(1) Introduction

Sellars’s Case against Foundationalism in EPM and Related Articles

Thomas Vinci

Part of what we hope to accomplish in this paper is a reinterpretation of the detailed structure of Sellars’s argument against the given in “Empiricism and the Philosophy of Mind” [1] (hereafter referenced in the body of the article as “EPM”) and other earlier articles, part is a criticism of the argument thus reinterpreted. The first we attempt in Part I, the second in Part II.

Part I: Interpretation

1. Introduction: A road map of sections 1-34 of EPM

Sellars case against the given in EPM occurs in the first 34 sections. This case is actually a series of cases against a series of foundationalist targets, eight in number. (1) A non-epistemic form of classical sense datum theory discussed and critiqued in section 3; (2) an epistemic form of classical sense-datum theory discussed and critiqued in sections 4-6, but especially 6; (3) a realist version of classical sense-datum theory discussed (briefly) in section 29; (4) a logical-positivist form of foundationalism discussed in sections 32-34; ( 5) a form of foundationalism in which the data of sense are formulated in terms of the language of “looks” discussed in sections 10-22; (6) a form of sense-data theory in which the data of sense are formulated in terms of perceptions of “surfaces,” perceptual objects abstracted from the facing surfaces of physical objects, discussed in section 23. (7) Sellars also discusses a view in section 9 that he attributes to Ayer, and (8) some historical views that he attributes to the British Empiricists in sections 26-29. In the present article we will be concerned with only the first four.

There are dependency relations among these critiques that take some care to unravel. We begin with the critique of Logical Positivism in sections 32-34. (All passages quoted just below are from EMP, 166-167.) Logical Positivism, as Sellars characterizes it, contains two main elements. First there are immediate, non-verbal propositional awarenesses, “awareness that something is the case, e.g., that this is green.” Second, there are acts of asserting a level of basic statements, statements that “express” the non-verbal awarenesses and derive their “authority” from the awarenesses. For these acts of assertion to be actions and to possess the kind of justification needed for knowledge, these utterances must be part of a rule-governed practice, which practice requires that “it is the knowledge or belief that the circumstances are of a certain kind, and not the mere fact that they are of this kind, which contributes to bringing about the action.” This last requirement is an expression of Sellars’s internalism, a feature of the present account that he carries over to his positive account developed in sections 35 and ff. Sellars’s objection to this version of foundationalsim is not directly given but implied by his observation that the first element of this account brings us “face to face with giveness in its most straightforward form.” The implication is that this is a bad thing, and that this is so has been amply demonstrated previously. But where is the demonstration?

The most closely preceding occurrence of the doctrine of non-verbal immediate propositional awareness occurs in Section 29. Here Sellars offers as an example of such an awareness, “awareness that they are red,” indicating that a philosopher endorsing such a thing is committed to a form of the given, “a realistic form, as in the classical sense datum theory.” [EPM, 160] Again, there is no separate critique of this theory, again the implication is that the critique has occurred previously. There is only one place where Sellars offers a critique of classical sense datum theory and that is in the first 6 sections. Of the two varieties of classical sense datum theory on offer there – a non-epistemic form and an epistemic form – only the latter would in principle be relevant to a critique of the non-verbal propositional awarenesses of section 29, and that critique is given in section 6. [2]

Sellars carries out this critique by attempting to show that the account is subject to a destructive trilemma, one horn of which is that “the ability to know facts of the form x is ǿ is acquired.” According to Sellars, the target empiricist of section 6 must accept this position because of “the predominantly nominalistic proclivities of the empiricist tradition.” (EPM, 132) But if we identify the target of Section (6) as the doctrine of non-verbal propositional awareness of sections 34 and 29, Sellars’s claim that this kind of empiricist must accept that knowledge is acquired is, in a word, baffling. This is because, as we shall see, the main grounds that Sellars offers why the target form of empiricism in section 6 (“Section-6 SDT” we will call it) treats knowing as an acquired ability is that it treats knowing as a linguistic ability. This characterization is precisely what the Section-29 version of sense datum theory, “Section-29 SDT” as we will call it, as well as Section- 34 Logical Positivism, say is false of their version of foundational awarenesses!

Despite the fact that Sellars calls both versions of the sense datum theory “classical” versions, we think that the only resolution to this exegetical conundrum is to deny that Section-6 SDT and Section-29 SDT are the same theory. This creates two further problems. The first relates to Section-6 SDT: what precisely is it, what motivates its introduction by Sellars in section 6, and of what interest to foundationalists is the fact that Sellars is able to construct a destructive trilemma against it? The second relates to Section-29 SDT: where, if at all, does Sellars give his critique?

Our answer to the second question is that Sellars does give a critique of the Section-29 SDT, but not in EPM. In (our) Section 3 we provide a reconstruction of the critique, identifying a crucial premise upon which it rests, and in (our) Section 4 we propose a source for the critique, and locate textual grounds outside of EPM for attributing that critique to Sellars. The premise is that all perception is propositional perception (“perception-that”) and that locutions in which non-propositional perception are apparently expressed must be reduced by paraphrase to propositional forms. This reconstruction forms the point of departure for our own critique of Sellars’s critique, which we develop in Part II.

Our answer to the first question takes the form of a reconstruction of the argument of section 6, uncovering its underlying assumptions. Once uncovered, they reveal a doctrine that is a somewhat ungainly hybrid, partly reflecting traditional empiricism, partly reflecting Sellars’s own views on language and knowledge. This hybrid is essentially Sellars’s own creation, arrived at by an internal dialectic from which the reader is kept largely in the dark. The refutation of this creation by destructive trilemma is both unsurprising and, we think, of little probative consequence in the quest for a refutation of foundationalism in any of the forms it has actually taken.

2. The Trilemma Argument.

In Sections 2-5 Sellars disposes of the possibility that sensory foundations might involve a sensing of particulars. In section (6) Sellars then moves to his most explicit formulation of the case against treating sensory foundations as a form of knowing. It is this argument which we call "The Trilemma Argument.”

But if a sense-datum philosopher takes the ability to sense sense contents to be unacquired, he is clearly precluded from offering an analysis of x senses a sense content which presupposes acquired abilities. It follows that he could analyze x senses red sense content s as x non-inferentially knows that s is red only if he is prepared to admit that the ability to have such non-inferential knowledge as that, for example, a red sense content is red, is itself unacquired. And this brings us face to face with the fact that most empirically minded philosophers are strongly inclined to think that all classificatory consciousness, all knowledge that something is thus and so, or in logical jargon, all subsumption of particulars under universals, involves learning, concept formation, even the use of symbols. It is clear from the above analysis, therefore, that classical sense-datum theories ... are confronted by an inconsistent triad made up of the following three propositions:

A. X senses red sense content s entails x non-inferentially knows that s is red.

B. The ability to sense sense contents is unacquired.

C. The ability to know facts of the form x is F is acquired.

Once the classical sense-datum theorist faces up to the fact that A, B and C do form an inconsistent triad, which of them will he choose to abandon? (EPM, 132; our emphasis.)

In the emphasized sentence Sellars identifies “classificatory consciousness” with knowledge, then identifies classificatory consciousness with the subsumption of particulars under universals, only then arguing for his conclusion that the subsumption of particulars under universals is not given but is actively accomplished, drawing upon learning, concept formation, even symbols. It is this identification which is so problematic. Treating the phrase “all knowledge that something is thus and so” in apposition to the phrase “all classificatory consciousness”, as Sellars does here, is to identify the two concepts, thus to foreclose the question whether knowledge is a classificatory form of awareness at all. The casualness of this identification is astounding, justified only by reference to some unspecified “empirically minded philosophers” who are said to think these things. Sellars takes this casual approach presumably because he thinks that his readers will agree to this -- will see that empiricists typically at least will endorse these assertions.

In the Cartesian tradition knowledge and judgment are acts that fall on opposite sides of the division between passions (acts of perception) and actions (acts of the will).[3] The British Empiricists follow Descartes in this.[4] How then are we to take the notion of knowledge in light of this tradition? If we take it as a passive, non-judgmental act of perceptual consciousness, the connection with classification is then lost for empiricists in this tradition like Locke, since classifying for Locke is an activity, a rather busy one judging from the following description: “Such precise, naked appearances in the mind, without considering how, whence, or with what others they came to be there, the understanding lays up (with names commonly annexed to them) as the standards to rank real existences into sorts, as they agree with those patterns, and to denominate them accordingly.”[5] But if we take it as an act of judgment, thus making an empiricist connection with classification, this identification of classificatory consciousness with judgment and Sellars’s identification of classificatory consciousness with knowledge that something is thus and so (subject-predicate knowing) amounts to an equation of judging that something is thus and so with knowing that something is thus and so . Even if we allow that subject-predicate knowing entails subject-predicate judging, the converse does not hold (one can judge that something is thus and so without knowing that something is thus and so), so the equivalence apparently asserted by Sellars fails. This is not a minor point, since the doctrine of the given is a doctrine about knowledge, either directly so – what is given is a kind of knowing itself – or indirectly, what is given provides the kind of justification needed to turn judgments into knowledge. The first of these is the form that seems to be at issue in section 6, the second in sections 32-34. (We are about to argue, of course, that Sellars does not actually endorse both directions of the equivalence.)

Part of the way forward is seeing that Sellars’s use of the phrase “classificatory consciousness” is perhaps not to be understood in empiricist terms but in his own terms: mental classifying just is subject-predicate judging,[6] judging is asserting and so what corresponds to a theory of “mental classification” is a theory of the conditions for properly making assertions of this kind. Learning, concept formation, even the use of symbols are among these conditions. Here the concept of a linguistic practice is basic. It is the combination of this account of classification with traditional empiricism that makes for a hybrid position held by a hypothetical philosopher we might dub “Sellars-Locke.” The other part is reformulating the Trilemma Argument so that all of its horns can be seen to be acceptable to Sellars-Locke. This can be done by explicitly making the connection between knowing and judging to be a one-way entailment, as follows:

Reformulated Trilemma Argument

1. X senses red sense content s entails x non-inferentially knows that s is red. (As A in original)

2. X non-inferentially knows that s is red entails x non-inferentially judges that x is red.

3. The ability to sense sense contents is unacquired.

4. The ability to make judgments of the form ‘x is F’ is acquired.

We now need to consider Sellars’s claim that no satisfactory way to break the trilemma (inconsistent quartet in the reformulation) exists for classical sense-datum theorists.

It is difficult to assess Sellars’s contention that no way consistent with “classical sense datum” theory is available to break the inconsistent quartet since Sellars is here concerned with only a certain version of classical sense datum theory, the Section-6 version. Indeed, we have remarked above that Sellars himself will later countenance what seems to be a new version of classical sense datum theory, the Section-29 version. We will see below that this version does break the dilemma (at horn (2)). With the version at issue here, however, we can say the following: (1) is definitive of the second horn of the Sense Datum Dilemma which we are currently investigating, so it has to stay. We shall stipulate that (2) is true by definition of the type of empiricism at issue here, as is (3): they both stay too. What of denying (4) – (C) in the original? Sellars says, “But to abandon C is to do violence to the predominantly nominalistic proclivities of the empiricist tradition.” (EPM, 132.) Brandom takes the reference to “nominalist proclivities” here to be a reference to the doctrine of concept anti-innatism. [7] Now, one can find anti-innatism in the British Empiricists and this might suggest that they are the target in section (6). We are skeptical of this since Sellars has an extensive discussion of the British Empiricists in later sections (sections 26-39), with no indication that the doctrines he there describes underlie the type of empiricist under discussion in section 6.[8] Accepting that Locke-Sellars is the target, as we suggest, then the acquired status of linguistic capabilities prevents the abandonment of (4), (C) in the original.

We now turn to Section-29 SDT. Sellars introduces this version, in passing, in his discussion (in section 29) of the doctrine he calls “Psychological Nominalism,” the culmination of his discussion of empiricism in its Early Modern form:

“Even more obviously, if the formation of the association [of words with classes of resembling particulars, a theory he derives from Hume] involves not only the occurrence of red particulars, but the awareness that they are red, then the conceptualist form of the myth has merely been replaced by a realistic version, as in the classical sense-datum theory.” (EPM, 160; our emphasis; our interpolation)

We are now in a position to see definitively that Section-29 SDT is inconsistent with Section-6 SDT as it is characterized in the Reformulated Trilemma Argument of Section 6: the former denies, while the latter affirms, clause (2): “X non-inferentially knows that s is red entails x non-inferentially judges that x is red.” This is so since the awarenesses of the Section-29 SDT are the same as the non-verbal propositional awarenesses of section 34, and, by Sellars’s own lights, the latter do not entail judgments: they are causally related to judgments, they authorize judgments, they do not entail them. This means that section-29 SDT is a version of the classical doctrine of immediate knowledge but Section-6 SDT is not. (Without calling it “knowledge,” the non-conceptual awareness of facts is what the proponents of Positivism (discussed by Sellars in section 34) treat as a foundation for knowledge.)

Sellars’s objection to Section-6 SDT is that it is inconsistent, as shown by the Reformulated Trilemma Argument. We of course accept this criticism. Our objection is that Section-6 SDT does not constitute a version of foundationalism that anyone, as far as we know, has espoused. What objection does Sellars see to the classical doctrine of immediate knowledge represented by Section-29 SDT? The only indication that he gives in section 29 is that with such a view we have “the givenness of facts”, and therefore are “back with an unacquired ability to be aware of repeatables…” (EPM, 160). “Repeatables” is Sellars’s term for properties, applying to specific properties – this shade of red – as a well as general properties, introduced in connection with his discussion of the British Empiricists in sections 26-29. There are three difficulties that Sellars’s exposition here presents. The first is that it does not include an explanation why a doctrine of immediate knowledge should depend on a doctrine of the givenness of properties. The second is that it is not clear why we are “back with an unacquired ability to be aware of repeatables”: we were never there to begin with. Sellars does not specifically say here where the doctrine of the givenness of properties has been refuted but the implication is that it is in section 6: (we are “back with an unacquired ability to be aware of repeatables…”. ) The third difficulty, consequent on the second, is that we are without any indication of how Sellars would propose to argue that an unacquired ability to be aware of repeatables is disallowed: section 6 is the place where Sellars refutes another version of the sense datum theory, the Section-6 version by showing that the position, as he stipulates it, is internally inconsistent. He makes no such argument for the Section-29 version, nor is it obvious what argument could be made.

In the next section we propose a reconstruction of an argument that Sellars could make against the Section-29 version of the sense datum theory and that we think he would have made in EPM if he were pressed to make it. The argument draws on some other doctrines of Sellars’s that are related to the doctrine of Psychological Nominalism, though not specifically asserted as such in section 29. As we have seen, Sellars’s text provides few data points for a reconstruction, so much of the reconstruction will be speculative, though speculative in a way that conforms to the spirit of Sellarsian doctrine in EPM, but does not occur there as such. Then, in section 4, we consider articles written around the time of EPM to show that Sellars there actually gives arguments that fit the reconstruction offered in section 3.1 and 3.2.

3. A Reconstruction of Sellars’s case against the Section-29 Version of Classical Sense Datum Theory.

The reconstruction needs to address the first and the third difficulty identified with the exposition in section 29. It needs to show, first, why the doctrine of immediate knowledge depends on a doctrine of the givenness of repeatables, and, second, what is wrong with this doctrine. What follows in section 3.1 is an argument presented from Sellars’s point of view, in italicized font, beginning with a characterization of the concept of immediate knowledge, showing why it depends on a doctrine of the givenness of repeatables. Section 3.2 addresses the second point: what is wrong with a doctrine that countenances the givenness of repeatables. Commentaries on the reconstruction are offered in brackets and are not italicized.

3.1 The dependence of immediate knowledge on the awareness of repeatables

Immediate knowledge in the classical sense is acquaintance with facts, e.g., the fact that that apple is green, the fact that that sense content is red, rather than acquaintance with particular things. On behalf of the Section-29 sense datum theorist, we shall take facts to be complex entities consisting of an object (or objects), properties (or relations), and the instantiation of the former by the latter. Facts, thus understood, are taken to depend on properties realistically construed so we shall call these facts r-facts to distinguish them from facts otherwise construed. We shall take properties to be “repeatables” in my language, (“universals,”[9]) and repeatables are multiply-instantiatiable entities. We shall take specific-property nominalism to be the doctrine that there are no repeatables. For those who think that facts are true propositions rather than complexes of this kind, the latter of which are states of affairs, they may substitute “actual states of affairs” for “facts” in what follows.

Since facts, as we have been construing them, contain properties as elements, the dependence of r-facts on properties is clear. Since properties are constituents of facts, the awareness of facts may reasonably be supposed to depend on the awareness of properties. This gives us the first element of the argument: the awareness of facts depends on the awareness of properties. The doctrine that there is awareness of properties is the doctrine of the givenness of repeatables – the Myth of the Given in the form of Classical Property-Realist Sense Datum Theory.

[Our investigation of Sellars’s argument against immediate knowledge has taken us fairly quickly to the question of Sellars’s philosophical treatment of properties in EPM, a treatment that rests on a surprising reading of the British Empiricists in sections 26-29.

The British Empiricists are generally regarded as nominalists - Locke, for example, says that “all things that exist [are] particulars”.[10] So one might expect that Sellars’s purpose in discussing the Empiricists would be to accept an account of Locke as a nominalist (an anti-realist about universals) as a starting point and then move to his own version of nominalism. Not so, for we are soon told that the British Empiricists believe in the existence of multiply-instantiated entities --"repeatables" in Sellars’s terminolgy, “universals” in more standard usage. They are not specific-property nominalists. According to Sellars, Locke’s claim that everything is particular means that everything is specific, including properties. What Locke regards as problematic is generic properties - red in general as opposed to a specific shade of red. (section 26) If we now introduce a special notion of “universals” that applies only to generic properties, since the Empiricists deny the existence of universals in this special sense, they can thus can be called “nominalists” in a special sense – general-property nominalists. On this interpretation the empiricists develop a line of thinking that reaches a dénouement in Hume's idea that general properties are similarity-classes of specific properties grouped by association with specific words (section 29). This account then serves as the inspiration for Sellars's own account of universals,[11] a nominalist account that sees universals as similarity classes of predicates grouped by the linguistic roles they play (sections 30-31). What is of interest to us is not so much what Sellars says about universals but what he does not say about the Empiricists's position that specific properties are the immediate objects of sense perception: he does not say how this position goes wrong.

Why does he not do so? In their otherwise illuminating discussion of this portion of Sellars’s argument de Vries and Triplett suggest an explanation:

Sellars is concerned with classical empiricist accounts of being aware of x as F - that is with accounts of conceptual awareness. Thus the question of whether there might also be a form of non-conceptual awareness does not enter into the issues at hand.[12]

This is not a satisfactory answer since one of “the issues at hand” from the discussion of classical empiricism is the Property-Realist version of the classical sense datum theory. The question whether there is awareness of specific properties of some kind is relevant to that issue, and the issue whether there is awareness of specific properties of some kind is a species of the issue whether there is non-conceptual awareness of specific repeatables.

One way to argue against this possibility is to attack the existence of specific repeatables directly and then argue that if there are no repeatables, there is no awareness of repeatables, so the Property-Realist version of the classical sense datum version of the Myth of the Given would be refuted. On this interpretation Sellars would be giving an independent anti-realist argument for repeatables in section 29, but he does not appear to do this: if anything, the argument is in the reverse order. Since there can be no awareness of repeatables, there is no compelling reason to believe in them, at least not in the sense that the classical empiricists understood them, so we should not believe in them. This leaves room for an alternative analysis of all property attributions, specific or general. The next part of the reconstruction adheres to this order of argument .)

3.2 Refuting the Givenness of Repeatables

A classical empiricist will defend the existence of awareness of properties on the grounds that it is intuitively obvious that we perceive properties. Saying, “I see the whites of their eyes,” is intuitively natural and commits us to the existence of the whites of their eyes. The whites of their eyes are repeatables, repeatables are properties, so there are properties. Because of their surface logical form, locutions of this kind assert a relation between a perceiver and a property. If their surface form is allowed upon analysis to stand as their ultimate form, then there will indeed be a strong intuitive case for the doctrine of the given in its classical sense datum Property-Realist form. But it cannot be allowed to stand: locutions of the form “Rxy” whereR is a perceptual relation, x is a person and y a property are analyzable into a different form which neither implies that properties are themselves objects of a perceptual relation nor even that quantifies over properties.

[Ordinary language appears to be against Sellars here: we think of seeing as a form of awareness and we speak of seeing colours as well as seeing objects and seeing that objects are a certain way. There is evidence that Sellars would say that this way of describing things is incomplete: when we see an object we necessarily see it as something, as being something, and that means that seeing objects entails seeing propositionally. This can be seen in his characterization of what even the sense-datum theorist (with his wits about him) must claim, namely that, “…it is logically necessary that if a sense content be sensed, it be sensed as being of a certain character…” (EPM, 129). Does this also hold for seeing properties? Sellars has an analogous reply: we do not see colours simpliciter, we always see them as instantiated in something.[13] So "seeing colours" is an episode that cannot, after all, stand on its own. Once this has been established Sellars can then apply contextual definition to eliminate expressions referring to properties, finally concluding that the success of this procedure shows that there are no properties. The reconstruction now concludes as follows.]

What we call “the seeing of objects” is implicitly the seeing of objects as being a certain way, and implicitly what we call “the seeing of properties” is implicitly the seeing of properties as instantiated by something, both of which amount to the same thing: seeing that something is a certain way. And that is propositional perception, the canonical formulation of which does not require the mention of properties. Thus, for the statement “I see whiteness in the snow” which mentions the property whiteness, a canonical propositional paraphrase, “I see that snow is white” can be produced which contains no name for whiteness; thus, what we call “the perception of properties,” when properly formulated, does not commit us to awareness of properties; thus, does not commit us to the doctrine of the given in its classical Property-Realist sense datum form.

4. Versions of this argument in other writings

We have not thus far offered much exegetical evidence that Sellars actually does reason in the way just reconstructed in any of his writings. However, in sections 7 and 8 of his paper, “Is There a Synthetic ‘a priori’?”[14] Sellars presents arguments that seem to us to contain some of the key elements of our hypothetical reconstruction. This paper is a “revised version” of a paper first presented in 1951,[15] thus antedating EPM by five years, and contains themes and arguments that are strikingly similar to those developed in EMP. The arguments of Section 7 (“Concept Empiricism: The Conservative Approach”) cover some of the same ground as the Trilemma Argument but takes a Section-29 version of classical sense-datum theory as its target form of empiricism, the arguments of Section 8 (“Concept Empiricism, Syntactics, Semantics and Pragmatics”) seem to anticipate the arguments of sections 32-34 in EPM. The target empiricisms in each case “agree, however, in concluding that the basic concepts in terms of which all genuine concepts are defined are concepts of qualities and relations exemplified by particulars in what is called ‘the given’ or ‘immediate experience.” [16]

Our focus will be on the former, conservative version. According to Sellars,

…the concept empiricist of this brand conceives of such symbols as ‘red’ and ‘between’ as acquiring meaning by virtue of becoming associated with such abstract entities as redness and between-ness, the association being mediated by our awareness of these entities. His attention is thus focused on the question, “How, and in what circumstances, do we become aware of abstract entities.” [17]

This kind of empiricist makes “an underlying presupposition of a distinction between the pure awareness of an abstract entity, on the one hand, and the linguistic…expression of this pure awareness on the other.” Sellars himself used to be a “convinced concept empiricist,” as he tells us at the outset,[18] so we might take the story that he tells of the fate of concept empiricism in these passages as part of a dialectic with himself that he also tells, with more twists and complexities, in EPM itself. Here, as in EPM, Sellars does not explain what, exactly, is wrong with the idea that we can have awareness of universals like redness that is not linguistic awareness – “In view of the widespread acceptance of the thesis in question, there is little need to construct one more argument in its defence”[19] – but he does give what may be part of an argument that fits our hypothetical reconstruction above. He does this in the course of characterizing the conservative concept empiricist as one who thinks that “the process whereby we come to be aware of universals is played by particulars which exemplify these universals.” Sellars continues the story thus:

In its more coherent form, the primary ground of this conviction seems to have been a metaphysical conviction to the effect that that abstract entities exist only in rebus, that is, in particulars, so that only through particulars could mind enter into relations with them…all awareness of universals is derived from the awareness of instances. [20]

The conservative concept empiricist is saying that we cannot be aware of redness unless we are aware of something’s being an instance of redness. It is here that Sellars makes his argument that all propositional awareness is a form of linguistic awareness. According to our reconstruction he does so in two key additional steps. (1) He provides a logical paraphrase of the statement we are aware of something’s being an instance of redness as “we are aware that something is red” thus eliminating the grammatical basis for affirming the existence of the universal redness. He does not give this step in the present paper but does give it, as part of a general theory of ontological commitment, in “Grammar and Ontology”[21] delivered as two lectures at Yale in 1958,[22] the same year as “Is There a Synthetic A Priori?” was published. (2) He argues that statements which do not name universals but introduce them grammatically only as adjectives – subject-predicate statements - must be understood as assertions in a language game rather than as some kind of immediate grasping of a proposition. The following passage is from section 7 of “Is There a Synthetic A Priori?” but could just a well as come from the Trilemma Argument of EPM.

Let us assume, then, that the situation which obtains when it is true to say that Jones is aware of a quality or relation or possibility or, even, a particular, can (in principle) be exhaustively described in terms of and dispositions relating to the use of linguistic symbols…. In other words, we are committed to the abandonment of what has happily been called the metaphor of the mental eye…[23]

It seems to us that the argument of Section 7 of “Is There a Synthetic A Priori?”, lightly reconstructed, is the intended argument against the Section-29 version of classical sense datum theory that we expect to find, but did not, in section 6 of EPM. (This is the argument that we have summarized in the italicized passage given earlier in this section.) We now turn to a criticism of those arguments.

Part II: Critique

5. Introduction

EPM serves as a watershed in the generally downward fortunes of foundationalism in the latter half of the twentieth century. However, starting with the growth of interest in externalist epistemology in the 1980's, there has been a resurgence of interest in foundationalism.[24] In the last few years there have even been attempts to revive classical foundationalism.[25] Sellars's own epistemology is, of course, thoroughly internalist in the sense that if anything is to serve as epistemic grounds for anything else both have to be the objects of justified belief. Some defenders of foundationalism[26] have maintained that Sellars’s argument against foundationalism rests on this premise. There is some plausibility to this claim but we shall not further investigate it here.

In a dilemma introduced at the outset of EPM by Sellars, he offers proponents of foundationalism a choice between taking what is sensed as being a particular or as being a fact. (EPM, 128-29) If the former, then there is not a logical, hence not an epistemic, relation between sensing and believing that p. So that choice must be rejected. If the foundationalist chooses the only other option – sensing is of propositional form – then sensing is knowing Sellars will argue, and cannot fall into the category of what is given. This is one of Sellars’s main conclusions, taken in Section (6). We have argued, however, that this is so because of an assumption of Sellars’s that knowing is in the category of judging, and judging is a language-dependent activity that depends on other activities and skills that no one, especially no empiricist, maintains is simply given in experience. But what of understanding knowing-that on the model of immediate awareness of a fact as the notion is classically understood? Sellars emphatically denies that this is possible in many places in EPM but, in addition to the many as-I-will-shows and as-I-have-showns, he does not actually show it. We have maintained that his key argument against it is a form of property anti-realism and originates in his discussion of a position he calls “Concept Empiricism” in his article “Is There a Synthetic A Priori ?” The argument ultimately depends on taking language expressing propositional-perception as fundamental and language expressing non-propositional-perception as somehow incomplete and derivative from the former. This is the ultimate basis on which Sellars rejects the classical doctrine of immediate knowledge and with it the doctrine of the given.

Before proceeding, we wish to forestall an objection. In saying that, for Sellars, all locutions expressing non-propositional perception, e.g., “Smith sees (what is in fact) the last Ford off the assembly line in 2004” must be paraphrased in propositional-perceptual locutions, we do not mean that he thinks that the paraphrase is a simple conversion of the initial sentence to a propositional version, “Smith sees that that is the last Ford off the assembly line in 2004.” Rather , the paraphrase is to the effect that there is something x that is the last Ford off the assembly line in 2004 and Smith sees that x, under some description or demonstrative reference, has some observable property P. The locution, “x has some observable property P” is then itself paraphrased in terms of a suitable subject-predicate statement eliminating ontological commitment to P.

We note that this account requires a way of demarcating those uses of “sees that” locutions which are not directly perceptual from those that are directly perceptual. We think that this is the Achilles heel of Sellars’s argument against immediate knowledge for we maintain that Sellars has no way to successfully draw the distinction if he is successful in maintaining that non-propositional perception is reducible to propositional perception. Since it is intuitively clear that there is a distinction between correct applications of the indirectly-perceptual and the directly-perceptual language of propositional perception, this means that the reduction-thesis must be rejected. Since this thesis is a key premise in Sellars’s case against immediate knowledge, that case falls to the ground. This is our critique of Sellars’s case against the given.

Our own suggestion how to distinguish the directly-perceptual from the indirectly-perceptual applications of locutions like “z sees that x is F” begins with a semantic account: “z sees that x is F” means something like “z knows that x is F on the basis of visual evidence.” We understand the basing relation to encompass both conscious inference (from one propositional awareness to another) and unconscious inference (from something that may or may not be propositional awareness to propositional awareness). Our suggestion for the visual evidence is that it be expressed in terms of non-propositional perception-ascriptions. In order for such ascriptions to be useable as evidence in conscious perceptual inference, two conditions need to hold: (1) the non-propositional statement cannot be given a Sellarsian propositional paraphrase, and (2) we must have a conceptual grasp of the proposition expressing the non-propositional perception. Since non-propositional perception is expressed in terms occurring within the common sense framework, no objection in principle to (2) comes from fundamental aspects of Sellarsian doctrine. Regarding (1), rejecting the paraphrase confronts the challenge of the Sellarsian argument for the paraphrase given above in section 3.2, but we believe that this challenge can be met, though we do not propose to meet it here. One feature of this account is that it assigns a single meaning to “sees that” locutions used in both direct and indirect applications. The difference between the two applications comes in how the evidence is specified: in the direct case the evidence requires seeing the object referenced in the content of the belief, in the indirect case it requires seeing an object distinct from that referenced in the content of the belief. We regard this feature as a virtue of our account on the grounds that one should not multiply meanings beyond necessity.

6. Is all perception propositional perception?

The key premise in Sellars’s attack on Section-29 SDT is the doctrine that there is no irreducible notion of perception that does not take a propositional object. Call this “Sellars’s propositional doctrine.” Our textual case for this has so far rested on material drawn from works other than EPM, but there is in EPM a text where Sellars says of a non-propositional formulation of seeing, “seeing a red and triangular replica,” that it is a cognitive state, and cognitive states for Sellars are ultimately propositional. (The passage is part of Sellars’s discussion of the “replica” model for sensing in section 60.) Here is the passage:

…the model for an impression of a red triangle is a red and rectangular replica, not a seeing of a red and rectangular replica….And by taking the model to be seeing a red and triangular replica, it smuggles into the language of impressions the logic of the language of thoughts. For seeing is a cognitive episode which involves the framework of thoughts, and to take it as a model is to give aid and comfort to the assimilation of impressions to thoughts, and thoughts to impressions which, as I have already pointed out, is responsible for many of the confusions of the classical account of both thoughts and impressions . [EPM, 191; Sellars’s emphasis in italics; our emphasis in bold]

Is he right about the propositional doctrine? We don’t think so.

Our argument begins with the observation that "sees that" has a perfectly standard non directly-perceptual use as well as a directly-perceptual use. An example of the non directly-perceptual use is, “I see that the sky is cloudy” made on the basis of a report in the weather section of the local paper that I am reading before I rise from my bed and look at the day outside. Of course another legitimate use of the locution “I see that the sky is cloudy” occurs when I do look outside and see that the sky is cloudy. Our objection to Sellars’s propositional doctrine will be that it prevents him from giving an informative explanation of the difference between these two uses. This is a fatal difficulty for Sellars’s theory of perception since it shows that the theory fails to explain what, within the commonsense framework, is essentially perceptual about propositional perception. If he abandons the doctrine , thus allowing for ineliminable non-propositional forms of perception-ascription, then he can provide the needed explanation but loses the key premise in his case against the Section-29 SDT version of the doctrine of the given. Either way, the combination of Sellars’s theory of knowledge and his theory of perception in EPM is unsatisfactory.

The difference between the directly –perceptual and the indirectly-perceptual use of “sees that” seems intuitively to be that in the indirect case the things referred to in the propositional content are not themselves seen whereas in the direct case the things referred to in the propositional content are seen. In our example, although I see a sentence on the page of the newspaper stating that the sky is cloudy, I do not see the sky itself, whereas I do see the sky itself when I look out my window. Now the locution “I see the sky itself” is a non-propositional perception verb, ascribing a relation between the sky and myself. For us this is the end of the story but for Sellars it is not, since he is now committed to paraphrasing non-propositional-perception locutions as propositional-perceptual locutions. In our example the paraphrase of “I see the sky itself” would be something along the lines of: “I see that the sky …” where the ellipses are filled in by a predicate of some kind, for example , “is cloudy.” But this returns either the very same sentence with which we began (“I see that the sky is cloudy”) or another sentence with the same subject and a different predicate. And so the problem recurs: what makes this use of that locution a directly-perceptual use as opposed to an indirectly-perceptual use? Intuitively we want to say that it is that we see the sky itself in the former but not in the latter case. But this response must lead to an infinite regress for Sellars and is therefore unsatisfactory.

It might be argued that this need not be a fatal problem for Sellars, since it may be possible to give a direct characterization of the difference between directly-perceptual and indirectly-perceptual uses of the same “sees that” locution in terms of the difference in the qualitative character of the experiences in each case. The essentials of Sellars’s account of experience in EPM is given in Section 22, the text in which he introduces the notion of “descriptive content”.[27] To arrive at this content Sellars starts with a “rough and ready” characterization of “looks” as follows,

‘x looks red to S’ has the sense of ‘S has an experience which involves in a unique way the idea that x is red and involves it in such a way that if this idea were true [and if S knew that the circumstances were normal] the experience would be characterized as a seeing that x is red.’ (EPM, 151)

and proceeds to identify the descriptive content by subtracting from the content of “sees that” a degree of endorsement of the propositional content common to both seeings-that and lookings. (In the case of seeings-that, the endorsement of the idea that that is red is categorical, with lookings it is tentative.) These “experiences” still have some residual content in common: this is the “descriptive content”. This content is what (two kinds[28] of) lookings-that and seeings-that share, but seeings-that and mere believings do not. Concerning the descriptive content Sellars says the following:

Now, and this is the decisive point, in characterizing these three experiences, as respectively, a seeing that x, over there, is red, its looking to one as though x, over there, is red, and its looking to one as though there were a red object over there, we do not specify this common descriptive content save indirectly, by implying that if the common propositional content were true [and if the subject knew that the circumstances were normal], then all these three situations would be cases of seeing that x over there is red. (EPM, 152)

The characterization of the descriptive content is a functional characterization keyed to the common propositional content between seeing-that and looking-as-though and it is this common propositional content that allows Sellars to identify, though not to intrinsically describe, a common experiential content between the two experiences. Thus, Sellars will attribute the same descriptive content to “I see that the tie is green” as to “it looks as though the tie is green” because they both express the same propositional content: “the tie is green.” Our question is whether we can adapt this idea to explain the difference between directly perceiving that the sky is cloudy and indirectly-perceiving that the sky is cloudy. The prospects for doing so dim when we consider that while there is here the same propositional content, “The sky is cloudy” the need is to explain a different experiential content, that between coming to endorse the proposition that the sky is cloudy because of what I see when looking at the sky and coming to believe this because of what I see when looking at words on the page of the newspaper. This way of putting things suggests that the difference between these two experiences is causal: the sense in which direct-perception is direct and indirect-perception is indirect is causal. We now consider whether a causal account of the difference conceptualized pre-scientifically (within the Manifest Image) is available to Sellars.

It is clear that there is a causal difference between direct and indirect perception in the kind of cases we are considering. In both cases we have been considering there is a series of events beginning with electromagnetic radiation from the sky eventually triggering a chain of electro-chemical events in S’s brain which results in S believing that the sky is cloudy. In the indirect case the chain is not from sky to S’s eye directly and contains brain states that somehow represent something other than x (the words on the page in our example) whereas in the direct case the chain is directly from sky to S’s eye and contains brain states that do represent x itself (the sky in our example.) This is a rough and ready scientific account of the causal difference between direct and indirect perception which we may, for the sake of argument, assume to be correct. It is not, however, an account that Sellars can rely on since this account does not offer an analysis of what we understand by the difference between these two experiences within the conceptual framework of commonsense (the Manifest Image) but is, rather, an excursion into the Scientific Image, a major transgression of Sellarsian method.

But perhaps Sellars’s position is that within commonsense there is the idea that objects can cause beliefs (or candid utterances of sentences) in some intuitively standard and direct way and that this commonsense idea will suffice to differentiate the directly-perceptual applications and the non-directly perceptual applications of sees-that locutions. Thus, when I say “I see that the sky is cloudy” in the former applications I think that the sky causes the utterance in some standard, direct way but in the latter applications I do not think this. Let’s grant that there is this notion within the common sense framework. The proposal, formally put, is that there are two senses of seeing-that locutions, a direct and an indirect sense, and when we say that S sees that x is F in the direct sense the proposition expressed entails that x causes the belief that x is F in a direct, standard way whereas when we say that S sees that x is F in the indirect sense, the proposition expressed does not entail this.

We think that there remain two problems with this account. The first is that the only “standard, direct way” of understanding causation available within the Manifest Image that will do for this purpose is non-propositional seeing – clairvoyance, for example, another direct form of belief causation, will not do – but for the reasons given above, Sellars cannot appeal to this notion. Second, Sellars’s proposal assumes that there are two distinct senses of sees-that locutions, one direct, one indirect, in our commonsense understanding. But we dispute this. We agree of course that there are different applications (direct vs. indirect) of the locution -- this is the distinction we are trying to explicate -- but different applications of a univocal locution and different senses of an ambiguous locution are not the same thing. [29] Intuitively (for us) when one says “I see that the sky is cloudy” when seeing the sky and when one utters these words when seeing the words in the newspaper, one is using that locution with the same meaning. Previously, we have offered a univocal analysis of sees-that locutions that allows for an explanation for the two types of application.

There is one other way to give a causal account of the difference between direct and indirect perception within the commonsense framework that we will consider. We note that one of the differences between the two applications is that in the indirect case we are usually aware of arriving at the belief by inferring it from some other, intervening belief whereas in the direct case we do not. Thus, in our example, the causal chain leading one to believe that the sky is cloudy contains an inference from the belief that the newspaper says that the sky is cloudy, whereas no intervening belief occurs in the direct case. The difference between inferential and non-inferential knowledge is well established within EPM, asserted, indeed, in its very first sentence. (EPM, 127.) However, although the presence of conscious inference in the indirect case and its absence in the direct case are usual, they are not inevitable and do not provide an account of the essential difference between the cases. Thus, it is possible for someone to look at the words in a newspaper “The sky is cloudy” and to come to believe immediately, that is, without conscious inference from the belief that those words appear on the page, that the sky is indeed cloudy. Conversely, it is possible to look at the sky, say to oneself that one sees cloud-shaped and -colored objects overhead and come to believe on this basis that the sky is cloudy. This is a case of seeing that the sky is cloudy directly but it is also a case where the belief-component of “seeing that” is arrived at by inference.

Our conclusion is that the difference between direct and indirect applications of “sees that” cannot be satisfactorily explained causally by appeal to the inferential status of the belief-component of “seeing that.” Indeed, we believe that no causal account occurring within the Manifest Image and not involving an ineliminable sense of non-propositional seeing can satisfactorily be given. This is because the difference between the two cases is essentially a difference in the sense experiences we have in each case, not a difference in non-experiential causal mechanisms.

We have already seen that Sellars’s account of experience as descriptive content does not provide an answer to our problem. There are two other places to look for an account of experience that would do the job. One is in an account of phenomenal-looks talk, for example, for locutions like “that looks red.” But Sellars rejects such an account for reasons that we cannot go into here.[30] The second is in the texts where Sellars gives his account of how the functional characterization of the intrinsic content is to be given intrinsic specification. This account is given in the final part of EPM, sections 60-62.

Sellars treats the specification of the intrinsic properties of the descriptive content not as a problem of philosophical analysis or introspective attentiveness but of the introduction of theoretical entities. (EPM, 192) Typically, theoretical entities are introduced in science by means of models (analogies) and Sellars maintains that doing the same here will prove helpful.

This time the model is the idea of a range of ‘inner replicas’ which, when brought about in standard conditions, share the perceptible characteristics of their physical source….Thus the model for an impression of a red triangle is a red and triangular replica not a seeing of a red and triangular replica. (EPM, 191).

Sellars’s idea seems to be that the differences between the experential (descriptive) content of seeing that p and seeing that q are isomorphic to the differences between the standard causes of these experiences when these causes are taken as physical objects. The physical objects are presumably those referred to in the conceptual component of the seeings-that. Thus if I see that the tie is green and I see that the apple is red, the standard causes of the descriptive component in each case are, respectively, a green tie and a red apple. If (1) we took the set P of all of the different propositions that we can see to be the case and that look to be the case, and (2) we took the set E (“E” for experience) of all of the descriptive contents associated with sees-that- and looks-that-pairs taking P as content, and (3) we took the set C of all of the standard causes for each sees-that element of E, we would, says Sellars, (4) find an isomorphism between the members of E and C. Sellars then reasons that intrinsic descriptions of the members of E should match the intrinsic description of the members of C. How they match is explained through the model: the theoretical entities (“sense impressions”) are understood to be analogous to entities with the color and shape properties of the objects in C. Put another way, it is as if sense impressions were objects with the superficial spatial properties and colours of the ordinary physical objects causing our sense impressions, though they are in fact from a different category altogether. Adopting this model assigns intrinsic properties to sense experiences in such a way as to explain our intuitive sense of how the members of E differ from one another. At the same time, Sellars can explain the difference between mere believings and seeings/lookings: in the case of believings the sense impressions do not occur.

There is one main difficulty with employing the theoretical apparatus of sense impressions to help distinguish direct and indirect applications of propositional perceptual ascriptions. This difficulty derives from the fact that the explanandum must be based on intuitive conceptual awarenesses of the descriptive content characterized by Sellars in terms of the general difference between propositional perceptual states and believings and of the general similarity between seeings and lookings when the propositional content is the same. This is a difficulty for Sellars’s proposal as a whole. He summarizes his proposal thus:

It will be remembered that we had reached a point which, as far as we could see, the phrase “impression of a red triangle” could only mean something like “that state of a perceiver – over and above the idea that there is a red and triangular physical object over there – which is common to those situations in which

(a) he sees that the object over there is red and triangular;

(b) the object over there looks to him to be red and triangular

The difficulty is that Sellars has no account of what conceptualization legitimately available in the Manifest Image might underwrite the similarity- and difference-judgments at issue in this passage. We remind ourselves that Sellars cannot appeal to a conceptual awareness of the sense impressions themselves – this is precisely what Sellars has been arguing against throughout EPM.

In making this criticism we are assuming that comparisons of similarity and difference are not immediate, that is, that they require that there be either a determinate or a determinable concept under which the person making the judgment can correctly subsume the objects being compared. For example, if we judge of two table settings on two tables that we observe in a restaurant that they share something in common, there must be some concepts correctly descriptive of the arrangements, e.g. the smaller of the blue plates are always set to right of the larger of the white plates, available to us and employed by us in making the comparison. Even if we may lack specific concepts we must at least have determinable concepts – the same color, a different shape – when we make comparisons. There must, in short, be some intrinsic descriptions available within a conceptual framework F and applicable to a pair of objects o and o’ for someone operating within F to apply similarity- or difference-comparisons to the pair o and o’. On Sellars’s own account, this is just what we lack in making a comparison between seeing that there is a green tie before us and its looking as though there is one. So this is our problem. We say that there is something in common here. But what? Under what description do we cognize the two states so that we might infer their similarity? As far as we can see, Sellars’s only response, offered with no apparent argument, must be that we do not infer the similarity from the same intrinsic descriptions but that we cognize it immediately – we just see that there is this similarity. But this is the doctrine of the given all over again – in its most objectionable form – and it has ensnared Sellars himself here.

We are not, certainly, denying that there is something in common between perceptual seeings-that and lookings-that and that we can cognize the similarity within the manifest image. There is, and we do: we see the objects non-propositionally, and we believe (are conceptually aware) that we see them non-propositionally. This intuitive answer explains our awareness of the difference between seeing that the tie is green and seeing that the apple is red: in the first case we see a green tie and believe that we see a green tie, in the second case we see a red apple and believe that we see a red apple. It also explains our awareness of the difference between seeing that the tie is green and just believing that it is – in the former case we see the tie and believe that we do, in the latter we do not see the tie and believe that we do not.

It is, admittedly, a bit more difficult to apply this explanation to the similarity we are aware of between its looking as though the tie is green when we believe that there is no tie and our seeing that the tie is green, but even here an intuitive answer is available: when we think that we see that the tie is green, we think that we see a real tie-shaped green object and when we think that it looks as though the tie is green when there is no tie, we think that it is as though we see a tie-shaped green object even though no such object is present. Like Sellars, we appeal to a model here (in this case an intuitive model) but unlike Sellars’s theoretical model the model is seeing a tie-shaped and colored object, a “replica” in Sellars’s terminology, rather than the replica itself.[31] We think that the former is the intuitive explanation, an explanation not available to Sellars since it depends on an analogical appeal to ineliminable non-propositional perception-ascriptions, something ruled out by Sellars’s Propositional Doctrine. Abandon that doctrine and the intuitive answers become possible for Sellars but at the price of giving up the main premise in his argument against the realist version of the classical sense datum doctrine of the given. Sellars can have a theory of perception adequate to the Manifest Image in EPM or he can have an epistemology adequate to refute the doctrine of the given in its Section-29 (immediate-knowledge) form in EPM; he cannot have them both.

Notes