The notes that follow were developed over the years to help my graduate and advanced undergraduate students at the University of Pittsburgh to see their way through the textual trees to the Sellarsian forest. They are meant to provide only a first take on the material, to indicate the most general outlines of the structure of the essay and of the thought behind it. To that end, many philosophically interesting issues and discussions have been brushed past. In particular, I have sedulously avoided discussing genuinely esoteric issues—such as the philosophical significance some have professed to find in the distinction between ‘red’ paragraphs and ‘green’ paragraphs. The formulations and characterizations that are provided are not intended to be definitive or authoritative. They aim to provide a place to start in reading this rich and difficult text.
The idea for such a document, and the notes to the concluding sections, had their origins in a handout Rorty circulated for similar purposes when I was a graduate student at Princeton in the '70s. I am grateful to my colleague John McDowell, and to our former student Danielle Macbeth, for many suggestions and improvements. It should be noted, though, that where their comments evidenced substantive disagreements about what Sellars is (and ought to be) saying—concerning in particular the intricacies of ‘looks’ talk in relation to reports of the presence of secondary qualities, and the various theses and commitments involved in scientific realism—I have stuck to my own readings. The errors that remain, both those of omission and of commission, should be charged to my account alone.
Note: Section numbers of “Empiricism and the Philosophy of Mind” are indicated in square brackets: . On the rare occasions where sections of this guide must be referenced, I use double brackets: [].
Wilfrid Sellars' "Empiricism and the Philosophy of Mind"
Part I - : An Ambiguity in Sense-Datum Theories
Section 1: Sellars announces that his project is to attack "the whole framework of givenness". By this he does not mean to be undercutting the distinction between judgments we arrive at noninferentially, paradigmatically through perception, and those that are arrived at as the conclusions of inferences. Indeed one of the positive tasks of the essay is precisely to tell us how to understand noninferential reports without insensibly sliding into the constellation of philosophical commitments Sellars calls “the Myth of the Given”. Sense-datum theories, his immediate target, are important only as prominent and influential instances of the appeal to givenness. We will have to learn to recognize such appeals in many less obvious guises.
In these opening sections, the Myth of the Given shows up in the guise of the idea that some kind of non-epistemic facts about knowers could entail epistemic facts about them.1 Epistemic facts about knowers are in the first instance facts about what someone knows (though we will come to see that facts about what one merely believes are equally ‘epistemic’ facts in Sellars’ sense). One of Descartes’ signal innovations was to define the mind in epistemic terms: for a state to be a mental state is for being in that state to entail knowing that one is in that state (transparency, ruling out ignorance) and for believing that one is in that state to entail being in that state (incorrigibility, ruling out error). The mind is the realm of what is known immediately, not just in the sense of noninferentially, but in the stronger sense that its goings-on are given to us in a way that banishes the possibility both of ignorance and of error. (Descartes’ thought was that if anything is known to us mediately, that is, by means of representations of it, then something—some kind of representations—must be known to us immediately, on pain of an infinite regress.) Sellars will try to show us that the Cartesian way of talking about the mind is the result of confusion about the distinction between epistemic and non-epistemic items, and the roles they can play in various sorts of explanation.
In its most familiar form, the Myth of the Given blurs the distinction between sentience and sapience. This is the distinction between being aware in the sense of being merely awake (which we share with nondiscursive animals—those that do not grasp concepts), on the one hand, and being aware in a sense that involves knowledge either by being a kind of knowledge, or as potentially serving to justify judgments that so qualify. The “idea that a sensation of a red triangle is the very paradigm of empirical knowledge” , is a paradigm of the sort of conflation in question. The Myth of the Given is the idea that there can be a kind of awareness that has two properties. First, it is or entails having a certain sort of knowledge—perhaps not of other things, but at least that one is in that state, or a state of that kind—knowledge that the one whose state it is possesses simply in virtue of being in that state. Second, it entails that the capacity to have that sort of awareness, to be in that sort of state, does not presuppose the acquisition of any concepts—that one can be aware in that sense independently of and antecedently to grasping or mastering the use of any concepts (paradigmatically through language learning).2 The conclusion of Sellars’ critical argument is that these two features are incompatible: only what is propositionally contentful, and so conceptually articulated, can serve as (or for that matter, stand in need of) a justification, and so ground or constitute knowledge. Davidson expresses a version of this thought with the slogan “nothing can count as a reason for holding a belief except another belief”. Sellars’ thought is better captured by changing this to “nothing can count as a reason for endorsing a believable except another believable,” where believables are the contents of possible beliefs, that is, what is propositionally contentful.3
Sellars understands propositional contentfulness, what is epistemic in the sense of being a candidate for knowledge, in terms of role in what he calls “the game of giving and asking for reasons”. “In characterizing an episode or state as that of knowing, we are not giving an empirical description of that episode or state; we are placing it in the logical space of reasons, of justifying and being able to justify what one says,” . To treat something as even a candidate for knowledge is at once to talk about its potential role in inference, as premise and conclusion. Because a crucial distinguishing feature of epistemic facts for Sellars is that their expression requires the use of normative vocabulary, to treat something as a candidate for knowledge is also to raise the issue of its normative status. The Myth of the Given eventually appears as “of a piece with the naturalistic fallacy in ethics”—the attempt to derive ought from is.4 This is because talk of knowledge is inevitably talk of what (conceptually articulated propositional contents) someone is committed to, and whether they are in various senses entitled to those commitments.
Section 2: Here Sellars distinguishes between the act or episode of sensing, on the one hand, and the content of that act, what is sensed, which is called a sense content, on the other. When one hallucinates a pink elephant, doing so is sensing, and the sense-content is what makes it an of-a-pink-elephant hallucination, rather than for instance an of-a-green-Norway-rat hallucination. In ordinary perception, the contents sensed must be carefully distinguished from the external objects sensed (which are entirely absent in the case of hallucinations).
Section 3: Now consider the suitability of sensings of sense contents as foundations of knowledge and justification on the Cartesian model.
The general idea of a foundation for knowledge can be sketched as follows. Our beliefs constitute knowledge only insofar as they are not only true, but justified—lucky guesses don’t qualify. One claim or belief can justify another to which it is inferentially related. If one is justified in a commitment to the claim that p, and q may be inferred from p, then one may for that reason be justified in a commitment to the claim that q. To say this is to offer a mechanism whereby justification can be inherited. But, the thought is, not all commitments that are justified can have inherited that status inferentially from others. There must be some other mechanism for acquiring positive justification status, to give the inheritance mechanism something to pass along. If p1 inherits its status from p2, and p2 inherits it from p3, and so on, then either:
at some point a claim is repeated (some pn is identical with a pm for m
there never is a repetition, in which case an infinite regress arises, in which each pn has the anomalous status of an unjustified 'justifier', which is not itself justified until an infinite number of other claims have been justified.5
The conclusion is that there must be some way of being justified without having to be justified. We ought to distinguish two senses of ‘justification’, one indicating a status (being justified), and the other making reference to a process (justifying) that can result in possession of the status.6 Then the conclusion is that there must be some other way of acquiring positive justificatory status besides justifying it in the sense of offering a justification. Besides inferential inheritance, there must also be some noninferential acquisition mechanism for this epistemic status.
So far, so good. Descartes concluded from this line of thought that there is a kind of claim or belief, call them basic beliefs, that form the foundation of all other beliefs in the sense that they are the font from which the justificatory status of all the rest flows inferentially. This does not follow, but Sellars will not contest it.7 Descartes believed further that unless those beliefs were certain (the ultimate positive justificatory status), none of those inferentially based upon them could even be probable (as C.I. Lewis put it in Mind and the World Order). Descartes gave philosophy a decisive epistemic turn which was, at least until Kant, confused with a subjective turn. The latter is a consequence only of Descartes' peculiar and optional way of working out the former. For he defined the mind by its epistemic status, as what is best known to itself by falling within the reach of the subject's incorrigibility and local omniscience. This epistemic definition is what motivates the assimilation of events whose contents are structured like sentences, such as thinking that Vienna is a city in Austria, and events whose contents are structured like pictures, such as imagining or seeming to see a red triangle inside a green circle.
To return to the idea of using sensings of sense contents as a foundation of knowledge, then, a process is pictured something like this:
Sensings of Sense Contents
In the standard perceptual case, it is because there is a red object with an octagonal facing surface in front of me that I find myself with a sensing of a red-and-octagonal sense content. It is because I have such a sense content that I acquire the noninferential belief that there is a red and octagonal object in front of me. And it is because I have this belief, together, perhaps, with other beliefs, that I am justified in the further inferential belief that there is stop sign in front of me.
The point to focus on is the nature of the 'because's. The first (arrow 1) can be understood as a causal notion, perhaps the sort studied by students of the neurophysiology of perception. As such, it relates particulars describable in a nonnormative vocabulary. This is a matter-of-factual, nonepistemic relation. The final 'because' (arrow 3), on the other hand, indicates the sort of relation Sellars calls 'epistemic'. It is an inferential notion, relating sententially structured beliefs (or believables) which are repeatable abstracta—a matter of reasons rather than causes. This justificatory relation is not a natural one, but a normative one; it is not the empirical scientist, but the logician or epistemologist who has the final say about it..
The question is: what sort of relation is the middle one (arrow 2)? Does it belong in a box with the first, causal relation, or in a box with the third, inferential relation? How are the sensings of sense contents to be conceived as related to (potentially foundational) noninferential beliefs? Here is where the distinction between the epistemic and the nonepistemic, between particulars specified in the language of causes and believables specified in the language of reasons, comes into play.
Suppose that one understands the sensing of a sense content to be the existence of a nonepistemic relation between one particular, the sense content, and another, the person doing the sensing. (This is the position Sellars himself eventually endorses.) If so, then it is hard to see how the sensing of a sense content could entail or justify a claim, for instance a noninferential belief. For only things with sentential structure can be premises of inference, not nonepistemically specified particulars. For this reason sensings, understood in terms of nonepistemic relations between sense contents and perceivers, are not well suited to serve as the ultimate ground to which inferentially inherited justification traces back. Since the occurrence of such a sensing does not entail commitment to any claim, it would be possible to have one without coming to believe anything, and certainly without coming to know anything (for this latter requires positive justificatory status). So it seems the foundationalist who wants to appeal to sensings as foundational must take the sensing of a sense content to be an epistemic fact about the sensing agent. But if so, what becomes of the particular?
Sections 4 & 5: The sense datum theorist can treat sensings as epistemic noninferential beliefs, from which inferences may be made and justification status inherited, so that sensings can perform their foundational function. To retain a role for the mental particulars that are sensed (sense data, sense contents conceived as a kind of sensed object), that theorist must be willing to say the following: "The primitive notion is believing that sense content x has property F. To sense the sense content x is to believe that it has some (no matter what) characteristic F. The sense content, which is a particular, is the intentional object of the epistemic sensing." The important thing to notice about this analysis is that epistemic notions are presupposed, not accounted for in terms of a supposedly antecedently understood nonepistemic notion of sensing a sense-content (thought of as a relation between a subject and a sense content, both of them particulars). In fact Sellars believes that no such reduction of the epistemic to the nonepistemic is possible, even in principle—though his arguments will not depend on this claim.
Section 6: At this point a further consideration is introduced: the ability to stand in the passive causal relations to the physical world envisaged by the fans of givenness is not something that must be acquired through experience or training. Organisms of the right sort get it just by being awake. But the capacity to have beliefs of the form 'x is F' involves classifying unrepeatables or particulars under repeatables or universals. It is natural to think that the capacity to classify is acquired, since one must learn by experience and training what the boundaries of the classes are. This line of thought results in the inconsistent triad of claims the sense datum theorist is committed to, and would like to be entitled to:
A. ‘S senses red sense content x,’ entails ‘S noninferentially believes (knows) that x is red.’
B. The ability to sense sense contents is unacquired.
C. The capacity to have classificatory beliefs of the form 'x is F' is acquired.
If A is given up, the sensing of sense contents becomes a nonepistemic event, which can at best be a logically necessary condition of knowledge or noninferential beliefs, not a logically sufficient condition of it. To take this way out would be to discard the line pursued in  & . If B is given up, the sense datum theorist must either claim that we need practice to feel pain, hunger, itches, and so on when we are infants, or claim that feeling these things is not sensing. But then what is sensing? If C is given up, a story must be told about what universal concepts are innate (unacquired, inborn, wired-in) and which are not. This would require much more than even latter-day innatists such as Chomsky have claimed, since substantive concepts like red and tall, not merely grammatical forms, would have to be innate. A is the Myth of the Given, in one of its forms, and Sellars will give it up. He'll then owe (and provide) us a new account of both thoughts and sensations, and of the origins (both in the order of causation and the order of justification) of knowledge.
Section 7: Sellars' diagnosis, which is not yet a treatment for the conceptual illness of givenness, is that it results from confusing two trains of thought, the first derived from an attempt to give a scientific account of perception and the acquisition of empirical information, and the second from an attempt to give a foundational epistemological account on the Cartesian model canvassed above in the discussion of :
The idea that there are certain inner episodes -- e.g. sensations of red or of C# which can occur to human beings (and brutes) without any prior process of learning or concept formation; and without which it would in some sense be impossible to see, for example, that the facing surface of a physical object is red and triangular, or hear that a certain physical sound is C#.
The idea that there are certain inner episodes which are the noninferential knowings that certain items are, for example, red or C#; and that these episodes are the necessary conditions of empirical knowledge as providing evidence for all other empirical propositions.
The first class consists of particulars, picked out by their causal role. The second consists of claimings structured like sentences, picked out by their inferential or justificatory role. Sellars will offer an account (starting in ) of the genus, inner episode, to which these two species belong. He will call the first kind ‘sense impressions’, and the second kind ‘thoughts’, and will describe the roles they play. Finally, he will explain how they are related in human knowledge. (I have talked about belief so far, where Sellars talks about knowledge, in order to emphasize that the question of the justification of or warrant for noninferential beliefs has yet to be discussed.)
The result of running together these two lines of thought is “the idea that a sensation of a red triangle is the very paradigm of empirical knowledge.” That idea is subject to precisely those related ‘perplexities’ Sellars has pointed out:
Should we think of the sensation in question as a kind of particular (structured like a triangle), or as a kind of belief (structured like a sentence)?
Is the capacity to have empirical knowledge like this acquired by experience, or prior to experience?
Is it prior to the rest of our knowledge in the order of causation, or in the order of justification and evidence?
Part II - : Another Language?
Section 8: This and the next two sections (both marked in the original as Section 9!) are in one way an aside. The main thread is picked up again in . The excursus is used to introduce some important ideas that will be discussed further along. The topic here is one possible form a sense datum theory might take to avoid the nonepistemically-specifiable-particular vs. only-epistemically-specifiable-sententially-structured-premise dilemma Sellars is constructing for it. One might give up entirely on the nonepistemic side of things, and embrace the foundational noninferential belief side. Thus Ayer sees sensing-of-sense-data talk as equivalent to and derivative from talk about how things look or seem to a subject. The suggestion comes in three parts:
There is a class of noninferential beliefs that form a justificatory basis for the rest of our empirical beliefs. [Note that this would be sufficient to respond to the regress argument sketched above in [], though as suggested there, it is not a necessary condition for a response.]
Three nested descriptions of a phenomenon. First, a platitude: I may be mistaken that there is a red triangle in front of me. It is not possible for me to be mistaken about there seeming to be one. Next, a reifying move: an application of the Cartesian principle that although appearance must be distinguished from reality since subjects can be in error about the latter, on pain of an infinite regress it cannot be that one might be mistaken about the former also. Finally, a foundational claim: The class mentioned in (a) consists of beliefs that would be expressed by sentences used to make perceptual reports, prefixed by a special operator “It looks to me now that…,” "It seems to me now that..." or "It now appears to me just as though….”
Sentences of the form "S is having (or is aware of) a sense datum that is F" (say, red and triangular) are by definition equivalent to sentence of the form "It seems to S that he senses something F". On this understanding, there are no particulars that are sense data -- the apparently referential singular terms that give the contrary impression must be understood contextually, like the 'it' in 'it is raining'.
Section 9: Here Sellars offers an observation about this approach, and then formulates a dilemma for it. The observation regards merely generic lookings. Something can look polygonal without there being any determinate number of sides that it looks to have. But nothing can be polygonal without there being a determinate number of sides that it has. (This contrasts will be explored in .) So the inferences one is permitted to make in sense datum talk as introduced by the equivalence asserted by (c) are not the same as those licensed by the sense datum theorist’s talk of sense data as particulars (for which the above 'inference to further determination' goes through). Thus the code is misleading.
Section 9 bis: The dilemma presents a more serious objection. If sense datum talk is just a code, it is redundant (insofar as it is not misleading). So what good is it? It can't explain anything about seemings or appearance. To do that it would have to be a theory of appearings, explaining them by relation to a certain kind of particular, namely sense data. (Sellars begins to explain how he thinks about theoretical explanation in  and . We then hear a lot more about this topic in the second half of the essay, beginning at -.) But this would reintroduce the strand of thought (1) above (in  and []), which the code theory is formulated precisely to avoid. The lesson is that that strand of thought is not altogether mistaken. The account Sellars will offer provides a theory of appearings, and will embrace and reconcile (1) and (2), properly understood. So (c) is not a way to avoid the problem. It allows us to look, however, at the assumptions (a) and (b) to which it was conjoined. Sellars’ conclusion is that this line of thought is committed already at step (b) to the possibility of inferring from claims exclusively about how things seem to claims about how things actually are. But if, as (a) and (b) assert, all empirical evidence ultimately derives from how things seem, it is clear that such an inference cannot be warranted empirically, by inductive correlation of appearances and realities. The alternative seems to be to find a definitional reduction according to which "ordinary discourse about physical objects and perceivers could (in principle) be constructed from sentences of the form 'There looks to be a physical object with a red and triangular facing surface over there'." Since commitment to (a) and (b) is much more widespread than commitment to (c), it is important to see what is wrong with the view they express—why the reduction they presuppose is impossible. To that end Sellars turns to the logic of ‘looks’ or ‘seems’ talk.