Robert Brandom

Wilfrid Sellars is the greatest American philosopher since Charles Sanders Peirce. He is the most profound and systematic epistemological thinker of the 20th Century. When intellectual historians look back at the progress of philosophy, our century will appear as a time when, after three hundred and fifty years, we finally saw through Descartes. It is in our time that the collection of puzzles and problems that have collected around the Cartesian dualism of body and mind has been supplanted by those associated with what now appears to be the more fundamental Humean-Kantian dualism of fact and norm, which appeared only darkly and misleadingly in its Cartesian guise. Heading the list of names they will associate with that conceptual sea-change are Wittgenstein and Sellars.

Our century will also appear as having placed language at the center of the semantic stage that categories of thought used to occupy. Concern with ideas has been succeeded by concern with meanings and sentences. Heading the list of names they will associate with that shift of problematic are those of Wittgenstein, Dummett, Quine, and Sellars. Alone among this distinguished group, Sellars never allows his recoil from Cartesian models of the mind to drive him into behaviorist hostility to inner episodes as such -- he has always taken seriously the fact that "all is not dark within".

The last hundred years have been a time of unprecedented progress in formal logic. When the historians look back to the crucial philosophical appreciations of the expressive role played by formal logistical idioms and its lessons for understanding the structure of the scientific enterprise and its products, heading the list of authors will be Frege, Russell, Carnap, and Sellars.

I could go on. When you keep this sort of company, being a dead philosopher is not such a bad job. For we don't just read the mighty dead, we live and talk with them, we converse with and are confounded by them daily. In just this way, Wilfrid lived his life with Kant, as a constant conversational companion -- learning from and correcting him, questioning and challenging, and being questioned and challenged by him. We philosophers eat our dead. They are our nourishment and they keep us alive and surely as we do them.

Rather than simply praising Sellars in general terms, I'd like to say something about the content of his views. I want to display a piece of the system. I'll sketch just one central strand of thought that occupies a strategic position with respect to the philosophy of language, the philosophy of mind, epistemology, and philosophy of science. The restriction to expounding one aspect of Sellars' account is not just a result of time constraints. None of us is deep enough to see to its bottom, nor broad enough to see it whole. I'll talk about his views on three topics: perception and observation, immediate awareness, and the postulation of theoretical objects.

One way into the first topic is to ask what the difference between observation and mere differential responsiveness consists in. What is the difference between your capacity to perceive and report red things and that of a photocell that responds differentially to red things? As an intermediate case, we might consider a parrot trained to say "That's red" when confronted with red things, or even a prelinguistic infant. A point commonly made in the tradition is that there is no consciousness that is barely of something. Consciousness is always of something as something. This is the classificatory theory of consciousness common to abstractionism and its opponents. But the photocell does classify stimuli, precisely by responding differentially to them. In just this sense a chunk of iron classifies its environments, distinguishing two classes accordingly as it rusts in some and not in others. Sellars says that perception is different in that it has a conceptual component. Perception, and so observation, requires not just differentially responsive classification, but differentially responsive classification under concepts. The difference between you and the photocell is that you do, and it does not, grasp the concept red, which you are differentially disposed to apply. The difference consists in the understanding, or grasp, of contents, and it is this understanding that characterizes an observer.

But what is it to grasp a concept? Descartes paints a picture of a spooky kind of metaphysical light going on, but does not have much helpful to say about what it consists in. Sellars, by contrast, has an account of concepts in terms of inference. Your grasp of the concept red is more than the mere capacity to respond differentially to red things, in that you know what the inferential significance of applying the concept red is -- you know what follows from something's being red, for instance that it is colored, not a prime number, and so on. The response you are differentially disposed to produce is inferentially articulated. It occupies a place in a network of reasons that can be given for applying it and concepts it can be a reason for applying (or not applying). Inference, for Sellars, is a normative affair. To say what is a reason for what is to say something about what judgments one ought to make, not necessarily what judgments one does in fact make. Causally reductive accounts of the fundamental semantic relations of inference and incompatibility, by contrast, ignore this normativeness, and have a hard time explaining how it is possible to have incompatible beliefs, or to believe a premise but not the conclusion it entails. Isn't there a circularity in appealing to our 'knowing' what follows (i.e. what it would he correct to infer) from a claim and what it follows from? No, because the knowledge in question is know-how, a practical ability to discriminate, that is, differentially respond to, good and bad inferences. The responses that you are differentially disposed to produce in this case will involve the application of further concepts, whose inferential articulation must also be mastered in practice. The understanding that distinguishes your perception from the photocell's differential responsiveness and the parrot's, is an inferentially articulated normative status. Sellars' account of such statuses is social, and linguistic.

Related to this story about inferential articulation and perception is a counter-Cartesian account of the relation between perceiving how things are and how things seem. For it might be tempting to say in response to this way of making out the difference between us and the parrot: "Agreeing that the parrot does not grasp the concept of red, because its responses do not for it play a role in a normative structure of inference and incompatibility (as opposed to the non-normative incompatibility of, for instance, being unable to respond simultaneously to something as red and as green), still there is some kind of perception that the parrot has, in common with us, and that the photocell does not. At least the parrot can understand looking-red, a recurring 'that way'-ness in its awareness."

Not so, says Sellars. The two opposing views line up like this. Descartes noticed that we can be wrong about how things are, but not about how things seem. He concluded that how things seem constitutes what we know or understand immediately, without dangerous inferences that go beyond what is given in awareness. He went on to use such 'seemings' as the model for understanding the mind in general, the result being his epistemic definition of the mind. He also concluded that such seemings could serve as a foundation for knowledge and understanding, because grasp of seemings presupposes no nonimmediate (inferential) knowledge or understanding.

Sellars' (contrary) view is that perception of how things are requires both reliable differential responsive dispositions and the capacity to respond with a judgement, the endorsement of a claim, that is, by adopting a normative attitude toward an inferentially articulated claim. Mastering the concepts applied in reports of how things look or seem involves withholding the usual endorsements of responsive inclinations. Thus if I have good reason to believe that the thing in my hand is not red, but am responsively inclined (by my training and wiring) to call it red, I can express the responsive disposition without endorsing it by saying that it looks red. The extent of my willingness to endorse such dispositions, for instance in a case of bad light or brief exposure may vary. This explains the possibility of merely generic, or qualitative lookings (e.g. how a polygon can look merely many-sided, without there being any definite number of sides it looks to have), which are hard to assimilate on other approaches.

This account explains the incorrigibility of 'seems' judgements, from which Descartes takes his departure. There have been many critics of this Cartesian line besides Sellars, but only Sellars has offered an alternative account of the fundamental datum. Sellars is a systematic philosopher, and as such is not content to diagnose and treat an explanatory disease. His concern is not solely therapeutic in a negative sense. He wants to reinfect the patient with a better explanatory idiom or conceptual scheme -- not just to bring about a return to some hypothetical antecedent mean of health, but positively to invigorate and empower. Thus, according to Sellars' story, the incorrigibility of 'seems' judgements simply reflects the fact that they withhold the usual endorsements of 'is' judgements. There can be no mistake where there is no endorsement. The claim is already taken back. By the same token, this incorrigibility is revealed as trivial, and totally unsuited to being foundational. In particular, one must be able to use inferentially articulated concepts (and so must endorse some inferences) even to grasp the concept looks-red (to have the know-how regarding inferences in the vicinity of the endorsements one is withholding). Thus a 'seems' judgement is not an immediate cognition. Perception of how things look, as of how things are, requires mastery of inferentially articulated concepts. The lesson is: One can not have a language or conceptual scheme in which all judgements can only be arrived at noninferentially. One can have a language or conceptual scheme in which all judgements can be arrived at both noninferentially (on some occasions) and as conclusions of inferences (on other occasions). "The fabric is red," may express either a commitment arrived at noninferentially by observation (perception) or one inferred from one expressed by the "The fabric is scarlet." A point to keep in mind is that it is the same claim that can be made either as a report or as the conclusion of an inference. The content does not differ in the two cases. If it did, perception would be cognitively idle.

This story about the relation between inferential and noninferential applications of concepts (the idea of one and the same item being at once a causal response to the environment and a normatively significant item in a network of reasons) is then extended, in Sellars' account, to include theoretical concepts. Philosophers have worried over the ontological status of these unobservable, or merely theoretical objects. Insofar as they are cognitively accessible, they are so only to reason, not to the senses. What kind of spooky thing are they? How are they different from the ordinary kind of object that can be seen and felt and so on? Instrumentalists even deny that there are any such objects. Theoretical claims are just make-believe, not really about anything at all. Such talk is just an instrument, a calculational device for keeping straight claims about observable objects (the only kind there really is). Sellars understands theoretical claims as purely inferential claims, that is, claims one can only become entitled to as the conclusions of inferences. This is an optional stratum of language -- there need not be any theoretical vocabulary in an empirical language. But to talk about how we can become entitled to a claim is one thing. To talk about what kind of thing the claim is about is another. There is no systematic difference between the kinds of things we reason about and the kinds of things we observe. The difference between theoretical objects and observables is not an ontological difference, but a methodological difference. It is a difference not in the kinds of things, but only in their relation to us. The planet Pluto was postulated to explain perturbations in the orbit of Neptune. At that point the only way to become entitled to claims about, say, its mass, was inferentially. When instruments improved to the point that we were able to observe it, it changed status from a theoretical to an observable object. But it didn't change, only its relation to us changed. One might say in response to this that in that case Pluto was not an unobservable, but only unobserved, and similarly for genes (finally observed as DNA long after Mendel postulated them). But Sellars agrees with the instrumentalist (against many realists) that there are only observable objects. For once you have adopted his theory (already discussed) of what is needed for observation, and so discarded the notion that something is immediately or directly observable, it follows that nothing is in principle unobservable. De facto unobservability remains, but this is no more serious than the fact that some things are, in fact, unobserved.

Sellars resolutely turned his face against the pervasive Platonic prejudice that corresponding to different kinds of knowledge must be different kinds of object. Noninferential, inferential, and purely inferential claims can all be about the same kind of object. I think in the present circumstances we can take some comfort in this robust realism. Just as the object of knowledge, admiration, and affection does not change fundamentally when our mode of cognitive access to it shifts from noninferential to inferential, it does not alter when, as is the case with Wilfrid, that access shifts from the faculty of perception to that of memory.

The noumenal Wilfrid has always been his thought. That's still here. <\b>