Kant's Transcendental Proof of Realism sans phrase

From an e-mail, Duck:

[Quoting a post of mine]

"Kant's Transcendental Proof of Realism" still doesn't deserve to be neglected like it appears to be.

Maybe you can post on it? That's a provocative title (although not as provocative as "K's proof of Transcendental Realism" would be). I'm happy to hear about why Kant is not an idealist, but I don't know KW's particular take on the matter.

I actually have covered some of the material in it in earlier posts; I just didn't cite Westphal when I summarized his arguments, since I don't have the book anymore. It's worth pointing out that Notre Dame Philosophical Reviews covered the book a while back; my summary will overlap NDPR's at points, but I'm more favorably disposed to the book than the Notre Dame reviewer was.

Westphal's book is divided into four parts: A critique of Henry Allison et al's "epistemic" version of transcendental idealism (hereafter TI) by means of the defense of a "metaphysical" reading of TI; two separate ways in which Kant's argument for TI fails, "the transcendental affinity of the manifold" as a "material transcendental condition" and the shift from "each event has a cause" to "each event has an external physical cause" in the Analogies; and then a concluding section wherein Westphal argues that TI was not required to "make room for freedom" and that a TI-less version of Kant's system can be developed which avoids the problems Westphal diagnosed while still yielding interesting results. The book as a whole is devoted to defending "transcendental arguments" as not being wed to TI (and so, contra Barry Stroud, as being able to "reach all the way out" to reality), with an aim of suggesting that Kant/Hegel have plenty of good transcendental arguments lying around for us to make use of. In particular, Westphal is concerned with a Kantian argument for "mental content externalism" -- for the thesis that our thoughts being contentful is dependent on the external world being some particular way, such that changes in the world would imply changes in our thoughts.

The first part was not anything particularly new to me; starting a study of Kant by saying "Allison is wrong and here's why" is something of a commonplace by now, and I've read enough of Karl Ameriks's stuff to find the "metaphysical" reading of TI almost common-sensical. (Bader over at the Transcendental Idealism blog has done a fine job covering these issues.) If Kant didn't understand TI in this way, then there's an awful lot of incomprehensible remarks floating around in places like his "Lectures on Metaphysics" and his ethical/religious works. Allison's views are certainly interesting, and undeniable Kantian, but they just aren't Kant's own views.

Westphal also does do a nice job handling a standard objection to Kant's TI (which goes back to Jacobi), that it demands we both apply the category of "cause" to the thing-in-itself (as cause of our representations) and cannot apply the category of "cause" to the thing-in-itself (or else TI is false). Kant's position doesn't fall that easily; Kant doesn't need to apply the schematized concept of "cause" (the category of causation) to the thing-in-itself, but merely the "logical" concept of "cause", of an implicative relation in general. The relation in which the thing-in-itself stands to my representations is analogous to the relation between the sun and the warmed rock, to use Kant's exemplar of a causal relation. Kant doesn't think we can say more than this about the "cause" of our representations, for the same reason we can't say much of anything about the causal power of our free wills -- to do either would be to exceed the limits of the understanding. So Kant has to leave the relation of our representations to things-in-themselves as "mysterious", but it's not a crippling mysteriousness; Kant's TI allows for this sort of indeterminate talk of things in themselves, though it denies that we can have any determinate knowledge of them (because the categories do not apply to them). Kant's TI, read metaphysically, is in the main coherent. If it fails (and Westphal thinks it does), the devil's in the details. And to the details he proceeds.

The first issue Westphal considers is one that doesn't get much mention in the literature, "the transcendental affinity of the manifold". Westphal argues that this should be identified with Kant's "logical principle of genera". I'll quote a passage concerning the latter, from A653/B 681:

Suppose (a case that is readily thinkable) that among the appearances offering themselves to us there were so great a diversity -- I will not say in form (for in that regard appearances may be similar to one another), but in content, i.e., in the manifoldness of existing beings -- that even the very keenest human understanding could not by comparing appearances with one another discover the slightest regularity. If that were so, then the logical law of genera would have no place at all; and even a concept of genus, or any general concept whatsoever, would have no place -- nor, indeed, would even an understanding, which deals solely with such concepts. Hence the logical principle of genera, if it is to be applied to nature (by which I here mean only those objects that are given to us), presupposes a transcendental one. According to this transcendental principle, homogeneity is necessarily presupposed in the manifold of possible experience (although we cannot a priori determine the degree of this homogeneity); for without homogeneity no empirical concepts and hence no experience would be possible.

Kant argues that what is given to us in the manifold of sensation must have at least some minimal amount of order to it; if nothing is at all similar to anything else in sensation, then we could not take notice of anything given to us in sensation, for there would be no determinate objects or states for us to discriminate among (for if they were determinate, they would be determinate in some determinate way relative to other determinations), and hence our thoughts of them would likewise be wholly indeterminate; "without homogeneity no empirical concepts and hence no experience would be possible". The problem this causes for TI is that Kant specifies that this is a transcendental requirement on the matter of what is given to us, not on the form. But according to TI, all transcendental requirements are formal -- the matter is given ab extra, and then the mind imposes the intuitional forms of space and time on it, with the rest of Kant's transcendental requirements falling out of this imposition of spatio-temporal form. This was how the "Copernican turn" was supposed to work, by conforming objects to thought rather than t' other way 'round. But with the principle of genera/the affinity of the manifold, we have a requirement on how the matter of the object must be for thought to get a grip on it at all. If this too is contributed by thought, then there is no longer a sense in which the matter of intuitions is given ab extra; the thing-in-itself can only be (at most) an occasion for the mind to generate "representations" for itself out of whole cloth. This sort of idealism is something Kant's transcendental idealism was supposed to be tailor-made to avoid -- it's empirical idealism, subjective idealism. But if Kant wants to maintain the "Copernican" notion that objects conform to thought because of the transcendental application of forms upon what is given ab extra, then he has to fall back into it. Kant has a problem.

Westphal traces the root of TI back to the Transcendental Aesthetic, where Kant claims to establish that space and time are nothing in themselves, nor are they properties of objects in themselves, nor are they properties of objects in relation to one another, but space and time are merely the forms under which we intuit objects. The forms are due to us, not the objects. God could intuit objects without our forms of intuition (and indeed he must, if God is to play the practical role Kant lays out for him, in the way Kant lays it out). Kant's later arguments for the restriction of knowledge to appearances all refer back to the arguments of the Aesthetic; the schematized categories can't apply to things-in-themselves because, qua schematized, they apply to objects extended in space and time, and there are no things-in-themselves in space and time.

Kant's argument in the Aesthetic doesn't work, however. The argument works as a long disjunctive syllogism: Space and time must be A, or B, or C, or... and cannot be A, and cannot be B, and cannot be C, and... hence space and time are transcendentally ideal, and do not apply to things in themselves, as this is the sole remaining possibility. But Kant ignores a possibility when he lays out this argument: (At least some) objects are in themselves extended in space and time, and only those objects which are extended in space and time can be given to us in intuition, and then only insofar as they are extended. This allows space and time to be "formal requirements" on any object which can be given to us, but without requiring any transcendental meddling. (If these same objects have some non-spatio-temporal properties, or if there are non-spatio-temporal objects, the objects cannot be given to us under those aspects. But nor can we so much as imagine what these would be like, though they remain logically possible. And what can be given to us is the objects themselves, as they are in themselves.) (I need to look at McDowell's "Radicalization of Kant" essay more closely at some point; McDowell seemed to end up endorsing a stronger thesis, that space and time are the forms of objects generally, when this extra step isn't needed to secure thought's grasp on its objects. There can be things we can't notice without this sliding into "We can't notice anything". McDowell seems to be making the same overly-strong claim as Hegel here, at least.)

A quick glance at the Aesthetic (which is a lot shorter than I recalled it being) gives me this:

How, then, can the mind have an outer intuition which proceeds the objects themselves, and in which the concept of these objects can be determined a priori? Obviously, this can be so only insofar as this intuition resides merely in the subject, as the subject's formal character of being affected by objects and of thereby acquiring from them direct presentation, i.e., intuition, and hence only as form of outer sense in general.

The problem is in the "obviously". That's Kant's argument against the position Westphal considers, so to speak. It just doesn't occur to Kant. Kant holds that we have a pure, formal intuition of space a priori, and so he thinks space must be imposed by the mind ("outer intuition... resides merely in the subject, as the subject's formal character of being affected by objects"), and thus space cannot be the form of objects in themselves. But, as Westphal shows, this doesn't follow. Our pure formal intuition of space might be what allows us to be presented with objects without it having to distort them in the process of doing so. Kant recoils too far from the "abstractionist" picture of space as "derived from experience" into a picture of space as forced onto experience. (The objection of a "neglected alternative" in the Aesthetic is common, but Westphal takes it in a different direction than most. It's generally posed as part of a broadly anti-Kantian project of criticism, which is not Westphal's aim. Westphal aims to show that the Aesthetic can be shown to fail without this vitiating the aims of the Critique as a whole.)

The second failing, the sliding between "every event must have a cause" and "every event must have an external physical cause" appears to again just be an oversight on Kant's part -- which he caught later on. Trying to fix the argumentative "gap" here was one of the goals of Kant's "Metaphysical Foundations of Natural Science", and then later the "Opus Postumum". Westphal picks apart the opening chapter of the MFoNS, on "Phoronomy", and shows how (among other things) Kant doesn't have a good reason to reject the possibility of hylozoism, and so Kant failed to patch up his system. Kant wants to somehow argue from the mere notion of matter as "the movable in space" to the thesis that each physical event must have an external physical cause, and the prospects for success here are dim. But Kant needs to somehow vindicate the stronger form of the causal thesis, because this is what the analogies required, and trying to wheedle it out of the bare concept of "the movable in space" is all his "critical metaphysics" allows. This part of the book is more nit-picky, it seemed to me, and doesn't have a great deal of bearing on TI generally; Westphal spends most of his time arguing against Michael Friedman's reading of the MFoNS, since Friedman's the big name when it comes to the MFoNS, but this doesn't seem to have much to do with the rest of the book. Westphal could've gone from the affinity of the manifold straight to his closer and the book would've been a better read.

The conclusion of Westphal's book is that Kant's transcendental arguments can be reworked into a form that does show that the categories (or at least the categories of causation) can be applied in possible experience (and not outside it), and that this (sound) argument doesn't establish anything like transcendental idealism. Westphal calls it "realism sans phrase", to distinguish it from empirical realism, internal realism, virtual realism etc.. Westphal argues that it should've been Kant's own position, but for some unfortunate oversights on Kant's part; the sorts of arguments Westphal makes use of have parallels in Kant (which Westphal is quick to draw attention to). Realism sans phrase both is and is not the "transcendental realism" Kant rejects; the latter is a confused view that would take maxims like "Each event has a cause" and try to draw conclusions from them without consideration of the context within which such a maxim can be held true by us (viz., within possible experience -- in application to objects in space and time); the "transcendental realist" is thus accused of making a mess of freedom and nature (A 543/B 571). Kant specifically says that the "transcendental realist" does this to time and space by trying to measure them -- are they finite or infinite? (A 491/B 519). But the way Kant initially introduces the "transcendental realist" (in A 369-372) is more anodyne: the transcendental realist holds that objects which we intuit as extended in space are also extended in space in themselves -- "apart from the senses" in the sense that our mind does not effect their spatiality. Kant slides from this characterization to the characterization that the transcendental realist must regard objects as extended in space "apart from the senses" in a different sense -- objects are in space, and we intuit objects in space, and these two are independent* of one another. Even if objects were not extended in space in themselves, we would still intuit them spatially, and if they are extended in space in themselves, then we still might intuit them nonspatially. Which means that our "spatial intuitions" can't tell us about any objects in space; we have to infer the one from the other "as cause from effect". Thus "transcendental realism" is wed to "empirical idealism" (and TI is wed to "empirical realism"). So Kant's condemnation of "transcendental realism" is ambiguous: it's a condemnation of empirical idealism (good), of the use of the categories outside the boundaries of possible experience (good), and of the notion that objects are in themselves as we intuit them to be (bad). Westphal means to affirm the third position, while retaining a Kantian opposition to the first two.

Westphal's final chapter is devoted to showing that (contra Kant) TI is not needed to "make room for freedom" in Kant's system. Westphal shows that Kant's arguments in the analogies and the paralogisms jointly rule out determinism in psychology. As Kant points out in the introduction to the MFoNS, psychology deals with no objects of outer sense; the soul is not a physical object, and so it cannot be treated of by recourse to physical, causal laws. The paralogisms make the point even sharper: We have no knowledge of the soul as a substance at all; it is only equivocations which make it seem that there must be a thing underlying all my thoughts to give them unity. And if my mind cannot be licitly considered as a thing, then a fortiori it cannot be considered as a physical thing, which it would have to be for Kant's argument for determinism in the phenomenal realm to apply to it. Because the soul is not a physical substance, the analogies have no bearing on it; we cannot affirm knowledge of any causal relations involving a soul. Kant does explicitly affirm psychological determinism in the first Critique (in the response to the third antinomy), but Westphal notes that the section on the antinomies was written before the paralogisms/analogies/MFoNS, and that while the paralogisms were completely rewritten for the B edition, the antinomies were left unchanged. So Kant appears to have not noticed that his views in the paralogisms implied that psychological determinism is not known to be true, and so TI was not needed to allow for thought to be ungoverned by natural laws. Without recourse to a "noumenal self", Westphal is able to get to the conclusion Kant wanted from the first Critique: It is not known that my thoughts follow upon one another according to a law. I am inclined to not rest satisfied with all of the conclusions Westphal ends with, but they strike me as being the most faithful to Kant's texts out of all the commentator's I've read.

*That reality and truth are "independent of our beliefs" is the slogan that Davidson, in "The Structure and Content of Truth", says it is pointless either to accept or to reject. (p. 305 in the relevant issue of "The Journal of Philosophy".) Davidson says that the only sense which we can give to this slogan depends on correspondence, which cannot be made intelligible. But it seems to me that Davidson simply ignores the most plausible sense of the slogan: Reality can be a certain way, a certain sentence can be true, without anyone believing that it is so. This is not to say that all of reality could be this way, or that all sentences could be this way, or that all of our beliefs could be false. Reality and truth are independent of our beliefs**, but our beliefs cannot be generally independent of reality and truth. If reality/truth alter substantially enough, our beliefs have to follow. Only some of our beliefs can be wrong; most are true. But most true sentences would not cease to be true if all rational animals suddenly perished, and thus all belief came to an end.

It appears to me that Kant makes a similar gaff -- he conflates the notion that spatial objects are spatially extended independent of our perceiving them as spatially extended with the notion that our intuiting an object as spatial is independent of there being spatial objects. We are able to intuit objects as spatial because there are spatial objects given to us; if there were no spatial objects, we would have no intuitions at all, a fortiori no spatial ones. So again the independence only goes one way. Getting a proper view of this independence is just to reject the transcendental ideality of space and time while still recognizing the formal role space and time play in our epistemic life.

**excepting, of course, the parts of reality which are our beliefs, and the parts of truth which are sentences talking about our beliefs.