The task of “transcendental logic” is to explicate the concept of a mind that gains knowledge of the world of which it is a part. The acquisition of knowledge by such a mind involves its being acted on or “affected” by the objects it knows.1
Sellars’s characterization of transcendental logic presupposes his commitment to transcendental naturalism. Transcendental naturalism seeks to identify the general features any conceptual system must possess in order to know the nature of which it is a part. It is bound by a minimal constraint of immanence: the mind’s immanence to nature. Our minds are a part of nature. But what ‘nature’ is and what ‘minds’ are is not yet fully determined. The constraint of immanence has the following consequence: if our minds are a part of nature, then what we know about our own minds cannot differ fundamentally in kind from what we know about other parts of nature. We will see later that this has interesting ramifications for understanding the appeal to ‘immanence’ by many contemporary philosophers.
Understanding exactly how known objects “act upon” or “affect” the knowing mind is the fundamental problem. Although the mind is embedded in nature, it is not a mirror of nature. Sellars’s rejection of the myth of the categorial given—the belief that the categorial structure of reality imprints itself upon the mind as a seal imprints itself on wax—means rejecting the presumption of a preestablished harmony between knowing mind and known world. This is effectively to rule out intellectual intuition as a means of accessing the fundamental structure of reality. Human understanding is discursive, not intuitive. Its medium is what Kant called judgment and what Brandom calls assertion, which always stands in a variety of justificatory relations to other judgments or assertions. To characterize a creature as minded is not to give an empirical description of it but to recognize it as capable of participating in the game of giving and asking for reasons.
The converse of the claim that human understanding is discursive is the claim that human intuition is sensible, not intellectual. To say that our intuition is sensible is to say that our discursive understanding of nature as the totality of objects existing in space and time is conditioned by the way nature appears to us in both space and time. This is the problem of representation, first formulated by Kant. If nature is cognizable to the extent that it is representable, then nature is the system of true representations or actual states of affairs. But as Sellars points out, “[N]ot every system of empirical representables constitutes nature, but only that system of empirical representables, the representings of which would be true.”2 Transcendental logic aims to uncover both what it is for something to be an “empirical representable” and what it is for something to be a “true representing.” Understanding the former involves grasping the way in which conception interacts with sensation. Understanding the latter requires grasping the nature of the connection between truth pertaining to what is conceptually represented, and truth pertaining to nonconceptual representings. In Kantian parlance, this is the question of the relation between understanding and sensibility. However, it cannot be reduced to the bald contrast between concepts and intuitions, for part of Sellars’s revision of Kant involves extending the reach of the concept to intuition and acknowledging the indispensable role of conceptual intuition, whose operation is aligned with yet fundamentally distinct from that of nonconceptual sense-impressions. As we shall see, the latter are postulated in accordance with the requirements of transcendental logic to explain how conceptual intuition is guided by nonconceptual factors.
Intuition and Sensibility
The distinction between representing and represented is not the difference between two separate things but the formal (or transcendental) distinction between the reality of a thing insofar as it is represented and the reality of a thing independently of its being represented. This is the distinction between objective reality—reality immanent to the represented—and reality an sich (in itself), i.e., non-represented reality.
In Sellars’s naturalistic revision of Kantianism, the distinction between represented and non-represented is contained within the immanent distinction between representing and represented. Thus four things need to be distinguished:3
- Non-representings in themselves
- Represented non-representings
- Representings in themselves
- Represented representings
But as there are two types of representings, conceptual and nonconceptual, we must also distinguish between:4
- Conceptual representings in themselves
- Conceptual representeds
- Nonconceptual representings in themselves
- Nonconceptual representeds
However, the distinction between conceptual and nonconceptual representings does not map on to the orthodox distinction between concepts and intuitions. It turns out that intuitions have conceptual form. Thus the relevant contrast is between intuition and sensibility. Intuition turns out to be conceptually informed, but conceptual intuition is determined by the nonconceptual structure of sensibility.
On the traditional account, intuitions deliver particulars which are perceived as being thus and so. Consider the following perceptual report:
I see this as a red rectangular brick
The traditional view maintains that my beliefs about the brick are distinct from my seeing the brick:
I see this as a red rectangular brick and I believe that this a brick with a red and rectangular facing surface
Here we have a distinction between intuitive perceptual taking, or “seeing as,” and belief proper, which has propositional form. My seeing this as being thus and so and my belief that it is (or is not) thus and so are distinct, with the latter presupposing the former.
This traditional account falls prey to the myth of the categorial given: the assumption that to be aware of X is to be aware of it as X. In this version, the myth fuses thinking and sensing: it assumes that things present themselves to sensory consciousness already endowed with categorial form. We sense something as something before superimposing onto it our belief that it is thus and so:
I see this as a red rectangular brick and I believe it is too big for the job at hand
What is wrong here is the assumption that objects cause us to be in certain sensory states, and these sensory states are already endowed with the categorial form that allows them to play a justificatory role in empirical knowledge. The object causes me to see it as what it is and this justifies my subsequent beliefs about it and its relations to other objects. Causation and justification are illegitimately fused. By separating them, we distinguish between the sensory states which objects cause perceivers to be in, and the perceptual states in terms of which perceivers respond to their sensory states. In order for these perceptual responses to play a justificatory role in empirical knowledge they must already be endowed with categorial form: they must be seeings as (or hearings as, tastings as, touchings as, smellings as). This is to say that they must involve categorially formed conceptual intuitions of sensible particulars. Once we acknowledge the fundamental role of conceptual intuition in empirical perception, we can subsequently distinguish between what we see of objects, and what we see objects as. What we see of an empirical object is a function of our embodied, perspectival relation to it; but this perspectival relation already presupposes the conceptual intuition of the object as something thus and so. What is intuited is never a bare particular; rather, it is a condition of our ability to intuit particulars that they be conceptually intuited as something. In the following passage, Sellars explains why conceptual intuition is an epistemically irreducible type of representation:
Consider the statements
This is a pyramid
This pyramid is made of stone
The first has the explicit grammatical form of a sentence.
So does the second. But notice that the grammatical form of a sentence is lurking in the subject of the second sentence.
From the standpoint of transformational grammar we would think of it as derived from the deep structure
This is a pyramid and it is made of stone
One might be tempted to think of ‘this’ as a pure demonstrative having no other conceptual content than that involved in being a demonstrative. Kant does think of an act of intuition as a demonstrative thought, a Mentalese ‘this.’ However he does not think of this Mentalese demonstrative as a bare Mentalese ‘this.’ An example of an act of intuition would be the Mentalese counterpart of
This cube facing me edgewise
where this is not to be understood as, so to speak, a Mentalese paraphrase of
This is a cube which faces me edgewise
The role of an intuition is a basic and important one. It is the role of bringing a particular object before the mind for its consideration. Thus, though there is a close relationship between
This cube facing me edgewise . . .
This is a cube which faces me edgewise.
the former is an irreducible kind of representation. It is a demonstrative representation which has conceptual content and grammatical form. As noted above it contains the form and content of the judgment “This is a cube.” Thus for Kant intuitions are complex demonstrative thoughts which have implicit grammatical (and hence categorial) form.5
Intuitions are conceptually formed ‘this-suches’: “this cube facing me edgewise,” “this stone pyramid,” etc. To say that conceptual intuitions are irreducible representations is to say that they provide the fundamental data for perceptual experience and that they deliver the ultimate subjects of predication for empirical judgments.6 It is because intuitions are representations endowed with conceptual content and grammatical form that they can play this fundamental role in empirical knowledge. Thus perception cannot be decomposed into the sensing of bare particulars coupled with propositionally structured beliefs about those bare particulars. What is intuited is categorially determined and thus already available for propositionally structured belief.
But how do intuitions relate to sensibility, or what Sellars calls “sense-impressions”? Our perceptual reports are primarily about physical objects; they are not about not the sensory states caused by those objects and responded to by our reports. We perceive physical objects as thus and so and we deliver perceptual reports about those objects, although those reports are in part responses to the sensory states caused by those physical objects. It is important to realize that our perceptual vocabulary is developed in response to the perceptible characteristics of publically accessible physical objects and that we first have to master that vocabulary before we subsequently learn to deploy it to make perceptual reports about our own sensory states. The ability to perceive our own sensations presupposes the ability to perceive publically accessible objects. Thus the properties of sense-impressions are the postulated counterparts of the properties of physical objects, but counterparts whose properties are modifications of their models: the sense-impression of a red rectangle can be described as a “red rectangular impression” even though it is understood that strictly speaking the sensation itself is neither red nor rectangular. ‘Redness’ and ‘rectangularity’ are the models for the properties of the representing (the sensing) through which they are represented as the properties of intuited particulars (e.g., this red rectangular brick). These counterpart properties are the properties of nonconceptual representings, or sense-impressions.
Sellars’s account of conceptual intuition lets us see how intuited perceptual content, i.e., that which is empirically represented, is conditioned by the forms of intuition, i.e., by space and time as forms of what is conceptually represented (which is not to equate space and time with concepts of the understanding), as well as by the forms of sensibility, i.e., the nonconceptual representings whose spatiotemporal structure is analogous to but categorially distinct from intuited space and time. Thus Sellars distinguishes between space and time as forms of conceptual intuition and their theoretical counterparts, the sigma and tau dimensions as the forms of sensory, i.e., nonconceptual, representings.7 Crucially, it is these nonconceptual forms that guide true representings. Thus truth at the level of represented content is anchored in something akin to truth, which Sellars calls “correct picturing,” at the level of nonconceptual representing. But how does nonconceptual form condition conceptual content, which is to say, meaning? In order to understand, we must first get clear about Sellars’s ‘non-relational’ theory of meaning.
Meaning and Picturing
Meaning statements such as
‘Le chat est sur le paillasson’ (in French) means the cat is on the mat (in English)
are taken to be informative insofar as the English sentence “the cat is on the mat” means what it does because it expresses the nonlinguistic thought or proposition that the cat is on the mat. But for Sellars, meaning statements do not correlate linguistic items with nonlinguistic items (‘meanings’ understood as nonlinguistic entities, whether thoughts, propositions, or states of affairs). Rather, they correlate the linguistic function of an item in an unfamiliar language with that of a linguistic item in a familiar language—in this case, a sentence saying that p. Meaning statements such as:
‘Rouge’ (in French) means red (in English)
correlate linguistic items across two different languages by saying that they play an equivalent role in the two languages
‘Rouge’ in French is a ●red● in English
says that the mentioned sign design plays the same linguistic role in French as ‘’red’ does in English. ‘Red’ here is not being mentioned but used in a special way: not as it is ordinarily used in English (as meaning the color red) but as an illustrating sortal in a metalinguistic assertion.
Similarly, in statements such as
a ‘rouge’ is a ●red●
a ‘triangulaire’ is a ●triangular●
‘rouge’ and ‘triangulaire’ function as distributive singular terms rather than abstract nouns.
Sellars’s crucial contention is that metalinguistic properties picture nonlinguistic properties via the syntactical configuration of sign-design tokens. This link between metalinguistic form and nonlinguistic structure is utterly decisive for Sellars. It provides him with a way of dispensing with appeals to abstract entities in accounting for meaning. He does this by reconstructing the semantic role played by relational expressions and empirical predicates without hypostatizing them as abstract entities.
In accounting for relational expressions, Sellars’s chief inspiration is Wittgenstein’s claim in the Tractatus that we say that aRb by placing the names ‘a’ and ‘b’ in a certain dyadic relation. This dyadic relation is a pattern of inscription. It is the inscription that shows how a and b are related by inserting the symbol ‘R’ between the names ‘a’ and ‘b’. But the relation itself is not an object. And the token ‘R’ that relates a and b is not a name. Thus what ‘R’ does in the statement ‘aRb’ could be done without using a symbol. Consider the statement “a is larger than b.” We could adopt a convention whereby the graphic properties of the inscriptions ‘a’ and ‘b’ say what the statement “a is larger than b” says. For example:
This inscription states what “a is larger than b” states without using the expression “is larger than.” But it is crucial to note that nothing in the above inscription plays the role (allegedly) played by “is larger than.” That b is below a is essential to the meaning of this statement. But this graphic feature does not correspond to the role played by the expression “is larger than.” Rather, in the inscription above, b’s being below a plays the role played by a and b having “is larger than” between them. Thus both the “is larger than” and b’s being below a are functioning here as inscriptions, which is to say graphic objects, rather than as signifying expressions. This insight extends to the semantic role played by empirical predicates. The statement “x is red,” which means that object x has the property red, could be written x. Here it is the way in which the name ‘x’ is inscribed that tells us what property the object x has. The inscription x has two relevant features: it features a token of the name ‘x’ which refers to object x, and it is has a specific graphic characteristic, i.e., being inscribed in bold type. Fundamentally, Sellars’s claim is that predicates do not play an independent role within linguistic expressions: “Not only are predicative expressions dispensable, the very function played by predicates is dispensable.”8 Consequently it is a mistake to abstract the role played by predicates from the role of the expressions in which they occur. It is this abstraction of a fragment of function that encourages the mistaken idea that predicates designate conceptual properties or metaphysical attributes. The predicative role should not be reified and turned into an abstract entity called a ‘property’ that exists independently of sentential contexts. Still less should the conceptual property supposedly expressed by the predicate be hypostatized and turned into an ontological attribute that exists not only independently of language—as conceptual properties are alleged to—but independently of thought. As Sellars asserts: “[T]he extra-linguistic domain consists of objects, not facts. To put it bluntly, propositional form belongs only in the linguistic and conceptual orders.”9 The philosophically decisive consequence is the following: conceptual functions are linguistically incarnated in sign-designs whose material characteristics picture objects as being somehow. This ‘somehowness’ is shown not said by the manner in which names are uttered or inscribed.
An utterance or inscription by itself is of course not a statement. It is a physical pattern (phonemic, graphic, or gestural). Sellars’s naturalism requires that although semantic function is logically irreducible to causal function, it is causally dependent upon it. In other words, semantic function is inoperative independently of its physical incarnation. Thus rule-governed conceptual activity, i.e., thinking, is embodied in pattern-governed regularities, i.e., physical behaviors (whether or not this embodiment is necessary, and the precise nature of its necessity are questions we cannot pursue here). Hence Sellars’s “norm-nature meta-principle,” according to which “espousal of principles is reflected in uniformities of performance.”10 But crucially picturing itself is not a semantic relation or function. It is a “second-order isomorphism” between objects in the natural order. It does not consist in a relation of resemblance between representation and represented; it consists in the structural equivalence between properties of relations among representations considered as natural objects and properties among represented objects. Sellars’s suggestion is that conceptual properties do not designate attributes or ways of being but are nevertheless rooted in acts of representing that picture reality in ways that can be said from within the conceptual order to be more or less adequate.
Thus concepts do not represent, but conceptual function is embedded within a representational function through which representational systems map the worlds they inhabit. In other words, the roles of conceptual categories are embedded in and conditioned by the mapping function. Sellars illustrates this in his account of “robot picturing” in “Being and Being Known.”11 The robot’s wiring diagram determines transformations from sentences to other sentences in accordance with mathematical and logical principles. In addition, it must also contain the equivalent of inductive generalization such that if its tape contains sentences pairs like
Lightning at p, t thunder at p+Δp, t +Δt
And no sentences pairs like
Lightning at p, t peace at p+Δp, t +Δt
Then it prints sentences such as
Whenever lightning at p,t, thunder at p +Δp, t +Δt
In the conceptual order, which Sellars calls the order of signification, the tape pattern ‘::’ signifies lightning and the pattern ‘::, 9, 15’ signifies lightning at place 9 and time 15. Here we have established a functional equivalence between the Robotese sign-design ‘::’ and the English sign design ‘lightning,’ as well as one between ‘::, 9, 15’ and ‘lightning at place 9 and time 15.’
But in the real order—i.e., the spatiotemporal order in which both the robot’s representings and the objects represented by it exist—it is possible to establish a systematic correlation between certain ‘matter-of-factual’ properties of its representational states and certain ‘matter-of-factual’ properties of the objects that it represents. The correlation has to be established as relations between matters of facts—which is to say, in terms of a set of cognitively discernible pattern-governed regularities—because it is not God-given: there are of course any number of more or less arbitrary ways in which one could establish such a correlation (which is why Rorty and other left Sellarsians reject Sellars’s picturing constraint on meaning altogether). But Sellars’s claim is that the correlation is constrained by the fact that representational systems are products of their environments—thus the ways in which they can represent their environments are delimited by certain fundamental features of those environments. For Sellars, the correlation is generated by the mapping function through which natural selection obliges representational systems to generate more or less adequate pictures of their environments. Of course, the crucial questions are whether we can identify one and only one relevant mapping function and what the proper criterion of pictorial adequacy might be.
If we grant its existence, the mapping function will account for the fact that there is a systematic correlation between tokenings of ‘::’, ‘éclair’, and ‘lightning’ in Robotese, French, and English, and instances of lightning in the world. There are matter of factual properties relating particular occurrences of these inscriptions and vocalizations to particular occurrences of lightning. This system of relations constitutes a pattern in the causal order and it is this pattern which incarnates the rule. But crucially for Sellars, the regularities in the real order, which is to say, the regularities at the level of picturing, are the condition for the functional equivalences that obtain at the level of signification:
Isomorphism in the real order between the robot’s electronic system and its environment is a presupposition of isomorphism in the order of signification between robotese and the language we speak.12
This is to say that while espousals of principle are logically irreducible to regularities of performance, such regularities provide the causal conditions for these espousals. Thus it is not linguistic competence that provides the criterion of pictorial adequacy, but rather pictorial adequacy that furnishes the criterion for linguistic competence. This is to say that the correctness of a picture provides the criterion for gauging the correctness of a linguistic performance:
Linguistic picture-making is not the performance of asserting matter-of-factual propositions. The criterion of the correctness of the performance of asserting a basic matter-of-factual proposition is the correctness of the proposition qua picture, i.e., the fact that it coincides with the picture the world-cum-language would generate in accordance with the uniformities controlled by the semantical rules of the language. Thus the correctness of the picture is not defined in terms of the correctness of a performance but vice versa.13
What does Sellars mean by “the world-cum-language”? If espousals of principle are reflected in uniformities of performance, then “the world-cum-language” is the set of uniformities or pattern-governed regularities generated within the natural order through the semantic rules espoused by language-using animals. But note that it is the espousals that generate the regularities, not the rules themselves: Sellars cannot grant causal efficacy to rules without hypostatizing norms as abstract entities and thereby violating his own naturalism, which forbids recourse to supernatural causation. If rules are constituted through the espousals of language-using animals, and espousals are the result of training, i.e., of animals learning to conform to the rules of criticism through which they are inducted into the normative order, then whatever causes the espousal can always be explained as the effect of a regularity rather than a rule. If so, Sellars’s fundamental distinction between rule-governed activity (i.e., reasoning) and pattern-governed behavior (i.e., conditioning) threatens to collapse. And without it, the attempt to ground the correctness of assertion in the correctness of picturing becomes otiose.
Sellars is well aware of this difficulty and provides some of the resources required to address it. The key to his response is the idea that rule-governed conceptual competence is itself a kind of practical know-how, but one generated through cultural rather than biological conditioning. The distinction between pattern-governed behavior and rule-governed activity is not a difference in kind; rather, rule-governed activity is a species of pattern-governed behavior: a recursive loop generated through the interaction between complex patterns. It is a patterning of patterns; but a patterning executed through the same kinds of causal mechanisms that generate patterns in general. Sellars illustrates this idea by modeling the distinction between pattern and rule in terms of the distinction between game and metagame:
Pattern governed behavior of the kind we should call “linguistic” involves “positions” and “moves” of the sort that would be specified by “formation” and “transformation” rules in its meta-game if it were rule obeying behavior. Thus, learning to “infer,” where this is purely a pattern governed phenomenon, would be a matter of learning to respond to a pattern of one kind by forming another pattern related to it in one of the characteristic ways specified (at the level of the rule obeying use of language) by a “transformation rule”—that is, a formally stated rule of inference.14
The metagame states the rules governing the game. The rules of a language consist of the formally stated rules of material inference specifying the proper function (i.e., the inferential role) of linguistic expressions. Such rules can only be stated at the metalinguistic level. The rules of an ordinary game specify the permissible ways pieces in the game can be moved. These rules are explicitly stated in the metagame; they are not part of the game itself (they are not pieces in the game). But competence in the game requires competence in the metagame.
The relationship between game and metagame can be illustrated through the following diagrams. In Sellarsian parlance they represent the language-entry transition from game to metagame (perception); the intra-language transition within the metagame (inference); and the language-exit transition from metagame to game (action):
Perceiving a specific configuration of♗ et de ♚ shaped pieces of wood as a bishop checking a king; inferring “If one’s king is threatened by a bishop, interpose a pawn,” and interposing one’s pawn are all rule-governed practical competences akin to those involved in perception, reasoning, and action more generally. But perception, reasoning, and action must enable language users to find their way around in the world and satisfy their needs. If linguistic competence makes a difference in the world, language must be articulated with the world despite the fact that its conceptual structure does not directly reflect the structure of reality. Inferential competence is constrained by the need to map the world correctly.
But what is the criterion of cartographic, which is to say pictorial, adequacy? It is formulated using our extant conceptual categories, and as such is internal to our signifying scheme and dependent upon our available predicative resources. Yet it can still be used to track the correlation between conceptual order and real patterns. Sellars’s theory of picturing is an attempt to articulate the logical, i.e., normative-inferential, powers of concepts together with the empirical or ‘matter-of-factual’ characteristics of the linguistic items in which these powers are incarnated:
[…T]he ultimate point of all the logical powers pertaining to conceptual activity in its epistemic orientation is to generate conceptual structures which as objects in nature stand in certain matter-of-factual relations to other objects in nature.15
This is the point at which the Rortyan objection raises its head. How could the “world-cum-language” or the various matter-of-factual characteristics (shape, size, color, internal structure, etc.) in terms of which we correlate linguistic and nonlinguistic items provide a criterion of correctness for linguistic assertion? Since picking out these empirical facts will depend on our conceptual resources, which are norm-governed, all we are doing is comparing facts with other facts; specifically, facts about linguistic objects with facts about nonlinguistic objects. But the criterion of correctness will be internal to our system of linguistic conventions in both cases. Facts about pictorial adequacy are just that: facts. And like all facts, they will depend on historically circumscribed fact-stating resources, just as they will be intelligible only within the extant space of reasons in terms of which we justify all assertions concerning matters of fact. Thus picturing fails to provide a truly independent, which is to say, nonnormative criterion of adequacy for the alleged correspondence between the normative and real orders. Since any mapping function correlating the factual properties of linguistic items with those of nonlinguistic items will be more or less arbitrary, we cannot use it to establish a criterion determining the degree of pictorial adequacy between linguistic assertions and nonlinguistic reality. Given the arbitrariness involved, we could just as reasonably proclaim an ever-increasing divergence, rather than convergence, between our linguistic pictures and nonlinguistic reality.16
Such considerations doubtless underlie Rorty’s skepticism about Sellarsian picturing. But the objection misses something important. Sellars’s claims that logical powers have a “point” and that conceptual activity is endowed with an “epistemic orientation” need to be taken seriously. What we know about the world is always accompanied by what we know about our knowing about the world. Empirical science is not just the accumulation of facts about the world but also (and increasingly) the accumulation of facts about how we know the world. These facts help us orientate ourselves: they contribute to a narrative of our cognitive evolution that develops as part of our ongoing understanding of our biological and social history. Cognitive progress is not only charted in terms of knowledge of facts, but also through facts about knowing. And knowledge does not only develop in the dimension of cumulating facts about the world but also in the dimension of integrating facts about knowing into our knowledge of the world. The veritable telos of cognitive enquiry is not exhaustive description but practical transformation: the integration of knowing and doing such that what we know about the world and our place in it allows us to transform both it and ourselves in order to realize our various purposes; purposes which are not fixed but perpetually redefined in light of what we come to know.
Thus knowledge of matters of fact is rooted in logical powers (powers of inference), but these logical powers must also be understood as rooted in objects in nature bearing certain factual relations—and hence natural capacities—vis-à-vis other objects in nature. The criterion of adequacy for picturing is internal to the signifying order because it is through reasons that we formulate our purposes. We do not measure this adequacy by stepping outside the conceptual order and comparing its degree of correspondence to the nonconceptual order. Rather, we use the conceptual resources of matter-of-fact discourse to try and make material mode statements about the second-order isomorphism between the properties connecting representings and the properties connecting the objects they represent. The mapping function does not preexist this discursive activity; it is constituted in and through the discourse that seeks to capture it. The attempt to uncover causal invariances between certain features of assertions and certain features of objects is part of the activity that contributes to the determination of the function. Thus the criterion of pictorial adequacy is also practical, not just theoretical: it is formulated in terms of the degree to which what we know about the correlation between representing and represented allows us to realize our purposes in the world. The adequacy at issue here is practical and transformative, not theoretical and contemplative. For Sellars, as for Hegel, the ideal is not an inert supernatural phantasm, but something that actualizes itself in and through the real. Because we are norm-governed creatures, our performances can be judged in terms of the principles they ought to embody. Even the claim that our performances fail to embody these principles presupposes the authority of the ideal on which we have defaulted. In this regard, empirical facts about the systematic uniformities between linguistic items and nonlinguistic objects are still facts about objects, not about concepts. That we need norms to state facts does not entail that all the facts we state are ultimately about norms. The Rortyan objection moves from the premise that all factual properties are norm-governed to the conclusion that all factual properties are normative properties. The premise is true but the inference is invalid and the conclusion is false. Thus there is nothing intrinsically incoherent about Sellars’s claim that we can use concepts to determine the degree of pictorial adequacy that concepts bear to nonconceptual reality.
The Immanence of Representation
What can we conclude from this account? First, a lesson concerning immanence. What is transcendentally immanent is the difference between representables and things-in-themselves, not the fusion of sensing and being proclaimed by philosophies of immanence (Bergson, Michel Henry). The transcendental difference between representables and things-in-themselves is not a two-world theory (sensible/supersensible), but a double-aspect theory about a single, immanent world. The distinction between the sensible and the supersensible is methodological, not ontological. The manifest world of intersubjective experience—encompassing both the public and private domains—is empirically real in the only acceptable sense of ‘empirical.’ What is immanent is our corrigible, justifiable, and shared knowledge of ourselves and our world. But this means that those philosophies of immanence which begin from an experience allegedly lying beneath or beyond judgment, categorization, and representation, begin from an abstraction. The way towards absolute knowing does not lie in plunging deeper into the alleged ineffability of subjective immediacy. It starts with the reflexive stratification of immanence into representing and represented, and the gradual recognition that what we know about the latter (the represented) is conditioned in ways we don’t yet know by the former (our representings). Objective knowledge remains incomplete unless supplemented by knowledge of objectivating structure. This structure is spatiotemporal in a transcendental rather than empirical sense.
Thus there are two dimensions of spatiotemporal structure: the one which we represent, and the one in which our representing unfolds. The goal of cognitive enquiry consists in incorporating ever more facts about the structure of representing into every represented fact. This would be the naturalization of the involuted spiral of absolute knowing. In this sense, spatiotemporal location provides the transcendental coordinates for our species’ collective world story:
The “transcendental” or epistemic function of spatio-temporal concepts as forms of representing must be distinguished from their empirical function in matter-of-factual judgments about historical fact.
In linguistic terms this means roughly that spatiotemporal predicates are essential not only to object-language statements, but to the metalinguistic statements that ascribe logical (epistemic) powers to linguistic forms.17
It is not only what is represented that is represented as existing at a particular location in space and a particular point in time; representing itself is located in space and is actual in time—but a noumenal space and time that, although conceived as partially analogous to the space and time proper to perceptual experience, possess their own distinctive structures to be uncovered through some future alliance of physics and neurobiology. What this amounts to is the claim that the logical powers of the concepts through which we apprize spatiotemporal reality are themselves spatiotemporally conditioned. As it progresses, the history of what we know incorporates within itself more and more facts about the empirical structure of knowing. The limit of this movement would be the point at which empirical (sigma-tau) facts about the structure of knowing are incarnated in the structure of empirical (spatiotemporal) facts.