The philosophy of Wifrid Sellars is an integrated and complex metaphysical system. Because of this, much of his work is quite obscure to many philosophers, even those who have tried diligently to penetrate beyond a rudimentary grasp of seemingly isolated points. This lack of knowledge of Sellars' work is a serious deficiency, I believe, because Sellars is one of the very few philosophers today who both attempts to produce a unified view of the world and man, and is also at home with the devices and techniques developed in the various branches of analytic philosophy during the twentieth century. There is, I find, much to be learned by ontologists and by linguistic philosophers through an understanding of his work.

One thing that would profit both the frustrated readers of Sellars and Sellars himself would be a careful attempt to explicate and evaluate critically the many interrelated theses stated and defended by Sellars. But, so far as I know, there has been little work of this kind done. I know only of two fine reviews by Keith Lehrer and Gilbert Harman,¹ and a very helpful expository article by Richard Bernstein² that deal directly and in some detail with Sellars' work. This is not to mention, however, the many philosophers, including myself, who have been influenced by his work. What I believe would be most helpful at this point would be a critique of some central thesis in Sellars' work, that is, an attempt to raise objections to this thesis and to refute the replies that can be anticipated, based upon other of Sellars' claims. This is what I shall try to do in this paper. My hope is that it will [418] further the understanding of his complex philosophy, and help carry forward discussion of the philosophical issues involved.

I SELLARSIAN SCIENTIFIC REALISM

Sellars subscribes to a hypothesis he calls scientific realism, and he attempts to justify this hypothesis the way one justifies a scientific hypothesis by showing that nothing implied by his hypothesis is dubious, and that some of what alternative hypotheses imply is dubious. He is, then, attempting to vindicate a leading hypothesis rather than to validate a conclusion, this is one of the ways some of the seemingly isolated points he makes are related. We can begin to understand Sellars' hypothesis through his claim that "in the dimension of describing and explaining the world, science is the measure of all things, of what is that it is, and of what is not that it is not (SPR, p. 173).³ However, because there are different species of scientific realism, we must make finer distinctions to discover which species Sellars espouses. We can begin by rough characterizations of three kinds which, for ease of discussion, I shall name as follows:

(1) Minimal Scientific Realism: all pure theoretical terms, such as "electron" and "nuclear fission," that are required [419] for the best scientific explanation of the observable behavior of physical objects and persons, refer to (often unobservable) objects or properties. (This is a contrary of scientific instrumentalism, i.e., all pure theoretical terms of science are merely nonreferring symbolic devices that help to warrant inferences from observation premisses to observation conclusions.)

(2) Moderate Scientific Realism: all physical objects have as constituents and properties only the objects and properties referred to by certain of the pure theoretical scientific terms that are required for the best scientific explanation of the observable behavior of the objects.

(3) Extreme Scientific Realism: all physical objects and all persons have as constituents and properties only the objects and properties referred to by certain of the pure theoretical scientific terms that are required for the best scientific explanation of the observable behavior of the objects and persons.

Many philosophers who accept moderate scientific realism would agree with W. V. Quine that sensations are pure theoretical entities and can therefore be eliminated, because there is no explanatory need for sensation-terms. All human behavior can be explained, in principle at least, using only the terms of [420] neuro-physiology and other physical sciences.⁴ Consequently, they would find it unjustified to accept moderate scientific realism and reject extreme scientific realism as I claim Sellars does. The reason he does so, however, is not because he disagrees with Quine about what is required to explain human behavior, but because he thinks that for persons, unlike physical objects, there is more than behavior to be explained. Persons have perceptual propositional attitudes about and conceptual representations of the world around them, and these nonbehavioral facts must be explained in addition to the facts of observable behavior. According to Sellars, then, a person is sometime "under the visual impression that (visually taking it to be the case that) there is (or of there being) a red and rectangular physical object in front of one" (SM, p. 14). These "impressions that" must be explained if science is to explain fully, and, for Sellars, describe fully everything that occurs. And, he says, although sense impressions are not needed to explain the "white rat type descriminative behavior" (IA, sect. 49) we have in common with other animals, they, or at least their theoretical counterparts in scientific theory, are needed to explain such propositional attitudes. Sellars' kind of scientific realism is, then, neither extreme nor merely moderate. We can define it as follows:

Sellarsian Scientific Realism: moderate scientific realism; and all persons have as constituents and properties only the objects and properties referred to by certain of the pure theoretical scientific terms that are required for the best scientific explanation of the observable behavior and the perceptual prepositional attitudes of the persons.

Sellars disagrees with Quine. then, not about what is required to explain human behavior, but with Quine's implication that "correlative physical states'' will be able to take over the explanatory role of all mental entities. More specifically, he claims that they will not be able to take over the role of certain theoretical "inner" particulars, sensa, in the scientific explanation of propositional attitudes. Consequently, being a scientific realist Sellars [421] holds that there are sensa, and they cannot he eliminated for the reasons that postulated theoretical entities, such as the ether, have been eliminated. As a result, he disagrees with a materialist such as Rorty who argues for the elimination of sensations.⁵ He also disagrees with identity theorists such as Feigl and Smart,⁶ because he claims there are sensa and none are identical with brain entities.⁷ Thus although he is a scientific realist, Sellars is neither an eliminative nor a reductive materialist. For many, this would seem to be at best an odd combination of views.

In the rest of this paper I wish to examine Sellars' reasons for thinking that materialism is mistaken and the replies he might make to certain critical objections to his position. In so doing, I shall draw material from a variety of Sellars' writings with the hope that the integrated result will help illuminate the relationships between certain of his seemingly isolated theses. Sometimes, I shall have to extrapolate and interpolate, because in certain central places I find unclarities in statement and invalidities in argument. By doing this, I hope to pinpoint a crucial place in Sellars' work where he must considerably clarify and strengthen his argument for Sellarsian scientific realism.

II A SELLARSIAN ARGUMENT AGAINST MATERIALISM

We can extrapolate the following argument from Sellars' work to prove that scientific realism, at least as he conceives it, implies that materialism is false: [422]

1. If scientific realism is correct, then sense impressions are particulars.
2. It is necessary to assume sense impressions for scientific explanations of certain propositional attitudes of persons.
3. If scientific realism is correct, then if it is necessary to assume entities of a certain kind for scientific explanations of some facts, then there are entities of that kind.
4. No particulars which are sense impressions or sensa are identical with physical phenomena.
5. If materialism is true, then if there are sense impressions or sensa they are identical with physical phenomena.

   Therefore

6. If scientific realism is correct, then materialism is false.

Only the third and fifth premisses of this argument are clearly acceptable. I shall, however, grant premiss II for our purposes here. That is, I shall not debate here what is needed for an adequate scientific explanation of propositional attitudes. This leaves premisses I and IV, neither of which I find to be justified by Sellars. And, although premiss IV is at least plausible and thus not unreasonable to accept, I shall argue that it is more difficult for a scientific realist to justify it than for certain of his antagonists, such as certain naive or common sense realists. It is, however, premiss I that I shall claim is the weak link in the argument. Most of what follows will be a discussion and ultimately a rejection of premise I and various attempts to support it that I have gleaned from Sellars' writings.

A . EXAMINATION OF PREMISS IV

Many identity theorists, such as Feigl, claim that entities, such as "raw feels" or sense impressions, are identical with certain brain phenomena. This claim is usually debated independently of the scientific realism-scientific instrumentalism issue. But, as might be expected, this is not true of Sellars, because his argument against the identity theory depends on the truth of minimal scientific realism. We can see this by noting that his argument is a [423] species of the property objection to the identity theory that depends on physical objects having theoretical constituents. He claims that a red sense impression, for example, has the property of being "ultimately homogeneous," but brain entities, being conglomerations of scientific particles, lack that property. He says that the "feature which we referred to as 'ultimate homogeneity' and which characterizes the perceptible qualities of things, e.g., their colour, seems to be essentially lacking in the domain of the definable states of nerves and their interactions" (SPR, p. 35). This way of putting his objection makes it seem to depend on the argument that because each macro-part of a brain is composed of discrete particles and is thus discontinuous, given the truth of minimal scientific realism, no macro-part of the brain has any homogeneous property. This is a weak argument, because if each of a group of discrete particles was the same color, then it would certainly seem the macro-entity they constitute would have the property of appearing homogeneously colored to normal perceivers under normal conditions. And this, it is plausible to argue, is one clear meaning of the claim that a physical object is homogeneously colored. What is important, however, is not that homogeneity is lacking if brain parts are composed of discrete particles, but that none of the discrete constituent particles are colored. Indeed, Sellars says that "it doesn't make sense to say of the particles of physical theory that they are coloured" (SPR, p. 35). Consequently, argues Sellars, because if none of the particles are colored then no group of them is colored, no expanse consisting of such particles is a homogeneous or even a heterogeneous colored expanse. Thus no red sense impressions are identical with brain phenomena.

An immediate objection to this argument can be raised against the premiss that if none of the constituent particles of something is colored, then the thing itself is not colored. But why could not some group have a property none of its constituents have, especially if, as seems reasonable, we construe being colored as presenting a colored appearance to a normal perceiver under normal conditions? Sellars is aware of this objection, for he says,

There is nothing immediately paradoxical about the view that an object can be both a perceptible object with perceptible qualities and a system of imperceptible objects, none of which has perceptible qualities. Cannot systems have properties which their parts do not have? Now the [424] answer to this question is "yes." if it is taken in a sense of which a paradigm example would be the fact that a system of pieces of wood can be a ladder, although none of its parts is a ladder. . . . It does not seem plausible to say that for a system of particles to be a pink ice cube is for them to have such and such imperceptible qualities, and to be so related to one another as to make up an approximate cube. Pink does not seem to be made up of imperceptible qualities in the way in which being a ladder is made up of being cylindrical (the rungs), rectangular (the frame), wooden, etc. The manifest ice cube presents itself to us as something which is pink through and through, as a pink continuum, all the regions of which, however small, are pink. (SPR, p. 26)

1. An Argument for Premiss IV. A Principle of Reducibility

Sellars has, however, a stronger argument based on a principle of reducibility:

R1 If an object is in a strict sense a system of objects, then every property of the object must consist in the fact that its constituents have such and such qualities and stand in such and such relations or, roughly, every property of a system of objects consists of properties of, and relations between, its constituents (SPR, p. 27).

Sellars goes on to conclude: "With something like this principle in mind, it was argued that if a physical object is in a strict sense a system of imperceptible particles, then it cannot as a whole have the perceptible qualities characteristic of physical objects in the manifest image" (SPR, p. 27). We can see, therefore, that Sellars [425] requires for his refutation of an identity theory not only the thesis of minimal scientific realism, but also a principle of reducibility.

We have been granting to Sellars the thesis of minimal scientific realism, but granting him his principle of reducibility is another matter. As he employs it in this preceding argument, it is equivalent to:

R2 For any object, O, that is a system of objects, if O has a kind of property, K, then K is a kind of property the individual constituents of O have.

R3 For any object, O, that is a system of objects, if O has a property, P, of kind, K, then either K is a kind of property the individual constituents of O have, or O having P is reducible to some (at least) of the individual constituents of O having some other properties or relations.

One difference between R3 and R1, which is worth noting but which is not relevant to our present purposes (but see p. 451), is that R3 does not state that property P is reducible to some other property or relation. It states that O having P is reducible. Thus although a gas having a certain temperature would seem to be reducible to some group of molecules having a certain average kinetic energy, the question is here left open about whether this requires that the property of having a certain temperature is reducible to the property of having a certain mean kinetic energy.

It might be objected here that although R3 can easily accommodate the temperature example, Sellars cannot adopt it because the conditions that are necessary for the justification of his version [426] of scientific realism require that he adopt R2. It is true that Sellars thinks that his version of scientific realism requires some principle of scientific reduction, but I believe that R3 is sufficient to meet his requirement. For Sellars, scientific realism is justified only if what he calls "the scientific image" (SPR, pp. 18-24) provides a better picture or conception of what there is than "the manifest image" (SPR, pp. 6-14), that is, only if it provides a better picture than the one resulting from using common sense, observational terminology to describe man and the world. The scientific image will attain this status only when there is one complete explanation and description of what there is that uses only scientific theoretical terms. But there will be one complete scientific description only when all sciences are unified. This requires that all descriptions true at some scientifically nonbasic level, such as chemistry, biology, neurophysiology, and including the observation level, have counterpart descriptions true at the scientifically basic micro-physical level, that is, at the level of scientifically basic or simple entities (SPR, pp. 19-21; IA, sect. 52). Thus "the scientific image of a man turns out to he that of a complex physical system" (SPR, p. 25). It certainly seems there would be a complete set of such counterparts once all objects meet the reduction requirements of R3. For, if 'O has P' is true at some nonbasic level, and O having P is not only correlated via correspondence rules with, but also reduced to the constituents, C, of O having Q, then 'The C's of O have Q' would be the basic level counterpart of 'O has P' (cf. TE, pp. 71-78).

If the preceding is correct, it seems that R3 is sufficient to meet what Sellars requires for the scientific image and thus for scientific realism, and because R3 does, but R2 does not, accommodate the temperature example. Sellars should adopt R3 instead of R2. But if he does that, then his preceding argument that sensa are not identical with anything physical fails because R3, unlike R2, is compatible with something that appears homogeneously as red being reducible via correspondence rules to some brain phenomenon consisting of certain scientifically basic physical particles having some quite different kind of property or relation. This kind of reduction would allow the basic science to explain, at its level, using only the kind of predicates that apply to physical particles. what, at another level, would be explained by assuming sense [427] impressions. This would allow in turn for there to be a complete explanation and description of everything without requiring unique variables ranging over a different kind of basic entity, sensa.

Before moving on we might ask whether Sellars is right in thinking that his version of the scientific image actually requires a principle as strong as R1 or R3. It is not clear that it does, even if we grant that the Sellarsian scientific image is justified only if all macro-objects that exist are identical with groups of basic particles, each of which particles has only the nonrelational properties ascribed to it in the basic science. It is not clear that this scientific image requires the reduction of these group properties to properties that are part of the subject-matter of the basic science. Nor is it clear that, even if such a reduction is required, it cannot be done by postulating properties in the basic science that apply to groups of basic particles but not to individual particles. If, as it seems, either of these two ways of accommodating nonbasic group properties is compatible with Sellars' scientific image, then not only does it not require any principle of reduction like R1, but there would also be additional reason to reject the preceding Sellarsian argument against the identity theory.

It might be claimed that although we have found three ways a Sellarsian scientific image can accommodate the identity theory, neither of the two ways that deny the kind of reducibility expressed in R1 are compatible with reductive materialism, that is, a materialism that reduces mental phenomena to physical phenomena by means of the identity theory. Both the view that there are groups of basic particles with group properties nor expressible at the basic level, and the view that such properties can be accommodated at the basic level by adding theoretical group properties at that level, lead to the conclusion that some groups of particles have scientifically irreducible phenomenal properties, namely, those phenomenal properties of the sensations identical with certain groups of particles. Such group objects would be neither purely physical nor purely mental (as are raw feels) but perhaps some "neutral" stuff, because they would have both physical and phenomenal properties. It is true that a reductive materialist can accommodate sensations in neither of these two ways. He requires a reduction such as the kind expressed in R1 or R3. But Sellars is not a [428] reductive materialist; nor need an identity theorist be one. Thus all three ways to handle sensations are open to both of them.

2. Another Argument for Premiss IV. The Colors of Brains and Sense Impressions

Although we have found reason to reject Sellars' justification of premiss IV: "No particulars which are sense impressions or sensa are identical with physical phenomena," it might seem that there is a species of the property objection different from Sellars' that suffices to establish IV. If a sense impression is identical with a brain phenomenon, they have all the same properties. But some sense impressions, such as afterimages, are yellow, and no brain phenomenon is yellow. Therefore at least some sense impressions are not identical with brain phenomena. Although this argument, suitably supplemented, appears decisive for certain naive realists, it is not for a scientific realist such as Sellars. For such naive realists, each brain part has one occurrent sensuous color which is a nonrelational property that is not be analyzed as appearing colored to a normal perceiver in normal observation conditions. But the brain phenomena which are candidates for identity with yellow afterimages are not yellow, and nothing both is and is not occurrently yellow in this sense. The scientific realist, however, can claim that brain parts are colored only in the relational sense that they appear colored to normal observers in normal conditions. But a sensum is not yellow in this sense, because it does not appear yellow to normal observers in normal conditions. Thus on this view it is possible that one entity is "sensum-yellow" and is not "physical-yellow" at the same time. Consequently this second species of the property objection to identifying sense impressions with brain parts is, in its present form at least, unable to help a scientific realist such as Sellars. Nor can I find anything else in his writings that he could use for this purpose. We can conclude, therefore, that Sellars has provided no way to justify premiss IV is the argument to disprove reductive materialism that I have constructed from his work. Whether IV can be justified is, of course, another matter. [429]

B. EXAMINATION OF PREMISS 1

When we turn to premiss 1: "If scientific realism is correct, then sense impressions are particulars," we shall see that Sellars has much more to say relevant to it than we found relevant to premiss IV. But because it is by no means clear that Sellars would want his statements used as I shall use them, the arguments I shall present will, once again, he extrapolations from and interpolations into his writings. I emphasize this partly because of the complexity of the issues and of Sellars' statements about them, and partly because I shall claim not only that Sellars' defense of 1 is unsatisfactory but also that 1 is false.

To begin the discussion of premiss 1, let us agree with Sellars that sense impressions, for example, impressions of a red and triangular object, which he calls "raw feels" and "nonconceptual episodes," are used to explain perceptual propositional attitudes, such as being under the impression that there is a red and triangular object in front of one (IA, sects. 22-23). As Sellars says:

It is therefore crucial to my thesis to emphasize that sense impressions or raw feels are common sense theoretical constructs introduced to explain the occurrence not of white rat type discriminative behavior, but rather of perceptual propositional attitudes, and are therefore bound up with the explanations of why human language contains families of predicates having the logical properties of words for perceptible qualities and relations. (IA, sect. 49)

Sellars seems to have no doubt that sense impressions are required if there are to be explanations of perceptual propositional attitudes. He claims that having an impression that implies having an impression of. As he says:

Thus, the fact that a person is under the visual impression that a certain [430] stick in water is bent is taken to imply that he is having a visual impression of a bent object. I shall assume that this is true. (IA, sect. 21)

we are not committed to saving that there is a thing -- an appearance -- of which the wool "white," in its sensible use, designates a property. We are saying, rather, that there is a certain state or process -- that of being appeared to, or sensing, or experiencing -- and we are using . . . the adverb "whitely," to describe more specifically the ways in which the process occurs.⁸

Sellars has not overlooked this appeal to the adverbial theory, however. Indeed, he thinks that the adverbial theory is correct for the manifest image, or common sense picture of man, in which persons are construed "as subject of conceptual episodes proper" (SM, p. 166) rather than bundles of particles as in the scientific image. As he understands the manifest image, impressions of are states of perceivers rather than particulars. According to him, for the manifest image, [431]

the correct view is that to have a sensation of a red rectangle is to sense a-red-rectangle-ly, or strictly speaking, because "sense" is not a complete verb in itself, such as might take an ordinary adverb of manner, it is to a-red-rectange-ly-sense, which alone is the complete verb. (SM, p. 168)

My point in all this is that the only ultimate logical subject involved in a person's having a sense impression of a red rectangle is the person, though, of course, impressions are derivative logical subjects, in the sense in which smiles as logical subjects are derivative from people smiling and waltzes from people waltzing. (SM, p. 169)

if scientific realism is correct, it the end of the road somehow the phrase "a red rectangle" will lose its adverbial status and by a final transposition, will become once again a common noun for particulars, though not the particulars with which the story began. (SM, p. 172

The question before us is not whether this story of concept transformation will prove to be correct, even in part. It is why Sellars thinks that a stage in which persons as groups of basic physical particles in certain states, such as white-sensings, is ultimately unsatisfactory for the scientific image, and why he thinks [432] his scientific realism requires the move from such states to a new kind of particular, sensa. In what follows I shall construct from Sellars' writings three reasons to defend the claim that scientific realism, as he construes it, and what he calls the "scientific image" require the counterparts of sense impressions in the scientific image to be particulars. I am far from certain, however, that Sellars would agree with these reasons. Indeed. I am sure he would reject the first. I include it because a discussion of it will help locate and illuminate central areas of Sellars' thought.

1. First Reason: Sense impressions are Analogous to Physical Objects

We have seen that Sellars claims it is necessary to assume sense impressions to explain certain propositional attitudes. He thus construes them as theoretical entities even though he admits that we can have direct knowledge of them. Sensation-terms, then, are both theoretical and reporting terms (SPR, p. 195), or, as I have called them. "theoretico-reporting" terms. But being theoretical, these terms are, like all theoretical term, formed by analogy, in the sense that they are "analogical extensions of concepts pertaining to the public and intersubjective world of things and persons" (SPR, p. 48). Sellars' view, then. is that in the order of concept formation it is the language of public, observable objects which is basic, and all other concepts that apply to imperceptible entities, such as electrons and sense impressions, are formed by analogy with the basic ones (SR, sects. 19-37). But, because he also claims that "priority in the order of concept formation must not be confused with ontological priority" (RA, p. 296 n.), Sellars can consistently maintain his theory of concept formation with his scientific realistic position that those manifest image-concepts, basic in the order of concept formation, actually apply to nothing at all (SPR, pp. 126, 173). Nothing is literally red in the public occurrent sense of 'red' that is applied to physical objects, although some things, sense impressions, are occurrently "red" in a different but analogically derived sense. [433]

One argument for premiss 1 that it might seem Sellars would accept is stated by Bruce Aune in his "Comments" on Sellars. Aune argues that Sellars view of the analogical nature of the formation of sensation-concepts requires that sense impressions be inner particulars rather than states, such as sensing whitely. He considers Sellars' view that "an impression [of a red rectangle] is conceived as analogous in certain respects to objects that are red and rectangular on their facing side,"¹⁰ and concludes that "if anything were to be conceived as analogous to a facing surface, it would presumably be the kind of thing that philosophers have called a 'sense datum' or 'sense content', something which is, plainly enough, a peculiar kind of particular, not a state of having something.¹¹

Must the analogy required for concept derivation be an analogy of objects with objects? Why not one of objects with states of a perceiver? Sellars argues that it can be the latter (RA, p. 289), because, as he says, "the essential feature of the analogy is that visual impressions stand to one another in a system of ways of resembling and differing which is structurally similar to the ways in which the colours and shapes of visible objects resemble and differ" (SPR, p. 193). And states of color-sensing can vary in ways structurally similar to and functionally dependent on the ways colors of visible objects vary.¹² There seems, then, to be nothing wrong with a "trans-category analogy" (SPR, p. 93) between objects and states for the purposes of concept derivation.

It might be objected, however, that although an object-state analogy is close enough to help explain the derivation of concepts, it is not close enough to help explain the occurrence of impressions that some object is red. That is, how could the state of red-sensing, which is not sensing an object, explain being under the impression that there is an object and it is red? Or, again, how can a person come to believe that there are objects if none of the perceptual states from which his beliefs arise include any objects at all? This objection stresses beliefs about objects rather than, as [434] Sellars does, beliefs about colors, and its force is that states are much less like objects than sensed "color" is like colors of physical objects. As far as I know Sellars has not replied to this objection but he might base his answer on his view of the manifest image, the image corresponding to most of our ordinary beliefs. He says that "there is an important sense in which the primary objects of the manifest image are persons" (SPR, p. 9), and that "in the construct which I have called the original' image of man-in-the-world, all 'objects' are persons, and all kinds of objects ways of being persons" (SPR, p. 10). Using this, I shall speculate, Sellars might reply that our impression that there are objects does not arise solely from perceptual impressions, but also involves our conception of ourselves as persons acting in the world. And, applying this person-picture to perceptual episodes results in impressions that there are observable objects rather than merely perceptual states.

Our conclusion about the first argument for interpreting sense impressions as inner particulars rather than states of perceivers is that there is no reason to reject the object-state trans-category analogy either for explaining the formation of sensation-concepts or for explaining the impression that there are perceptible objects. And, even if there were good reason to reject the analogy, Sellars could not consistently use the resulting argument to make his point. Let us. then, turn to another argument, which, I believe, Sellars does accept.

2. Second Reason: In the Scientific Image Persons are Aggregates

The second argument is considerably more complicated than the first. The essential argument here is that the adverbial interpretation of sense impressions requires persons to be single logical subjects, and while this is compatible with the manifest image, scientific realism and the scientific image requires them to be pluralities of logical subjects. Thus although in the manifest image sense impressions can be states of perceivers, they must be some kind of inner particular in the scientific image. The most relevant passage from Sellars is:

By "identifying" in the above manner a person with a plurality of [435] logical subjects, i.e., the constituent parts of the "computer," we have undermined the logic of sense impressions. For whether these parts be construed as material particles or as nerve cells, the fact that they are a plurality precludes them from serving either jointly or separately as the subjects of the verb "to sense red-rectangle-wise." We must therefore either introduce another logical subject (an immaterial substance) to do this work, or turn each sensing into a logical subject in its own right, i.e., introduce a new category of entity ("phantasms" or "sensa" we might call them) with predicates the logical space of which is modelled on that of visual impressions, as the latter was modelled on that of coloured and shaped physical objects. (SPR, p. 101)

1. If s is a logical subject of 'to sense red-ly', then s is a single logical subject.
2. No single logical subject is a plurality of logical subjects.
3. If scientific realism is correct, then all persons are pluralities of logical subjects.

   Therefore

4. If scientific realism is correct, then if s is a person then s is not a logical subject of 'to sense red-ly'.
5. If no person is a subject of 'to sense red-ly', then sense impressions are inner particulars (rather than states of persons).

   Therefore

6. If scientific realism is correct, then sense impressions are particulars.

Here (6) is premiss 1 of the original argument.

In order to evaluate these premisses, we need to clarify the phrases containing 'logical subject'. I think the following rough definitions will suffice for our purposes:

s is a logical subject of 'to sense red-ly' = _df. is an individual which is a true value of 'x' in 'x senses red-ly',

s is a single logical subject = _df. s is a logical subject and s does not consist of a plurality of logical subjects, i.e., s is a simple logical subject.

Using these definitions, we can grant that premisses (2) and (3) are true. There is doubt about (5), however, because, as seen from the preceding quotation, postulating Cartesian egos as simple logical subjects of 'to sense redly' is an alternative to positing sensa. Nevertheless let us grant premiss (5) for our purposes, because even without it we get in (4) that the adverbial interpretation of sense impressions is false if scientific realism is correct. I wish to concentrate on premiss (1). On the face of it, (1) seems clearly false. Persons are subjects of the verb 'to sense red-ly' whether they are single or pluralities of logical subjects.

a) A Defense of Premiss (1): Primitive Monadic Predicates and Single Logical Subjects

It is not clear how Sellars would defend (1) from such an attack, nor is it even clear why he thinks (1) is true. One line of defense, however, can be derived from his discussion of color words. It seems to be Sellars' view that if an aggregate is the logical subject of a primitive term, then the term must be an n-adic predicate where n equals the number of items in the aggregate. (See RA, p. 299.) But 'red' is a primitive monadic predicate. Therefore no aggregate is the subject of 'red'; its subjects must be single logical subjects. The same would be true of 'to sense red-ly'. The argument here can be put as follows:

7. If a term is a primitive monadic predicate, then whatever is a logical subject of the predicate is a single logical subject.
8. 'to sense red-ly' is a primitive monadic predicate.

Therefore

1. If s is a logical subject of 'to sense red-ly', then s is a single logical subject.

b) A Variation of Premiss (1): The Scientific Image and Basic Level Descriptions

We have not yet, I believe, exhausted what can be gleaned from Sellars' writings which is relevant to this topic. We have already discussed one of his requirements for the scientific image of the universe to be correct. This requirement is that science be unified in the sense that there be one complete description of the universe at the basic level of science to compete with the manifest or common sense image. If no such unification occurs, then each unreduced scientific theory will present a partial and fragmented picture, and the manifest image will triumph. This unity of science in turn requires that all descriptions of objects of the universe that are true at some nonbasic level of science, must have a corresponding true description at the basic level of science. That is, corresponding to each true nonbasic description there must be a true description using only predicates that have individual scientifically basic entities as subjects. If this requirement is conjoined with two additional claims, then we shall be able to derive a variation of (7) which with (8) will yield a variation of (1) we can use to derive (4). The first is that the second of the above requirements of the correctness of the scientific image implies that all primitive predicates have individual scientifically basic entities as subjects. The second is that according to the scientific image all scientifically basic entities are single logical subjects. We can lay out this new argument as follows:

9. If the scientific image is correct, then for every description that uses a monadic predicate with a nonbasic entity as subject and that is true at a scientifically nonbasic level, there is a corresponding description true at the scientifically basic level that uses only predicates with individual scientifically basic entities as subjects.
10. If the antecedent and the consequent of (9) are true, then all monadic predicates true of nonbasic entities are definable by primitive predicates true of scientifically basic entities. [438]

    Therefore

11. If the scientific image is correct, then no monadic predicates true of scientifically nonbasic entities are primitive predicates, i.e., then s is a logical subject of a primitive monadic predicate only if s is a scientifically basic entity.
12. If the scientific image is correct, then if s is a scientifically basic entity, then s is a single logical subject.

Therefore

(7a) If the scientific image is correct, then if s is a logical subject of a primitive monadic predicate, then s is a single logical subject.

(1a) If the scientific image is correct, then if s is a logical subject of 'to sense red-ly' then s is a single logical subject

c) A Defense of Premiss (10): Primitive Predicates and a Tractarian Language

I believe that Sellars holds in something like (10) and I think that what lies behind it is his acceptance of Wittgenstein's conception of language in the Tractatus as his model for an ideally perspicuous language, the kind of language Sellars seems to find required for the best picture or description of what there is. (See SPR, pp. 207-215; and SR, sects. 18, 89-90). In such a Tractarian [439] language each primitive subject term names one ontologically basic entity and each primitive monadic predicate ascribes one property to one ontologically basic entity. All other names and monadic predicates are incomplete symbols, that is, symbols defined using logical terms with primitive names and predicates.¹³ Consequently each nonbasic, or nonatomic, sentence is logically equivalent to a molecular sentence consisting of atomic sentences joined by truth functional logical connectives, and each nonbasic entity is to be understood as a logical construction out of basic entities. Given that something like a Tractarian language provides the best or most perspicuous picture of what there is, and, as Sellars also seems to believe, that the scientific image requires the theoretical terms of science to provide, in this sense, the best picture of what there is, then (10) is justified. For, given the above, the scientific image is correct only if scientific theoretical terms that predicate properties of individual scientifically basic entities are primitive predicates. All others, whether theoretical or observation terms, are defined by means of these primitive predicates.

This is a plausible and I think a Sellarsian defense of (10). But I find at least two reasons for thinking that it fails. First, for the scientific image to triumph over the manifest image, and all others, it is only required that for any language structure with any reasonable degree of perspicuity, the theoretical terms imbedded in that structure provide, with only the addition of the logical constants of that structure, a more accurate and comprehensive picture, or description, of what there is, than any other set of terms imbedded in the same structure. But just as someone who claims that the terms of a phenomenalistic language or those of the observation language provide the best picture need not show he can imbed his terms in a Tractarian language structure, neither is this required of someone who champions the scientific image. That is, the scientific image must be best in the sense of being better than all its alternatives, but need not be best in the sense of being the ideal picture of what there is. At most it would be necessary to imbed the terms in a Tractarian language if an [440] opponent has succeeded in doing it. But even in that eventuality, I think such a success by one competitor would prove sufficient reason for rejecting another, only if all else is equal. This is seldom the case.

Second, past failures give reason to think that no program of logical constructionism, an integral part of imbedding any set of terms in a Tractarian language structure, will succeed. For example, there is good reason to think that many sentences using only physical object terms are not logically equivalent to any sentence using only sense-data terms, and thus there is good reason to think that some physical objects are not logical constructions out of sense data. Similarly, there is reason to think that the relationships among pure observation sentences and pure theoretical sentences is logically contingent, and so neither observable entities nor theoretical entities are logical constructions out of the other. But the preceding good reasons to reject both a phenomenalistic program of logical construction and a scientific realistic program of logical construction do not provide good reason to reject either of the conflicting claims about which kind of terms provide the best picture of what there is. Thus neither the scientific image nor a phenomenalistic image being best requires that its nonbasic terms are definable by its basic terms. But this would be required if being the best picture entailed being imbedded in a Tractarian language. The preceding defense of (10) fails.

d) Another Defense of (10): Correspondence Rules as Definitions

I can find just one more way someone might attempt to justify (10). One thing is clear. The scientific image, by requiring that there be one complete description of the universe at the basic level of science, also requires that all nonbasic terms used in true descriptions be either directly or indirectly related to basic level theoretical terms by correspondence rules. But, according to the present argument, once all such correspondence rules are established, they will, in effect, provide redefinitions of the nonbasic terms they contain. Consequently the scientific image implies such definitions, and premiss (10) is true. This argument depends on construing correpondence rules as definitions or, as Sellars says of [441] certain correspondence rules, at least as "anticipations of definitions . . . the implementation of which in an ideal state of scientific knowledge would be the achieving of a unified vision of the world" (TE, pp. 77-78). There are, however, different views about the kind of definitions correspondence rules are or may become. The first view is that of Hilary Putnam who claims of the correspondence rule "Light is electromagnetic radiation," that it is a definition but that in calling it a definition he does not "mean that the statement is 'analytic'. But then 'definitions', properly so called, in theoretical science virtually never are analytic."¹⁴ If one adopts this view of correspondence rules as logically contingent definitions, then although the task of showing observable entities to be logical constructions out of theoretical entities will not be completed in the strict sense which requires logical equivalence, what results would seem to be sufficient for scientific purposes and perhaps also for producing a unified scientific view of what there is. The second view of correspondence rules as definitions stems from Sellars' claim that a correspondence rule can become at least a part of a redefinition of the observation term it contains once all correspondence rules needed to relate the theoretical with the observational are established. Furthermore, according to Sellars, these would be analytic definitions. As he says: "The force of the 'redefinition' must be such as to demand not only that the observation-sign design correlated with a given theoretical expression [be] syntactically interchangeable with the latter, but that the latter be given the perceptual or observational role of the former so that the two expressions become synonymous by mutual readjustment" (SPR, p. 125). Thus once the ideal state of scientific knowledge is reached, all that is needed for a Sellarsian program of logical construction would be available. That is, as Sellars says, "correspondence rules would appear in the material mode as statements to the effect that the objects of the observational framework do not really exist -- there really are no such things. They envisage the abandonment of a sense and its denotation" (SPR. p. 126; and see SR, sects. 59-75). The second view goes beyond Sellars' claim about observation terms, however, because it requires that all [442] correspondence rules, including those relating sensation-terms to basic theoretical terms, will become definitions of the nonbasic terms they contain. As we shall see later (p. 447), there is some reason to think that Sellars would reject this extension of his claim, because once all observation terms are theoretically redefined, sensation-terms will have to carry all sensory meaning and thus will not be theoretically definable by the correspondence rules in which they appear. But if he should argue this way, then, because he would be rejecting the last reason for accepting the consequent of (10), he should reject it. And, because he wishes to maintain (9), he should either reject the scientific image or deny (10). I feel sure he would do the latter.

Although it is a point worthy of discussion, we need not here decide the issue of whether in an ideal state of science correspondence rules would or should be considered analytic definitions. But again 1 think that it is the view that ontological reduction requires logical constructionism which prompts someone to argue for the definitions to be analytic.¹⁵ The relevant point for our purposes is that if we grant that either of these two views of definition by correspondence rules is correct, then we should also grant that once 'to sense red-ly' and other adverbial phenomenal terms are "defined" via correspondence rules relating them to scientific basic entities, they no longer will be primitive, and premiss (8) will be false. Our conclusion, then, is that if both views about correspondence rules being definitions are wrong, then, as previously argued, (8) seems true but (10) is false. But if either view is correct, then someday it may be that (10) will prove to be true. Then (8) would he proven false. In either case this justification of (la) fails. And. because I can find no other plausible way to justify (la), I conclude (1a) is dubious and the argument using it to justify premiss 1 fails. The result is that this Sellarsian attempt to show that the scientific image requires the move from sensing as states of persons to sensa as a special kind of particular, also fails. We have not yet found reason to reject sensings for sensa. [443]

3. Third Reason: Sensing Red-Ly Is An Emergent Property

There is, I believe, one more argument against sensing red-ly and for red sensa that can be extracted from Sellars' writings. As in the previous argument it centers on two consequences of Sellars' conception of the scientific image we have already discussed. The first is that if the property of sensing red-ly is a property of something, then it is a property of persons and is, consequently, a property of groups of basic entities. The second is the requirement that all properties of groups of basic entities are reducible to properties or relations of individual basic entities. In addition the argument is also based on the premiss that no scientifically basic entities sense red-ly. But, so goes the argument, if a group of basic entities has a property but individual basic entities do not. then the property is an emergent group property. And, if something has an emergent group property then that property is not reducible to properties or relations of individuals in the group. Consequently, if the scientific image is correct, then, as in the previous argument, no person senses red-ly. And, as in the previous argument we can use the premiss: "If no person senses red-ly, then sense impressions are inner particulars,' to conclude again that if the scientific image is correct sense impressions are particulars.

Once again I shall try to put the crux of this argument in a more perspicuous form.

1. If the scientific image is correct, then sensing red-ly is a property of something only if it is a property of a group of scientifically basic entities.
2. No scientifically basic entities have the property of sensing red-ly.
3. If no basic entity has a property of a certain kind, but a group of basic entities does, then the property is an emergent group property.

   Therefore

4. If the scientific image is correct, then sensing red-ly is a property of something only if something has an emergent property of a group of scientifically basic entities. [444]
5. (5) If something has an emergent group property, then it has a group property that is not reducible to properties or relations of the individuals that compose the group.
6. If the scientific image is correct, then all properties of groups of scientifically basic entities are reducible to properties or relations of individual scientifically basic entities.

   Therefore t

7. If the scientific image is correct, then nothing, and a fortiori no person, has the property of sensing red-ly.

We can accept both premises (1) and (2) because the former expresses an obvious consequence of Sellars' scientific image and the latter expresses a truth about basic physical particle. Premiss (3) is less obvious but I think reasonable, ami also acceptable to Sellars. Although he discusses emergence (see CE), he does little to explicate the concept of emergent property except to say that 'emergence' connotes two or more levels and to imply that emergent properties are nonbasic level properties. (See CE, pp. 246-247.) This seems true, which gives us reason to accept, and to assume Sellars accepts (3). We are left, then, with (5) and (6). While we shall see that Sellars might reject (5) it is clear he accepts (6), as can be seen by reviewing his principle of reducibility, R1 (p. 424). As mentioned previously, however, it is not clear that his scientific realism requires (6). We shall consider (5) first, because what I wish to say about (6) depends on the results of the discussion of (5).

a) Examination of Premiss (5): Reducible Emergent Properties

Premiss (5) concerns reduction of properties, and although it is far from clear what are the requirements for property-identity, let alone property reduction, Sellars thinks that one requirement for both is that the predicate expressing the property must be defined by, or "become'' synonymous with, the predicate expressing the reducing property. (See IA, sects. 8, 37, 41, 46.) This is indeed a sufficient condition for property-identity. If, as Sellars thinks, it is also a necessary condition, it would seem to [445] bolster (5), because there seems to he no way to define emergent properties, such as phenomenal properties, by physical properties. This is not the case, however, because using another claim by Sellars we can show that a definitional reduction of some properties is plausible and thus (5) would be refuted. To see this, let us consider three different definitions, the first two of which are derived from definitions of physical predicate' stated by Sellars (see I A. sect. 45), and the third is modeled on the second. None of them are ultimately satisfactory but they will do for our purposes here.

P is a physical₁ property = _df. P is a property belonging to a spatio-temporal nomological framework of scientific explanation.

P is n physical₂ property = _df. P is a property expressed by one of a set of predicates adequate to the theoretical description of nonliving matter.¹⁶

P is a physical₃ property = _df. P is a physical₂ property expressed by one of a set of predicates adequate to the theoretical description of the individual basic level particles of physics.₁₆

E is an emergent_n property = _df. 1) E is a property of an entity which is either simple or composed of physical_n entities, and 2) E is not a physical_n property.

It is a reducibly_n emergent property = _df. 1) R is an emergent_n property of some entity, and 2) for any entity, a, with property R, a having R is identical with some of the physical_n constituents of a having a physical_n property or relation.

As often interpreted, psychological properties such as thinking or sensing red-ly, are emergent in the sense using 'physical₁' and thus are also emergent in the other two senses. Certain physiological properties of human brains would be emergent in the sense using 'physical₂', and many observable properties, such as having a rise in temperature, would be emergent only in the sense using 'physical₃'. For our purpose of examining premiss (5) I shall claim that it does not matter which sense is used, because there is no reason to accept to accept (5) for one subscript on 'physical' and not for another. But this falsifies (5), because the temperature of a gas is emergent₃ and, if another claim by Sellars is correct, reducible to the average kinetic energy of the molecules of the gas. As we have seen, Sellars claims that in a completed science the correspondence rules tying temperature to micro-particles could become definitions of temperature-terms, just as Sellars believes can be done for all empirical terms. He claims that "it makes| sense to speak of turning empirical predicates -- and in particular observation predicates -- into definitional abbreviations of complex theoretical locutions, and that being so defined, observation terms could continue "to play their perceptual role as conditioned responses to the environment" (TE, p. 77). It follows that temperature, although emergent, is definitionally reducible, and premiss (5) is false. Furthermore, given this conclusion about temperature, it seems reasonable to accept that there are also reducible emergent properties of the other two kinds. Consequently the properly of sensing red-ly, while emergent₂, might well be reducible to physical₂ or even physical_3properties.

I think Sellars might agree with the preceding reason to reject premiss (5) because empirical properties, such as temperature, are reducible emergent₃. (See TE, pp. 71 ff.) Nevertheless he might reply that phenomenal properties, such as sensing red-ly, are at least emergent₂ but not reducible₂ and that premiss (5) would be true when amended to apply only to emergent₂ sensation-properties. He would, then, disagree with the final inference of the [447] preceding argument which assumes an analogy between the temperature case and the sensing red-ly case. The point is that correspondence rules would not provide definitions of sensation-terms in this "ideal" state of science, for, in that state, all the counterparts of sensible physical colors, among others, will have been relegated to the realm of sensation. This is necessary if there are to be definitions of empirical terms by physical₃ terms in correspondence rules. All aspects of the meanings of empirical terms involving sensible qualities, such as felt temperature and seen color, must be expunged if observation terms, such as 'gas' and 'table', are to be defined by physical₃ theoretical terms. But some terms must involve some kind of counterpart of sensible physical qualities, according to Sellars, because something relevantly analogous to sensible physical qualities is needed to explain, among other things, being under the impression that a table is red. Consequently, at this final state of science, sensation-terms, although related to certain physical₃ terms by correspondence rules, are not definable by those physical terms, because there is no further limbo to which the sensible quality aspect of their meaning can be relegated. According to this argument, then, if the scientific image is correct, sensing red-ly is an emergent₃ property of a group of basic or physical₃ entities and it is not reducible to any physical₃ properties or relations. Taking this conclusion with premiss (6) , it seems that we can derive (7) and are driven to assume sensa as the basic level counterparts of sensible physical qualities and sensings.

The inference to (7) is too quick, however, because the way we must amend premiss (5) also requires a change in (6), if (7) is to be validly deduced. Sellars' claims about the definability of empirical terms by theoretical terms of the physical sciences is plausible enough to cast doubt on (5), but his claim about the nondefinability of sensations-terms suggest a plausible replacement.

(5a) If something has the property of sensing red-ly, then it has an emergent3 group property of scientifically basic individuals that is not reducible to properties of individuals that compose the group.

The change in (6) required to yield (7) with (4) and (5a) results in a statement in need of examination:

(6a) If the scientific image is correct, then all properties of groups of scientifically basic entities are reducible to physical₃ properties of individual scientifically basic entities.

b) Examination of Premiss (6a): The Scientific Image and Nonmaterialistic Reduction

The crucial difference between (6) and (6a) is that where (6) merely requires of the scientific image the reduction of group properties to properties of basic level individuals, (6a) requires reduction to physical₃ properties. Even granting that the scientific image requires a definitional reduction of properties, which there is some reason to doubt, nothing in the Sellarsian argument that the emergent₃ property of sensing red-ly is not definitionally reducible to physical₃ properties prohibits the definition of sensing-terms by theoretical terms that apply to scientifically basic individuals where there is an adjustment in the meanings of both sets of terms. (See SPR, p. 125.) One way for such an adjustment to occur would be for some basic-level terms to acquire some counterpart of the sensible quality meaning of sensation-terms. Thus the property of sensing red-ly might in this way be definitionally reducible to properties of scientifically basic entities. Each sentence predicating sensing predicates of a person might become synonymous with a sentence predicating either a relationship among certain basic entities or some nonrelational property the basic entities have in common. In this way, for example, 'John hurts' might become synonymous with 'John's C-fibers are firing', which may then become synonymous with some statement about certain molecules in John's brain, which may finally become synonymous with some statement about basic entities in John s brain. And, although these properties or relations of basic entities to which sensings would be reduced would not be physical₃ or physical₂ but merely physical₁, because of the change in meaning of the terms expressing them, they would be properties of basic-level [449] individuals.¹⁷ Such a reduction to the basic level but not to a physical₃ level is certainly compatible with scientific realism and the scientific image even if both require a definitional reduction of properties. Thus even if (6) is acceptable, (6a) is not, and the revised version of the argument fails as did the original version with premiss (5).

c) One More Version: No Nonmaterialistic Reduction of Sensing

There is, however, one more move Sellars might make. In order to avoid the preceding objection to (6a) he could maintain (6) and search for a revision of (5) that avoids the objections to it. Let us here grant that (5a) is true, because the property of sensing red-ly is not definitionally reducible to any physical₃ properties or relations. Sellars might go on to claim that it is not reducible to any properties or relations of scientifically basic entities, whether the properties or relations are physical₃ or merely physical₁. That is, no properties of basic entities, even if they become required only when explaining the behavior of living things, serve to reduce the property of sensing red-ly. Sellars might then replace (5) with:

(5b) If something has the property of sensing red-ly and this is an emergent group property of scientifically basic individuals, then something has a property that is not reducible to any properties of individual scientifically basic entities.¹⁸

The preceding objection to (6a) does not apply to (6), but it is easily adapted to (5b) . Once we are allowed to consider any sort of property of basic entities, there seems to be no reason to [450] reject even the definitional reducibility of sensing red-ly. It might be that the same predicates that now express physical₃ properties will change so that a certain sentence using them will become synonymous with 'John hurts'. Sellars has given us no reason to reject this possibility and I have found none. Thus we can conclude that (5b) is as dubious as (6a), and the third version of the argument fails as did the second.

Although we have found reason to reject three versions of the argument based on the emergent property of sensing, it might be thought that the results of our examination of the argument provide the grounds for a sound argument against materialism. If materialism is true, then because we cannot eliminate both sensa and sensings and cannot reduce₂ sensa. we must reduce₂ sensings. But we have just seen that the property of sensing red-ly is not definitionally reducible to a physical₂ property. Therefore, materialism is false. There are three objections to this argument. The first is that even if it is true, as we have been assuming while discussing Sellars' argument, that the only way open at this point to save materialism requires a reduction of the property of sensing red-ly to a property that is at least physical₂, it is not clear why a reduction, which entails identity, could not also be contingent.

I would prefer, however, not to rest the burden of refutation on this first objection, because there is a problem about the requirements of property-reduction. If the second objection is correct, however, that problem need not be faced. Although there maybe a form of reductive materialism that entails the reduction of properties, it does not seem that all forms do. This conclusion is bolstered by remembering that we have construed reductions of particular states and events, such as some person sensing red-ly, to be what are essential to reductive theories, and it is not clear that such a reduction entails a reduction of the properties, such as the property of sensing red-ly. The third objection is that there are other forms of materialism, such as those that attempt to eliminate sensations rather than reduce them, which do not require any sort of reduction.¹⁹ Thus a refutation of reductive materialism [451] is not sufficient to refute materialism. Incidentally, in spite of Sellars, I also believe that the scientific image does not require a reduction of properties. If, as I claimed in discussing principles of reducibility (p. 425), the reduction requirement of the scientific image can be met by the reduction of entities having properties to basic-level entities having basic-level properties or relations, and if, as also seems true, such reductions do not entail property reductions then property reductions are not required by the scientific image. Thus not only premiss (6a) but also (6) would be dubious.

III CONCLUSION: THE SELLARSIAN ARGUMENT AGAINST MATERIALISM FAILS

We have seen that none of the arguments that I have constructed based on Sellars' writings, even those assuming that the scientific image is correct, suffices to justify premiss 1 of the Sellarsian anti-materialism argument: "If scientific realism is correct, then sense impressions are particulars." Indeed it seems that sensings meet the requirements of the scientific image as well as sensa, therefore, sensa are not required for Sellarsian scientific realism and premiss 1 is false. We have also seen that Sellars provides no way to justify premiss IV: "No particulars which are sense impressions or sensa are identical with physical phenomena." In the latter case, however, we did not conclude that the premiss is false. Nevertheless either the conclusion about 1 or about IV is sufficient to refute the argument as I have constructed it from Sellars' writings. This is not to say, of course, that he cannot avoid these problems. Indeed, it is my hope that he can and that this paper will be of some help in his attempt to do so.

University of Pennsylvania.

* The research for this paper was done under National Science Foundation Grant 2083.

1 K. Lehrer, "Review of Science, Perception, and Reality," The Journal of Philosophy (1966), pp. 266-277; Gilbert H. Harman. "Review of Philosophical Perspectives,'' The Journal of Philosophy (1969). pp. 133-144.

2 R. Bernstein, "Sellars' Vision of Man-in-the-Universe," The Review of Metaphysics, Vol. XX, nos. 1 and 2 (March, June, 1966), pp. 113-143, 290-316.

3 Throughout this article I shall use the following abbreviations for Sellars' works:

CE: "The Concept of Emergence." in H. Feigl and M. Scriven, eds., Minnesota Studies in the Philosophy of Science, Vol. I (Minneapolis, 1958), pp. 238-252.

IA "The Identity Approach to the Mind-Body Problem." The Review of Metaphysics, Vol. XVIII, no. 3 (September 1965), pp. 430-451.

RA: "Rejoinder" (to Aune), in H. Castaneda, ed., Intentionality, Minds, and Perception (Detroit, 1967).

SM: Science and Metaphysics (London, 1967).

SPR: Science, Perception and Reality (London. 1963).

SR: "Scientific Realism or Irenic Instrumentalism," in R. Cohen and M. Wartofsky, eds., Boston Studies in the Philosophy of Science, Vol. II (New York, 1965).

TE: "Theoretical Explanation," in B. Baumrin, ed., Philosophy of Science, The Delaware Seminar, Vol. II (New York, 1963).

4 See W. Quine, Word and Object (New York, 1960), pp. 204-265; and The Ways of Paradox (New York, 1966), pp. 208-214.

5 See R. Rorty, "Mind-Body Identity, Privacy, and Categories," The Review of Metaphysics, Vol. XIX, no. 1 (March, 1965), pp. 24-54.

6 See H. Feigl, "The 'Mental' and the 'Physical'," in H. Feigl and G. Maxwell, eds., Minnesota Studies in the Philosophy of Science, Vol. II (Minneapolis, 1958); and J. Smart. "Sensations and Brain Processes," The Philosophical Review (1959), pp. 141-156.

7 Strictly speaking, Sellars would reject this identity claim only if the relevant brain entities are what he calls physical₂ entities. He claims that there is a sense in which sensa are brain entities (IA, sects. 28-32), and a sense in which they are physical, i.e., physical₁ (see p. 445). The latter sense will not help Feigl and Smart however

8 R. M. Chisholm, Theory of Knowledge (Englewood Cliffs, N. J., 1966), p. 96.

9 See my "Mental Terms, Theoretical Terms, and Materialism," Philosophy of Science (1968), p. 96.

10 B. Aune, "Comments" (on Sellars), Intentionalily, Minds and Perception, p. 275.

11 Ibid., p. 276.

12 Cf. R. Chisholm, Perceiving (Ithaca, N.Y., 1957), pp. 113-119.

13 However, see Sellars' comparison of PMese and his own Jumblese (SPR, chap. 7). In Jumblese there are no predicates, but merely ways of characterizing individual constants and variables.

14 H. Putnam, "Minds and Machines," in S. Hook, ed., Dimensions of Mind (New York, 1961), p. 157.

15 I have discussed ontology and analysis in some detail in Metaphysics, Reference and Language (New Haven, 1966).

16 In those two definitions, 'adequate' must be interpreted as 'minimally adequate'. If not, then if a set (P₁___P_n) is adequate to explain something, then so is any consistent set (P₁___P_n+1, and any property consistent with such a set of explanatory properties would be physical₂ or physical₃. We can say a set (P₁___P_n) is minimally adequate to explain something if and only if (P₁___P_n) is adequate to explain it, and no proper subset of (P₁___P_n) is adequate to explain it.

17 Incidentally, this kind of definitional reduction of one property "to" another property is not the one kind of reduction Sellars considers, because it does not require that the predicates expressing the two properties be synonymous. (See IA, sects. 8, 37, 45.) It is more like a definitional reduction of the average family to ordinary families.

18 The third version, perhaps more amenable to Sellars than the first two, was suggested to me by Vicki Levine.

19 See Rorty, and for a discussion of his position, see my "On the Elimination of 'Sensations' and Sensations," The Review of Metaphysics, Vol. XXII, no. 1 (September, 1968), pp. 15-36.