CHAPTER 7

FOUNDATIONALISM

      My purpose in this chapter is to locate Sellars' stance in the Foundationalist/Non-Foundationalist controversy. He is reputed to be a Non-Foundationalist -- even the father of Non-Foundationalism -- because of his rejection of "the given"; and it is held that his essay "Empiricism and the Philosophy of Mind" (originally titled "The Myth of the Given") is the locus classicus for Non-Foundationalist arguments. I will take Laurence Bonjour's book The Structure of Empirical Knowledge to be a good representation of this view.

      The only commentators that I am aware of that have dissented from calling Sellars a straightforward Non-Foundationalist are Cornelius Delaney, James Cornman, and William Alston.{1}

      Delaney argues that Sellars' position is a dialectical overcoming of both foundationalism and coherentism through a rejection of their common presupposition that there is only one kind of inference at issue; where in fact there are really two kinds -- a primary level and a metalevel inference. Coherentists are right that justification of basic empirical reports requires inference, but they are wrong in thinking that it requires primary level inference. Foundationalists, on the other hand, are correct in rejecting primary level inference, but they are wrong in rejecting metalevel (trans-level) inference.

      Cornman, in his own right, is also correct in pointing out that "Quine and Sellars can be foundationalists."{2} His point is that the argument in Sellars' "Empiricism and the Philosophy of Mind"{3} is directed only against foundationalism construed as claiming that reports about physical objects are basic (what I will call Pseudo-Foundationalism). And, he notes correctly that this argument is not against foundationalism construed as reports about phenomenal reports (what I will call Genuine Foundationalism).

      Alston too is wary of presenting Sellars as a straightforward Non-Foundationalist:

It may be doubted that Sellars can be counted among the foes of immediate knowledge, for he is wont to present his position as a sort of synthesis of foundationalism and coherentism.{4}

      Both Delaney and Cornman are right, but they fail to carry the reasoning far enough. Bonjour carries the reasoning a step further in the right direction by entertaining the possibility of an a priori defense of Foundationalism. In speculating on various defenses of Foundationalism, he attributes an a priori justification to Lewis, Chisholm, and Firth. He even speculates that if Foundationalism is to be justified at all, it will have to be justified on a priori grounds. His two shortcomings are his failure to include Sellars in this group, and his failure to spell out that the form of an a priori justification should be a transcendental argument.

      I will argue that Sellars has an a priori defense (transcendental argument) for a version of Pseudo Foundationalism. Simply put, the justification for the reliability of reports about physical objects is that if they were not reliable for the most part, then we wouldn't have empirical knowledge. Their reliability is a necessary condition for the very possibility of empirical knowledge. This has been missed by most expositors of Sellars' epistemology -- including C. F. Delaney and L. Bonjour.

      To put matters in perspective, I will start with the classical analysis of (empirical) knowledge -- which Sellars accepts -- as 'justified true belief'. I will state the regress problem for a justification of knowledge, and indicate how Foundationalism is introduced as a solution. Next, I will point out a scope difference between the classical and the contemporary approaches. And I will distinguish Pseudo Foundationalism from Genuine Foundationalism. I will agree that Sellars has refuted a version of Pseudo Foundationalism, but not Genuine Foundationalism -- a position I will defend with a transcendental argument. Finally, I will present objections to Sellars' argument against Genuine Foundationalism.

I. ANALYSIS OF KNOWLEDGE

      The theory of knowledge has its explicit origin in Plato's attempt in the dialogue Theaetetus to distinguish true opinion from knowledge. Starting from this intention, the distinction has ossified into the formula that knowledge is true judgment (opinion, belief) with a "logos," "account", "warrant", or "justification". As Bonjour puts it: "'A knows that P' if and only if '(1) A must believe confidently that P, (2) P must be true, and (3) A's belief that P must be adequately justified'."{5} Sellars agrees with this analysis: "The explication of knowledge as 'justified true belief' . . . remains the orthodox or classical account and is, I believe, essentially sound."{6}

      Bonjour adds that a theory of adequate justification must also provide a metajustification, or vindication:

it is incumbent on the proponent of such an epistemological theory to provide an argument or rationale of some sort to show that his proposed standards of justification are indeed truth-conducive . . .{7}

      Most of the discussion has centered on providing an unpacking of "adequate justification." Since the concern with the justification of knowledge is ultimately a concern with how to minimize error and how to maximize truth or, at least, verisimilitude, an 'account' is intended to be a specification of a reasonable guide to these ends. And the metajustification (vindication) is to apply to any such proposed guide.

      Broadly speaking, we may say that the methods which secure truth and minimize error are deductive and inductive (including abductive) reasoning. Deductive reasoning secures true conclusions relative to the truth of the premises; while inductive reasoning operates with the less secure truth preserving criterion of reasonableness (warrantedness, probability). Bonjour's demand for metajustification would require a vindication of deduction and induction. Let us assume that both deductive and inductive methods are reliable. Still the philosophically interesting question is to probe for the ways that the truth of the premises (including hypotheses) of deduction and induction are themselves justified. The immediately obvious available answer is that the premises are justified on the basis of previous deductive or inductive arguments. But we can again repeat our request for the justification of the truth of these prior premises, and we are off on an infinitely regressive path of justifications. How to terminate this kind of questioning constitutes the regress problem concerning the justification of knowledge. There seem to be only a few possible answers. Bonjour formulates these nicely in the following passage.

1. The regress might terminate with beliefs which are offered as justifying premises for earlier beliefs but for which no justification of any kind, however implicit, is available when they are challenged in turn.
2. The regress might continue indefinitely "backwards," with ever more new empirical premise-beliefs being introduced, so that no belief is repeated in the sequence and yet no end is ever reached.
3. The regress might circle back upon itself, so that if the demand for justification is pushed far enough, beliefs which have already appeared as premises (and have themselves been provisionally justified) earlier in the sequence of justificatory arguments are again appealed to as justifying premises.
4. The regress might terminate because "basic" empirical beliefs are 3 reached, beliefs which have a degree of epistemic justification which is not inferentially dependent on other empirical beliefs and thus raises no further issue of empirical justification.{8}

      The other alternative, in view of the E. Gettier problem,{9} is to be skeptical. Few contemporary philosophers have taken the skeptic stance; but a few have -- notably Peter Unger and Keith Lehrer.{10} Their skepticism amounts to the claim that given the concept of knowledge as requiring a certainty which guarantees truth, there is no empirical 'knowledge', strictly speaking; instead there are more or less justified or justifiable beliefs. In this sense, Sellars, too, is a skeptic. The more modest pursuit is to find the most reasonable beliefs, despite fallibilism. So even if it is granted (which it is not by Unger) that there is a distinction between justified (reasonable) and unjustified (unreasonable) beliefs, the regress argument still has its bite. The problem becomes how to spell out these conditions of reasonableness, and to determine whether some set of beliefs has a privileged place. Most philosopher who believe that there is a rational answer to the regress problem have opted either for "foundationalism" or "coherentism."

II. THE CLASSICAL VIEW OF KNOWLEDGE

      The traditional analysis of knowledge assumes two terminological conventions. For Plato the terminological convention was that (1) the object of "knowledge" must be unalterable, and that (2) it is possible to make a voluntary judgment about the truth or falsehood of a proposition about the object. This terminological convention presents problems in talking about a knowledge of truths about sense perception. Plato took it for granted that sense perception is about a Heracleitean flux, and that such 'cognition' is involuntarily indubitable. Hence such cognition, because it violates the two terminological criteria -- by being about an alterable, and by being involuntarily indubitable -- is not called 'knowledge'.

      A similar terminology about 'knowledge' is preserved in Sextus Empiricus. He contrasts reports of sense phenomena (appearances) with (empirical) judgments: "Skepticism is an ability, or mental attitude, which opposes appearances to judgments in any way whatever. . . ";{11} and "By 'appearances' we now mean the objects of sense-perception, whence we contrast them with the objects of thought or "judgments."{12}

      Sextus provides the marks of our cognition of sense phenomena whereby they are distinguished from knowledge: "we do no overthrow the affective sense-impressions which induce our assent involuntarily; and these impressions are 'the appearances'"{13} and

The criterion, then, of the Skeptic School is, we say, the appearance, giving this name to what is virtually the sense-presentation. For since this lies in feeling and involuntary affection, it is not open to question. Consequently, no one, I suppose, disputes that the underlying object has this or that appearance; the point of dispute is whether the object is in reality such as it appears to be.{14}

The problem, as the Skeptics see it, is this. Cognitions present themselves as being of two types: those to which we find ourselves assenting involuntarily and those to which we can voluntarily decide to assent or not to assent. Those cognitions which force assent from us are the appearances; those to which we can give voluntary assent to are judgments. Knowledge, if it is to be found, for the Skeptic, is to be found within the realm of voluntary judgments. Knowledge is equated with justified true voluntary judgment, or, using the classical Skeptic terminology, knowledge is said to be true judgment according to the criterion. The criterion is that which is needed to warrant rational assent. This formulation, it should be pointed out, was intended by the Skeptics for only voluntary judgments. Since sense perception is involuntary, it was placed outside the realm of 'knowledge'. This use of 'knowledge' may be viewed by us as simply a terminological convention or proposal.

      The same terminological convention is to be found in Descartes, who talks of error in those areas of cognitions where we can voluntarily assent, dissent, or withhold both assent and dissent: "Now as to what concerns ideas, if we consider them only in themselves and do not relate them to anything else beyond themselves, they cannot properly speaking be false . . . "{15} and, "if I consider the ideas only as certain modes of my thoughts, without trying to relate them to anything beyond, they could scarcely give me material for error."{16}

      We come across this Platonic-Skeptical-Cartesian terminology also in C. I. Lewis, who refrained from using the words 'knowledge' and 'judgment' in relation to sense presentations or 'the given'.{17}

Apprehension of the presented quale, being immediate, stands in no need of verification; it is impossible to be mistaken about it. Awareness of it is not judgment in any sense in which judgment may be verified; it is not knowledge in any sense in which "knowledge" connotes the opposite of error.{18}

To forestall the objection that perceptions of objective facts, such as, for example, that what is before me is a pink ice cube also present themselves as involuntarily given, we must distinguish this kind of given which is dubitable from the kind of given which is indubitable.

      It seems to me that the so called argument from "illusions" -- a misnomer for all the ways that cause unveridical perceptions -- was intended to draw this distinction between the dubitably and the indubitably given. The indubitable is the cognition which is couched in the phenomenal appearance terminology about phenomenal properties. This demarcation, I believe, coincides with that of the Platonic-Skeptical-Cartesian-Lewisian tradition which demarcates appearances from judgments.

      The traditional analysis equates knowledge with justified true belief (i.e., voluntary judgment) or, using the Skeptic terminology, knowledge is true judgment according to the criterion. The criterion is whatever is needed to warrant assent to, dissent from, or suspension of judgment. This formulation, it should be remembered, was pertinent only to those beliefs which it is possible to judge voluntarily. Since sense presentations are involuntary, they are, in the classical terminology, outside the realm of 'knowledge'.

      Two terminological conventions have changed in the course of time from Plato to now. We talk of empirical knowledge, and the criterion of voluntary assent has been dropped.

III. FOUNDATIONALISM

      An examination of the literature on the foundations of knowledge quickly reveals a sort of tower of Babel, where talk is at cross-purposes. The reason for this is that the concepts of foundationalism and coherentism varies from author to author.

      The term 'foundations' is a metaphor suggesting the foundations of some structure, usually a house; sometimes a pyramid. To have a contrasting metaphor for 'coherentism' the images of circles and rafts have been used. Here are Sellars' images:

One seems forced to choose between the picture of an elephant which rests on a tortoise (what supports the tortoise?) and the picture of a great Hegelian serpent of knowledge with its tail in its mouth (where does it begin?). Neither will do.{19}

The foundations of the pyramid (or tortoise) is the metaphor for the justification of basic premises. How this justification is to be described varies in classical foundationalism. The proposed marks or criteria of basic premises have been certainty, incorrigibility, infallibility, indubitability, and/or self-warrantability.

      By contrast the image of the raft (or serpent) will mean the requirements of consistency and coherence. How 'coherence' is to be understood has, like the notion of foundational security, several interpretation. For example, in C. I. Lewis we find the term 'congruence' and in R. Chisholm the term 'concurrence' used for this purpose; while in Sellars we come across the notion of 'explanatory coherence'. The notion of coherence and these related notions are meant, I suspect, to express the requirement of systematic logical, temporal, spatial, and causal relations.

      But are foundationalism and coherentism contradictory alternatives? Since the defense of coherentism has been by way of an attack on foundationalism, i.e., a defense of non-foundationalism, in order to understand the foundationalist/non-foundationalist debate we need some minimal thesis which is in dispute. Let me, then, try to formulate it. If we view the situation abstractly and metaphorically, the foundationalism versus non-foundationalism debate could contain the following claims:

(EC) The raft exists; there is no pyramid.
(EF) The pyramid exists; there is no raft.
(M) The pyramid and the raft both exist.
(MC) The raft is a necessary condition for the pyramid, but the pyramid is not the necessary condition for the raft.
(MF) The pyramid is the necessary condition for the raft, but the raft is not a necessary condition for the pyramid.
(W) The pyramid and the raft entail each other.

      I will call (EC) Extreme Coherentism and (EF) Extreme Foundationalism -- no one holds such a view; (MC) is Moderate Coherentism, while (MF) is Moderate Foundationalism; (W) can be viewed either as Weak Coherentism or as Weak Foundationalism.

      To decide which of these alternatives is defended by so called foundationalists and non-foundationalists, let us look at the writing of a self-proclaimed foundationalist -- Roderick Chisholm; and the writings of the self-conscious non-foundationalist -- Laurence Bonjour. I am not sure where to place others, including Sellars. And since I am writing about Sellars, I will shortly explain why there is a difficulty in placing Sellars at either end of these poles. Let me first turn to Chisholm. He defends a version of foundationalism which has a foundationalist and a coherentist aspect. He writes: "There are two moments of epistemic justification, one of them foundational and the other not."{20} As Chisholm sees the epistemic situation, both sets of metaphors will do: they express two aspects of knowledge. Now, this is a significant thesis. It indicates that foundationalism does not deny coherence as a necessary condition for securing knowledge. And if this is the pronouncement of a foundationalist, which seems to fit the alternative (W) of Weak Foundationalism, then a non-foundationalist, who opposes Chisholm, must be defending either (EC) or (MC). I am not going to pursue the structural problem of how various types of beliefs are tied together as, for example, in various modes of induction or coherence; so I am not interested is dealing with the superstructure of the foundational-coherentist issue, save for one respect. I want to defend a necessary condition for foundationalism.

      I will begin with the following characterization of Foundationalism by Laurence Bonjour:

(F)

the central thesis of epistemological foundationalism as understood here, is the twofold thesis: (a) that some empirical beliefs possess a measure of epistemic justification which is somehow immediate or intrinsic to them, at least in the sense of not being dependent, inferentially or otherwise, on the epistemic justification of other empirical beliefs; and (b) that it is these "basic beliefs," as they are sometimes called, which are the ultimate source of all justification for all of empirical knowledge.{21}

      This formulation needs some clarification and qualification. First, Bonjour's formulation (F-b) is not quite satisfactory. It is misleading and it says too much. It is misleading by using the word 'source'. Elsewhere when he explains himself, he writes: "if inference from them [basic beliefs] is to be the sole basis upon which other empirical beliefs are justified . . . " [my emphasis].{22} My objection is to the word 'from': it should be 'to' on the construal of the justification of empirical knowledge which I have in mind. Let me explain the distinction. Bonjour is thinking of foundationalism on the model of an axiomatic system, such as, for example, Euclidean geometry.{23} On this model, the axioms are the "foundations" and the theorems are the "superstructure." If applied to empirical knowledge, this position would require the following reasoning. Suppose we are talking about how something appears, using the form 'D appears F to S', then, on Bonjour's construal of "foundational" justification, S can infer from 'D appears F' to 'D is F' or to 'D will appear G'. But this formulation is ambiguous. (i) Either we are talking here of an Aristotelian inference in the sense of it being mediated by a premise, or (ii) a Humean inference, i.e., a causal association between 'D appears F' and 'D is F'. If it is an Aristotelian inference, then we need a mediating premise, something like 'If D appears F under normal conditions to normal observers, then it usually is F'. But if this is the line of reasoning required, then obviously other empirical knowledge is being presupposed; namely, that the conditions and the observer are normal. This interpretation of foundationalism, obviously won't do since it requires other empirical knowledge.

      The other line of reasoning (ii), would have us go directly from 'D appears F' to 'D is F' or to 'D will appear G'. But if there are no mediatory premises (as is assumed), then these are non-rational events which we find ourselves suffering.

      Bonjour seems to have the latter line of reasoning in mind, and this sort of inference -- a Humean inference -- is, as Sellars would say, a non-epistemic fact. And in that case, it cannot serve as an epistemic foundation.

      The alternative idea of justification to Bonjour's which I have in mind is based on an inference to, not from, an appearance. In this sense, beliefs about how one is appeared to can be used for confirmation or verification. And if we use the hypothetico-deductive model{24} for this kind of justification (i.e., confirmation or verification), then the situation is one in which we have as premises a lawlike statement and a statement of the initial conditions, and from these we deduce an appearance statement:

Lawlike statement
Observation reports (initial conditions)
Therefore: Appearance statement

      If the deduced appearance statement is true, then assuming other truths, the basic (appearance) statement confirms to some degree the truth of the premises. From a logical point of view, the relation of the foundational basic empirical statement to that which it supports is: that which is supported entails the foundation, and not the converse.{25}

      On this analogy, perceptual knowledge about physical objects is taken to be a hypothesis, which together with other information entails a statement about how things will appear. And it is the entailment from the perception (as a hypothesis) to the appearance which partially justified perceptual knowledge. It is the logic of this situation which was in dispute, for example, in the debate between Chisholm and Lewis,{26} and is the relevant kind in the question of empirical foundational justification.{27} To ignore this understanding of foundationalism is to attack either some historical position or a straw man.

      Second, Bonjour's (F-b) says too much by writing "all justification." The implication is that basic beliefs provide the sole type of justification for empirical knowledge. Perhaps the better formulation would be to say that basic beliefs are the ultimate necessary conditions for all empirical knowledge. And here we would make contact with something that was sought by the 'verifiability theory of empirical meaning'. After all a Foundationalist need not dismiss other modes of justification simply by insisting on a justification by empirical confirmation. As I pointed out, a foundationalist such as Chisholm insists on the criteria of consistency and congruence in addition to empirical confirmation.

      Clearly (F-b) entails (F-a), but not conversely. I think it suffices to be a non-foundationalist by simply rejecting thesis (F-b). We may then view non-foundationalism to be Weak if it also accepts thesis (F-a), or as Strong if it also rejects thesis (F-a). For example, a Weak non-foundationalist might reject (F-b) in favor of:

(b') that it is these "basic beliefs" which are the ultimate source of some justification for all of empirical knowledge.

      My concern is not with the structure of a foundationalist view, i.e., thesis (F-b) or (F-b'), but only with thesis (F-a), which is compatible with both foundationalism and non-foundationalism. However the denial of (F-a) is incompatible with any version of foundationalism. So a good argument against (F-a) is ipso facto a good argument against foundationalism.

      Let us call empirical beliefs formulated by (F-a) 'basic empirical belief'. Note that according to (F-a) basic empirical beliefs are justified either inferentially or noninferentially, and are not justified by other empirical beliefs.

      Is there a way to further characterize these basic empirical beliefs? On Sellars' construal, basic empirical beliefs must be directly (non-inferentially) apprehended, i.e., spontaneously suffered, and presuppose "no knowledge of other matter of fact, whether particular or general."{28} If we combine Bonjour's and Sellars' understanding of basic empirical beliefs, we get the following formulation:

(BEB)

if p is a basic empirical belief, then
(i) p is caused noninferentially,{29} and
(ii) p is justified either inferentially or noninferentially, and
(iii) p is not justified by other empirical beliefs.

      If we are to include traditional discussions under this formulation, then I think we have to quickly make some verbal bridges. In this formulation, the 'noninferential' in (BEB-i) is to be taken as equivalent to the 'directly apprehended', to the 'intuited', and to the 'given'.

      Let me backtrack a bit because there is ambiguity in Bonjour's formulations. In (F-a), immediate (intrinsic) epistemic justification is characterized as excluding inferential or other epistemic dependency on other empirical beliefs. This requires

(a') an exclusion of all dependency (inferential or not) on other empirical beliefs, particular or general.

      But elsewhere he writes that weak foundationalism "holds that there are basic beliefs having some degree -- though a relatively low one -- of noninferential epistemic justification" [emphasis added].{30} Now this entails

(a'') an exclusion of any kind of inferential dependency on any other belief whatever.

      If Bonjour means to hold (a''), then he is denying our (BEB-ii). In that case, (BEB-ii) would have to be modified to read:

(ii') p is justified noninferentially.

IV. EXTERNALIST FOUNDATIONALISM

      Accepting (ii'), still leaves room for the foundationalist to maneuver. This move, it seems, must be toward what is called an Externalist Foundationalism: a position held, by among others, D. Armstrong, F. Dretske, Alvin Goldman, R. Meyers, and W. Alston.{31} The difference between Internalist and Externalist justifications of knowledge hinges on whether 'justified' is understood as 'justified by what I am aware of' or as 'justified by the circumstances which I may not be aware of', respectively. Since some of these circumstances may be reliable causal chains, the label Causal Theory of knowledge is an apt characterization of such an Externalist position.

      W. Alston, for example, argues for an Externalist justification of knowledge as better fitting the ordinary use of the term 'knowledge', as when we attribute knowledge to animals, children, and non-reflective adults. Bonjour argues against this view.{32} Both agree that knowledge is justified true belief, but Alston uses the word 'justified' as meaning 'justified internally or externally', while Bonjour uses the word in the sense of 'justified internally' only. Alston, in his defense, offers examples where there is only external justification, and appeals to ordinary language and 8 practice in calling such cases instances of knowing; commenting, "If terms like 'knowledge' are confined to the cognitive achievements of critically reflective subjects, we shall have to find a new term for the territory in its full extent";{33} while Bonjour, I take it, would dismiss the authority of ordinary language and practice, and refuse to describe such cases as instances of knowing. At best their disagreement is a normative one concerning the benefits of using 'knowledge' in these respective ways. But I am not interested is this kind of dispute. These seem to be simply different concerns based on different presuppositions. So, I will restrict my discussion to an exploration of the ramifications of assuming the Internalist view of knowledge, which means working with (BEB-ii) rather than (BEB-ii').

V. PSEUDO AND GENUINE FOUNDATIONALISM

      But before I go any further I want to make an important distinction concerning the objects of basic empirical knowledge. There are two candidates which are normally considered: physical objects and appearances (percepts, sensings, sense data). I will call the view which takes reports about physical objects to be basic 'Pseudo-Foundationalism', and the view which takes reports about appearances to be basic 'Genuine Foundationalism'.

Pseudo-Foundationalism, as I use the phrase, is the claim that reports about physical objects (i.e., observation reports) are the foundations of knowledge. The attack against Foundationalism is in many cases really an attack against Pseudo-Foundationalism. This is true, specifically, of Sellars' attacks. An example which Sellars gives is of a report that there is something green in front of him. Such a report, argues Sellars, cannot be foundational because it presupposes other empirical knowledge. Specifically, it presupposes the general knowledge that things have the colors they seem to have under normal illumination, it presupposes the particular knowledge that the illumination is now normal, it presupposes the general knowledge that things have the colors they seem to have to normal observers, and it presupposes the particular know- ledge that I know that I am at this moment a normal observer. Any report, then, such as the seeing of something being green satisfies requirement (BEB-i), but not (BEB-iii). And it is precisely Sellars' argument that observation reports about physical objects, because they do not satisfy the requirements of (BEB), cannot be foundational. Such reports presuppose general and particular empirical knowledge about the conditions of the person making the report and about the context in which the report is made.

      I am satisfied by this argument, and believe that Pseudo-Foundationalism is false. But if this argument against Pseudo-Foundationalism is accepted, and no other kind of justification for empirical knowledge is offered, then it seems to entail a dilemma. If knowledge claims about physical objects presuppose particular and general empirical claims, these in turn require further justification of the same sort, and these in turn other empirical justifications. This seems to entail either an infinite regress, a knowledge of everything, or a circle of justification.

A. GENUINE FOUNDATIONALISM

      Genuine Foundationalism, by contrast, must minimally claim that the basic empirical beliefs are expressed by, what I will call, 'phenomenal' reports -- what Lewis called 'expressive language', and what Chisholm means by 'noncomparative appear statements', and what Moore, Russell, Broad, Ayer, Price, and, more recently, Frank Jackson call reports about 'sense data', 'sensa', or 'qualia'.

      Sellars' position is that the required sorts of phenomenal reports are not available as analyses of claims about the observation of physical objects. But Sellars concedes counterfactually that

as long as looking green is taken to be the notion to which being green is reducible, it could be claimed with considerable plausibility that fundamental concepts pertaining to observable fact have that logical independence of one another which is characteristic of the empiricist tradition.{34}

      Sellars, I take it, is admitting that if (counterfactually) phenomenal reports are entailed as analyses of reports about physical objects, then they can serve as foundations of empirical knowledge.

      One philosopher who does seem to believe in the availability of phenomenal reports is V. W. O. Quine. He holds that empirical knowledge is justified by 'observation sentences' of which some 'occasion sentences' have direct 'stimulus meaning'. He writes:

Thus they are just the sentences on which a scientist will tend to fall back when pressed by doubting colleagues. Moreover, the philosophical doctrine of infallibility of observation sentences is sustained under our version.{35}

      My justification for the availability of phenomenal reports is given in the chapter on perception. Here I assume that they are available. But the discussion which follows can no longer be a confrontation with Sellars, since he rejects the assumption that phenomenal reports are available. However, it has been overlooked by Sellars' sympathizers that Sellars, in the passage quoted, had made the concession that it phenomenal reports were available, they would provide foundations for empirical knowledge. Heedless of this concession, they have tried to muster arguments from Sellars against the foundational character of phenomenal reports -- as if they were available. But given Sellars' concession, he could not possibly agree with such arguments assembled in his name.

      As an example of this misuse of Sellars, consider K. Lehrer's examination of Chisholm's claim that appearance statements (in the non-comparative sense) constitute the foundations of empirical knowledge. First, he grants Chisholm the availability of such statements: "We agree with Chisholm that there is such a thing as a non-comparative use of words."{36} Second, he claims that in describing one's state by using non-comparative appearance words, such language entails "the information needed to enable one to tell such a state from another."{37} Third, he concludes that a non-comparative use of words requires "independent information." Fourth, he says that this argument is similar to one used by Sellars.{38}

      I don't know if his argument is sound, but it is not (relevantly) similar to Sellars' in the cited passage. First, although Sellars recognizes non-comparative uses of appear words, he does not accept the adverbial analysis given to them by Chisholm. Nor does Sellars accept the phenomenological availability of adverbial states as analyses of perceptual reports, as does Chisholm. In any case, in the cited passage, Sellars is considering colors of physical objects, not of sensorial phenomena, as does Lehrer.{39}

1. JUSTIFICATION OF PHENOMENAL BELIEFS

      Before I discuss the justification of phenomenal beliefs, I would first like to make a distinction between a belief about phenomena (sense data, sensa) and a phenomenal (non-parroting) report. The belief could exist without the report, but not conversely. And I immediately concede that linguistic errors are possible in making a report. So the issue must center on the availability of basic empirical beliefs, and not on linguistic reports.

      In discussing the foundations of empirical knowledge, it is not necessary to claim that a specific basic empirical belief is certain, only that most of them have to be true. And if I know that most of them are true, then I may accept any specific basic empirical belief with confidence.

      Given such a weakening of the requirement for foundationalism, Bonjour correctly reconstructs any defense of basic empirical beliefs as requiring the following statistical syllogism:

(JBB)

(1) Belief B has feature F.

(2) Beliefs having feature F are highly likely to be true.

Therefore, B is highly likely to be true.{40}

      Let us concentrate on the first premise. It is clear from Sellars' criterion that a basic empirical belief must have as the feature F the properties (BEB-i) and (BEB-iii). (BEB-i) alone is not sufficient because any belief that we find ourselves having passively can have that status, including beliefs about observations of physical objects. But by criterion (BEB-iii), observational reports must be excluded as foundational. This is shown by Sellars' argument against a version of, what I have called, Pseudo Foundationalism. What other feature, then, is it necessary to add to (BEB-i) and (BEB-iii), if any, to have a viable form of foundationalism? The classical defenses of foundationalism have varied according to how F in (JBB) was interpreted. According to Bonjour, "[Basic beliefs] have been claimed to be not just adequately justified, but also infallible, certain, indubitable, or incorrigible."{41} The conclusion reached by Bonjour and others is that, as a matter of empirical fact, no phenomenal report is free from possible error. I accept such a conclusion. And my reason is simply that as an empirical fact, it is possible to suffer a momentary mental dysfunction. And such a mental dysfunction could in principle be objectively detected if a mind-brain correlation is assumed. Let me explain. A correlation could in principle be set up between verbal reports expressing basic empirical beliefs and brain states, which in principle is detectable by a cerebroscope. Suppose the report was "This is red", but the correlative brain state was "This is blue." The conclusion could be drawn that a mental dysfunction occurred.

      It may be conceded that a person's sense perceptions are psychologically indubitable, but from this it does not follow, so the argument goes, that sense perceptions are certainly true.

      Although I concede that phenomenal reports are not certain because of the possibility of overriding external criteria, I do not concede that phenomenal reports are ever uncertain because of internal evidence. In the following section I will focus attention on how the challenge to the certainty of appearances is made from internal considerations, and why it fails.

a. NELSON GOODMAN ON PHENOMENAL REPORTS

      We find the certainty of phenomenal reports challenged, for example, by Nelson Goodman in the following passage:

But the judgment that I made a few moments ago that a reddish patch occupied the center of my visual field at that moment will be dropped if it conflicts with other judgments having a combined stronger claim to preservation. For example, if I also judged that the patch occupying the same region an instant later was blue, and also that the apparent color was constant over the brief period covering the two instants, I am going to have to drop one of the three judgments; and circumstances may point to the first as well as to either of the others. If a statement may be withdrawn in the interest of compatibility and [sic] other statements, it is not certain in any ordinary sense; for certainty consists of immunity to such withdrawal.{42}

      Let us reformulate Goodman's problem as the following inconsistent triad:

(1) At time t1 I ostensively perceive x to be red.
(2) At time t2 I ostensively perceive x to be blue.
(3) At time t3 I judge that the color of x at t1 was the same as at t2.

      Goodman's point is that consistency requires that at least one of these judgments be dropped, and that it is possible to drop any one of them. His point is simply that despite the existence of [psychologically] indubitable sense perceptions, if these sense perceptions are not consistent with other claims, then we may regard an "indubitable" sense perception to be false. Hence the indubitable and the certain do not coincide.

(1) CRITICISM OF GOODMAN

      Goodman assumes that it is possible to reject any one claim of the triad. But is this true? I don't think so. Goodman's triad is badly constructed. First, note that all three claims are objective claims. The first two speak of how one and the same thing (the patch) appears. Report (2) is dubitable in claiming that it is the same patch that appeared in (1). Second, if these were sense data reports then they would have to be put in the form that something (unspecified) appears as a red patch. If we reconstruct Goodman's example in terms of expressive reports, it would have to be formulated as something like:

(1') At time t1 I ostensively perceive something x to be a red patch.
(2') At time t2 I ostensively perceive something y to be a blue patch.
(3') At time t3 I judge that at t1 and t2 I perceived one and the same thing, that x = y, and that its color at t1 was the same as at t2.

And now only the first two claims are expressive reports, and are the only cases of indubitable cognitions. The last claim is one which the Skeptics would call a judgment, and which could be wrong. Third, the last claim relies on memory, and if expressed as an indubitable claim would have to be put as:

(3'') At time t3 it seemed to me that at t1 and t2 I perceived one and the same thing, and that its color at t1 was the same as at t2.

      But put this way, there is no longer any inconsistency. In (3'') I am merely expressing how things seem to me now. And this is irrelevant to how things appeared before. My conclusion is that Goodman's triad relies on mixing appearance statements with objective claims. Goodman succeeds not in depicting that appearances are not certain, but only that objective judgments are not. Furthermore, Goodman has not shown how it is possible to withdraw an appearance report. He has missed a point about the nature of phenomenal reports -- a point which did not escape Sellars -- that they are logically independent of each other.

      An argument similar to Goodman's against the indubitability of phenomenal reports is offered by Bruce Aune in Knowledge, Mind, and Nature.{43} We take as a subject of an experiment a person who claims to have very vivid (eidetic) imagery. Aune proposes that we let him study the following letter-square:

e m f
r z a
o w p

We then withdraw the square from his vision, and let him consult his image. We ask him to do three things: (1) Read the image left to right, top to bottom; (2) Read the image right to left, bottom to top; (3) We ask him whether the image changed between doing (1) and (2). Suppose to these he replies, respectively: (1') e, m , f, r, z, a, o, w, p; (2') p, w, o, r, a, s, f, m, e; (3') The image did not change. There is a discrepancy between (1') and (2'). Aune concludes that one of these three reports must be false. This is correct. But it does not refute the indubitability of phenomenal reports.

      Aune's error is the same as Goodman's. Aune is mixing appearance statements with objective claims. (3') should have read: (3'') It seems to me that the image did not change. This is meant as a report of his psychological state at the time he is making it, and it is not to be construed as a report about objective facts.

      If error is to be detected, as I see it, it cannot be on the basis of comparing phenomenal reports, which, if properly formulated, are independent of each other (as Sellars recognizes). The error has to be detected, as I suggested before, by correlating the phenomenal report with a brain state. Once such a correlation is established, then the criterion of a correct report will be the relevant brain state. And the issuance of an anomalous report will then be considered criteriological evidence of a mental dysfunction.

b. MENTAL DYSFUNCTION

      A challenge not only to the certainty of phenomenal reports, but to any report whatever, is the possibility of mental dysfunction. This challenge, as far as I know, was first used in epistemological discussions by Descartes, who posits an evil genius who is bent on deceiving me in such a way that I cannot in principle discover the deception. However, O. K. Bouwsma{44} disputes that a deception of this kind is possible. To meet Bouwsma's challenge, we need to distinguish a weak and a strong sense of deception. The weak deception is the one that Bouwsma correctly analyzes to be in principle detectable. This would be the case if, for example, the evil genius tries to deceive me with a papier mache world. Such a deception on close inspection, as Bouwsma points out, is revealed for what it is.

      But there is a stronger sense of deception in which the evil genius perverts my cognitive functions in such a way that I make, among others, calculational and inferential errors. I am referring to the second-order or "metaphysical" doubts about the trustworthiness of our mental functions as, for example, in adding small integers. And this is the strong sense of deception, which amounts to mental dysfunction and perhaps insanity. Can I in principle detect this kind of deception?

      My immediate response is: Would it be coherent to speak here of a deception? It seems that if x is going to deceive y, x has to perform some action A, such that A causes y to have an erroneous belief B. Thus a deception presupposes a belief of some sort on y's part. But if the evil genius is causing a mental dysfunction in y, then it may not be proper to speak of having a belief to begin with. A deception through mental dysfunction -- if the dysfunction is severe enough -- may be an impossibility.

      This kind of skepticism based on the possibility of mental dysfunction was something new for epistemology. I am not familiar with any such possibility being discussed by Descartes' predecessors nor by any of the other classical philosophers. However Descartes' gambit enters into the speculations of current philosophers. For example, Bruce Aune, who I take to be a Sellarsian disciple, writes:

good reasons can actually be given for doubting the allegedly infallible character of immediate awareness . . . Consider, first, the verbal behavior of hebephrenic schizophrenics . . . Verbally, at least, these people are totally confused. Is there any reason to think that their thoughts are less chaotic than their words?"{45}

I have no doubt that these people suffer mental dysfunction. But the curious question is: Why did most of the great classical philosophers fail to discuss the machinations of Descartes' evil genius or its analogue of mental dysfunctions?

(1) PRESUPPOSITION OF MENTAL HEALTH

      The Cartesian problem of the possibility of mental dysfunction was never seriously entertained because, I believe, it was taken as obvious that philosophical reflection is done on the assumption that the philosopher who is speculating is not only sane but also quite rational. Even Descartes himself, while recommending his own philosophy, qualifies his remarks by saying "there are truths that can be known in every matter sufficiently to satisfy fully the curiosity of healthy minds" (italics mine).{46} A similar caution is introduced by Chisholm, who writes: "We may suppose, once again, that we are dealing with a rational person" (italics mine).{47}

      Does this have any bearing on the indubitability of phenomenal beliefs? If the question is, "Can I judge that you are wrong in your indubitable claims?" then the answer is: yes. I can always find you to be mentally insane. But the question can also be, "Can I judge myself wrong about indubitable claims?" and here the answer assumes complexity for the following reasons. What must be the case about my mental powers and my claims to knowledge if this is a meaningful question for me? The answer that has to be given is this: If I am sane, then the condition of being sane is the inerrancy of most of my beliefs. Indeed it presupposes Kant's transcendental conditions of experience.

      Descartes' cogito argument -- transformed into the Kantian perspective -- could be reinterpreted as an answer to this question of sanity. The answer would be that in order to appreciate the very question (which is a form of thinking), presupposes the possession of a coherent empirical conceptual framework. And to possess such a framework is to be sane.

      The anti-foundationalist may want to identify the empirically given (or self-evident) with the incorrigible. I suggest that to do so is a mistake. The self-evident need not be incorrigible. It is better understood as a type of proposition which, when the burden of proof is taken into consideration in a particular skeptical inquiry, is to be presumed to be true. The self-evident is to be construed as relativized to a context of evidence, in which the circumstances of the person making a claim is taken into account.

      Something like this is suggested by Sellars himself: "Having [epistemic] authority . . . is a 'defeasible' or 'forensic' matter. One knows unless there are reasons to suppose one doesn't."{48} If S says that it is self-evident to him that p, and it is indubitable for him that p, this does not entail that p is true. There are two reasons I have for saying this: the first is that a person may misuse the sentence 'p'; the second is that the person may be insane. In his correspondence with Castañeda, Sellars wrote:

let me remind you that to know (by seeing) that there is an object over there which is green and rectangular on the facing side involves background knowledge that the circumstances of perception are normal. I wish now to emphasize that the phrase 'circumstances of perception' must be construed to include not only the physical circumstances and the functioning of the sensory apparatus, but also the functioning of what might be called the 'conceptual apparatus' of the perceiver. Philosophers should pay more attention to the concept of madness, particularly its more extreme forms.{49}

Sellars' claim amount to this:

'S knows that there is a D, and that D is F' entails 'S knows that the physical, sensorial, and conceptual circumstances are normal'.

      Note that what Sellars is talking about here are cases of perception of physical objects and their properties. He is not talking about phenomenal reports. But obviously some of the considerations of talking about perceptual claims will have a bearing on phenomenal reports as well.

      Sellars claims that justification of knowledge claims requires knowing that the physical, sensorial, and conceptual conditions are normal. To know these things presupposes that such knowledge is possible. I will assume that such knowledge is possible for physical and sensorial normality, but it is not clear to me what he has in mind for conceptual normality. Sellars suggests that philosophers pay attention to the concept of madness, particularly the more extreme forms. But in view of the writings of Thomas Szasz, especially his The Myth of Mental Illness,{50} we must be cautious here. Much of what is considered mental illness may be abnormal -- though, under the circumstances, suitable -- coping mechanisms. I want to leave such cases aside.

      Let us concentrate rather on extreme forms of dysfunction of the conceptual apparatus, such as Aune had in mind, namely hebephrenic schizophrenia, or Wernicke's aphasia. These are cases in which the conceptual apparatus is chaotic. If this is what Sellars has in mind, then he may be thinking of the dysfunction of the necessary conditions for the existence of language and conceptual thought, which for Sellars are the abilities to make language entry transitions, intra-language transitions, language exit transitions, and the possession of auxiliary positions. I assume that for Sellars there are analogous distinctions for thoughts. Now any of these may be disrupted either singly or jointly. And perhaps Sellars is getting at the idea that the most important of these functions is the intra-language inferential process. If this process completely fails, then we have something like hebephrenic schizophrenia or Wernicke's aphasia. But Sellars probably would not include these as cases of the sort of conceptual dysfunction that can be detected by the person suffering them.

      So what would count as a case of a conceptual dysfunction that is detectable? He must be thinking primarily of the other linguistic and thought functions (transitions). For example, it is possible that the intra-linguistic inference and language exit patterns are relatively coherent, but the language entry apparatus becomes abnormal -- resulting in an inability to recognize, identify, or name things, the so called 'agnosia'. These vary from the inability to recognize colors, musical sounds, shapes, or faces to the inability to recognize various physical objects.{51} Related to such failings are various forms of amnesia. Also, the thought-language connection may fail as in aphasia, or the ability to carry out actions may fail (apraxia). There is a host of such separately identifiable functions that may fail. What is significant is the way such failures are connected or disconnected with each other. For example, the thought-language connection may fail at either end. In Broca's aphasia or anomia, there is language failure without significant thought failure; whereas in Wernicke's aphasia the language function is preserved to some extent but without coherent thought.

      Given this kind of clinical data, can a person know that he is conceptually dysfunctional? The answer is: yes. The obvious cases are those of failure to recognize or name objects or events, or where hallucinations occur. People suffering anomia, amnesia, or agnosia may be aware of their affliction, as may people who hallucinate. Also those who suffer Broca's aphasia may be aware of their debilities.{52} What about people suffering intra-linguistic dysfunctions? Here there seems to be room for various possible shortcomings, ranging from failure to understand, or a failure to follow an argument; to fallacious reasoning and deluded beliefs. Can a person be aware of such shortcomings? Yes, from external evidence or introspectively, provided he can have the appropriate meta-thoughts. But, as Sellars put it: "To lose the tendency to have appropriate meta-thoughts is to cease to be rational or to lose one's mind."{53} What is Sellars getting at with the requirement of meta-thoughts? Sellars thinks the following:

To think

I have the thought that it is raining

Is to presuppose that one's conceptual apparatus is working properly. And, since the proper working of one's conceptual apparatus involves . . . reacting to having MO [meta-thought] when (ceteris paribus) and only when one has O [thought], one is, if the presupposition is correct, in a position to make the translevel inference.

I am under the impression that I have the thought that-p.
So, I have the thought that-p.

{54}

      Castañeda objected that the requirement of having a meta-thought to justify a thought would lead to an infinite regress. To this Sellars replied: "I have never claimed that to have direct noninferential knowledge one must actually draw a translevel inference. I have simply said that one must be able to draw it."{55} The key point is that to have a meta-thought that one is having a though of some type is to presuppose that one's conceptual apparatus is working properly. And "To lose the tendency to have appropriate meta-thoughts is to cease to be rational or lose one's mind."{56}

      It seems that this follows as a corollary from Sellars' description of formal and material principles as occurring in a meta-language. To lose the ability to have meta-thoughts would be to lose a necessary condition for a conceptual language. It is always possible that no matter how things seem to me, I am from an external point of view insane. But in an extreme case I cannot know that I am. It may be appropriate to include in the concept of insanity either the inability to understand what insanity is or the inability to apply such a concept to oneself. The only attitude in view of this concept of insanity is to assume that one is sane. One mark of sanity, for those who can read and understand these words, is that in thinking that they understand, they are sane.

      Speaking more specifically, anyone who can even raise the question of their own sanity in an epistemological inquiry must be presumed to be sane. Chisholm, for example, as I have already mentioned, in his discussion of the justification of knowledge not only presumes sanity, but rationality: "We may suppose, once again, that we are dealing with a rational person."{57} Does Sellars disagree? Not at all: "in the absence of particular reasons for thinking that something has gone wrong, we are entitled to suppose that our conceptual machinery is functioning properly."{58}

c. TRANSCENDENTAL ARGUMENT FOR BASIC EMPIRICAL BELIEFS

      Sellars' early essays are attempts to specify the necessary conditions for empirical knowledge. He looks for guidance to Carnap's work in syntactics and semantics, which specified the formal conditions for mathematical languages. In order to include normative formal conditions for empirical languages of the natural sciences, Sellars expands the syntactical and semantical metalanguage to include a 'pragmatics'. This requires answering the neo-Kantian question: How is an empirical language possible?

      His answer is to expand on the formation and transformation rules to include what he called 'conformation rules', which he later called, following Carnap, 'P-rules', and which he later preferred to call 'material rules of inference'. Of course, the use of material rules of inference presupposes other concepts such as 'meaningful', 'verified', 'confirmed', 'true', and so on. And the normative study, Pragmatics, was intended to formulate in a metalanguage the necessary conditions for an empirical language. Suppose we spell all of this out and the result is a schema for a possible empirical language. We must remember that such a conception is arrived at through an analysis of the concept of an empirical language: it is an a priori concept of an empirical language -- which is to say that Empiricism is an a priori concept. The concept of an empirical language is governed by the a priori requirements of consistency, coherence, applicability, and adequacy (as Whitehead formulated his requirements for cosmology).{59}

      Suppose now that we have this conceptual structure and we are seeking the tokens of this type. Prima facie two answers are possible: the empirical languages are tokens of our pure pragmatics; or some experiences are themselves the tokens. Sellars' position here is to accept the first possibility and to reject the second one. But Sellars' rejection would be one of degree only. He would not deny that the 'language of experience' is analogous to a conceptual (conventional) language. But to recognize that such experiences are tokens of a type, would require the concepts of tokens and types; and these are available only through a conventional language. The upshot is that though such a pre-linguistic symbolism may approximate a language, yet it lacks the concepts to reflectively appreciate this fact. So to properly speak of 'knowledge' requires being is a full-blown language.

      The Foundationalism/Non-Foundationalism issue, as I have defined it, rests on the question whether there are basic empirical beliefs such that their acceptance as being true or false does not rest on the truth or falsehood of other empirical beliefs -- general or particular.

      Sellars' ultimate approach to this problem is from a neo-Kantian perspective, which relies on a transcendental argument of how a Pure Pragmatics is possible. Sellars, like Kant, is looking for the conditions for the possibility of empirical knowledge. His reasoning is that the possibility of empirical knowledge requires the possession of a language.{60} And to possess a language requires the possession of propositions which play roles in the language entry-inference-departure roles. In this structure, the language-entry propositions (observation reports) are generically foundational in the sense that inferences and intentions (language-departure propositions) are ultimately concerned with observational propositions. With this in mind, he wrote:

There is clearly some point to the picture of human knowledge as resting on a level of propositions -- observation reports -- which do not rest on other propositions in the same way as other propositions rest on them.

But went on to add,

On the other hand, I do wish to insist that the metaphor of 'foundations' is misleading in that it keeps us from seeing that if there is a logical dimension in which other empirical propositions rest on observation reports, there is another logical dimension in which the latter rest on the former.{61}

And what is this "another logical dimension"? It is simply the idea that isolated observation reports are not sufficient to provide either concepts, or language, or empirical knowledge. As he put it, "We have to be in this framework [of entry-inference-departure transitions] to be thinking and perceiving beings at all," and then he adds "and acting beings."{62}

      From the perspective of an explication of the concept of empirical knowledge, it is an analytic truth that there must be observation reports and that these must be conceptually connected to various other statements. To have the concept of an observational report entails the possession of a conceptual framework in which the observational report fits via inferential connections.

      Can this transcendental reasoning meet Bonjour's strong requirement (JBB-i)? In formulating the foundationalist position as (BEB), we left open the following possibility:

(iv) p is justified by nonempirical beliefs.

      And an a priori argument could supply such a nonempirical belief. One commentator, Robert Meyers, understood that an a priori argument is not ruled out by Sellars as a possible type of argument for the justification of basic empirical beliefs:

Sellars does not say so explicitly, but it is clear that the only sort of presupposed knowledge the foundationalist rules out is empirical knowledge; a priori presuppositions would be perfectly acceptable.{63}

      In considering the possibility of an a priori argument, Bonjour simply writes, "it is hard to see how a particular empirical belief could be justified on a purely a priori basis."{64} He is correct in his skepticism about "particular" empirical beliefs, but such a skepticism should not be extended to "kinds" of empirical beliefs. He goes on to attribute an a priori justification of basic empirical beliefs to Lewis, Chisholm and Firth. Specifically, he attributes to them the belief that

basic or foundational beliefs are warranted by virtue of possessing the conjunctive property [F] of (i) being confined in their subject matter to a purported description of my present experience, and (ii) being accepted by me now.{65}

      Bonjour does not see how this is an a priori argument since this formulation implies two empirical claims. The first is that I do, in fact, have the belief B; the second that this belief B has, in fact, the property F. His obvious rejoinder to such empirical implications should be to demand justification for these two claims. But he draws back arguing,

If the original issue is whether I am justified in holding a certain belief, then the raising of that issue seems to presuppose that I do in fact hold it, so that it becomes inappropriate to demand justification for this latter claim.{66}

The result is that Bonjour is conceding a reflective knowledge of psychological states, such as knowing that I am in the state of believing that p. In fact he concedes that most of my beliefs about my psychological states must be presumed to be true. He calls this the 'doxastic presumption'.

      In any case, he has not conceded that the empirical contents of beliefs have to be presumed to be generally true as well. In other words, he concedes that I can presume to know that I believe that p, but I cannot presume that p is true. And this is the sort of empirical knowledge which matters. Are there, then, a priori justifications for these? One could propose the following seemingly a priori argument to satisfy Bonjour's demands. Let 'B' stand for some belief about the property of a physical object, such as 'I see that this physical object is red'. 'B*' stands for 'This is an observation report'. Substitute for (1) of (JBB) the claim:

(1') Belief B is a token of type .B..

and then change (2) to

(2') Beliefs which are tokens of type B* must be highly likely to be true (if there is to be empirical knowledge).

      The conclusion is:

(3') Therefore, B must be highly likely to be true (if there is to be empirical knowledge).

      Premise (1'), I believe, must be granted: it is an instance of Bonjour's doxastic presumption. It must be assumed that people are able to apply concepts to experience, if there is to be empirical knowledge.

      Premise (2'), however, presupposes that .B. can be true. How 17 is this possible? On Sellars' reasoning, this will be the case if the tokens of .B. stand in a picturing relation to physical objects. Sellars puts this as: X is a reliable symptom of Y.{67} Thus if a person is to know that, for example, 'This is red', he must also know that observing that 'This is red' is a reliable symptom of the fact that indeed this is red under normal circumstances. In short, he must have knowledge of the following necessary truth that

x is red. iff .x would look red to standard observers in standard conditions

      It seems that to be justified in claiming (2'), one needs not only general knowledge of this sort, but also the particular knowledge that the person himself and the conditions are indeed normal. This is to say that a person has observational knowledge if he is capable of such a trans-level inference. Bonjour takes this as the coup de grace for foundationalism:

the most fundamental and far-reaching objection to foundationalism -- namely that there is no way for an empirical belief to have any degree of warrant which does not depend on the justification of other empirical beliefs.{68}

      To this line of reasoning, Castañeda pointed out to Sellars that this would involve an infinite regress. For example, to observe that this is red, would require the ability to know that the illumination was, for example, from sunlight. But to observe that the sun was shining would require other general and particular knowledge, and so on to infinity.{69}

      Sellars answered that for a person to have observational knowledge he must be only capable of providing such additional knowledge, and not expressing it at the time of the observation.{70} Thus if Jones' claim to see that this is red is to count as knowledge, he need not at the time of the claim have the thoughts occur to him expressing general and particular knowledge, he simply must be capable of formulating such knowledge as justification. So the regress does not appear to be vicious.

      Since making observation reports clearly implies having other empirical knowledge, the attempt to give an a priori justification fails for particular empirical claims. However, the a priori argument, according to Sellars, justifies the foundational status of observational reports of one kind or another. This is to say that the requirement of observational reports does not give to any particular set of observational reports an incorrigible or immutable status. To believe otherwise is to embrace, according to Sellars, a form of the Myth of the Given. Here is how Sellars put it:

To reject the myth of the given is to not commit oneself to the idea that empirical knowledge as it is now constituted has no rock bottom level of observation predicates proper. It is to commit oneself rather to the idea where even if does have a rock bottom level, it is still in principle replaceable by another conceptual framework in which these predicates do not, strictly speaking, occur. It is in this sense, and in this sense only, that I have rejected the dogma of givenness with respect to observational predicates.{71}

      I believe Cornman got it right when he wrote:

Sellars is opposed to observation statements being given and thus unalterably at the foundation . . . But none of this requires that he reject foundationalism. Indeed, his rejection of "the" given is not a rejection of foundationalism.{72}

      And indeed it is a Sellarsian faith that our present observation reports can be replaced by theoretical reports of an ideal science.

He [the philosopher] must also envisage the world as pictured from the point of view -- one hesitates to call it Completed Science -- which is the regulative ideal of the scientific enterprise. As I see it, then, substantive correspondence rules are anticipations of definitions which it would inappropriate to implement in developing science, but the implementation of which in an ideal state of scientific knowledge would be the achieving of a unified vision of the world in which the methodologically important dualism of observation and theoretical frameworks would be transcended, and the world of theory and the world of observation would be one.{73}

      I challenge this thesis in my chapter on Animal Realism.

      But since it can be presumed that our empirical knowledge does rest on observation reports, cannot we also presume that most of these reports are reliable (if we are to have empirical knowledge)? And if this is the case, then it also seems reasonable to hold each observation report to be reliable unless there is reason not to do so. Chisholm, for one, believes such an approach is reasonable:

This is the principle that anything we find ourselves believing may be said to have some presumption in its favor -- provided it is not explicitly contradicted by the set of other things that we believe. The principle may be thought of as an instance of a more general truth -- that it is reasonable to put our trust in our own cognitive faculties unless we have some positive ground for questioning them.{74}

      This seems to be acceptable to Sellars on transcendental grounds, and constitutes, I believe, a very weak foundationalism -- but foundationalism, nonetheless.

(1) PHENOMENAL REPORTS

      But notice that Sellars' justification of empirical knowledge by a trans-level inference applied only to observation reports as a class about physical objects. What about phenomenal reports?

      I propose the following a priori argument to satisfy Bonjour's demands. Let 'B' stand for some phenomenal belief about how one is appeared to, such as, 'This looks red' ('this' is used demonstratively but indefinitely as to category; 'red' is used in a non-comparative, phenomenal sense). Substitute for (1) of (JBB) the claim:

(1*) Belief B is a token of type B*.
(2*) Beliefs which are tokens of type B* must be highly likely to be true (if there is to be empirical knowledge).
(3*) Therefore, B must be highly likely to be true (if there is to be empirical knowledge).

      Suppose we grant the existence of phenomenal reports (call them 'intuitions'), is there any further difficulty? Bonjour presents the following dilemma. The intuitions are either cognitive or not. If they are not cognitive, they cannot give nor receive epistemic justification. If they are cognitive, then they can give and receive epistemic justification. "In either case, such states will be incapable of serving as an adequate foundation for knowledge. This at bottom, is why the empirical givenness is a myth."{75}

      We can answer Bonjour's dilemma by taking it by the horns. Bonjour is conflating two notions of 'epistemic justification'. There is empirical and there is a priori justification. My present contention is that there is an a priori justification, and that this is compatible with foundationalism.

      If phenomenal reports were available for the justification of observational knowledge, then they would fall into the class of introspective reports, such as reports of thoughts and after-images. And what kind of justification does Sellars require for introspective reports? I believe he would accept the above transcendental argument with the following qualification:

In the absence of facts about its occurrence which would make it non-authoritative (and these are of a limited and -- in principle -- ascertainable variety) it is authoritative. Having authority . . . is a 'defeasible' or 'forensic' matter. One knows unless there are reasons to suppose one doesn't.{76}

      However introspective knowledge, for Sellars, is conceptually secondary, and presupposes knowledge of physical objects. But this is not an objection, since conceptual dependence is compatible with epistemic independence. So Sellars concedes that introspective knowledge has to be regarded as incorrigible:

As in the case of our direct knowledge of our sensations and feelings, the incorrigibility [about thoughts] is in large part accounted for by the 'minimality' of the cognitive claim that is being made. There are no conceptual ways to go wrong, unlike the case of "There is an inkstand on 19 the table".{77}

      From this it is obvious that Sellars would consider that observational knowledge would have specific rather than generic foundations if it entailed phenomenal reports. This would be the case because phenomenal reports would be logically atomic -- not requiring any other empirical knowledge for their justification, and they would be justified by the above transcendental argument. But Sellars denies that observation reports entail phenomenal reports; so this justification is otiose for Sellars.

      A philosopher who does use phenomenal reports as foundations of empirical knowledge is R. Chisholm. Phenomenal reports, for him, describe phenomenal properties, which are the proper and common sensibles of Aristotle. Let us refer to a phenomenal property by the letter 'F', then Chisholm's foundationalism can be formulated in the following way:

(1) I know that this is F because it appears F.
(2) It is directly evident that it appears F.
(3) I know that it is evident that it appears F because it does appear F.{78}

Sellars never confronts Chisholm directly on this kind of epistemic justification. [He does in The Metaphysics of Epistemology (AC 1996)] The most extensive coverage is given in his "Structure of Knowledge," but here he simply points out that Chisholm talks about introspective reports as being 'directly evident' and goes on to discuss the issue of the 'indirectly evident'. Sellars must take this evasive tack because he denies Chisholm's presupposition (1) that observational reports entail phenomenal reports.

      But step (3) in Chisholm's reasoning may be amenable to a Sellarsian treatment. In his Perceiving, Chisholm's considers how a phenomenal appearance statement could be justified, and answers: "But if he wished to describe his own justification for asserting the appear statement, he could do little more than to repeat his statement."{79} He calls this "defense by repetition" of phenomenal reports.{80} In his Theory of Knowledge, Chisholm gives basically the same answer:

To the question "What justification do I have for thinking I know, of for counting it as evident, that something now looks red to me, or tastes sour?" I could reply only by reiterating that something does now look red or taste sour.{81}

      I have always wondered why Chisholm resorts to this argument, and despite such theological language as that appear statements are 'unmoved movers' or 'self-moving prime movers',{82} (3) seems to be an expression of a tautology. Bonjour seems to agree with my assessment when he comments, "that extremely paradoxical remark hardly constitutes an explanation."{83} I think one can do better than Chisholm in justifying phenomenal reports. There are at least two better ways to defend phenomenal reports. The first way is Sellars' transcendental defense of introspective reports along the lines I have already given. The second is the argument to the best explanation, which would run as follows. The explanation for why something appears F is simply that something does appear F. This looks exactly like Chisholm's formulation, but the meaning is different. Chisholm uses the formula as a reiteration of the original claim. My formula is not simply a reiteration but an appeal to the hypothesis that how things phenomenally appear is due to the existence of reified appearances or sense data. That Chisholm does not attach this meaning to his reiteration is obvious from his rejection of the sense-datum inference which my approach entails.{84}

d. CONCEPTUAL VS. NON-CONCEPTUAL REPRESENTATION

      I believe that there is a way to bridge the gap between Sellars and Lewis-Firth-Chisholm about premise (1) of Chisholm's argument by using Sellars' distinction between Language, or L, and a Representational System, or RS. Suppose we identify Lewis' 'sense meanings',{85} Firth's ur-concepts,{86} Chisholm's 'phenomenal reports',{87} Everett Hall's 'natural language of perception',{88} and Romane Clark's 'sensuous judgments'{89} with an RS. All of these may, in a refined way, have the following theses in common with an RS: (i) propositional form, (ii) Humean inferences, and (iii) no logical connectives. I say in 'a refined way' because I am not clear that each of these authors sees the necessity of (iii) in their analogues of an RS. Clark is clear when he writes: "I do not think of the logical connectives as part of perceptual judgments. Certainly not on the primitive level of basic perceptions."{90} Lewis's concept of 'sense meaning' is hard to pin down. But a few things are clear. Sense meaning is prior to and independent of language,{91} even animals possess it,{92} and it requires anticipatory imagery.{93} These remarks suggest at least the acceptance of (ii). Lewis also suggests that sense meaning may be connected with Kant's 'schema'.{94} At this point there is a connection between Lewis and Sellars. In "The Role of Imagination in Kant's Theory of Experience," Sellars introduces the notion of an 'image-structure', 'sense image', or 'image-model' which, according to Sellars, is the work of Kant's productive imagination. There is according to Sellars as 'image-model world'{95} prior to and independent of logical connectives (categories). A 'schema' for Kant, according to Sellars, bridges concepts and such image-models. A schema is the construction of an image-model by the productive imagination in accordance with a concept.{96} In saying this, Sellars may be in disagreement with Lewis over the interpretation of Kant. In any case, I think that both Sellars and Lewis agree that a schema is a recipe for construction of an image-model. Their disagreement, I suspect, is over whether there is a requirement for conceptual guidance. But that doesn't matter here, it suffices that Sellars acknowledges pre-conceptual image-models, which, as far as I can see, are the same thing as RSs.

e. RECONCILIATION?

      Now my attempt to reconcile Sellars with these other philosophers would take the form of reinterpreting Chisholm's

(1) I know that this is F because it appears F.

where 'it appears F' can be understood as 'I am rs-aware of an image-model rs-F'.{97} And if we generalize, we can formulate the thesis that a perception of a physical object entails an rs-perception of an rs-(physical object), which in turn entails an rs-awareness of an rs-appearance.

      Would this be acceptable to Sellars? The answer is: no. His reason is this:

although the objects of which we are directly aware in perceptual consciousness are image-models, we are not aware of them as image-models. It is by phenomenological reflection (aided by what Quine calls scientific lore) that we arrive at this theoretical interpretation of perceptual consciousness.{98}

      Sellars apparently accepts an RS (image-model world) as a theory which explains our perceptual knowledge, but not as an analysis of it. Sellars is introducing two difficulties for a reconciliation. The first difficulty is that although in point of theoretical fact we do experience image-models, we do not experience them in propria persona{99} as image-models. The point of denying the propria persona principle is that physical objects or events are conceptually constituted by various logical and causal relations, and these are never given by any sense modality, but must be learned. Sellars' rejection of this principle is acceptable.

      The second difficulty for Sellars is that observational reports are not analyzable into, i.e., do not entail, phenomenal beliefs; rather, the image-world is accessible to phenomenological reflection. This is also acceptable. But Sellars seems to muddy the waters in this context by talking in the same breath about perceptual reflection and theory construction. When Sellars talks about theories he means postulational theory. I believe there is a difference between phenomenological reflection and theory construction. Sellars himself acknowledges this distinction when he writes:

sheer phenomenology or conceptual analysis takes us part of the way, but finally lets us down. How far does it take us? Only to the point of assuring us that Something, somehow a cube of pink in physical space is present in the perception other than as merely believed in.{100}

      When Sellars is doing phenomenological reduction, he is not theorizing. The trouble with the above reduction, however, is that it is not reductive enough. This was the point made by Firth in his reply to Sellars' Carus Lectures.{101}

      If we made the reduction go further we would get something like Chisholm's analysis. To express the fact that we are dealing with a phenomenological reduction, let's modify Chisholm's formula

(1) I see that this is F because it appears F.

by expanding it to read

(1') I see that this is F because I can perceptually reduce this to it appears F.

      We can then agree with Sellars that (1) does not express an analysis; rather it expresses a phenomenological or perceptual reduction.{102} And if physical objects do entail (perceptually reduced) phenomenal reports, then empirical knowledge does have foundations.

---

[Go to Chapter 8]

---