(Presented at a student philosophy conference in 1980 and published in Auslegung, Winter 1980. Footnote numbers are in square brackets.)

Last year, it was my privilege to address this conference on the subject of "Foundationalism, Coherentism, and Epistemic Ultimates." At that time, I contended that foundationalism -- the view that a necessary condition for the justification of any empirical proposition is that it rest, wholly or partly, on some epistemically ultimate "self-justifying" proposition(s) -- could not be correct, because the epistemically ultimate propositions it requires are impossible. Through some relatively straightforward arguments, I attempted to show that such propositions (a) could not be justified themselves without some reliance on still prior propositions, and (b) could not be adequate in content to provide for the justification of the rest of our empirical knowledge.

I come before you today, hat in hand -- not exactly to recant, but to admit that my foundationalist opponents had a better case than I was willing to give them credit for at the time. The key insight of foundationalism -- an insight to which coherentism must pay homage -- is that our empirical judgments must somehow be connected with our sub-judgmental sensory experience of the world. The fact that foundationalists have been unable to embody this insight in a defensible theory should not lead us to abandon the insight itself.

Coherentism, the view which I advocate in opposition to foundationalism, holds that propositions are justified by having a relation of logical coherence with a system of other propositions -- what I usually refer to as a "system of thought." Somewhat loosely, a proposition P is coherent with a system S if:

• it does not conflict with any other proposition(s) of S or any deductive consequences of the proposition(s) of S;
• it gives logical (not necessarily deductive) support to at least some of the other members of S; and
• it receives logical (not necessarily deductive) support from at least some of the other members of S.[1]

This means that coherentism goes against what most people would consider "common sense" in justifying empirical judgments. For example, if I said, "That is a cardinal on the branch yonder," most people think that my judgment is justified (if true) by reference to a certain non-mental "fact" in the world, i.e., the cardinal sitting on the branch. At what some would consider a more sophisticated level, philosophy students often think that the judgment is justified by reference to the sensations of the person making the judgment.

Blanshard and others have focused devastating criticism on both of these ideas, and I won’t repeat their arguments here.[2] The logical result, though, is that coherentism justifies such a statement by reference to lower-level perceptual judgments: not by reference to non-mental facts, and not even by reference to sense-experiences.

The question before us is: How can a theory which locates epistemic justification in a seemingly endless maze of judgments finally come to earth and connect with the sense-experiences that unarguably form the basis of empirical knowledge?

That there is some kind of raw experience, a "given" element which underlies our judgments, is undeniable. Moreover, this given element constrains the system of judgments we can correctly make. But how? How is it that when I feel a pain, I am "justified" in recognizing it as a pain but not as a tickle? The mental sequence seems to go like this:

Level 1. Sub-judgmental experience: pain.

Level 2. Elementary perceptual judgment: recognition of the pain as "pain."

Level 3. Explicit judgment (with or without verbal statement): "I have a pain."

In the adult mind, of course, this sequence is virtually instantaneous, but I think it captures the logical order of things.

It’s fairly clear that level (3), the explicit judgment that "I have a pain," is justified at least in part by reference to level (2), the implicit judgment in recognition of the pain as a pain. But how can we justify going from level (1) to level (2)? If we construe justification as a quasi-logical relation which can only obtain between judgments, then we simply cannot justify going from (1) to (2), because (1) is not a judgment.

Nevertheless, we must justify it somehow, or empirical knowledge will be completely severed from its connection with our sense-experience: which is tantamount to saying that empirical knowledge is impossible. There will be no more reason for me, upon having a "painful" experience, to judge that I am in pain than for me to judge that I am the king of Siam; and this is intolerable.

This is the point that foundationalists and truth-as-correspondence theorists have been trying to make: and they’re right. Our empirical knowledge must somehow be tied to our sense-experience. But how?

The most obvious way of attacking the problem is to revise our notion of the justification-relation. If we could allow elementary judgments of perception to be justified on the basis of sub-judgmental units of "bare experience," then our problem would be solved. On having a sub-judgmental experience of pain (level (1)), we would then be justified in making the perceptual judgment that recognizes the pain as "pain" (level (2)). We would have bridged the gap between our bare experiences and our system of judgments. This is the approach which, it seems to me, is unconsciously taken by most foundationalists.

A typical foundationalist formulation involving this type of justification is given by the (considerably under-rated and under-appreciated) Roderick Chisholm:

If a person believes, without ground for doubt, that he is perceiving something to be F, then it is evident (justified) for him that he perceives something to be F.

This sounds fine, and is even true, but it doesn’t solve our problem. "Perceiving something to be F" and "believing that one perceives something to be F" hardly qualify as bare experiences: they sound much more like perceptual judgments at levels (2) and (3). And yet Chisholm has captured, as well as possible, the foundationalist method of tying judgment to experience.

What ultimately thwarts Chisholm’s approach is that as soon as we bring the "bare experience" of F-ness into the realm of language, it ceases to be bare; as soon as we (explicitly or implicitly) tack the label "F" onto an experience, it has been classified and brought into a system of judgments -- i.e., it has been transformed into a judgment itself.

What Chisholm seems to be telling us is that if we have arrived at level (2), the level of implicit perceptual judgment, then we are justified in moving on to level (3) and making our implicit judgment explicit. But we already knew how to get from level (2) to level (3): that was never the problem. What we needed was to get from level (1) to level (2).

I don’t mean to pick on Chisholm, who in my estimation is one of the keenest minds to have worked on this problem. But no approach like the one he proposes can ever do the job of connecting our judgments to our experience: it can only connect higher-level judgments to lower-level ones. There is no way to set up a justification relation by which an ex hypothesi logically characterless bare experience could warrant or constrain a judgment in any way. We would always be left in the situation of making elementary empirical judgment X because ... because what? As long as the right-hand term of the justification-relation (the "what") is supposed to be bare experience, there can be no answer, because any answer we give will inevitably be a judgment.

Now, we come to the interesting part. If we cannot set up a direct justifying link between judgment and experience, then we had better try to set up an indirect link. Otherwise, empirical knowledge is impossible and we are living in a dream world.

I have on occasion shocked people at parties by saying that Wittgenstein, in the latter part of his life, was a crypto-coherentist and an incipient Absolute Idealist.

Of course, what shocks people the most is that I’d bring up such a dry and esoteric matter when all decent party-guests are trying to get happily drunk. But second after that, at least for anyone still sober enough to remember who Wittgenstein was, is the suggestion that he could have embraced such notions. And yet, some of Wittgenstein’s ideas (both his early work around the time of the Tractatus Logico-Philosophicus and his more mature reflections from the 1930s to the 1950s) give us a way out of our predicament -- along decidedly coherentist lines.

Wittgenstein said a lot of really strange things, even a few at parties, but two of the strangest -- and for our purposes, the most important -- are the following:

• That I cannot know that I am in pain; I can, however, know that other people are in pain.
• That initial indoctrination in a language-system is not education, but training.

What I wish to maintain, taking my cue from these theses of Wittgenstein, is that what links bare experiences to elementary perceptual judgments is not judgment, but association. I would suggest that elementary perceptual judgments are judgments only in a special sense, which will be explored in what follows. We will even discover – 60 years late, perhaps, but better late than never – the real significance of Wittgenstein’s much-criticized "picture theory of meaning" from the Tractatus.

Consider the way that we learn language in the first place. Our parents hold a red thing before us and say the word "red;" this procedure is repeated over and over, and when we finally utter the sound "red" upon being presented with a red object, we are rewarded with a sweet. In this way, a linguistic "peg" is established. We do not yet know that a red thing is red, but we associate the quality of redness with the linguistic peg.

Before proceeding any further, I should say a few words about what I mean by "association" in this context. One might object to my linguistic-pegs thesis by arguing that a child would need to recognize a red thing for what it was before he or she could associate it with the sound "red." Then, the child would remember that in the past, this particular color had been been presented at the same time as a particular sound, viz, "red." So, it might be asserted, even my invocation of association and linguistic pegs does not really solve the problem of getting from level (1) to level (2) -- because it presupposes that we are already operating with level (2) implicit perceptual judgments.

This objection can be met by distinguishing between ordinary memory and what I call "quasi-memory." Ordinary memory is conscious and propositional; quasi-memory is unconscious, automatic, and sub-judgmental. For example, when I am remembering a melody, I do not need to think ahead and explicitly "remember" what the next note is supposed to be: my consciousness glides effortlessly and automatically from one note to the next.

Association, as it applies to sense-qualities and linguistic pegs, is a form of quasi-memory. It is a relation which exists between two contents of consciousness when the presence of one immediately elicits the idea (here meant in Hume’s sense of a faint mental "image" of the appropriate perceptual type) of the other. Thus, the association between a sense-quality, e.g., red, and a word-sound, "red," need not depend on any prior judgments, implicit or otherwise. When a child (or an adult) is presented with a red object, his or her consciousness glides automatically -- with no thought required -- to the idea of the word-sound "red."

Now, at the extremely low level of a child who has just learned the word "red," judgment is not yet present. The child does not know that a red thing is red; all the child "knows" is that saying the sound "red" when presented with a red thing resulted in getting a sweet; and even this, he or she knows only by quasi-memory.

But as the child continues to experience and learn, his or her world becomes richer. The child acquires linguistic pegs for more experiences, and begins to learn linguistic pegs for the relations between these experiences -- e.g., that this red-experience is "behind" that green-experience; or that a red-experience never occurs at the same time and position in the perceptual field as a green-experience.

Eventually, mirroring in consciousness the structure found in experience, the child realizes that a red-experience is distinct from a green-experience -- i.e., the child acquires rudimentary concepts of identity and difference. These rudimentary concepts are built, as before, on quasi-memory associations and the absence of such associations.

It is within the context of the child’s rudimentary concepts of identity and difference, presence and absence, the the essential form of judgment -- that one thing is another -- arises.

Let’s now return more specifically to the question about our most elementary judgments of perception. What I wish to maintain is that in one respect they are judgments, but that in another respect, they are associations. As the child’s awareness of the world broadens, the child arranges the mental representations of the word-sounds he or she has learned into a formal structure isomorphic to that found in experience. This formal structure is a system of thought in embryo.

With the advent of the formal mental representation of the structure of the child’s experience, the word-sound "red" has come to be associated with two things. On the one hand, it is associated with the red-experience. But on the other hand, it has now also become associated with a place in a formal system of mental representations of word-sounds -- that is, with a linguistic peg. It is in this latter association that the elementary judgment of perception really becomes a judgment.

But consider another case:

• association #1: this-red
• association #2: this-same-that
• association #3: that-green

But in experience, a red and a green cannot occupy the same position in a perceptual field at the same time. Thus, the second set of associations is in conflict, i.e., is incoherent.

We now have, no longer a mere association of an experience with a word-sound, but an elementary judgment of perception that can conflict with other elementary judgments of perception. We have a mental unit which is tied to "bare experience" but which can also be tested for truth or falsity by virtue of its position in a formal system. And -- to no one’s surprise, inasmuch as I’m doing the talking -- the truth-test involved is coherence.

Thus, we have done what we set out to do. We have provided a link between knowledge and experience such that elementary judgments of perception are justified, albeit not quite in the way we had expected.

1. Of course, the more comprehensive and coherent the system, the more justification a proposition can derive from it.

2. For example, see Blanshard, The Nature of Thought, Vol. 1, Chapters I-III (on the relation between sensation and perception) and Vol. 2, Chapter XXV (on the tests of truth).

3. This is the actual insight at the root of Wittgenstein’s "picture theory" of meaning in his Tractatus Logico-Philosophicus, although Wittgenstein was not able successfully to articulate it at the time.

4. Note that because it is now part of a formal system, the word-sound "red" has become an actual word.