Central European analytic philosophy in the first half of this century focused on the relation between science and experience. Controversies extended beyond the philosophy of science, and shook philosophy to its foundations, affecting every part of the discipline: from aesthetics and philosophy of religion to metaphysics and epistemology. The issue of how experience relates to thought and language had two components: First, how does it relate to meaning? Second, how does it relate to knowledge? Controversy raged over the empiricist criterion of meaningfulness, which set the positivists against all others but brought them together as their emblem. On the second issue, by contrast, they were divided. What is the epistemological bearing of experience on scientific knowledge? On this question Otto Neurath and Moritz Schlick had a famous controversy. According to Neurath, our ship of knowledge must be rebuilt at sea, and only its coherence really matters. This Schlick rejected, however, as "an astounding error." Schlick saw coherentism as adrift, and insisted on experiential moorings.

This controversy, though hardly novel with the Vienna Circle, much exercised them, and soon involved Rudolf Carnap and Carl Hempel and, eventually, Hans Reichenbach. With these it crossed the Atlantic, and with Karl Popper the Channel. In the second half of the century it has been a central issue in epistemology, where it has attracted Quine, Sellars, Chisholm, Davidson, Rescher, Rorty, and many others.

It is this second issue that I would like to take up: What is the epistemological bearing of sensory experience on our knowledge?

The idea of epistemically foundational status appears already in Carnap's The Unity of Science (1932), where a "protocol" statement is defined as a "direct record of a scientist's experience." A "primitive" protocol is then said to be one that excludes "... all statements obtained indirectly by induction or otherwise."

In the middle decades of our century, the doctrine of foundationalism took a radical form, a throwback to Cartesian epistemology. Schlick, for example, required indubitable and incorrigible foundations. Soon thereafter , C.I. Lewis also adopted that requirement, thus recanting his earlier "conceptualistic pragmatism."

Many others also adopted a radical foundationalism of certainty. "This doctrine," says Nicholas Rescher, "insists on the ultimate primacy of absolutely certain, indefeasible, crystalline truths, totally beyond any possibility of invalidation." But Rescher himself rejects the quest for absolute foundations as "quixotic."²

Eventually a more moderate foundationalism is suggested by Hempel, who writes: "When an experiential sentence is accepted 'on the basis of experiential evidence', it is indeed not accepted arbitrarily; but to describe the evidence in question would simply mean to repeat the experiential statement itself. Hence, in the context of cognitive justification, the sentence functions in the manner of a primitive sentence."³

If put in terms of beliefs rather than sentences, Hempel's point appears thus:

(H) Beliefs held on the basis of direct experiential evidence are not arbitrary. Yet to

state the evidence for such a belief is just to voice the belief. Hence, in the context of cognitive justification, these beliefs function as primitive or basic.

Among the foundations defended by Chisholm, early and late, are sensory foundations-or apprehensions of the given-as well as knowledge of one's own beliefs and other propositional attitudes, which are also said to satisfy Hempel's conditions for being primitive or basic.

Such foundationalism was attacked famously by Wilfrid Sellars, another philosopher influenced by German philosophy, from Kant and Hegel through Carnap and logical positivism. Two main issues divided Sellars from Chisholm and fueled their long and widely followed controversy. Their disagreement involved, first, the relation of thought, or intentionality, to language, and, second, the relation of experience to empirical knowledge. On this second issue they thus continued the controversy of their predecessors from Central Europe: Neurath, Schlick, Carnap, and Hempel.

Sellars seems right in saying that Hempel's thesis points to inferences of the form:⁵

(a) It is a fact that a is F;

So, it is reasonable (for me) to believe that a is F.

But his objection is now that any such argument as a will do its job only if the premise has authority, only if it is something which it is reasonable to believe. According to Sellars, this leads from a to the following alternative argument schema:

(b) It is reasonable to believe it to be a fact that a is F; So, it is reasonable to believe that a is F.

And this is of course, just as Sellars says, "obviously unilluminating."

That, again, is Sellars's critique. But the sort of problem he raises is not unique to his critique. A main theme of Richard Rorty's attack on foundationalism is the alleged "confusion of causation with justification" that he attributes to Locke and others. Donald Davidson also adds his voice:

As Rorty has put it, "nothing counts as justification unless by reference to what we already accept, and there is no way to get outside our beliefs and our language so as to find some test other than coherence." About this I am, as you see, in agreement with Rorty.⁶

Hempel's insight hence appears to survive the sort of objection urged by Sellars, Rorty, and Davidson. But does it yield an acceptable foundationalism, as Chisholm believes? More specifically, does it adequately explain how it is that experience bears on knowledge? Shall we say that experientially given facts justify beliefs directly, merely through their truth? Shall we say, for example, that the mere fact that I have a headache suffices to justify my belief that I do?

What is the alternative to such foundationalism concerning how one is justified in believing that p? Circular or regressive reasoning will not adequately explain how that belief is justified. How could the belief that p derive its justification entirely from such reasoning? There is much to be said about this, but little time to say it here. So let me just gesture towards the problems involved by asking how, in either a pure regress or a pure circle, justification ever enters in the first place. Inferential reasoning serves to transfer justification but this presupposes that justification is already there in the premises of the reasoning, and that is precisely what neither the pure circle nor the pure regress is able to explain.⁷ A full explanation of how one's belief that p gets to be justified must apparently take us back to ultimate premises that do not get all their justification from further premises yet. Must there not be ultimate premises that somehow get some of their justification by means other than reasoning from further premises? If not, it is hard to see how justification ever appropriately enters the line of reasoning (regressive or circular) that leads to one's being eventually justified in believing that p.

That sketches an argument in favor of the appeal to foundations. But note well the highly determinable character of these foundations. All we have a right to suppose about our foundational beliefs, on the basis of our elimination of the circle and the regress, is that they are justified noninferentially. Any more determinate and positive thesis would require further defense. In particular, we are in no position to conclude that the foundation must be constituted by direct apprehensions of the experiential given.

We have been considering how sensory experience bears on our empirical knowledge and justification. And we have found two opposing positions. On one side are Neurath, Sellars, Rorty, and Davidson, among others. According to this side, experience bears causally on our beliefs, but it is a serious mistake to confuse such causation with justification. Experience bears at most causally on our beliefs about external reality, even on our simplest perceptual beliefs. We do not infer such perceptual beliefs from beliefs about our sensory experience, nor is the justification for such beliefs a matter of their coherence with appropriate beliefs about our experience.

On the other side are Schlick, Hempel, C.I. Lewis, and Chisholm, among others. For these it is an "astounding error" to suppose that the mere coherence of a self-enclosed body of beliefs might suffice to confer justification on its members. And it is hard to see what, other than sensory experience, could serve to supplement coherence appropriately so as to explain empirical justification. Accordingly, they prefer rather to postulate beliefs about such experience, the takings or apprehensions of the given, through inference from which, or by coherence with which, one must attain one's empirical justification. But this side notoriously fails to find foundations contentful enough to found our rich knowledge of an external world.

As so often in philosophy, this controversy leaves middle ground untouched. Our coherentists and foundationalists share an assumption:

(A) Experience can bear epistemically on the justification of belief only by presenting itself

to the believer in such a way that the believer directly and noninferentially believes it to be present, and can then use this belief as a premise from which to reach conclusions about the world beyond experience.

But does that serve to provide foundational justification for perceptual beliefs? Take a perceptual belief prompted appropriately by a corresponding experience. Take a belief that this is white and round, one prompted by visual experience of a sunlit snowball in plain view. Is that perceptual belief foundationally justified simply in virtue of its causal aetiology? When Sellars inveighs against the myth of the given, he targets not only the radical version of the myth involving direct apprehensions of given experience. He objects also to the more moderate version that postulates foundational knowledge through perception. Indeed the key passage that encapsulates his opposition to a foundational epistemology targets not a foundation of introspective direct apprehension but a foundation of perception.

Here I avoid issues about the nature of thought and its relation to language and society. So I will take the liberty of transmuting Sellars's argument into one pertaining directly to belief, justification, and knowledge, leaving aside whether to understand these in terms of moves in a language game governed by social rules. I am not denying that our main epistemic concepts are to be understood thus in terms of language and society. I am simply not joining Sellars in affirming it. Thus my preference for the transmuted argument that does not prejudge these issues. So transmuted, here then is the Sellarsian refutation of the epistemology of foundations:

We have seen that to constitute knowledge, an observational belief must not only have a certain epistemic status; this epistemic status must in some sense be recognized by the person whose belief it is. And this is a steep hurdle indeed. For if the positive epistemic status of the observational belief that this is green lies in the fact that the existence of green items appropriately related to the perceiver can be inferred from the occurrence of such observational beliefs, it follows that only a person who is able to draw this inference, and therefore has not only the concept green, but also the concept of an observational belief that this is green -- indeed the concept of certain conditions of perception, those which would correctly be called 'standard conditions' -- could be in a position to believe observationally that this is green in recognition of its epistemic status.⁹

All the view I am defending requires is that no belief by S now that this is green is to count as observational knowledge unless it is also correct to say of S that he now knows the appropriate fact of the form X is a reliable symptom of Y, namely that the observational belief that this is green is a reliable indicator of the presence of green objects in standard conditions of perception. And while the correctness of this statement about Jones requires that Jones could now cite prior particular facts as evidence for the idea that such belief is a reliable indicator, it requires only that it is correct to say that Jones now knows, thus remembers, that these particular facts did obtain. It does not require that it be correct to say that at the time these facts did obtain he then knew them to obtain. And the regress disappears.¹⁰

And there is a further problem. Our later access to earlier observational reactions is an exercise of memory. But memory itself seems to require, no less than perception, some meta-awareness of its reliability when exercised in circumstances of the sort in which it is now exercised.¹¹ And if there was a problem of regress attaching to the exercise of perception there would seem to be an equally disturbing problem of regress attaching to the exercise of memory. Perhaps the response would be that just as earlier proto- perceptions can become data supportive of generalizations about our perceptual reliability, generalizations that underlie later perceptual knowledge; so, similarly, earlier proto-memories can become data supportive of generalizations about our memorial reliability, generalizations that underlie later memorial knowledge. Perhaps, but this raises even more poignantly an objection akin to that raised earlier about perceptual knowledge: namely, that we cannot plausibly be said to remember particular earlier exercises of memory constitutive of a data bank which can later support our underwriting generalizations.

For a better solution we must go back, ironically, to a philosopher long miscast as the archetypal foundationalist and givenist. It is, I suggest, in Cartesian epistemology that we find a way beyond our regress or circle.

In the barest sketch, here is how I see Descartes's epistemological project. First he meditates along, with the kind of epistemic justification and even "certainty" that might be found in an atheist mathematician's reasonings, one deprived of a world view within which the universe may be seen as epistemically propitious. Descartes's reasoning at that stage can be evaluated, of course, just as can an atheist mathematician's reasoning. After all, atheist mathematicians will differ in the worth of their mathematical reasonings. Absent an appropriate world view, however, no such reasoning can rise above the level of cognitio. If we persist in such reasoning, nevertheless, eventually enough pieces may come together into a view of ourselves and our place in the universe that is sufficiently comprehensive and coherent to raise us above the level of mere cognitio and into the realm of higher, reflective, enlightened knowledge, or scientia. There is in none of that any circle that vitiates the project.¹²

Ancient skepticism, as represented by Pyrrhonism, and modern skepticism, as presented by Descartes, have been regarded as radically different. How plausibly? Descartes does raise a certain skeptical problem that is limited by comparison with the radical skepticism of the ancients: namely, the problem of the external world. But this is by no means the only skepticism of interest to Descartes. It is obvious in the Meditations that his concerns are much broader, as when he wonders how he can know the truth even when he adds three and two or when he considers how many are the sides of a square. It is precisely the radical skepticism of the ancients that mainly concerns Descartes (and not only Hegel, who is emphatic on the point). Moreover, this skepticism is best seen in the light of the epistemic problematic found already in Aristotle's Posterior Analytics,¹³ where it is given a foundationalist resolution, and, more famously, in the five modes of Agrippa. To the latter incarnation of that problematic the Stoics, in kinship with Aristotle, offer a foundationalist response. Where Aristotle appeals to rational intuition as a way to found scientific knowledge, the Stoics appeal to natural, animal perception as a way to found ordinary and empirical knowledge.

The Pyrrhonists find such externalism unacceptable because it dignifies mere "groping in the dark" with the title of knowledge.¹⁴ The Pyrrhonists highlight enlightened knowledge, which must be acquired and sustained in awareness of one's epistemic doings. Only this is "knowledge" worthy of the title. Sadly, they would prefer in their own practice to suspend judgment in specific case after specific case, with few if any exceptions, partly because they reject blind foundations. In their view, moreover, any attempt to move beyond foundations only misleads us into circles or regresses, viciously either way.

Descartes's response is balanced and sensitive to this Pyrrhonian dialectic. It grants the truth in foundationalism by allowing room for an inference-independent epistemic state of cognitio. Intuition gives us foundational cognitio, as suggested by Aristotle, and such unreflective knowledge is open even to the atheist mathematician. There is however a higher state of knowledge, reflective knowledge. Attaining such knowledge requires a view of ourselves -- of our beliefs, our faculties, and our situation -- in the light of which we can see the sources of our beliefs as reliable enough (and indeed as perfectly reliable if the scientia desired is absolute and perfect).

It is important to recognize, in assessing this Cartesian strategy, that while we do need to underwrite, at the later stage, the reliability of our faculties, what enables us to do so is the appropriate use of those very faculties in yielding a perspective from which reality may be seen as epistemically propitious. But we need not restrict ourselves to the use of rational intuition and deduction as the only faculties of any use in that endeavor. Descartes himself surely needed memory as well. And memory, by definition, operates over time. It is not a present-time-slice faculty. Nor, indeed, is deduction itself such a faculty, except where the whole proof can be seen in a flash. So memory, as a cognitio-level mechanism can join cognitio-level intuition and perception in yielding the pieces that, once present with sufficient comprehensiveness and coherence, can boost us to the level of reflective scientia able to underwrite all such faculties. This means that we need not later exhume from memory any particular cases of reliable perception or reliable memory in order to support inductively the generalizations about the reliability of our faculties. It is enough that such generalizations be present because of the combined operation of past perception and memory (and, perhaps, a gradual "induction" over time, and/or appropriate innate principles). If through such cognitio-level cognitive processing enough of a coherent and comprehensive picture comes together, such a picture can still underwrite the continued use of those very faculties, now with reflective assurance, and now at the level of enlightened scientia.¹⁵

We have gone beyond the mythology of the given, first by rejecting the assumption that experience can bear on the epistemic justification of our beliefs only by providing premises yielding knowledge of a world external to experience. Here is a better way to think of the epistemic efficacy of experience. Visual experience as if this is white and round may cause belief that this is white and round in the absence of any special reason for caution. That can yield perceptual knowledge that this is white and round, with no need to postulate any inference from one's experience to what lies beyond. Maybe there are such inferences, lightning inferences unconsciously or subconsciously yielding our perceptual beliefs as conclusions. But we need not enter that issue. It is enough that experience cause belief in some appropriate, standard way. Whether it does so via a lightning, unconscious inference we can leave open. Whether it does so or not, it may still endow the perceptual belief with appropriate epistemic status to constitute perceptual knowledge.

Nevertheless, a mere thermometer reaction to one's environment cannot constitute real knowledge, regardless of whether that reaction is causally mediated by experience. It is not enough that one respond to seeing white and round objects in good light with a "belief" or "proto-belief" that there is something white and round. Suppose one asks oneself "Do I know that this is white and round?" or "Am I justified in taking this to be white and round?" and has to answer "Definitely not" or even "Who knows? Maybe I do know, maybe I don't; maybe I'm justified, maybe I'm not." In that case one automatically falls short, one has attained only some lesser epistemic status, and not any "real, or enlightened, or reflective" knowledge. The latter requires some awareness of the status of one's belief, some ability to answer that one does know or that one is epistemically justified, and some ability to defend this through the reliability of one's relevant faculties when used in the relevant circumstances. But this leads to a threat of circle or regress, a main problematic, perhaps the main problematic of epistemology. Surprisingly, already in Descartes himself, in the founder of modern epistemology, we find a way beyond that problematic.¹⁶

¹Paper for a conference sponsored by the University of Pittsburgh's Center for Philosophy of Science: "German-American Interaction in Scientific Philosophy After 1933," and held in March of 1996.

²Methodological Pragmatism (New York: NYU Press, 1977), p. 210. Rescher proposes to replace such foundations with "presumptions" accepted prima facie and subject to refutation. And he goes on to distinguish the rationality of practice from the rationality of theory, and to argue that the former permits a "...presumption in favor of established methods...[that] tilts the burden of proof in this context against a sceptical opponent" (p. 233). Rescher has developed this idea into an important version of pragmatism, but one that differs in kind from those I take up here.

³"Some Theses on Empirical Certainty," Review of Metaphysics 5 (1952); 621-29; p. 621. Carnap had spoken already in 1936 of the "confrontation of a statement with observation," and had proposed "acceptance rules" for such confrontation: "If no foreign language or introduction of new terms is involved, the rules are trivial. For example: 'If one is hungry, the statement "I am hungry" may be accepted'...." (From "Truth and Confirmation," in Readings in Philosophical Analysis, ed. Herbert Feigl and W.S. Sellars (Appleton, 1949), p. 125. These claims appeared first in "Warheit und BewŠhrung," Actes du congres internationale de philosophie scientifique, Vol. 4 (Paris, 1936), pp. 18-23

⁴In "The Theory of Knowledge," in Philosophy: The Princeton Studies: Humanistic Scholarship in America (Garden City, NJ: Prentice-Hall, Inc., 1964), Chisholm cites Hempel's work, and also an earlier paper by C.J. Ducasse, "Propositions, Truth, and the Ultimate Criterion of Truth," Philosophy and Phenomenological Research 1 (1939): 317-40.

⁵See Sellars's "Epistemic Principles," in his lecture series, "The Structure of Knowledge," in H.N. Castaneda, ed., Action, Knowledge, and Reality (Indianapolis: Bobbs Merrill, 1975), pp. 337-8.

⁶"A Coherence Theory of Truth and Knowledge," in Kant oder Hegel?, ed. Dieter Henrich (Stuttgart: Klett-Cotta, 1983), reprinted in Ernest LePore, Truth and Interpretation (Oxford: Blackwell, 1986), pp. 307-20; p. 310.

⁷Again, I do not pretend that this is an adequate analysis of our issue (which I myself regard as quite complex); but here it will have to serve.

⁸Actually this sort of approach is defended already by Thomas Reid.

⁹"Empiricism and the Philosophy of Mind," Minnesota Studies in the Philosophy of Science, Vol. I, ed. H. Feigl and M. Scriven (Minneapolis: University of Minnesota Press, 1956), pp. 253-329. Reprinted in R.M. Chisholm and R.J. Swartz, Empirical Knowledge (Englewood Cliffs: Prentice-Hall, 1973), pp. 471-541; p. 512.

¹⁰Ibid., p. 513. In the version of Science, Perception, and Reality published by Ridgview Press in 1963, Sellars added the following footnote, on p. 169: ÒMy thought was that one can have direct (non-inferential) knowledge of a past fact which one did not or even (as in the case envisaged) could not conceptualize at the time it was present.Ó

¹¹This requirement is defended in my Knowledge in Perspective (Cambridge University Press, 1991), pp. 280ff.

¹²Among the pieces that need to come together in order to raise the belief that p above the level of cognitio, to the level of scientia, may well be found appropriate cognitio that one enjoys cognitio that p. In discussion at the Pittsburgh conference it was objected that comprehensiveness and coherence are matters of degree while it is very hard to see how to draw a line above which lie the degrees of comprehensiveness and coherence that suffice for knowledge, though it was also suggested that we might do better by appealing to practical considerations and not just to comprehensiveness and coherence. However: (a) it is not clear how appeal to practical considerations will really help with the problem of drawing a line. Moreover: (b) compare a concept like that of being tall. That is presumably to be defined in some such way as this: being sufficiently taller than the average. Presumably someone just infinitesimally taller than the average is not tall. One has to be taller than the average by some margin, one has to be "sufficiently" taller than the average. But how do we define that margin? Is there, even in principle, some way to capture our actual concept of tallness by means of some such definition? There seems no way. Yet we do surely have and use a concept of tallness, do we not? Why can't we view epistemic justification similarly in terms of "sufficient" comprehensiveness and coherence?

¹³A 3.

¹⁴Sextus, in particular, invokes similes that illuminate our issue, such as the following: "Let us imagine that some people are looking for gold in a dark room full of treasures....[N]one of them will be persuaded that he has hit upon the gold even if he has in fact hit upon it. In the same way, the crowd of philosophers has come into the world, as into a vast house, in search of truth. But it is reasonable that the man who grasps the truth should doubt whether he has been successful." (Against the Mathematicians, VII 52, in the Teubner text, edd. H. Mutschmann (Leipzig, 1914).)

The combination of coherence and comprehensiveness comports with a concept of epistemic justification that is "internal." But it remains to be seen just where to draw the relevant boundaries: At the skin? At the boundaries of the "mind"? At the present- time-slice? At the boundaries of the subject's lifetime? Using some combination of the above? If so, which? And why? And why do we and should we care whether people are thus "internally" justified? My own answers would rest on a subject-centered conception of epistemic justification as intellectual virtue, and on the importance to a social species of keeping track of the epistemic aptitude or ineptitude of oneself and one's fellows, especially where it is possible to exercise some measure of control, however indirect.

¹⁵Many others since Descartes have groped for a similar way: from Hegel through Sellars; much work on epistemic circularity has also appeared of late, and I try to come to terms with some of the most important in "Philosophical Scepticism and Epistemic Circularity," Proceedings of the Aristotelian Society, supp. vol. 68 (1994): 268- 90. In "How to Resolve the Pyrrhonian Problematic: A Lesson from Descartes," forthcoming in Philosophical Studies, I develop the suggestion here that Descartes shows us the way beyond that problematic.