I have to apologise that its taken so long to get this third part up. I had section 7 written when I posted the last part, but a number of things came up at the beginning of this week which have made it difficult for me to finish section 8. Anyway, it’s done, and this caps off my response to Levi’s posts. I had originally wanted to say more about Levi’s claims about Kant, specifically regarding the bits of Kant that he claims to take up, but I need to get on with other things.
Also, Levi has since posted a response to part 2 (here). I don’t want to tackle the points he makes in the detail I’ve gone into below, again, because I need to get on with other things, but I think there are perhaps four quick points that can be made:-
1) Levi now claims that my criticisms of his account of withdrawal can be circumvented by means of his distinction between first order and second order observation. In essence, this is a perspectivalist solution to the problem of how to understand direct and indirect access. The claim is effectively that because we can observe that other systems lack our own particular sensitivities to the environment, we can see that there is some loose sense in which they are not accessing aspects of the environment that we are. We can then by analogy hold that there must be bits of the environment that we are not accessing. I think this will prove very problematic, but I won’t elaborate here.
2) At several points in his response Levi makes the claim that he can address problems I’ve raised for him in regional ontology. For instance, he claims that any problems I’ve raised for him regarding the differences between intentional and non-intentional systems can be handled at the level of the regional ontology of intentional systems. The important thing to point out here is that if Levi introduces new metaphysical resources to account for the intentional relations that we enter into, then he abandons what was supposed to be the real thrust of OOO, because this is tantamount to reintroducing special metaphysical relations that only humans (or intentional systems more broadly defined) can enter into in order to secure the possibility of knowledge. However, if what Levi means here by regional ontology doesn’t involve introducing such specialised metaphysical resources, precisely what does it involve, and how can it help?
3) Levi seems to think that my discussions of a ‘shared apparatus of meaning’ imply something like a static background of meaning available in advance as a condition of the possibility of communication. This couldn’t be further from the truth, indeed, the Brandomian position I adopt more often gets accused of being too dynamic, insofar as it denies that there are anything like analytic truths that fix the meaning of our claims (i.e., it is a form of semantic holism). There are two important upshots of this. On the interpersonal level, communication is less like the exchange of fully formed meanings than it is a co-operative activity in which we negotiate one another’s commitments, the meanings of which are determined by their relations to others. On the broader social level, the inferential norms (or concepts) which determine the relations between sentences (and thus their meanings) are subject to continuous revision, insofar as the process of revising our commitments just is the process of revising our concepts. The only thing which is fixed here is the fundamental norms governing these dynamic activities. Incidentally, Levi also at one point says that Brandom is insufficiently concerned with non-discursive practices. This misses the point that such practices are in fact Brandom’s answer to the objections that his approach is too dynamic. For Brandom, it is shared practices of talking about and engaging with things (what he calls ‘thick’ or ‘object-involving’ practices) that allow for the possibility of interpersonal communication and conceptual revision. To explain this in detail would require too much space (I also don’t think Brandom’s account of this is quite adequate even if it’s on the right track), but it’s important to see that Levi is well off the mark here.
4) Finally, Levi responds to my concerns about representation by invoking what he takes to be adverse connotations of the word. He thinks that focusing upon representation tends to produce epistemologies in which there is too much focus placed upon mental contents, and this tends to obscure the importance of concrete practices, along with the social and historical dimensions of knowledge development and retention. All I can really say to Levi here is that although there are a number of good historical examples in which these coincide (e.g., Descartes), that the connotations he finds say more about his own prejudices than anything else. Brandom’s approach to representation takes account of everything he thinks it would exclude: semantic holism (against self-subsistent mental contents), thick practices, and an account of how both social and historical dimensions of linguistic practice are necessary for representation. Much as was the case with the word ‘normativity’, I think Levi’s reading too much into the notion of ‘representation’, and he needs to get over this if he’s to deal with the variety of issues that it involves (and which I sketched in the last post).
Anyway, onto the main event once more. Here are sections 7 and 8.
7. Rationality, Normativity and Eliminativism
I can now start to clarify my own position, and in doing so address Levi’s criticisms of my use of the notion of normativity (here). However, it will be useful to first get to the core of Levi’s position, in order to reveal its underlying motivation. I’ll then try to show that this is similar to the motivation for my own position. To this end, the final paragraph of his post on my transcendental realism (here) is perfect:-
As an aside, I think philosophers really need to relinquish situating epistemological questions in terms of things like thought, propositions, and perhaps even knowledge. This sort of terminology suggests far too passive a relation to knowledge and invites metaphors of specularity or mirroring. Instead, we should focus on knowledge practices or what people actually do in producing knowledge. The problem with thought is that it cuts all of those practices out of the story at the outset, as if they can safely be ignored and we can just talk about consciousness, thought, representations, and proposition. I think a number of problems in epistemology are just poorly posed because of this tendency. It might sound strange to say that we should relinquish talk of knowledge in epistemology. However, my point here is that we should instead talk about inquiry. Knowledge has connotations of factoids you look up in an encyclopedia. The concept of inquiry gets at the real work involved in producing knowledge. Philosophers, in their way of talking about knowledge, seem strangely disdainful of the practices that actual knowledge-producers use in producing knowledge. We seem to like the results of that inquiry while simultaneously treating the process of inquiry as philosophically insignificant.
This sums up Levi’s approach to knowledge neatly. He rejects propositional accounts of thought and knowledge because these treat knowledge as products indifferent to the processes of production through which they are arrived at, and thus invite the introduction of illicit metaphysical relations such as representation.
It’s not hard to see that Levi is strongly influenced by Deleuze’s critique of such accounts of knowledge in ‘The Image of Thought’ in Difference and Repetition. In this chapter, Deleuze suggests that philosophy has traditionally been too concerned with knowledge as a product, and recommends refocusing upon the processes of learning through which such knowledge is produced. He also advocates a concomitant refocusing upon knowing-how in opposition to knowing-that. Deleuze takes these position because of his commitment to the strong version of the principle of univocity, which demands that he think of all things in the same metaphysical terms. This means he that he must be able to account for knowledge in the same terms as other system states, and that he must be able to account for its production in the same terms as the production of such states. Deleuze thus ends up with a position in which the world is composed of material processes that reciprocally constrain one another’s development, such that the sociological processes through which we produce new scientific theories are viewed as metaphysically on par with the processes through which species acquire new evolutionary adaptations. Although Levi’s metaphysics differs from Deleuze’s on a number of crucial points, we can see that the motivation for his position is largely consonant with Deleuze’s.
Now, as I’ve noted above, my own metaphysical position is largely Deleuzian. Importantly, I don’t disagree with the metaphysical conclusions that Deleuze comes to in ‘The Image of Thought’. From a metaphysical standpoint, I think that the evolution of species and the evolution of scientific paradigms should be thought in the same adaptational terms. What separates me from Levi then? Why do we come to such different conclusions about propositional content?
To answer these questions it’s useful to point out the similarities between Levi’s position and another familiar position that he’s often argued against, namely, the Churchlands’ eliminative materialism. There are of course plenty of differences between Levi and the Churchlands. He doesn’t accept the particulars of their neurophilosophy, and they wouldn’t accept the particulars of his broader metaphysical project. Nonetheless, they seem to be united on one particular point: that we must cease to understand human beings’ (and other organisms’) behaviour in terms of anything like propositional attitudes. This is the point of the Churchlands’ rejection of folk-psychology – the rejection of our common understanding of one another in terms of propositionally contentful states such as beliefs and desires (e.g., x believes that P, and x desires that Q). Moreover, despite the fact that they adopt a loosely representational approach (albeit a non-discursive one), the Churchlands are also driven toward a broadly pragmatist approach to the assessment of knowledge (i.e., ‘truth’ is abandoned in favour of some kind of adaptational criteria). However, the truly crucial similarity is that they both accept that the whole way in which we talk about knowledge, and thus epistemology as such, must be radically changed to avoid appeals to anything like propositional content. On this basis, I will characterise both positions as forms of radical eliminativism.
The problem for radical eliminativism is that it ends up being self-defeating. This isn’t because it involves believing that there aren’t any beliefs, or even that it can’t claim to mean what it says. Both think that there are phenomena correlated to the notions of ‘meaning’ and ‘belief’ that must be explained, they simply deny that they should be explained in propositional terms. The problem is rather that in doing so they undercut the practices of rational justification through which they are themselves to be justified. There are at least two ways they do this.
First, both approaches undermine the idea of normatively articulated inferential relations between the propositions expressed by assertions. This is to say that they undermine the very idea that there are good reasons for accepting any given assertion. They try to replace this with some pragmatic analysis of the practical effectiveness of accepting the assertion, explained in terms of some deeper informational dynamics. Leaving aside the general problems I pinpointed for Levi’s account of effectiveness above, and the general problems Ray Brassier locates for the Churchlands in the first chapter of Nihil Unbound, there is the more specific problem that it is entirely unclear that either theory meets this proposed standard, and thus should be accepted over and above the intentionalist approaches they are rejecting. Of course, one could try and come up with some general argument for the effectiveness of reasoning in order to shore up the arguments they have provided, but this would be subject to the next problem.
Second, both approaches make it impossible for us to explicitly specify the content of our commitments, insofar as separate what determines the ‘meaning’ of what we say from the ordinary ways in which we talk about what we mean. To explain, most debates generally involve each side making additional statements to clarify the meaning of other statements they’ve made (e.g., when I said that ‘the fattest man in the world lives next door’, I meant that ‘the man with the greatest girth lives next door’, not that ‘the man with the greatest weight lives next door’). This is an integral part of the process of negotiation through which we reach an acceptable level of understanding regarding which statements we agree about and which we disagree about. For this to be possible it must be in principle possible to effectively use ‘that’ clauses to explicitly use different sentences to mean the same thing (i.e., to express the same proposition). However, if we hold that sameness of meaning is really just congruence between deeper informational dynamics, then this kind of explicit negotiation becomes impossible. Until we start examining the deep structure of one another’s information processing mechanisms, we can’t effectively negotiate anything like common understanding of what something means. This is absolutely crucial if we’re going to be able to argue about the goodness of inferences, i.e. about what follows from what and why.
It’s important to point out here that we don’t need anything like discrete ‘meanings’ that are ‘traded’ in communication in order for this kind of negotiation to work. On the Brandomian account there’s a good sense in which no one completely grasps the content of any proposition, not even the one’s they are themselves committed to, insofar as inferential roles (which determine semantic content) are potentially infinite (what follows from what follows… from the proposition and all possible sets of auxiliary premises). The important point is simply that we can take ourselves to grasp the same proposition, even if our grasp of it is potentially divergent. It’s this ability to count as talking about the same proposition that enables us to negotiate the discursive terrain of argument, and to explicitly work out where we agree and disagree.
Anyway, the point is that these forms of radical eliminativism end up insisting that we abandon forms of discourse that are essential to the process of argument through which they are supposed to be justified and assessed. The challenge is thus to articulate and justify a position which is attentive to the concerns that motivate radical eliminativism: the empirical inadequacy of intentional forms of psychological explanation (the Churchlands) and the metaphysical illegitimacy of hypostatised forms of representation (Levi and Deleuze), while nonetheless avoiding the trap of invalidating the very institution of articulating and justifying positions. This is the goal of my own approach, which by contrast I will characterise as moderate eliminativism (see here for a detailed treatment of my opinions on eliminativism).
The crux of my position is this: I accept that in there is a good sense in which there are no such things as propositions or anything with propositional content (e.g., beliefs, desires, theoretical and practical commitments, norms, etc.), and thus that nothing like propositional or more broadly representational content should play any role in either empirical or metaphysical accounts of the world. In my preferred terms, the propositional is not real. Nonetheless, I think that not only can we still engage in propositional talk intelligibly, but that there is a good sense in which we must engage in such talk. Propositions and norms (among other things) are like elaborate fictions, but they are elaborate fictions that we can’t avoid. It is the idea that we can be bound to talk about things that aren’t real, and that the description of this talk plays some kind of foundational philosophical role, that most people find hard to understand let alone swallow. Because of this I usually end up being accused of giving propositions and norms some different kind of metaphysical status, rather than no such status at all. I’m now going to try and provide some ways of thinking about these ideas that should alleviate the most common fears they spark (a different treatment of the same issues can be found here). This will recapitulate some of what has already been said in previous sections.
In essence, I think that we are all playing a very complicated and ongoing game – that of giving and asking for reasons for what we think and do. Although it is best understood as a game, it is not for that matter something trivial. In this game we use sentence tokens to make moves which have the significance of assertions, thereby undertaking various theoretical commitments. There are various other types of moves we can make (e.g., questions, retractions, declarations of practical commitments, etc.), but all moves alter our score in some way, both by changing what we are committed to and changing whether or not we are entitled to those commitments. In playing the game, we must engage in what Brandom calls discursive scorekeeping. We keep track of the commitments and entitlements of ourselves and those we discourse with. The ways in which these moves alter the score are determined by rules describing the inferential relations between sentences. We say that two assertions express the same proposition just in case the sentences they use have the same inferential role, i.e., just insofar as they alter the score in the same way. This means that propositions are individuated by inferential norms governing the the proper use of sentences in making moves and keeping score within discourse, and they do not pre-exist this individuation. This elaborate practice of trading in reasons (and acting upon them) is just what rationality consists in.
Now, I don’t want to recapitulate the whole of Brandom’s scorekeeping pragmatics here. What I want to get across is the idea that rationality is matter of engaging in a social practice in which we ascribe to ourselves, others, and the tokens we use certain normative statuses. The salient point is that these normative statuses are socially instituted. Whether or not something has a certain normative status, and what the content of that normative status is, are not things which are independent of our attitudes about them (not just our individual attitudes, but the attitudes of the relevant social group). Moreover, the content of the normative status consists in a set of norms governing how we are to behave in relation to the thing in the context of the practice. To give an example from a different game, if we have a game of poker, we can stipulate that different coloured chips are to be used in place of money, and that these will have differing values (e.g., £1 for white, £5 for blue and £10 for red). For the sake of argument assume that there is no pre-existing convention governing which colour chip should be more valuable or what the value should be. The normative statuses that are ascribed to the chips thus specify how they are to be used in the context of the game, and also ultimately play a role in determining what the consequences of the game will be (i.e., how much money people will take away), but that the chips have these statuses, and what the statuses are is something which is entirely dependent upon an agreement we make at the beginning of the game. This is an example of normative statuses which are explicitly instituted, but there are also implicitly instituted statuses. It’s very common for practices of treating things in certain ways to slowly emerge in ways which only later become codified in the form of explicit rules (or not at all).
What I draw from this social institution of normative statuses is that even though they may be predicated of real things, they are not for that matter real properties (I’ve talked about this here too). It is not an objective matter of fact whether a poker chip has a normative status, or whether that status has a certain content, in the same way that it is an objective matter of fact that the chip is a certain size and colour. Normative statuses may have a certain amount of objective content, insofar as there are objectively assessable criteria for their application, but this does not make them objectively assessable. For instance, it is a condition of being a legal adult that one is over the age of 18, but although it is an objective matter of fact whether someone is over the age of 18 (and thus whether they count as a legal adult) the consequences of being a legal adult are not objectively discernible (i.e., without appealing to the attitudes of those who institute the laws governing legal adulthood). This latter claim requires a certain amount of additional justification.
Without going into detail about my account of objectivity (which can be found in my TR essay here), its possible to provide a basic justification of this idea by looking at a common objection to it. The objection is that it’s perfectly possible to study the behaviour of any social group in entirely objective terms, and that it is thereby possible to give a purely objective account of any social roles that are instituted by the group. For instance, such accounts would explain the status of the chips in the poker game by describing the way those chips tend to be treated. The crucial point is that there is an important difference between describing the way things tend to be treated and describing the ways they ought to be treated. There are at least two good reasons for this. On the one hand, there are plenty of cases in which behavioural tendencies run completely counter to the instituted norms. To give a fairly trivial example, the vast majority of games of Monopoly I have played have ended up abandoned part way through, with victory assigned to one player on the basis of arbitrary criteria, even if there is a good sense in which we should have kept playing until one player won in the proper fashion (on the basis of this sample, the tendency and the norm would diverge). On the other hand, normative statuses are supposed to determine the appropriate way to treat something in situations which have not yet occurred. The only option here is to identify what should be done with what some individual or group are disposed to do. The major problem is that these dispositions underdetermine what is correct. There will be cases in which there simply is no coherent disposition either way. What all this means is that behavioural tendencies fail to achieve normative closure – they do not completely determine correct and incorrect behaviour in advance.
This is the point at which another objection is usually levied at me. This is that the only way to achieve such normative closure is to posit norms as some special kind of entity that completely determine the appropriate behaviour in all cases in advance. However, this is precisely the approach I’m trying to avoid. Rather, the point is that if we are engaged in a given practice, and we come to a situation in which we disagree about what the normative statuses dictate we should do, then we can proceed to give reasons for our different interpretations so as to determine the right course of action. What this means is that if we are unsure about or disagree about what the content of a norm is, we can nonetheless proceed to make this content explicit through the process of interpretative argument.
The crucial point here is that there need not be some special metaphysical thing – the propositional content of the norm – which pre-exists our argumentative interpretation of it. What it is to treat things as possessing normative statuses is to treat them as if the correct way to treat them is completely determined in advance, in such a way that when we come up against indeterminacy we can engage in a rational process through which we progressively determine the content of the norm. Normative closure is not some strange metaphysical property, but rather is itself something which only makes sense within the context of the practices of rationality, i.e., in a context where we can talk about determinate propositional contents.
There is one last objection to pre-empt here, namely, that I’m arguing in a circle: I’ve explained rationality in terms of normative statuses and I’m now explaining normative statuses in terms of rationality. The answer to this is that I am indeed arguing in a circle, but that it is a virtuous rather than a vicious one. On this approach normative notions are taken to be explanatory primitives, insofar as we do not attempt to reduce normative statuses to anything else, such as causal dispositions. Nonetheless, how these primitives function is still explained in a way which usefully elaborates their relationships to a variety of other concepts. We do not simply appeal to an implicit understanding of normative notions, but try to make explicit this understanding. The virtue of the circle lies precisely in the fact that it makes this explicit (indeed, this is why Brandom’s masterwork bares the title it does). The strategy is to explain what one must do in order to count as saying anything, including what one must do to say what one must do.
I’ll conclude this section by returning to the claim made earlier in response to Levi’s points about the relation between meaning and normativity: the notions of meaning and normativity are inseparably bound up together. Propositional content is articulated by norms governing the use of sentences, but these norms themselves have propositional content that is made explicit through the process of using sentences within interpretative debate. This is all metaphysically licit because all of the actual behaviour that this consists in can be described in terms which make no appeal to mysterious entities like ‘propositions’ and ‘norms’ (i.e., the description of talk about ‘propositions’ and ‘norms’ need make no reference to them). It is only when we engage in this behaviour ourselves – when we play the game of giving and asking for reasons – that we must start talking about such things, because playing the game involves treating what we and others say (theoretical commitments), and what we and others must do (practical commitments), as having determine content. It is simply the case that the fact that we must talk about such things in order to play the game doesn’t imply anything about their reality.
8. Epistemology, Metaphysics, and Transcendental Realism
Now that I’ve cleared up the major differences between our positions, I can finally turn to the dispute over the relation between epistemology and metaphysics. This will also involve tackling the differences between the position which I call ‘transcendental realism’ and the position of the same name that Levi inherits from Roy Bhaskar. Unfortunately, as with Luhmann, I simply do not have the time to go away and steep myself in Bhaskar’s work before addressing the conclusions Levi draws from it. As ever, the consensus that we’re simply unable to read everything the other references will have to prevail here. I will base my analysis on several different posts Levi has made about what he takes from Bhaskar (including, but not limited to here, here and here). I’ll start by trying to briefly sketch Levi’s position, before pointing out the flaws I see in it, and then recapitulating the argument I provided in my original comment.
The major cornerstone of Levi’s position is something he takes from Bhaskar, which he calls the epistemic fallacy:-
The epistemic fallacy does not lie in engaging in epistemology. That would be absurd. Of course we should raise epistemic questions. The epistemic fallacy consists in the thesis that properly ontological questions can be fully transposed into epistemological questions.
This is also combined with a very particular (and, I think, peculiar) conception of what epistemology is:-
Epistemology, as I understand it, is a meta-inquiry into how we know. It doesn’t presuppose any particular object of knowledge. (from comment 10 here)
This leads to a very specific understanding of what it is to commit the epistemic fallacy:-
Epistemologically driven arguments will always pitch questions of what beings are in terms of our access to these entities.
Now, I’m happy to accept that reformulating questions of what things are in terms of our access to them is indeed fallacious. I’m also happy to concede that reducing metaphysical questions to epistemological questions is equally illegitimate. I have two problems here though. First, I don’t think that these two things are equivalent. This is because Levi’s definition of epistemology is far too narrow: not all epistemological questions are questions of access, or rather, questions of how we know. There are at least also questions of what knowledge is, along with various crucially interrelated notions such as justification, truth, and so on. Indeed, the question of what knowledge is should be the central question of epistemology, given that it is quite literally the science of knowledge. Second, to claim that one should not reduce metaphysics to epistemology is not yet to say anything positive about the relation between the two. For instance, it says nothing about which has methodological primacy. However, Levi seems to draw precisely these kinds of conclusions on the basis of the above claims:-
Moreover, it is ontology that is the condition of epistemology, not the reverse. The world must be a certain way in order for knowledge to be possible and these ontological conditions cannot be swallowed by an epistemological reduction to questions of what is given for or to consciousness.
These conclusions represent the core of the ‘transcendental realism’ he inherits from Bhaskar. This approach uses transcendental arguments to deduce ontological (or metaphysical) claims about the nature of the world from the fact that scientific inquiry is possible. This is to say that it deduces ontological conditions of the possibility of scientific practice. From such arguments Levi deduces various things, including the fact that the virtual structure of objects (or generative mechanisms) must be distinct from their qualities, and that objects must be distinct from their relations. At a glance, this seems to fit reasonably well with Levi’s approach to knowledge as I’ve so far described it, wherein knowledge must be understood in metaphysical terms (even if not all interactions should be understood as matters of knowing). However, there is a very serious problem with the conjunction of these ideas, and thus with his thoroughgoing subordination of epistemology to metaphysics.
The problem follows from the way that transcendental arguments work. This has nothing to do with the fact that Levi is using transcendental arguments to derive ontological results, rather than epistemological ones (i.e., that he’s talking about the world rather than thought). It comes from the fact that if one is to derive conditions of the possibility of anything, one must first have an initial account of that thing, i.e., one must have some idea of what it is one is locating the conditions of. This does not mean that one must have a complete account, as often the transcendental argument will extend the initial account of the thing by demonstrating that, if the thing conforms to the initial account, then it must also have certain additional features. However, arguments about the validity of the conclusions of transcendental arguments can always regress to disputes about the validity of the premises, and the initial account of what one is grounding functions as a premise here. This means that if one wants to derive the conditions of the possibility of experience, one must have some account of what experience is, and equally that, if one wants to derive the conditions of the possibility of scientific practice, one must have some account of what scientific practice is. For example, in trying to demonstrate the conditions of the possibility of mathematical knowledge, Kant actually had to provide a detailed (and substantively original) account of what mathematical knowledge is (i.e., of the synthetic a priori). Of course, in uncovering the conditions of the possibility of this knowledge he extended this provisional account (i.e., by grounding it in the pure forms of intuition and the productive imagination), but one of the best ways to dispute his conclusions (as many have) is to dispute his premises (e.g., by arguing that mathematical truths are analytic). Applying this idea then, we can examine another quote from Levi:-
In other words, the onticological thesis is that the world must be a particular way for certain practices like perception, experimentation, discourse, etc., to be possible and that the world would be this way regardless of whether we perceived, experimented, or discoursed about it. This is what is known as a transcendental argument.
The question here is this: what are perception, experimentation and discourse? Any account of what makes them possible is dependent upon a prior account of what they are, and these are eminently epistemological questions. If Levi’s metaphysics is to begin with these transcendental arguments, then he can’t depend on metaphysically loaded accounts of any of these in order to derive his ontological conditions of possibility. This means that either he needs an epistemology prior to his metaphysics (i.e., a non-metaphysical account of knowledge), or he needs to abandon this transcendental starting point.
Moving on, let’s examine one of the specific transcendental arguments Levi makes. This is the argument that is meant to establish the split between substance and qualities, or generative mechanisms and their local manifestations. The argument is roughly that the split is an ontological condition of both the possibility of experimental practice and its practical necessity for the acquisition of scientific knowledge. To quote Levi:-
if experiment is indispensable, then this can only be because objects do not manifest their powers or capacities under ordinary conditions. Objects do not manifest or “give” their powers under ordinary conditions. Rather, it is only under the highly structured and isolated conditions of the experimental setting that we are able to encounter– or better yet, dis-cover –the powers that lie within objects. As a consequence, passively given sensations are not the origin of knowledge. Ontologically, then, the condition under which experiment is both possible and necessary is only in a world where objects can act without manifesting their act in either nature or for a perceiving subject.
The first thing to note here is that the argument turns around an opposition between experimental conditions and non-experimental conditions, or between what Levi calls ordinary and isolated conditions. The second thing to note is that it is formulated in terms of givenness. The core idea is that there is something ‘given’ under experimental conditions that is not given in ordinary conditions, and that this is a matter of the mechanisms which generate the effects encountered in ordinary conditions being revealed (or dis-covered). I’ve got a number of problems with this.
First, how are we meant to draw the distinction between ordinary and experimental conditions? For Levi’s argument to hold up, he has to be able to draw it in a way which has nothing to do with the fact that we choose how the situation is composed in the experimental case, but must instead focus on differences in the composition of the situations themselves. Put another way, it must be entirely possible that what Levi is here calling an experimental situation could occur naturally. A good example of this would be the 1919 solar eclipse that created the conditions under which the effects of general relativity were observed for the first time. In this case, although there were obviously measuring instruments involved, there was nothing else in the way of experimental setup. Levi’s approach seems to be to argue that the number of interacting factors involved in an experimental situation is somehow less than that found in an ordinary situation. A good example of this might be working out the effects of a particular microbe on a certain plant by monitoring their interactions in a controlled environment in which all other plants and microbes have been removed (but can potentially be reintroduced in different arrangements). The problem is that, once we discount the fact that the experimental situation is controlled, it’s not clear that we can draw a clear distinction between ordinary and isolated conditions just on the basis of the number of factors involved. It’s obvious that there are situations in which there are more or less causal factors involved in producing the outcome, and thus that it makes sense to talk about relative isolation, but it’s not obvious that there is a point at which a quantitative difference in the number of such factors becomes a qualitative difference between types of situation, and thus that it makes sense to talk about absolute isolation. Unfortunately, it seems that the argument Levi is giving depends on such an absolute distinction.
Second, it’s unclear precisely what is supposed to be ‘given’ in experimental conditions that isn’t in ordinary ones, and what this ‘givenness’ is supposed to consist in. At first glance, it seems like what he’s saying is that whereas we only observe the actual effects of causal powers under ordinary conditions, we can somehow observe the powers themselves in experimental conditions. This goes against Hume’s point that we don’t observe modal aspects of things at all. Rather, we must always infer the modal features underlying the production of the actual (e.g., laws, capacities and tendencies) from the actual. One doesn’t need to accept the general empiricist suspicion of modal notions that follows from this, or Hume’s argument that all such inferences are invalid, in order to accept this basic point. One can still hold this even if one accepts, as I do, that observation claims are always implicitly modal, insofar as understanding them involves grasping certain counterfactual inferences one can draw from them, because these inferential relations are part of the content of the concepts that make the observations possible. In essence, our observation claims about the actual are implicitly modal, but this implicit modal dimension (i.e., the content of our concepts) is revised indirectly by drawing inferences from new observation claims, rather than through directly observing modality. Moreover, most experiments do not produce single observations that imply facts about generative mechanisms, but instead produce whole series of observations from which such facts are induced. In these cases it is especially clear that scientific discovery involves drawing inferences on the basis of a variety of evidence (some produced in controlled conditions and some not), rather than producing some special kind of event in which a generative mechanism manifests itself. Indeed, the reason for experimental controls is usually to make sure that one gets the right range of evidence (e.g., a full spectrum of the various possible ways in which the different variables can be altered in relation to one another).
Now, I don’t think that Levi actually means ‘observation’ when he talks about ‘givenness’ here. This is because he does at points talk about the way science gets at generative mechanisms by inferring them from what is given in observation. However, this raises further questions. First, it raises the question of how observation and inference are related. This brings us back to the above objection, insofar as this is an epistemological question if there ever was one. Second, it’s hard to see how he can give an account of how generative mechanisms are ‘given’ in a sense distinct from ‘observed’ without using the very metaphysical terms that this argument is supposed to justify. Third, even if he can get around the first two problems, this still produces a certain tension within his account of withdrawal. Take the following quote:-
No one has ever perceived a single object, but we do perceive all sorts of effects of objects. Traditional epistemology has confused these effects with the objects themselves. Fortunately we do occasionally manage to cognize objects through a sort of detective work that infers these generative mechanisms from their effects; without, for all this, ever exhausting the infinity of a single object. At any rate, if objects were not withdrawn in this way, the practice of experiment would be unintelligible.
Here he seems to indicate that it is the very fact that we must infer the generative mechanisms (or modal aspects) rather than observing them directly that constitutes the withdrawal of these mechanisms (or substance) from their effects (or qualities). Moreover, he seems to indicate that the excessive dimension of this withdrawal is in fact quantitative, rather than qualitative, i.e., that inference enables us to genuinely grasp the powers of things in part, just not in full. This differs markedly from Graham’s account of withdrawal, which is strictly qualitative (as evidenced by his distinction between real and sensuous qualities). However, this kind of description of withdrawal contrasts completely with other things Levi has said. For instance:-
OOO doesn’t deny that all entities encounter other entities in their own unique way. This is, in fact, a core hypothesis of OOO. What OOO does reject is the idea that somehow the manner in which one entity grasps, encounters, perceives, or interacts with another entity has anything to do with the withdrawn being of that entity.
How can generative mechanisms be ‘given’ in experiment, by a process through which they may be at least partially inferred, and this still have nothing to do with these mechanisms themselves (i.e., the withdrawn being of objects)? Levi proceeds from the premise that we do somehow produce knowledge of generative mechanisms, and from this argues that the ontological conditions of producing such knowledge somehow imply a limit upon it (withdrawal). However, he can’t formulate this limit in the same terms that Graham does (i.e., that we simply can’t know these mechanisms) without undermining the very premise from which he proceeds. Once more, this raises the question of what is common to Levi and Graham’s use of the term ‘withdrawal’. To what extent does this pick out a common metaphysical position, and to what extent is it a common metaphor or rhetorical gesture?
Leaving these objections to one side, I can now explain the central differences between my transcendental realism and that of Levi and Bhaskar. Before getting into the meat of the matter, I’d like to make one small point. In the second half of his original response to me (here), Levi said the following:-
(it’s very odd that Wolfendale continues to refer to his position as transcendental realism in, apparently, complete ignorance and opposition (!) to Bhaskar who coined the term)
Although I haven’t read Bhaskar, I’m pretty certain he didn’t coin the term. Kant originally coined the term ‘transcendental realism’ in order to opposite to (and thus better delineate) his own transcendental idealism. Reid spoke about this a bit in his paper at the Transcendental Realism Workshop (here). Kant’s sense of transcendental realism isn’t very deep – it simply involves transposing the conditions of thought onto the world itself (e.g., holding that the forms of intuition (space and time) provide the real structure of the world). Both Bhaskar and myself (among others) appropriate the term and give it a much a more determinate sense. Nonetheless, I do think it’s possible to say that one appropriation is better than another, and I will endeavour to show why I take my own to be superior.
The crucial point is that although Levi’s approach (and presumably Bhaskar’s) is both a form of realism, insofar as it makes claims about the real structure of the world as it is independent of our thought about it, and a form of transcendental philosophy, insofar as it arrives at these claims by means of transcendental arguments, its realism is not a conclusion of these transcendental arguments, but a premise of those arguments. Put another way, it assumes both that there is a world independent of thought, and that this world has a metaphysical structure independent of the structure of thought, rather than demonstrating either of these facts (by transcendental or other means). It simply uses transcendental tools to determine what this structure is.
By contrast, it is just these facts which my own transcendental realism attempts to establish by means of transcendental argument (see my essay for details). On the one hand, it attempts to show that the very structure of thought implies that there is an objective world, by showing that, if we are able to think and talk about anything, then the truth of some of our thought and talk must be absolutely independent of all our attitudes about it. On the other, it attempts to show that the very structure of thought also implies that there is a real structure of the world that is in excess of the structure of thought, by showing that objectivity cannot be limited to the particular questions of natural sciences, but extends to those most general questions that lie at their foundations (e.g., what are entities? what are properties? what is essence? what are part/whole relations? etc.). The former amounts to a refutation of most forms of epistemological relativism, and the latter amounts to a refutation of both metaphysical idealism, anti-realism about metaphysics, and what I have called deflationary realism. I don’t think these arguments are entirely water tight yet, but we are more concerned with the aim here rather than with its achievement.
There is a further contrast between these two positions which is best approached through Levi’s comments about what differentiates Bhaskar’s transcendental philosophy from more classical transcendental approaches:-
Bhaskar is engaged in a transcendental inquiry. However, what distinguishes Bhaskar’s transcendental inquiry so much from prior transcendental inquiries is that it does not have recourse to mind, culture, language, or the human in formulating its answer, but rather to the world. In effect, Bhaskar asks not what our minds must be like for science to be possible, but rather, in a jaw dropping and audacious move, what the world must be like for science to be possible.
I suspect that Levi would use this point to show that my position is regressive in comparison to Bhaskar’s, because my ‘transcendental realism’ only talks about the structure of thought about the world, rather than the world itself. There is a certain sense in which he would be right in claiming this. This is because my transcendental realism is not as much a metaphysical position as it is an account of what metaphysics is. My aim is to give an account of what it is to do metaphysics by providing an account of metaphysical discourse that both situates it within a more general account of discourse and shows why it is a necessary complement to natural scientific inquiry. However, the crucial point is that although this project does not itself produce positive metaphysical results, it does nonetheless critically delimit the range of viable metaphysical positions.
For example, I believe that my approach demonstrates a priori that numbers don’t really exist. Many will have problems with this, insofar as it excludes from metaphysics what was previously seen as an exemplary metaphysical debate. However, the metaphysical tradition has undergone a long process through which it has critically delimited its subject matter and the kinds of answers that are appropriate to it. It is enlightening to look at the earliest metaphysical positions we find in the presocratics, in which everything is fundamentally water, fire, love and hate, amongst other things. These are bold gestures towards giving an account of the fundamental nature of reality, but they don’t yet really have any idea about what it is to give such an account. We have since realised that these gestures just aren’t viable metaphysical positions. This isn’t to denigrate the achievement of the presocratics. Without them we would not have metaphysics at all. It is simply to recognise that they lie at the very beginning of the gradual development of metaphysical self-consciousness (i.e., the self-consciousness of what we are doing in doing metaphysics). We know that saying everything is fire just doesn’t answer the question, but in order to know why we need to properly understand what the question is.
This brings me to another of the arguments from my comment on Jon’s blog (here), which rearticulated a point I’d made at the end of an earlier post (here). I’ve so far provided an argument that shows specifically why Levi’s position requires some form of epistemology prior to ontology, but the argument I provided on Jon’s blog was a more general argument to the effect that epistemology has methodological priority in relation to metaphysics. Here is the argument quoted in full:-
i) If we are to be able to have a proper argument about which metaphysical position is correct, then we must be able to make explicit what we’re arguing about, i.e., we must be able to make explicit precisely what metaphysics is. Otherwise we are either talking past one another, or open to the objection that metaphysics is hopelessly confused and should therefore be abandoned.
ii) The questions regarding what metaphysics is are epistemological questions.
iii) We can’t define metaphysics in metaphysical terms without begging the question.
iv) Therefore (i, ii & iii), there must be at least some part of epistemology, sufficient to define metaphysics, that is independent of metaphysics.
v) This means that we must at least be able to legitimately discuss knowledge in non-metaphysical terms, and and any position which denies this thereby denies the possibility of adequately circumscribing metaphysics, and thus the possibility of genuine explicit metaphysical debate.
There is the possibility of arguing that the vicious circle presented in 3 is a hermeneutic circle, rather than a vicious circle, but this requires a genuine argument. Moreover, this argument would have to be an epistemological argument that did not deploy metaphysical assumptions without thereby making it into a genuine vicious circle. I think this is an impassible bind, but if you have an argument, I’d love to hear it.
Although Levi quoted this argument in his response, he didn’t actually address any of the specific points made in it. Instead, he claimed that the argument missed the point of Bhaskar’s transcendental realism. Since I’ve now gone to some lengths to unpack the significance of the latter (at least, as Levi presents it), I think I’m entitled to claim that none of it counters any of the points made in the above argument. Metaphysics is a form of inquiry, and thus one must give an account of inquiry in general if one is to give any account of metaphysics specifically. This is just to do epistemology. This means that if we are to be at all clear about precisely what we’re arguing about in metaphysical debate, we’ve got to be able to do epistemology, and if we’re going to do so without entering into a vicious circle, then that epistemology must be independent of metaphysics. This is precisely the approach that I adopt in articulating my transcendental realism.
To conclude, I think I can provide at least two reasons why my position is more deserving of the title ‘transcendental realism’ than Levi’s. First, as I’ve shown above, my approach is not simply a realism that happens to deploy transcendental arguments, but a transcendental defence of realism itself. Second, despite being fundamentally different from Kant’s transcendental idealism, my transcendental realism is more directly comparable with it, insofar as it represents not a particular metaphysical position but a reorientation of what metaphysics is. Kant’s critique of metaphysics was supposed to demonstrate that metaphysics is an a priori discipline concerned with unpacking the content of the categories and the rest of the transcendental machinery, so as to provide the structure of any possible nature. My approach carries out an analogous critique of metaphysics, but comes to the conclusion that metaphysics is an a posteriori discipline that is both continuous with and distinct from natural science. The results are very different, but the methods are importantly similar. The similarities and differences are much more intricate than this (as is detailed in the essay), but this point suffices to show why my position has a greater claim to the title.
This raises one final point, which serves as quite a good example of the superiority of the methodology I’m advocating. In his manifesto (here and here), Levi criticises Kant for making metaphysics (or just philosophy more generally) an a priori matter. On this point, we are agreed: metaphysics cannot just concern itself with our thought about the world, but must concern itself with the world itself. However, Levi has no substantial argument for this claim, and it seems to me that he cannot have any substantial argument for it without giving methodology priority to epistemology. How are we to argue against those who claim that metaphysics is a priori? One cannot prove a posteriori that metaphysics is a posteriori. The only solution is to beat Kant at his own game, and to argue a priori that metaphysics is a posteriori. This is precisely what I aim to do.