1
After Kant, Sellars, and Meillassoux
Back to Empirical Realism?
James R. OâShea
1. At the outset of his 2008 talk on âTime Without Becoming,â Quentin Meillassoux describes his key concept of correlationism, a concept that he had expounded in his influential 2006 book, AprĂšs la finitude, published in English in 2008 as After Finitude: An Essay on the Necessity of Contingency, in the following way. âI call âcorrelationismâ the contemporary opponent of any realism,â and he indicates that he particularly has in mind certain views shared by âtranscendental philosophy, the varieties of phenomenology, and post-modernismâ: namely, as puts it, the idea âthat there are no objects, no events, no laws, no beings which are not always already correlated with a point of view, with a subjective accessâ (Meillassoux 2014, 9). Transcendental (as opposed to post-modern) correlationism, in particular, is the view that âthere are some universal forms of the subjective knowledge of thingsâ (ibid. 10). âThe realist, by contrast, maintains âthat is possible to attain something like a reality in itself, existing absolutely independently of his viewpoint, or his categories, or his culture, or his language, etc.â (ibid. 9).
Meillassoux is here painting in very broad strokes, but in his more detailed realist arguments against âthe basic argument of these âphilosophies of accessââ he has been concerned to stress âthe exceptional strength of its antirealist argumentationâ (ibid. 10). In what follows, I examine how Meillassouxâs conception of correlationism in After Finitude, as I understand it, relates firstly to Kantâs transcendental idealist philosophy, and secondly to the analytic Kantianism of Wilfrid Sellars. I argue that central to the views of both Kant and Sellars is what might be called, with an ambivalent nod to Meillassoux, an objective correlationism. What emerges in the end as the recommended upshot of these analyses is a naturalistic Kantianism that takes the form of an empirical realism in roughly Kantâs sense, but one that is happily wed with Sellarsâ scientific realism, once the latter is disentangled from two implausible commitments that made such a reconciliation seem impossible to Sellars himself.
2. I want to begin by asking whether Kantâs empirical realism should be taken to be a âcorrelationistâ view in Meillassouxâs sense. The empirically real objects of possible experience, on Kantâs view, are (in the case of âouter senseâ) causally interacting material objects that persist independently of our perspectival encounters with them. But of course Kantâs empirical realism is inseparable from his âtranscendental idealismâ: such objects are necessarily the objects of certain a priori forms of sensible and conceptual representation, namely, space and time and the categories respectively. Meillassouxâs conception of âtranscendental correlationismâ is clearly intended to include this primary sense in which the empirically real, mind-independent material objects of possible experience, for Kant, are âalways already correlated withâ or correlative to certain a priori forms of representation âin us,â as Kant is wont to put it. While this certainly seems appropriate in general, everything hangs on just how one understands these elusive but fundamental Kantian conceptions, and in particular what sort of mind-dependence of the objects of experience or âphenomenaâ is thought to be entailed by Kantâs transcendental philosophy.
At the outset of After Finitude, Meillassoux presents these issues vividly in terms of what he calls âancestralâ realities, an example of which would be what he dubs an âarche-fossilâ: âI will call âancestralâ any reality anterior to the emergence of the human species,â and an arche-fossil or âfossil-matterâ are materials such as the radioactive decay of isotypes that indicate the existence of an ancestral reality or event (Meillassoux 2008, 10). In relation to correlationism, Meillassoux then formulates the philosophical âproblem of ancestralityâ in terms of the question, âhow are we to conceive of the empirical sciencesâ capacity to yield knowledge of the ancestral realm?â and thus in terms of âthe nature of scientific discourseâ:
how is mathematical discourse able to describe a world where humanity is absent; a world crammed with things and events that are not the correlates of any manifestation; a world that is not the correlate of a relation to the world? This is the enigmaâŠ.
(Meillassoux 2008, 26)
Meillassoux proceeds to argue that the very formulation of this question, âwhat is the condition that legitimates scienceâs ancestral statements?â has as âits primary condition ⊠the relinquishing of transcendentalismâ (ibid. 27). Given that I will follow Meillassoux in rejecting âspeculativeâ versions of correlationism, and will agree that, as I shall consider it, âcorrelationism is not a metaphysics: it does not hypostatize the correlationâ in terms of any âancestral Witnessâ taken to ground the ancestral (e.g., God, Absolute Mind, the Kantian Ego incorrectly interpreted, et al.), our primary concern becomes whether Meillassoux is right that the problem of the ancestral entails the rejection of transcendental correlationism of the Kantian varieties, however sophisticated (cf. the two transcendental âcorrelationist rejoindersâ that Meillassoux considers and rejects at 2008, 18â26).
3. The first transcendentalist rejoinder that Meillassoux considers involves the familiar attempt to account for the existence of unperceived or unwitnessed ancestral events by introducing âa counterfactual such as the following: had there been a witness, then this occurrence would have been perceived in such and such a fashionâ (2008, 19). In apparent support of this interpretive move Meillassoux might well have quoted Kantâs example in the Critique of Pure Reason about the existence of unwitnessed inhabitants of the moon that we could experience:
Accordingly, the objects of experience are never given [as things] in themselves, but only in experience, and they do not exist at all outside it. That there could be inhabitants of the moon, even though no human being has ever perceived them, must of course be admitted; but this means only that in the possible progress of experience we could encounter them; for everything is actual that stands in one context with a perception in accordance with the laws of the empirical progression. Thus they are real when they stand in an empirical connection with my real consciousness, although they are not therefore real [as things] in themselves, i.e., outside this progress of experience.
(A492â3/B521; interpolations added)1
Meillassouxâs reply to this first transcendentalist rejoinder, however, indicates that he is understanding Kantâs empirical realism and transcendental idealism in ways that I think we ought to reject (cf. OâShea 2012, 2016).
Meillassoux reconstructs the transcendentalist rejoinder as making the phenomenological point that what is already actually given in experience is always given against the background of aspects or regions that are not entirely given all at once in the experience, but which can or could be given by âadumbrationsâ (Husserlâs Abschattung) from what is actually given in the experience. On behalf of the rejoinder Meillassoux cites, for example, the existence of the unperceived backside of a perceived cube (2008: 19). Meillassoux responds, however, that this trivial account of the existence of unperceived objects or events in terms of our capacity to fill in further unwitnessed âlacunaeâ as implicitly present in what is already actually given (cf. 2008: 20, on the idea of âa lacunary givennessâ)âthis correlationist construction of mere unwitnessed lacunae in âan event occurring when there is already givennessââdoes not touch the deeper problem of the ancestral, which âdesignates an event ⊠anterior to givenness itself,â and which thus refers to âoccurrences which are not contemporaneous with any givenness, whether lacunary or notâ (ibid.):
More acutely, the problem consists in understanding how science is able to thinkâwithout any particular difficultyâthe coming into being of consciousness and its spatio-temporal forms of givenness in the midst of a space and time which are supposed to pre-exist the latterâŠ. [S]cience thinks a time in which the passage from the non-being of givenness to its being has effectively occurredâhence a time which, by definition, cannot be reduced to any givenness which preceded it and whose emergence it allows.
(Meillassoux 2008, 21)
Once we properly recognize that what is âat issue here is not the time of consciousness but the time of scienceâ (ibid.), then, according to Meillassoux, we can see that âevery variety of correlationismâ is exposed as an extreme idealismâ (2008, 18), whether Berkeleyâs or Kantâs, by the problem of ancestral realities that exist prior to âthe coming into being of consciousness and its spatio-temporal forms of givennessâ (2008, 21).
It should be clear from these passages, however, that at least as far as his response to this first rejoinder is concerned (and we shall see that the second rejoinder does not alter the essential issues at stake), Meillassoux understands the correlationism involved in Kantâs empirical realism cum transcendental idealism in a way that restricts its claims to the domain of lacunae-filling constructions from some real historical or hypothetically imagined witnessings involved in some actually given conscious experience. Then it will of course be impossible to make sense, for example, of Kantâs own pre-Critical nebular hypothesis concerning the origin of our solar system within a single spatiotemporal-causal material framework that pre-existed the emergence of any human consciousness. But even internal to the first Critique this is not, in my view, the correct way to understand Kantâs âformal idealism,â as he also calls it (B518â519 n).
Kant argues that any finite, temporally discursive cognition of a world requires or (as we might put it in this context) âcorrelatesâ objectively with certain a priori forms of conceptual and sensory representation in general. As such Kantâs empirical realism is not based on gap-filling constructions from some actually given experiences, whether real or imagined, contra Meillassoux. Meillassoux takes it to be an anti-Kantian point to suggest that, as science indicates, âgivenness could just as well never have emerged if life had not arisenâ (2008, 22). But Kantâs transcendental deduction, for example, concerns the lawful and structural forms, described functionally at the most abstract, second-order level, that must be true of any objective world that is to be the object of experience for any sensibly receptive yet potentially self-aware experiencer of that world; and these categorial forms of unity are then argued to be applicable, a fortiori, to our human spatiotemporal form of experience in particular. Kant does not start with particular, partial sensory experiences and then seek to account for our construction of a world of objects to fill the lacunae in those states of consciousness (whether real or hypothetical). Rather, he argues that the concept of any finite subjectâs being aware of a plurality of sensible states at all in the first place, and in particular being aware of it as such a plurality, entails that such a subjectâs world is already validly conceived by that subject as a directly perceived, empirically mind-independent, objective world of persisting physical realities in general, all of which, whether scientifically conceived or otherwise, must exist within a single, potentially infinitely extensive and divisible spatiotemporal universe. As I will put it, Kantâs correlationism is an objective correlationism in this highly abstract and formal sense, a sense which contrasts sharply with the historically conditioned and restricted subjective correlationism that one finds in Meillassouxâs descriptions of transcendental correlationism in the passages examined above.
4. This point can be further clarified and in fact reinforced by considering Meillassouxâs formulation of and response to the âsecond correlationist rejoinder from a transcendental perspective.â This second Kantian rejoinder claims that Meillassouxâs arguments concerning ancestral realities involve âan elementary confusion between the empirical and the transcendental levels of the problem under considerationâ (2008, 22). But, as with the first rejoinder, everything depends on how these distinctions are spelled outâin this case, the transcendental/empirical distinction. Meillassoux portrays the Kantian as responding this way:
You [Meillassoux] proceed as though the transcendental subjectâwhich is ultimately the subject of scienceâwas of the same nature as the physical organ which supports itâyou collapse the distinction between the conscious organ which arose within nature and the subject of science which constructs the knowledge of nature. But the difference between these two is that the conscious organ exists; it is an entity in the same sense as any other physical organ; whereas the transcendental subject simply cannot be said to exist; which is to say that the subject is not an entity, but rather a set of conditions rendering objective scientific knowledge of entities possible. But a condition for objective cognition cannot be treated as an object, and since only objects can be said to exist, it is necessary to insist that a condition does not existâprecisely because it conditions.
(Meillassoux 2008, 22â23)
In response to the Kantianâs âplay[ing] with the notion of condition in this way,â Meillassoux argues that, despite the above distinction between objects, which exist, and a transcendental condition, which (Meillassouxâs Kantian says) âsimply does not exist,â in the end âone still has to say that there is a transcendental subject, rather than no subject,â and âthat such a subject takes placeâ (Meillassoux 2008, 24). While following Kant in avoiding all speculative dogmatism, we must still grant that such a transcendental subject âtakes placeâ in the sense of being âposited as a point of view on the world, and hence as taking place at the heart of the world,â as a âposition in the worldâ (ibid. 24â25). Hence, as necessarily âlocalized among the finite objectsâ in this way, the transcendental subject âremains indissociable from its incarnation in a bodyâ (ibid.); and therefore all of Meillassouxâs earlier objections to Kantianism based on ancestral realities that exist entirely independently of the emergence of any such perspectival givenness remain sound, he contends.
It is crucial to see exactly why and how Meillassouxâs understanding of these central Kantian distinctions is mistaken, in my view. Meillassoux is certainly right about this important point: âGranted, the transcendental is the condition for knowledge of bodies, but it is necessary to add that the body is also the condition for the taking place of the transcendentalâ (2008, 25). But his argument involves the following fallacious inference: (i) [This is true:] Kantâs transcendental subject requires a perspectival, spatiotemporal embodiment, as âa necessary conditionâ; therefore (ii) [This is false:] Kantâs âcorrelationistâ arguments concerning the necessary forms of any potentially self-conscious, finite cognition, are arguments that depend essentially on premises concerning (or concepts pertaining to) the emergenceâthe âinstantiationâ or âappearanceâ in space and timeâof such embodied subjects; for example, on facts such as âthe eme...