in premise (1), the concept of necessity in premise (2) and in the conclusion (5) is also problematic. Premise (2) is crucial for Bhaskar’s transcendental arguments. The conclusion follows only if he manages to justify the claim that ‘the world is

20 E.g. Bhaskar, RTS, pp. 260-1; PN p. 20. 21 Ted Benton has also pointed out that Bhaskar’s conception of natural sciences is far too restricted. He also argues that this leads Bhaskar (contra his own claims) to adopt an anti-naturalistic rather than naturalistic conception of the relationship between natural and human worlds and, accordingly, between natural and human sciences. See Ted Benton, ‘Realism and social science: some comments on Roy Bhaskar’s The Possibility of Naturalism’, Radical Philosophy 27, 1981, pp. 13-21. 22 For more about these kinds of experiments see e.g. Lawson, Economics and Reality, pp. 205-6; and Hugh Coolican, Research Methods and Statistics in Psychology, London: Hodder & Stoughton, 4th edition, 2004. 23 Philosophical studies in experimental physics are, for example, , Representing and Intervening, Cambridge: Cambridge University Press, 1983; and Peter Galison, How Experiments End, Chicago and London: The University of Chicago Press, 1987; In the field of there are also many studies concerning experiments in physics and in chemistry, see e.g. Mario Biagioli, ed., The Science Studies Reader, New York and London: Routledge, 1999. 9

P1, . . .,Pn’ really is a necessary condition of the practice X. Now it may be asked: ‘What is the precise of the concept of necessity in premise (2)?’ In his postscript to RTS and in PN Bhaskar uses the term ‘transcendental necessity’ in this context.24 This term clearly connects his transcendental arguments to Kant’s transcendental argumentation. Therefore, it seems that he is using in premise (2) the Kantian concept of transcendental necessity. 25 The concept of transcendental necessity in Kant’s transcendental arguments includes at least two assumptions: (i) Transcendental necessities do not depend on our experience. In other words, propositions concerning transcendental necessities are knowledge a priori. (ii) Transcendental necessities are apodeictic necessities. In other words, if some proposition is a valid description of some transcendental necessity, then it cannot be reasonably doubted. Transcendental necessities are the features of the empirical world that our understanding imposes on it and propositions concerning these necessities are knowledge synthetic a priori. Thus, Kant stresses that propositions about transcendental necessities cannot be justified from an empirical perspective but only from the point of view of transcendental philosophy.26

If Bhaskar uses the Kantian concept of transcendental necessity in premise (2) of his arguments, then it is presupposed in his arguments that the truth of premise (2) is justified a priori. It follows that premise (2) is knowledge a priori and, thus, its truth does not depend on our experience or our empirical knowledge concerning the practice X. Now Bhaskar faces a dilemma. On the one hand, if he wants to justify premise (2) a priori, he has to commit himself to transcendental idealism. As I argued earlier, the possibility of justifying a priori propositions concerning reality (i.e. knowledge synthetic a priori) requires commitment to the doctrine of transcendental idealism. This route is not available to Bhaskar, because the whole point of his transcendental arguments is to defend transcendental realist ontology that is incompatible with transcendental idealism. On the other hand, if Bhaskar gives up the claim that premise (2) can be justified a priori, then his transcendental argument breaks down. It is not possible to justify a posteriori any propositions about transcendental necessities in the Kantian sense, because knowledge a posteriori is always merely hypothetical and hence fallible.

If we accept, for the sake of the argument, that Bhaskar somehow manages to give a priori justification to premise (2), then premise (4) becomes redundant. If we know a priori that P1, . . .,Pn really are the necessary conditions for the practice X, then the critique of claims Q 1, . . .,Q n does not carry any weight in Bhaskar’s arguments. In other words, the warrant of the conclusion (5) does not change, if premise (4) is removed from the argument. Therefore it is quite surprising how carefully Bhaskar criticizes competing ontological claims in RTS. It seems that he implicitly wants to give more weight to premise (4) than he is explicitly ready to admit. However, if more weight is given to premise (4), then premise (2) has to be modified or at least interpreted differently.

To summarize. Even though Bhaskar distinguishes his transcendental arguments from Kant’s in RTS, he still wants to retain some a prioristic elements from Kantian arguments. I have argued that these a prioristic elements (e.g. failure to distinguish descriptions from their objects and the concept of Kantian transcendental necessity) are problematic and that Bhaskar’s transcendental arguments in RTS, therefore, fail to justify his transcendental realist ontology. I have also pointed out that the problems of Bhaskar’s transcendental

24 Bhaskar, RTS, p. 259; PN, p. 7. 25 This Kantian interpretation can be supported by further references to Bhaskar’stexts. Bhaskar writes in RTS (p. 24) that ‘the function of philosophy is to analyze concepts which are “already given” but “as confused”’ and mentions Kant as a pioneer of this kind of philosophical . He also claims in RTS (p. 247) and in PN (p. 8) that the conclusions of his study are apodeictic. In PN (p. 6) Bhaskar writes also that ‘If philosophy is to be possible (and I want to contend that it is in practice indispensable) then it must follow the Kantian road.’ Furthermore, he states in PN (p. 8) that philosophy considers the same world as the sciences ‘from the standpoint of what can be established about it [i.e. the world] by a priori argument, where it takes as its premises generally recognized activities as conceptualized in experience’. 26 For the nature of Kantian transcendental necessity see Leppäkoski, Transcendental How, pp. 255-60. 10 arguments are mainly due to his attempt to combine two incompatible doctrines: Kantian transcendental arguments (which presuppose the framework of transcendental idealism) and transcendental realism. Bhaskar’s problematic adherence to Kantian a priorism is also incompatible with meta-philosophical naturalism and brings his philosophical position very close to foundationalist first philosophies. In his later writings he slightly modifies this position. In the next section I briefly analyze the relevant later texts.

Bhaskar’s qualifications of his transcendental arguments

Bhaskar writes in the postscript to RTS, ‘I think that the transcendental method employed [in RTS ] is, in general, sound. It is, however, certainly the case that the book contains no adequate defense, or meta- philosophical justification of the latter [i.e. transcendental method].’27 In his later writings, he tries to specify and defend his ‘transcendental method’. I briefly list the most important qualifications he makes of his transcendental arguments. I also consider one possible way of interpreting his transcendental arguments from the point of view of these qualifications. However, I maintain that in the last analysis his qualified transcendental arguments and his philosophical position remain problematic.

Bhaskar’s most important qualifications include the following claims:

 Social practices other than scientific ones can function as a premise of a transcendental argument.28  The premises (i.e. concepts or descriptions of social practices) and the conclusions (i.e. propositions concerning the ontological structure of the world) of transcendental arguments are contingent and can be contested.29  The epistemological status of the conclusions of transcendental arguments is knowledge synthetic a priori, although knowledge synthetic a priori is interpreted in a historically relative, hypothetical and domainspecific way.30  Transcendental arguments always involve immanent critiques of competing philosophical ontologies.31  Transcendental arguments are a species of retroductive argument that is also used in science and in of other social practices.32 Even though Bhaskar admits that the premises of his transcendental arguments are contingent and can be contested, he still assumes that it is not necessary to refer to empirical analysis of the scientific practices in order to justify descriptions concerning these practices. As I argued earlier, this is a problematic assumption. Moreover, he still maintains that his arguments concern the necessary conditions of certain scientific practices and that the conclusions are somehow conditionally necessary (although potentially contested) truths about the world. Therefore, in spite of the above qualifications, the problems concerning the concept of ‘necessity’ largely remain.

One confusing feature of Bhaskar’s terminology that appeared already in RTS but becomes even more conspicuous in his later writings is that he uses many Kantian terms without clarifying their precise meaning. He writes, for example:

On the conception of philosophy at work in this book [i.e. RTS ] the ultimate premisses and the conclusions of philosophical considerations are contingent facts, the former (but not the

27 Bhaskar, RTS, pp. 258-9. 28 Bhaskar, PN, p. 10. 29 Bhaskar, RTS, pp. 259-60; PN, pp. 6-7; SRHE, pp. 11-12; RR, p. 14, 183. 30 Bhaskar, RTS, p. 259; PN, pp. 6-7; RR, p. 14. 31 Bhaskar, RTS, pp. 259-260; PN, pp. 7-8; SRHE, pp. 14-15; RR, p. 182. 32 Bhaskar, SRHE, p. 11. 11

latter) necessarily social and so historical. It is only in this relative or conditional sense that philosophy can establish synthetic a priori truths (truths about world investigated by science). Philosophy operates by the use of pure reason. But it does not operate by the use of pure reason alone. For it exercises that reason always on the basis of prior conceptualizations of historical practice, of some more or less determinate social form.33

What is presupposed in any given scientific activity is at once a possible object of scientific explanation; so that what is apodeictically demonstrable is also scientifically comprehensible; that is, what is synthetic a priori is also (contingently) knowable a posteriori.34

Bhaskar also talks about demonstration and deduction of the transcendental necessity of certain conditions of the possibility of certain scientific practices.35 If these Kantian terms are taken seriously, then I think his position is either incoherent or incomprehensible. If it is interpreted according to the meanings that Kant gave to his special philosophical terms, it is incoherent. These Kantian concepts have their proper meaning only in the context of transcendental idealism and Bhaskar’s transcendental realism is incompatible with this Kantian doctrine. If Bhaskar gives the Kantian terms different meanings from Kant, his are incomprehensible, because he has not specified clearly enough what he means by these terms. It is not enough to claim, for example, that ‘synthetic a priori truths’ are historical, relative and contingent claims, if the terms ‘synthetic’ and ‘a priori’ are not specified, because the original Kantian term ‘synthetic a priori proposition’ refers precisely to propositions that are ahistorical, universal and transcendentally necessary.

I will ignore these serious terminological ambiguities and previously mentioned problems concerning Bhaskar’s descriptions of scientific practices for a moment and consider one possible interpretation of his arguments from the point of view of the above qualifications. His arguments might be interpreted as ‘naturalized transcendental arguments’. By this I mean arguments that combine a posteriori premises and a priori transcendental reasoning. On this interpretation, premise (1) of Bhaskar’s transcendental argument is understood as an a posteriori and, hence, fallible description of certain historical practice X. The in premise (2) from the practice X under our description to the necessary conditions of it’s possibility is interpreted as a piece of a priori reasoning. In other words, it is assumed that it is possible to infer without recourse to experience that the practice X under our description is impossible unless the ontological structure of the world contains certain features that are described in propositions P1, . . .,Pn. This inference might be to be justified a priori, because propositions P1, . . .,Pn refer to the ontological structure of the world only conditionally. In other words, propositions P1, . . .,Pn in premise (2) do not refer directly to the ontological structure of the world, but they rather identify merely a space of possibilities and it is contingent whether these possibilities are ever actualized. Accordingly, premise (3) might be interpreted as an a posteriori claim which says that the practice X under our description is possible, because such a practice really exists and our a posteriori description of it is empirically adequate or confirmed. Then the conclusion (5) of the argument says that, given that our description of the practice X is empirically adequate, our claim that the world really is as described in propositions P1, . . .,Pn has acquired some kind of warrant. In other words, we have reason to believe that propositions P1, . . .,Pn do not refer to mere possibilities but to real features of the world.

I think that this naturalized transcendental argument is rather a reinterpretation than a mere interpretation of Bhaskar’s transcendental arguments. However, it is interesting to consider it more closely. It should be

33 Bhaskar, RTS, p. 259. Italics mine. See also PN, pp. 6-7. 34 Bhaskar, PN, p. 8. Italics mine. See also RTS, p. 259 and SRHE, pp. 12-13. 35 Bhaskar, PN, p. 7. 12 noted, first, that the previous argument requires that the term ‘a priori’ be relativized to the historical practice X, and the description of this practice is thought to be knowledge a posteriori. However, even if this kind of ‘naturalized transcendental argument’ could be presented sufficiently clearly, I think that this kind of halfway position between naturalism and is unviable. One reason for this is that the concept of necessity in premise (2) is still problematic. Nevertheless, for the sake of argument, I assume that problems concerning the interpretation of this concept can be solved. Now it may be asked, ‘How is it possible to demonstrate a priori the claim that propositions P1, . . .,Pn in premise (2) really are the necessary conditions of the possibility of some practice X under our description?’ It is surely possible to present a plausible hypothesis about necessary conditions of practice X under our description, but it is another thing to demonstrate a priori that this hypothesis is true. I doubt whether these kinds of a priori demonstrations are possible at all, because it is always possible that there are a set of propositions, say Z1, . . .,Zn, that hypothetically describes the necessary conditions of the practice X under our description and that is incompatible with propositions P1, . . .,Pn. In this situation there is no a priori way to prove that one set of propositions really describe the necessary conditions of the practice X and the other does not. In other words, scientific and other practices do not determine what kind of ontological interpretation we give to them and particular practice X under our description is always in principle compatible with two or more incompatible ontological interpretations. This not to say that all ontological interpretations of practice X under our description are equally plausible from the point of view of apriori considerations. There might be good a priori reasons for repudiating some interpretations, because they are, for example, internally incoherent. Moreover, I am not here denying the possibility of comparing different ontological interpretations of the practice X with the help of further empirical knowledge concerning that practice or by evaluating them in the light of current scientific ontology. What I am denying is the possibility of inferring a priori from the description of the practice X to the true description of its necessary conditions of possibility that conditionally refer to the ontological structure of the world.

In addition to this problem, the whole idea to naturalize transcendental arguments can also be questioned both from the point of view of Kantian transcendental philosophy and from the perspective of meta- philosophical naturalism. The term ‘transcendental argument’ in its Kantian meaning refers to arguments that produce knowledge about our way of cognizing (or knowing) reality. Knowledge about our way of cognizing reality is, according to Kant’s transcendental idealism, also knowledge about the objects of our cognitive experience, because our way of cognizing reality partly constitutes the objects of our cognitive experience. Kant also emphasizes that it is impossible to combine empirical claims that are based on cognitive experience (i.e. empirical knowledge) with transcendental arguments, because the latter concern the very possibility of our cognitive experience (and empirical knowledge). In other words, transcendental and empirical perspectives for Kant are totally distinct and irreconcilable. Therefore, the attempt to reconcile these perspectives in ‘naturalized transcendental arguments’ is incoherent from the Kantian perspective. I am not saying that the Kantian perspective is tenable (I think it is not). I only point out that the attempt to naturalize transcendental arguments is not coherent from the Kantian point of view.

From the perspective of meta-philosophical naturalism it is important to focus on the a prioristic and foundationalist features of Kantian transcendental philosophy, which have been seriously criticized by naturalist philosophers. It might be said that it is precisely the denial of a priorism and that distinguishes meta-philosophical naturalism from the tradition of transcendental idealism and transcendental philosophy more generally. Now, if transcendental arguments are naturalized, it might be asked, from the point of view of meta-philosophical naturalism, ‘What distinguishes naturalized transcendental arguments from scientific arguments or scientific explanations?’ If the supposedly a priori nature of these arguments is offered as an answer, then it may be asked, ‘What justifies this kind of a priori philosophical reasoning given that most conclusions of a priori arguments concerning the nature of reality have turned out to be false or problematic in the course of the ?’ On the other hand, if the 13 a priori status of transcendental arguments is denied (i.e. the premises and conclusion of these arguments are interpreted as knowledge a posteriori), then there is no reason to call them transcendental. In other words, the epistemological status and the logical structure of this kind of supposedly transcendental argument turn out to be exactly the same as in some scientific arguments or explanations. For these reasons the effort to naturalize transcendental arguments fails also from the perspective of metaphilosophical naturalism.

I have argued that Bhaskar’s attempt to justify his transcendental realist ontology by using Kantian transcendental arguments fails. The problems in his arguments are not solely terminological in the sense that defining some terms more precisely can solve them. On the contrary, I have argued that, besides serious terminological ambiguities in the presentation of his transcendental arguments, these arguments also contain certain Kantian presuppositions that are not compatible with his transcendental realist ontology. Therefore, his transcendental arguments fail to justify his ontological doctrines and, accordingly, Bhaskar’s qualified neo-Kantian philosophical position is incoherent. I have also argued that the attempt to naturalize transcendental arguments in the context of transcendental realist philosophical ontology is problematic, because practices, which form the premises of these supposedly ‘naturalized transcendental arguments’, can (in principle) always be interpreted from the point of view of two or more incompatible ontological theories and there is no a priori way to decide which interpretation is true.36

Naturalism as a Meta-Philosophy

Even if Bhaskar’s transcendental arguments fail to justify his transcendental realist ontology, it might be possible to develop a different kind of argument to support that ontology. I think the perspective of meta- philosophical naturalism offers a promising alternative to Bhaskar’s problematic neo- Kantian meta- philosophy. Before considering the possibility of naturalizing transcendental realist ontology, I present some general features of the programme of meta-philosophical naturalism. I also illustrate what this kind of naturalistic attitude to philosophy has meant in the context of philosophy of science and briefly discuss some arguments that have been directed against meta-philosophical naturalism.

Meta-philosophy deals with the following questions: What is the object of philosophical claims and theories? What is the method of philosophy? What is the epistemological status of philosophical claims and theories? What is the nature of philosophical problems? What is the relationship between philosophy and the other kinds of social practices (e.g. science, art and religion)? These questions have been answered differently in the course of the history of philosophy. Sometimes philosophers have dealt with these questions explicitly and sometimes they have answered them only implicitly by doing a certain kind of philosophy without further explicating their conception of philosophy. Meta-philosophical naturalism is one possible way to answer these questions. The roots of naturalistic philosophy in the broad sense of the term can be traced back to ancient Greece, especially to ’s philosophy. However, meta-philosophical naturalism in its present form is a quite new phenomenon.

36 Justin Cruickshank interprets critical realist ontology as a fallible and a conceptually mediated ontological theory that is developed via internal critiques of the other philosophical positions. From the perspective of the above arguments, Cruickshank’s interpretation is problematic for two reasons. First, he thinks that it is unproblematic to use transcendental arguments (combined with internal critiques) in justification of critical realist ontology. Second, he does not pay enough attention to the problems that concern justification of the descriptions of certain scientific practices. It is not enough to refer to internal critiques of the other philosophical positions in order to establish the description of certain scientific practice. See Justin Cruickshank, ‘Critical realism and critical philosophy’, Journal of Critical Realism vol. 1, no. 1, 2002, pp. 49-66. 14

Willard van Orman Quine is often mentioned as a founding father of the present form of meta-philosophical naturalism.37 Quine formulated his position mainly in the context of , but his meta-philosophical naturalism includes also ontology and .38 There are many followers of Quine in the field of .39 In recent philosophy, naturalistic positions have also been defended in philosophy of ,40 in philosophy of natural sciences41 and in philosophy of social sciences42. Furthermore, there are vital naturalistic traditions in science studies.43

However, the meaning of the term ‘naturalism’ is controversial, and it is used differently in different contexts. Here I will interpret naturalism as a meta-philosophical research programme that has been applied in different areas of philosophy. Therefore, my concern is not to present any classification of the different forms of naturalism. I shall instead formulate one possible general characterization of the programme of meta- philosophical naturalism and discuss briefly how it differs from transcendental philosophy and other anti- naturalistic meta-philosophies. I think that following claims define the programme of meta-philosophical naturalism clearly enough for the purpose of this paper:

1. Philosophical claims and theories concern the same world as scientific theories. So there is no special supernatural or transcendent metaphysical realm that is supposed to be the object of philosophical theorizing. 2. The methods of philosophy are not qualitatively different from the methods of science. It is wholly legitimate and often desirable to use scientific methods in philosophy. Scientists also use some methods (e.g conceptual analysis and thought experiments) that have sometimes deemed exclusively philosophical. Therefore, methodological between philosophy and science is only a of degree. 3. Philosophical claims and theories are a posteriori and hence fallible. There are no necessary a priori philosophical truths—not at least outside the realm of logical and mathematical truths. 4. There is no special class of philosophical problems. It is in principle possible that problems that have been traditionally deemed philosophical (e.g. mind-body problem) will turn out to be scientific problems in future. However, problems that philosophers usually deal with are often more abstract and general than the problems of the empirical sciences. It is obvious that these kinds of problems cannot be solved by reference to in any straightforward way.

37 See e.g. Kornblith, ed., Naturalizing Epistemology, Cambridge, Massachusetts & London, England: The MIT Press, 1994; Quine’s influential text about naturalized epistemology is Williard von Orman Quine, ‘Epistemology naturalized’, in W.V.O. Quine, Ontological Relativity and Other Essays, Cambridge: Harvard University Press, 1969. This text is also reprinted in Kornblithe, ed., Naturalizing Epistemology, pp. 15-31. 38 Recent study about Quine’s naturalistic meta-philosophical position is Heikki J. Koskinen, From Metaphilosophical point of view: A Study of W.V. Quine’s Naturalism, Acta Philosophica Fennica, vol. 74, Helsinki: Societas Philosophica Fennica, 2004. 39 See Kornblith, ed., Naturalizing Epistemology. 40 See David Papineau, Philosophical Naturalism, Oxford UK & Cambridge USA: Blackwell, 1993. 41 See e.g. Richard Boyd, ‘Scientific realism and naturalistic epistemology’, in PSA 1980, vol. 2, 1982, pp. 613-62. Werner Callebaut, Taking the Naturalist Turn or How Real Philosophy of Science Is Done, Chicago and London: The University of Chicago Press, 1993; Ronald N. Giere, ‘Philosophy of science naturalized’, Philosophy of Science, vol. 52, 1985, pp. 331- 56. Ronald N. Giere, Explaining Science, Chicago: University of Chicago Press, 1988; and , ‘Normative naturalism’, Philosophy of Science, vol. 57, 1990, pp. 44-59. 42 See Harold Kincaid, Philosophical Foundations of the Social Sciences, Cambridge: Cambridge University Press, 1996. 43 See Barry Barnes, David Bloor & John Henry, Scientific Knowledge: A Sociological Analysis, Chicago: The University of Chicago Press, 1996; Mario Biagioli, ed., The Science Studies Reader; Callebaut, Naturalist Turn; Michael E. Gorman, ‘Psychology of science’, in William O’Donahue & Richard F. Kitchener, eds, The , London: Publications, 1996, pp. 50-65. 15

5. The relationship between philosophy and science is continuous. Scientific knowledge is relevant for creating and testing philosophical theories that usually concern more general and abstract issues than scientific theories. 6. Meta-philosophical naturalism is a research programme in the same sense as scientific research programmes. Therefore it has no special a priori philosophical justification and its claims (especially the claim 3 above) and methods are also applicable to itself.

According to this kind of meta-philosophical naturalism, philosophers do not posses any privileged view of the ontological structure of reality, nor can they offer any certain foundation to our knowledge about the world. Furthermore, it is held that there is no special philosophical method that guarantees a priori the truth of philosophical claims and theories, which are rather deemed to be as fallible as scientific knowledge. This is not to say that philosophers should abandon conceptual analysis and thought experiments. It is rather to emphasize that conceptual analysis and thought experiments are also used in special sciences and are therefore not specifically philosophical methods. Meta-philosophical naturalists also consider that conceptual analysis and thought experiments in philosophy should be combined in some way or another with empirical research in order to evaluate the truth of their theories, because there is no a priori way to prove the truth of any philosophical proposition. Therefore, the naturalistic attitude to knowledge and philosophy is thoroughly anti-foundationalist. I think that coherent meta-philosophical naturalism should also include some kind of reflexivity requirement which states that it’s methods and doctrines are applicable to itself. In other words, claims concerning the nature of philosophy and philosophical claims about the nature of reality should be interpreted as fallible theories that are in principle susceptible to scientific explanation. I think that this kind of meta-philosophical naturalism is also compatible with external realism and scientific realism.44 Therefore, I believe that there is no necessary link between and naturalism although some proponents of naturalism are empiricist. It should also be emphasized that although the relationship between philosophy and empirical science is understood as a matter of degree in metaphilosophical naturalism, this does not necessarily mean that philosophy is reduced to science or replaced with science. There are, for example, many general ontological and epistemological problems that are not studied by any special science and that cannot be solved by referring to the results of empirical research in any direct way. Therefore, there is still room for philosophy even though metaphilosophical naturalists deny a hard and fast demarcation between philosophy and empirical science. Nevertheless, it should be noted that there are different views among naturalist philosophers about the legitimate methods of science and the exact relationship between philosophy and science.

I think the above claims differentiate naturalistic meta-philosophy effectively from the tradition of ‘first philosophy’. Proponents of that tradition believe that the main aim of philosophy is to lay the necessary grounds for the understanding of reality and the justification of our knowledge, especially scientific knowledge. Therefore, they think that philosophy forms the ground of all other sciences. In this tradition philosophical knowledge is also thought to be knowledge a priori and, thus, independent of the knowledge produced by the empirical sciences. In the course of the history of philosophy either metaphysics or epistemology have usually occupied a position of ‘first philosophy’. For example, Kant’s transcendental philosophy and Hegel’s absolute idealism clearly belong to this tradition. The current naturalistic trend in

44 For example, Richard Boyd writes, concerning the relationship between scientific realism and naturalism, that ‘scientific realism is, by the lights of most of its defenders, the science’s own philosophy of science. Considerations of the significant philosophical challenges which it faces indicate that it can be effectively defended only by the adoption of a metaphilosophical approach which is also closely tied to science, viz., some version or other of philosophical naturalism.’ See Boyd, ‘Scientific realism’. 16 philosophy can be understood as a reaction against which also belongs to the tradition of ‘first philosophy’.45

In the context of philosophy of science, the naturalistic attitude has meant that philosophers of science have abandoned all normative theories about the universal Method of Science that was thought (by logical positivists and ) to define the concept of (scientific) rationality. concerning the power of formal (e.g. first-order predicate calculus) to function as an a priori method has also increased among naturalist philosophers of science. From the point of view of naturalism, the scientific methods and the standards of scientific evaluation are usually seen as historically evolving, and philosophical theories about these methods and standards are thought to be fallible. The naturalistic turn in the philosophy of science has—at least arguably—also meant that science and philosophy have moved closer to each other than in the days of logical positivism and Popperian falsificationism. Naturalistic philosophers of science usually begin their work by carefully analyzing real scientific practices and only after that they present general philosophical theories of science and normative considerations about science. Some even think it impossible to formulate any interesting ‘general theory of science’. They think philosophers of science should rather focus their attention on some specific scientific practices and conceptual problems in some specific field of inquiry (e.g. quantum , evolutionary biology or psychology). Some advocates of naturalistic philosophy of science have also largely abandoned the prescriptive attitude towards scientific practice that was earlier thought appropriate for philosophers of science. The naturalistic trend is also closely connected to the emergence of naturalistic science studies in which scientific practices and research processes are studied with the methods of different sciences (e.g. history, anthropology, sociology, psychology, cognitive science, neurosciences, biology).46

Although meta-philosophical naturalism has been a quite influential position in many discussions of recent philosophy, it has also been criticized widely. It has been argued, for example, that naturalistic philosophy does not pay enough attention to the traditional normative concerns of philosophy. In the philosophy of science, for example, it has traditionally been held that philosophers of science should present prescriptive claims about how science should be done, not just descriptions of how science has actually been done. Many critics of naturalism have also claimed that the arguments of naturalists are viciously circular. It has been argued, for example, that naturalistic theories about science based on scientific research already presuppose the validity of certain scientific practices and methods. Therefore, according to the critics, it is viciously circular to defend certain philosophical theories about science by referring solely to the results of empirical science. Some anti-naturalist or culturalist philosophers also believe that does not pay enough attention to human , language, culture, moral values and existential questions. In addition, it is argued that naturalistic philosophy is hard to recognize as philosophy, because it gives up philosophical argumentation in the traditional sense. The most scientistic and non-normative forms of naturalism, like Quine’s doctrine of naturalistic epistemology, have encountered the severest criticism.47

45 For more about tradition of ‘first philosophy’ and it’s relation to philosophical naturalism see Jon Jacobs, ‘Naturalism’, The Internet Encyclopedia of Philosophy (http://www. iep.utm.edu/n/naturalism.htm/, 2002). Historical context of the current metaphilosophical naturalism in the context of epistemology is discussed in Kornblith, ed., Naturalizing Epistemology and in the context of philosophy of science in Callebaut, Naturalist Turn. 46 For more about naturalist turn in philosophy of science and the relationship between science studies and naturalism in recent philosophy of science see Callebaut, Naturalist Turn. Ronald N. Giere also tries to combine cognitive studies of science and philosophy of science see Giere, Explaining Science. Larry Laudan defends a position that combines naturalism and normative aspect of philosophical theorizing see Laudan, Normative Naturalism. 47 For criticism of Quine’s doctrine of naturalistic epistemology see Jaegwon Kim, ‘What is ‘naturalized epistemology’?’, in Kornblith, ed., Epistemology Naturalized, pp. 33- 55; and Harvey Siegel, ‘Naturalism and the abandonment of normativity’, in O’Donahue & Kitchener, eds, The Philosophy of Psychology, pp. 4-18. 17

These questions are under debate and it is not possible to decide here whether meta-philosophical naturalism can be defended against these and other critiques. It is not clear either where its limits lie. In other words, it remains an open question in what areas of philosophy meta-philosophical naturalism can be fruitfully applied. Nevertheless, it should be noted that normative forms of naturalism have been developed in the fields of epistemology48 and philosophy of science.49 Advocates of normative naturalism in philosophy of science usually think that naturalist philosophers are also able to present normative claims concerning certain actual scientific practices, but they think these claims can be justified only by referring to some other actual scientific practices, which have been successful, and to scientific theories about our cognitive capacities and social processes of knowledge acquisition. They also admit that there is no non-circular way to defend theories concerning the criteria to be used in evaluating the success of scientific theories and practices.

In what follows I will apply the perspective of meta-philosophical naturalism to the area of philosophical ontology (or metaphysics) and to the area of scientific evaluation. I think that meta-philosophical naturalism can best be defended against previous critiques in these contexts. If we accept anti-foundationalist and fallibilist views of all (including ontological) knowledge, then anti-naturalist philosophical ontology is impossible, because the idea of ontological knowledge synthetic a priori that is also fallible becomes incoherent. I also think that normative naturalism concerning evaluation in science is a more plausible position than the traditional philosophical theories about it which claim that there is some fixed set of a priori justified criteria that can be used in every evaluation in science. I think that this kind of anti-naturalist position has become implausible, because it has been pointed out in many studies concerning the history of science that the criteria of evaluation that have been actually used in science have changed in the course of the history of science. Nevertheless, there is not enough space here to offer any detailed arguments for these positions. I also leave it open whether the perspective of meta-philosophical naturalism can be reasonably applied to areas of philosophy (e.g. semantics, ) other than ontology and evaluation in science.

Naturalization of transcendental realist ontology?

It might be possible to save at least some of the conclusions of Bhaskar’s problematic transcendental arguments by naturalizing not just transcendental arguments but the whole transcendental realist ontology. By this I mean that the re-interpretation of transcendental realist ontology as a hypothetical and fallible theory about the general ontological structure of the world that is justified by using naturalistic arguments. These naturalistic arguments should no longer be called ‘transcendental’ and the ontology that is justified by these arguments should be called ‘naturalized critical realist ontology’ rather than ‘transcendental realist ontology’. I sketch here one possible form such naturalistic arguments might take. My suggestions are, however, tentative and I admit that more work is required to naturalize critical realist ontology. Nevertheless, I hope to show that naturalistic arguments offer a promising alternative to Bhaskar’s problematic transcendental arguments.

In bare outline, the general form of naturalistic argument that might be used in justifying naturalized critical realist ontology is as follows.

1. X is a successful natural scientific practice described on the basis of empirical analysis of the practice. 2. It is hypothetically and explanatorily a necessary condition of the successfulness of X under our description that the ontological structure of the world really is as described in propositions P1, . . .,Pn.

48 E.g. Harold I. Brown, ‘Psychology, naturalized epistemology, and rationality’, in O’Donahue & Kitchener, eds, The Philosophy of Psychology, pp. 19-32. 49 E.g. Laudan, Normative Naturalism. 18

3. Propositions P1, . . .,Pn are compatible with the ontological commitments of current scientific theories which have stood the test of critical evaluation by the relevant scientific community. 4. The explicit ontological propositions or implicit ontological presuppositions of competing philosophical positions, say Q 1, . . . Q n, are incompatible with propositions P1, . . .,Pn. Therefore, the successfulness of X under our description remains impossible or unintelligible from the point of view of these competing philosophical positions and/or ontological propositions and presuppositions Q 1, . . .,Q n are internally incoherent and/or ontological propositions and presuppositions Q 1, . . .,Q n are incompatible with the ontological commitments of the well confirmed scientific theories. 5. The best explanation of the successfulness of X under our description is currently that the world is as described in propositions P1, . . .,Pn.

The premises and the conclusion of this argument are all hypothetical and fallible claims. The conclusion (5) is about the relative of the critical realist ontology, not about its supposedly conditionally necessary truth. However, I think that the above argument needs some qualifications.

The argument assumes that scientific practices exist. It will not therefore convince the skeptic who questions the existence of all scientific practices or the possibility of distinguishing scientific from other kinds (e.g. magical, religious) of practices. However, I do not think this is a weakness because the argument is not directed at refuting skepticism about science. It aims rather to defend a fallible theory about the ontological structure of reality. Other kinds of arguments are needed to refute skepticism about science.

It is also presupposed in premise (1) that scientific practice X under our description is successful. Now it might be asked, ‘How can the success of scientific practices be evaluated?’ I think a plausible answer is that the standards used in evaluating the success of practice X are the general standards of good science. From the perspective of naturalism these standards are not a priori, universal or unchanging. They might not provide clear demarcation criteria between science and non-science either, and might also be slightly different in different natural sciences, and develop as science progresses. However, I assume that it is possible to present fallible claims about the general standards of good science and test them empirically against the current natural scientific practices. Then the general theories about the standards of good science can be used to evaluate the success of the particular scientific practice X. For example, Richard Boyd,50 Harold Kincaid51 and Larry Laudan52 have advocated this kind of normative naturalism. I maintain here that this kind of naturalistic argumentation strategy is wholly legitimate and perhaps the only plausible way to formulate and defend theories about evaluation in science. However, I refrain from presenting a list of the general criteria of good science or scientific evaluation. For the purposes of this paper it suffices to point out that such a list can in principle be assembled and defended from the point of view of meta-philosophical naturalism—at least in the context of current science.

The function of premise (3) is to guarantee that ontological propositions P1, . . .,Pn are compatible with the scientific world-view. I admit that the meanings of the concepts ‘scientific world-view’ and ‘critical evaluation of scientific theories’ are not self-evident. However, I assume that they can be explicated from the point of view of meta-philosophical naturalism. Therefore, I do not commit myself to any view that says that there is a certain a priori universal algorithm that can be mechanically applied in every between competing theories or research programmes. I also think that the scientific world-view is not unchanging or wholly coherent. Nevertheless, I assume that scientific research forms the best basis for our attempts to build coherent representations of the different aspects of the world.

50 Boyd, ‘Scientific realism and naturalistic epistemology’. 51 Kincaid, Philosophical Foundations. 52 Laudan, Normative Naturalism. 19

The logical form of the above argument is inference to the best explanation (or abduction) rather than transcendental argument in the Kantian sense.53 It is not transcendental argument, because its premises are not knowledge a priori and none of its premises, nor its conclusion, is transcendentally necessary. Inference to the best explanation proceeds usually backwards from some observed and described phenomenon to a description of an unobservable structure or mechanism, which, if the description of it is true, causally produces the phenomenon or participates in its causal production. Inference to the best explanation also includes consideration and criticism of other possible explanations of the phenomenon in question. Therefore, the premise of the above argument is scientific practice X as it is observed and described, and the ontological propositions P1, . . .,Pn are descriptive claims about the ontological structure of reality, which, if true, explain the practice X under our description better than the alternative ontological propositions or presuppositions Q 1, . . . Q n. It is important to emphasize that this inference is not deductive and, thus, its conclusion is not apodeictic but only fallible. Nevertheless, to the best explanation are widely used in special sciences. For this reason it is plausible to assume that it can also be used successfully in arguments that concern the general ontological structure of reality.

This naturalistic argument does not lead to the epistemic fallacy in Bhaskar’s sense. By ‘epistemic fallacy’ Bhaskar means the view ‘that statements about being can always be transposed into statements about our knowledge of being’.54 Therefore, in the epistemic fallacy ontology is reduced to epistemology. Now, ontological propositions P1, . . .,Pn in the above argument, if they are true, refer to the ontological structure of the world, not just our ways of acquiring knowledge about this world. However, it is not assumed in the argument that claims about the ontological structure of the world can be justified a priori.

Nor does the argument presuppose a ‘God’s eye view’ of the world or that there is ‘One-True-Picture’ of the world.55 Ontological propositions P1, . . .,Pn refer, if true, to some aspects of reality described at a very abstract level. They do not contain claims about any particular entity.56 I think that it is wholly compatible with this approach that there are many different conceptual schemes or linguistic frameworks which can be applied in describing some parts of the world. The ontological structure of the world might well be so complex that it is impossible to describe it by using one conceptual scheme. However, it is always (or at least usually) possible to rationally compare different conceptual schemes. I think that this position presupposes commitment to the correspondence theory of truth interpreted as a definition (not criterion) of the concept of truth.57

Now it might be asked, ‘Why is naturalized critical realist ontology better than Bhaskar’s transcendental realist ontology, which he tries to justify by using qualified Kantian transcendental arguments?’ I think there are at least five benefits if we compare naturalistic arguments presented above and my previous interpretation of Bhaskar’s transcendental arguments. First, the description of the practice X in premise (1)

53 I think that retroduction can be understood as a species of inference to the best explanation of the kind that is used in historical explanations. However, I think that is a mistake to count transcendental arguments as a species of retroductive arguments, as Bhaskar does. See Bhaskar, SRHE, p. 11. 54 Bhaskar, RTS, p. 16. 55 Fay claims that critical realists commit themselves to these views. See Brian Fay, ‘Critical realism?’, Journal for the Theory of Social Behaviour vol. 20, no. 1, 1990, pp. 33-41. 56 Bhaskar also explicitly advocates this view. See Bhaskar, RTS, pp. 29-30. 57 This is, however, a quite complicated issue the details of which it is not possible to go into here. See e.g. Niiniluoto, Critical Scientific Realism; Searle, The Construction and Justin Cruickshank, Realism and Sociology. Anti-foundationalism, Ontology and Social Research (London and New York: Routledge Series in Critical Realism. Routledge, 2003), pp. 95-119. It should be also noted that Bhaskar’s argument against the correspondence theory of truth (RTS, pp. 249-250) is based on a failure to distinguish the definition of truth from the and his interpretation of the relation of correspondence (in the correspondence theory of truth) as a relation of resemblance. See Collier, Critical Realism, pp. 239-242. 20 of the naturalistic argument is based on empirical analysis of this practice and not some allegedly a priori conceptual nor immanent critiques of the other philosophies of science. Therefore, results of the empirical studies of science might also be consulted in order to establish premise (1) of the naturalistic argument. Description of the practice X is also clearly distinguished from the practice X itself in every phase of the naturalistic argument. Second, premise (2) in the naturalistic argument is a fallible hypothesis (not a priori truth), which says that it is explanatorily a necessary condition of the practice X under our description that the ontological structure of the world really is as described in propositions P1, . . .,Pn. In other words, the concept of necessity in premise (2) refers to conditions which are necessary in the explanatory (not transcendental ) sense. Third, premise (4) in the naturalistic argument is its most important step. In Bhaskar’s transcendental argument premise (2) already includes the claim that P1, . . .,Pn are the necessary conditions of the practice X. Therefore, it seems that premise (4) becomes redundant in his argument. This is of course problematic since premise (2) in Bhaskar’s argument is also problematic. Fourth, the conclusion of the naturalistic argument does not say that it is necessary (in any sense) that the ontological structure of the world is as described in P1, . . .,Pn. It rather states that, if we suppose that world really has a certain ontological structure, which is described in propositions P1, . . .,Pn, then we can explain the practice X under our description better than with some alternative ontological propositions or presuppositions Q 1, . . . Q n. So propositions P1, . . .,Pn are hypothetical and, thus, fallible descriptions about the ontological structure of reality. In Bhaskar’s transcendental argument, as I pointed out, the epistemological status of the conclusion is not clear. He seem to suggests that it is somehow conditionally necessary that the world is P1, . . .,Pn or, alternatively, that the truth of propositions P1, . . .,Pn about the ontological structure of reality is conditionally necessary. However, as I argued, Bhaskar is not able to justify this claim. Fifth, naturalization of the critical realist ontology breaks all the links to neo-Kantianism, which, as I pointed out, is clearly incompatible with critical realist ontology. Therefore, naturalized critical realist ontology is not a foundationalist first philosophy. It is rather a hypothetical theory about the ontological structure of reality that might at best provide some guidance to scientists engaged in scientific research.

Although I think that the naturalization of critical realist ontology is desirable, I am not claiming that meta- philosophical naturalism is a totally unproblematic doctrine. On the contrary. I have already pointed out some features that have been criticized in recent philosophical discussions. However, I think that, in the context of philosophical ontology, the problems of metaphilosophical naturalism are less severe than the problems of Bhaskar’s transcendental arguments and his qualified neo-Kantianism. I also think that naturalized philosophical ontology is a better underlabourer in relation to science than anti-naturalist philosophical ontologies, because the connection between naturalized philosophical ontology and science is very close. It is, however, the topic of another article to compare naturalistic and anti-naturalistic ontologies more closely.

Conclusion

I hope to have shown that Bhaskar’s transcendental arguments are problematic and, therefore, fail to justify his transcendental realist ontology. I have also argued that the attempt to naturalize his transcendental arguments by incorporating a posteriori claims and a priori philosophical theorizing is not viable. Nevertheless, it should be emphasized that the critique I have presented concerns the philosophical method Bhaskar uses to justify his ontological doctrines. I have not evaluated the substantial content of his transcendental realist ontology. Therefore, supposing that my arguments are valid, it might still be possible that there are other ways (than using transcendental argumentation) of justifying Bhaskar’s realist ontology. I pointed out that the alternative way to save at least some of the tenets of Bhaskar’s transcendental realist ontology might be the naturalization of the whole transcendental realist ontology. According to this proposal, Bhaskar’s realist ontology is reinterpreted as a fallible ontological theory justified by naturalistic argument. I sketched the general form of such an argument and pointed out that the epistemological status of its premises and conclusion is knowledge a posteriori. If the argument is developed further, it might be possible 21 to show that naturalized (and perhaps revised) critical realist ontology is currently the best explanation of certain scientific practices as apprehended in descriptions based on empirical analysis of the practices. I also pointed out that the logical structure of this naturalistic argument is exactly the same as in some scientific explanations. For these reasons it is not transcendental argument in the Kantian sense. Nevertheless, detailed formulation of this kind of naturalistic argument might also require revisions to some doctrines of Bhaskar’s ontology.

From the perspective of meta-philosophical naturalism, naturalized critical realist ontology should be developed by analyzing scientific practices empirically and criticizing competing philosophical positions. The credibility of this fallible ontological theory depends crucially on its capacity to explain successful scientific practices under our description, its compatibility with the current scientific ontology, and its capacity to point out weaknesses in the competing philosophical ontologies. This means that the relationship between philosophical ontology (or metaphysics generally) and scientific research becomes very close in naturalized critical realist ontology. I think that Bhaskar’s recent attempt to ‘dialecticize’ critical realism proceeds in the very opposite direction to my proposal for its naturalization.58 One symptom of this is that references to the work of other (especially contemporary) philosophers and to real scientific practices have become rare in Bhaskar’s recent writings and the quantity of his own neologisms have notoriously increased. It seems to me that Bhaskar is trying to develop with the help of supposedly a priori philosophical theorizing some kind of metaphysical system that explains not just science but the whole world and even more (God?). The epistemological status of his philosophical claims has become, at least to me, very difficult to understand. Therefore, I think that the naturalization of critical realism might also form a more secular and a more science oriented (although not necessarily scientistic) alternative to Bhaskar’s recent dialectical turn.59

58 See e.g. Roy Bhaskar, : The Pulse of Freedom (Verso: London, 1993); Roy Bhaskar, etc. The Problems of Philosophy and Their Resolution. (Verso: London and New York, 1994); Roy Bhaskar, ‘Introducing Transcendental Dialectical Critical Realism’, Alethia vol. 3, no. 1, 2000, pp. 15-21. 59 I would like to thank the two anonymous referees for their helpful comments on the earlier version of this paper. Responsibility for possible errors is of course mine.