(Published in the September 1982 issue of Idealistic Studies. Several arguments in this paper were considerably improved by Professor Blanshard’s kind comments. Copyright 1982, 1998 by Scott Palmer.)

Professor Blanshard’s views on these issues, as he remarks in his response to Rescher in the Schilpp volume, have "not stood quite still" in the years since The Nature of Thought was written. Impressed (as any rationalist must be) by the immense explanatory power of modern science, with its trans-experiential atomic particles, he no longer holds that coherence exhausts the nature of truth. The existence of atomic particles entails that "correspondence (in a sense that has never been satisfactorily defined) must enter into the nature of truth along with coherence." Moreover, he has retreated from the view that in a fully coherent system, every proposition will entail every other proposition. His response to Rescher’s criticisms, therefore, does not take a "hard line" in defense of his earlier views.

I intend in this paper to take just such a hard line. I see neither that Rescher’s criticisms are valid nor that Professor Blanshard needed to modify his earlier views on coherence. First, I will deal with Rescher’s attack on the argument that accepting coherence as the test of truth entails accepting it also as the nature of truth. Second, I will inquire whether modern science, with its postulated trans-experiential atomic particles, necessitates the abandonment of a thoroughgoing coherence theory of the nature of truth. Finally, I will discuss Professor Blanshard’s evolving conception of coherence and whether his view in The Nature of Thought really required the sort of amendments which have been made in it.

Does Coherence Pre-Empt Correspondence?

Blanshard, after demonstrating at length that coherence with a system of thought is our only defensible test of truth, offers an argument to show that this entails coherence must also be the nature of truth. He writes:

(I) Suppose that, accepting coherence as the test, one rejects it as the nature of truth in favour of some alternative; and let us assume, for example, that this alternative is correspondence. This, we have said, is incoherent; why? Because if one holds that truth is correspondence, one cannot intelligibly hold either that it is tested by coherence or that there is any dependable test at all.

(II) Consider the first point. Suppose that we construe experience into the most coherent picture possible … Would the mere fact that such elements as these are coherently arranged prove that anything precisely corresponding to them exists "out there"? I cannot see that it would … It is therefore impossible to argue from a high degree of coherence within experience to its correspondence in the same degree with anything outside.

(III) If you place the nature of truth in one sort of character and its test in something quite different, you are pretty certain, sooner or later, to find the two falling apart. In the end, the only test of truth that is not misleading is the special nature or character that is itself constitutive of truth.

(IV) If truth does consist in correspondence, no test can be sufficient. For in order to know that experience corresponds to fact, we must be able to get at that fact, unadulterated with idea, and compare the two with each other. And we have seen in the last chapter that such fact is not accessible. When we try to lay hold of it, what we find in our hands is a judgment which is obviously not itself the indubitable fact we are seeking.

Rescher reconstructs this argument as follows:

1. A coherence theory of truth must take coherence as the prime test of truth.
   
2. Now, if the definition of truth finds the nature of truth to reside in something other than coherence, something which – like correspondence – is not logically tantamount to coherence but can potentially diverge from it, then coherence cannot qualify as a failproof guarantor of truth.
   
3. But since a coherence theory must take coherence to be the prime test of truth (Premiss 1), it must see in coherence a failproof guarantor of truth.
   
4. But then, it follows (from Premiss 2) that a coherence theory of truth must take coherence to represent the nature of truth … For only what is essential to its very nature can provide a conceptually failproof guarantee for a thing.

How does Rescher’s version of this argument work, and how does he derive it from the passage quoted from Blanshard? Premiss (1), that a coherence theory of truth must take coherence to be the prime test of truth, seems unproblematic and almost tautologous. Premiss (2) simply makes the point that if one accepts coherence as the test of truth and correspondence as its nature, then the coherence test is in principle fallible, i.e., a proposition might be highly coherent with our system of thought and yet still be false (or vice versa), since the truth of the proposition would reside in something other than its mere coherence with the system. This premiss can be derived from passages (II) and (III). Premiss (3), which contends that taking coherence as the prime test of truth entails taking it as an infallible test, can be drawn from passage (III), though a good deal of "reading between the lines" is required.

The argument, then, is supposed to go like this. A coherence theory of truth must see coherence as the prime, and hence as an infallible, test of truth. But if the nature of truth is held to be other than coherence, then coherence cannot provide an infallible test of truth. Therefore, if we assume that coherence is in fact the correct criterion of truth, then we must also admit coherence to be the nature of truth.

Rescher finds the flaw in Blanshard’s argument, as he has reconstructed it, to lie in Premiss (3)’s insistence that coherence must be seen as an infallible criterion of truth. Why not, he asks, just accept coherence as a truth criterion which is generally reliable, and refrain from demanding absolute guarantees? He writes:

On Blanshard’s approach … the partisan of coherence as the criterion of truth is committed to regarding coherence as a guaranteeing criterion. He is committed to regarding the link from coherence to truth as inevitable and necessary. Now, subject to this presupposition, Blanshard’s position is unquestionably a strong one. But why need this presupposition be made? Why could not or should not the coherence theorist regard it as an authorizing rather than a guaranteeing criterion?

Premiss (3), then, has little to be said for it and ought to be rejected. But the rejection of Premiss (3) undercuts Blanshard’s argument. Hence, we can consistently hold that coherence is the proper criterion of truth and that correspondence is the nature of truth – having our metaphysical cake and eating it, too.

Now, subject to Rescher’s presupposition that the argument actually requires Premiss (3), his position is unquestionably a strong one. But why need this presupposition be made? The closest one can come to textual support from Blanshard’s own exposition is in passage (III), where he worries that our truth criterion and truth itself might sooner or later fall away from each other unless both are taken to be coherence. And unexpected support for Premiss (3) comes in Blanshard’s reply to Rescher in the Schilpp volume, where he expresses concern that a mere "authorizing" criterion of truth might be too loosely connected to truth itself:

… if the criterion A could be present while the B it was supposed to indicate was absent, and the criterion be absent when the thing indicated was present, the connection would be intolerably loose … What troubles me about Mr. Rescher’s proposed criterion is that it possesses this kind of looseness.

Such a passage as this, from the author of the argument in question, is not to be lightly shrugged off. Nevertheless, there is convincing textual evidence that Blanshard in The Nature of Thought did not regard coherence – at least, as we actually apply it – as a guaranteeing criterion. Moreover, his "test-nature" argument can be formulated without requiring coherence to be anything stronger than an authorizing criterion which gives us reasonable grounds for asserting truth. Let’s examine each of these points in turn.

What is the textual evidence? It is abundant indeed, but perhaps the plainest passage comes on the very next page after Blanshard makes his test-nature argument. Responding to the objection that taking coherence as the nature of truth leads to skepticism, he writes that:

… the theory does involve a degree of skepticism regarding our present knowledge and probably all future knowledge. In all likelihood, there will never be a proposition of which we can say, "This that I am now asserting, with precisely the meaning I now attach to it, is absolutely true." … This refusal to claim absoluteness for our knowledge appears even as a merit. For the road of history is so thick with discarded certainties as to suggest that any theory which distributes absolute guarantees is touched with charlatanism. (my italics)

If that does not qualify as a repudiation of the idea that coherence is an infallible guarantee of truth, then I find it difficult to imagine what would qualify.

Still, there is a possibility for confusion here. Perfect coherence, coherence of a proposition with a fully-developed system of thought (to be strictly correct, I should say the fully-developed system of thought) would constitute an infallible guarantee of truth. But the system of knowledge we have in hand is fragmentary and incomplete; such internal coherence as it has, and its logical relations to particular hypotheses we might wish to test, we can see only very dimly at best. Hence, coherence as we actually apply it cannot be taken as anything stronger than an authorizing criterion, even if in the ideal case, it would give us an infallible, guaranteeing criterion.

This is essentially the line of argument taken in another passage which shows that Blanshard did not regard coherence as a guaranteeing criterion of truth:

This answers by implication another objection to the theory. It is said that a truth once true must always be true, whereas on the coherence theory what was true may now be false, and what is now true may become false with expanding knowledge. That which coheres with the knowledge of an earlier time may conflict with the knowledge of a later time. Thus propositions may put on truth or falsity, and take them off again, which changing scientific fashions, which is absurd.

But note well the following: if Blanshard had taken coherence simpliciter as a failproof criterion of truth, this objection would be devastating. For then proposition X, coherent with the system of knowledge available in 1880, and proposition not-X, coherent with the system of knowledge available in 1980, would both be guaranteed as true; and Blanshard dismisses such a result as "absurd," as well he should.

Given, then, that Blanshard did not mean for coherence to be construed as an infallible guarantee of truth, what exactly is his argument to show that accepting coherence as the test of truth entails accepting it also as the nature of truth? A rough-and-ready outline might go like this:

1. Coherence is the (only defensible) test of truth.
   
2. Suppose that correspondence is the nature of truth.
   
3. Then by (1), the thesis that correspondence is the nature of truth must be testable by coherence; and the same applies to any other thesis which might be advanced, e.g., "That is a cardinal on the branch yonder."
   
4. By (2), the truth of the thesis that correspondence is the nature of truth consists of its correspondence with an extra-mental "fact."
   
5. Hence, in order to test the thesis by coherence, we must be able to compare it with the extra-mental "fact" to which it corresponds.
   
6. This, however, is impossible. Such extra-mental facts, even if they exist, are epistemically inaccessible. Indeed, this inaccessibility of extra-mental facts was one of the primary reasons why we adopted coherence as our truth criterion in the first place.
   
7. Furthermore, coherence is a logical relation which can obtain only between judgments – so that even if extra-mental facts existed and were somehow accessible, the correspondence between judgment and fact could not possibly be tested by coherence.
   
8. Therefore, if correspondence is the nature of truth, then coherence cannot be the criterion of truth. But this contradicts (1), which we regard as firmly established. Hence, correspondence is not the nature of truth.

This argument shows that if coherence is accepted as the test of truth, then correspondence cannot be the nature of truth; and yet it nowhere invokes coherence as anything more than an authorizing criterion.

We must in fairness note, however, that this particular version of Blanshard’s argument is not as powerful as the one proposed by Rescher. It shows only that if coherence is the test of truth, then correspondence cannot be the nature of truth. This is a weaker conclusion than the one Rescher derives, viz, that if coherence is the test of truth, then it must also be the nature of truth. The two versions of the argument can be made equivalent only by adding a strengthening assumption to the premisses of the weaker version, e.g., that coherence and correspondence are the only two possibilities for the nature of truth.

Moreover, Blanshard makes it clear that he thinks his argument has the power of Rescher’s version – not merely proving the limited negative thesis that truth is not correspondence, but proving instead the sweeping negative thesis that no alternative to coherence will do. He writes:

The argument is: assume coherence as the test, and you will be driven by the incoherence of your alternatives to the conclusion that it is also the nature of truth.

So Rescher’s construal of the argument as depending on coherence as a guaranteeing criterion is not based on mere caprice: it represents one strategy which can be taken to make the argument accomplish what Blanshard says it is supposed to accomplish.

There are, however, at least two other strategies available. The first is one about which we have already remarked briefly: it consists of taking coherence and correspondence as the only two alternatives for the nature of truth. Then, a proof that correspondence is not the nature of truth would lead, by disjunctive syllogism, to the conclusion that coherence is the nature of truth. Although this is the most profitable strategy to employ – an argument is needed to back it up, of course, and will be supplied shortly – it is clear that Blanshard himself does not employ it: he refers, for example, to the alternatives (plural) to coherence as the nature of truth.

The other possible strategy, which more complicated (and ultimately less successful) than the first two, has the merit of being in harmony both with Blanshard’s rejection of coherence as a guaranteeing criterion and also with his rejection of the thesis that coherence and correspondence are the only possibilities for the nature of truth. The avenue of attack is as follows.

Rescher makes the point, following Blanshard’s discussion of the issue, that the early coherentists did not clearly distinguish between the questions, "How do we know that a proposition is true?" and "What does it mean for a proposition to be true?", i.e., between the criterion and the nature of truth. Rescher writes that:

The earlier coherence theorists tended to view coherence as a characteristic mark of the truth without any very specific and definite commitment as to the nature of the "mark" at issue. Is coherence a somehow necessary feature of the truth – is it a part of the definition of truth or even the whole of it?

Blanshard expresses similar caveats about keeping the two issues separate. Nonetheless, one could easily maintain that there was more than mere carelessness behind the early coherentists’ apparent "slip:" that the distinction between nature and criterion is not as absolute as one might suppose.

Consider the sort of example that Rescher gives in his book to illustrate the difference between nature and criterion. You have a metal sample, and from its appearance and weight, you know that it could be any one of five different elements. However, you also know that of these five elements, only element X turns black when immersed in sulfuric acid. Then, turning black when immersed in sulfuric acid will be your criterion for determining if the sample is composed of element X. What you are doing is taking just one of the properties which make up the "nature" of element X, and using it as your criterion of the element’s presence. Which property you pick depends on two things: (a) the context, since your criterion must be a property not shared by the other elements which might compose the metal sample; and (b) the (number of) properties that element X has available to serve as a criterion of its presence. This works in the way it does because element X is, in fact, a mentally constructed complex of a great number of sensible and dispositional properties. The same applies to the other elements of which the sample might be composed.

This line of argument continues: But what are the properties that, added together, compose truth? Whatever else they might argue about, the proponents of different theories of the nature of truth agree on one thing: truth is logically and metaphysically unitary. For coherentists it is coherence with system, for correspondentists it is correspondence with facts, for pragmatists it is success in practice -- but for everyone who takes the question seriously at all, truth is just one thing. It is not a complex of different, isolable properties – like element X – which our thought can put together and take apart at will. To put the main thrust of the point somewhat loosely, then, truth has just one "property" which can be used as a criterion for its presence. This being so, whatever turns out to be the criterion must also be the single property which constitutes the nature of truth itself. And, as Blanshard has shown, this property is coherence with system. In the special case of truth (or any other logically unitary entity), criterion and nature must coincide.

The foregoing analysis, if acceptable, strengthens the argument Blanshard gives into a proof that coherence as the criterion of truth not only pre-empts correspondence as the nature of truth, but pre-empts every other alternative as well, showing thereby that if coherence is the criterion of truth, then it must also be the nature of truth.

Unfortunately, it is a bad analysis. However inclined we might be to accept the idea that truth is metaphysically unitary – whether owing to an antecedent monism, or to an affection for such rationalistic slogans as "there is only one truth" – we must face the fact that truth is not logically unitary in the sense required to make this argument work. It is simply not the case that the only thing we can say about truth is that it is constituted by coherence, or correspondence, or whatever our favorite notion of truth happens to be. Truth is what works; it is, as Hector Castaneda points out, the "psychologically preferred property" of judgments; it is that which sets us free; and so forth. These are all properties which can be correctly ascribed to truth, any one of which could – at least in principle – be chosen as a truth criterion. Hence, the argument fails to prove that criterion and nature of truth must coincide because truth has only one property in its nature to serve as a criterion.

It might ultimately be the case, and I happen to believe, that a perfect understanding of truth (or anything else) would find it to be a single logical whole, without the disparate, disconnected elements mentioned above. Seen in the light of this perfect understanding, truth would be identical with the Absolute of Bradley. But in our practical, day-to-day attempts to understand the world, we are unable to proceed at such a lofty plane.

The correct approach to strengthening Blanshard’s argument takes the line that coherence and correspondence provide the only possible accounts of the nature of truth. The point can be made simply. Regardless of our prior theoretical commitments, we must recognize that truth is a relation between a judgment and a "fact" which the judgment is about. Now, this fact is either a judgment (or system of judgments) or it is something else. If the fact is constituted by a judgment or system of judgments, then the relation between the original judgment and the fact will be one of logical coherence. If, on the other hand, the fact turns out to be something else – regardless of what that "something else" is – then we must suppose the relation to be some sort of correspondence. Hence, the question of whether a coherence or correspondence theory is correct rests on the character of the facts that judgments are about. But regardless of what sort of things facts turn out to be, we are still left with only two alternatives. If facts are judgments then a coherence theory is correct; if they are not judgments, then a correspondence theory is correct. There is no third option.

A pragmatist might object, of course, that truth is success in practice, and that, hence, there is a third option. Both Blanshard and Rescher have focused devastating criticism on this position, but here it should suffice to point out that success in practice is just not what we mean when we talk about truth. Imagine a dialogue between a man who claims that it is the nature of an elephant to have a trunk and another man who denies it. The latter does so on the ground that the true nature of an elephant is given by the pain experienced when it steps on one’s foot, and pain is not the sort of thing that can have a trunk. One need not be a devotee of ordinary-language philosophy to hold that our elephant-pragmatist’s objection is slightly beside the point. In the case of truth, while success in practice might result from its application, such success is not the same thing as truth itself, any more than a pain in the foot is the same thing as an elephant.

The same sort of rejoinder can be made to any other candidate for the nature of truth: when we talk about truth, we mean a relation between judgment and fact. The only matter left open to decide is what sort of thing the fact is. If we hold that it is a judgment or system of judgments, then we are committed to coherence as the nature of truth; if we hold that it is something else, then we are committed to correspondence. A person who denies this dilemma might indeed be talking about something, but he is not talking about truth.

Because coherence and correspondence give the only two possible accounts of the nature of truth, Blanshard’s proof that truth is not correspondence ultimately issues in the conclusion that truth is coherence.

The next task we set for ourselves was to inquire if modern science, with its postulated trans-experiential atomic particles, requires us to abandon a thoroughgoing coherence theory of the nature of truth. Before charging ahead, however, we must make clear just what the problem is supposed to be. What can the existence or non-existence of atomic particles have to do with theories of truth?

In his closely reasoned argument in The Nature of Thought, Blanshard shows that even in our most ordinary perceptual judgments (e.g., "There is a pencil on the desk"), the "fact" which our judgment is about, and which we use to test it, is actually just a mini-system of lower-level perceptual judgments (which are, of course, connected by innumerable logical lines to all the other judgments of the whole system of knowledge). Since what we really have, then, is a higher-level judgment about a system of lower-level judgments, it makes perfect sense to say that the truth relation involved is logical coherence between judgments.

But suppose that atomic particles exist and we can make judgments about them. Then, when we make such a judgment – for instance, "Uranium atoms are radioactive" – our judgment cannot be referred to a lower-level perceptual judgment, because atoms are by their very nature outside of our experience. The "fact" which our judgment is about is outside the mind and, hence, is not a judgment itself. Thus, the truth of our judgment consists in its relation to something which is not a judgment or system of judgments, and this relation cannot reasonably be construed as one of logical coherence. If it is anything, it would have to be some sort of correspondence. We must face the hard fact that if atomic particles qua trans-experiential entities exist, then both the coherence theory of the nature of truth and idealism itself will have to be rejected. Moreover, because Blanshard’s test-nature argument shows that there is a logical link between the criterion and the nature of truth, the existence of atoms and the consequent acceptance of the thesis that truth is correspondence will also require the rejection of coherence as our truth criterion.

Blanshard has recounted how his tutor at Oxford, H.H. Joachim, once turned to him as they were taking an afternoon walk and asked, "Do you suppose that there really are such things as atoms?" His answer at the time, prior to World War I, was probably much less affirmative than it is now. He thinks that, faced with the achievements and explanatory power of modern science, it is no longer reasonable to question the existence of atomic particles.

Perhaps he is right. But though I tremble inwardly at opposing such a "mighty engine of thought," as Blanshard himself once called his hero F.H. Bradley, I simply cannot go along with him on this point. The existence of such trans-experiential particles seems to me highly questionable for a number of reasons.

The most obvious reason is that scientists themselves do not now maintain that atomic particles exist in anything like the way that other "material things" exist: the very supposition that they do leads straight to paradox and incoherence. A recent Scientific American article on the nature of the electron, for example, made the following observations:

Indeed, the model [of electrons as tiny bits of matter] has grave flaws … It turns out that the rotational velocity at the surface of the electron is greater than the speed of light … All experimental data gathered so far are consistent with the idea that the electron is a point particle, entirely without extension. The arguments presented here then predict that the electron mass is infinite, a manifest absurdity.

Thus, if we take atomic particles – in this case, electrons – to be trans-experiential "material things," then we are driven to accepting the existence of material things which (a) can rotate faster than the speed of light, which contradicts the laws of physics, and (b) do not take up any space but have infinite mass, which contradicts any concept of "material thing" that makes sense to us.

But there is a still deeper and more important reason for rejecting the existence of trans-experiential atomic particles. The fundamental flaw in Blanshard’s contention that the existence of atoms requires an abandonment of coherence as the nature of truth is that it over-emphasizes the importance of our particular form of sense experience, and consequently fails to recognize the real function of our concepts and judgments.

Much of the work that has been done in metaphysics and epistemology seems to rely covertly on an unverifiable assumption. I call it:

The Assumption of the Inter-Personal Uniformity of Sense Experience:

For any inter-subjective empirical entity X, all human beings will experience X in the same way, except for occasional cases in which sense organs do not function properly, e.g., color-blindness.

This assumption has very far-reaching implications. Let’s take a brief look at what the world might be like if this assumption were true.

Case 1. Smith tells Jones that he owns a red sportscar. What is he telling Jones about the color of the car? Because both men experience "red" things in precisely the same way, Smith is telling Jones that when he (Smith) looks at the car he has a certain color experience, and that if and when Jones looks at the car, he will have the identical color experience. Moreover, the inter-subjective universal "red" is just that visual quality which is exemplified in the visual experiences of Smith and Jones.

Case 2. Smith tells Jones that "one cannot see atoms." Whenever either man looks at a physical object, he sees only an expanse of colors – a "solid thing" rather than an aggregation of tiny particles. Hence, "atom" refers to a trans-experiential particle.

There is no doubt that our world will be much neater if we can find a way to validate the Uniformity Assumption. This being said, however, it seems quite plain that the assumption is impossible to defend. We never have conscious access to the private sense experiences of others, and consequently can never know if they experience things in the same way as we do. Nearly everyone is familiar with the schoolboy puzzle about whether or not all people see the same color when looking at a "red" object, but the problem is even more serious than the puzzle suggests. We not only do not know that others see things in the same way as we do: we do not even know that they see at all in the same sense as we do. Consider the following case.

It is at least possible that a child should be born with its nervous system mis-wired. Specifically, consider a child whose optic nerves lead from its eyes to that part of the brain which controls auditory perception, and whose auditory nerves lead from its ears to that part of the brain which controls visual perception. Suppose further that the auditory perceptual field which results from stimulation of the child’s eyes is structured in this way: variations in what we would perceive as the colors of objects are perceived as variations in pitch; variations in what we would perceive as the spatial positions of objects are experienced as variations in loudness and timbre. An exactly analogous account is to be given of the visual perceptual field resulting from stimulation of the child’s eardrums.

Could such a child learn our common language? Of course. When presented with a red thing and the sound "red," the auditory experience elicited by the red thing would be associated in the child’s mind with the visual experience elicited by the sound "red," and eventually also with the kinesthetic experiences elicited by certain movements of the child’s mouth and vocal cords. He could identify items of different colors and in different spatial positions with complete accuracy, but his experience of them would be totally different from our own. No one would ever be aware of his strange mode of perception. Indeed, any one of us could be that child, without ourselves or anyone else ever being the wiser.

What this case reveals, whether or not anyone actually is mis-wired, is that our knowledge of the world is essentially structural, having nothing to do with the specific manner in which we experience reality. As far as content goes, the mis-wired child’s perception of a red firetruck will be completely different from our own, but on the side of structure it will be completely identical. The given elements which make up the firetruck will be arranged in exactly the same pattern, albeit transferred to different sensory modalities; and the "fact" of the firetruck will stand in exactly the same relations to the other facts which make up the child’s, and our, world. For the child, as for us, a red firetruck is red; red is a color; and the experienced color red is produced by a certain sort of stimulation of the eyes under certain conditions. So it would be incorrect to say that when the child looks at a red firetruck, he perceives something different from what we perceive. As far as the content of his perception goes, that is true, but absolutely unimportant. What he perceives, in terms of internal pattern and relations to the other facts which make up the world, is exactly the same as what we perceive. Red is still red, regardless of whether it is perceived as a color, a sound, or a tickle.

All of which does not, of course, show that the Uniformity Assumption is false: we cannot do that any more than we can show that it is true. What the above discussion does indicate, however, is that acceptance of the Uniformity Assumption tends to obscure the true character of our knowledge of the world: what our concepts are concepts of, and what our judgments actually assert. This being so, let’s reject the assumption for a moment, then see in what sort of world we find ourselves and what our judgments actually do in that world. We can return to the two cases discussed earlier.

Case 1. Smith tells Jones that he owns a red sportscar. What has Smith told Jones about the color of the car? Since we are now operating without the Uniformity Assumption, we cannot assume that he is communicating anything to Jones about his private experience of redness. The most information that he can convey is that the car is one term of a certain structure in experience, each member of which elicits (under normal conditions) the same experience as all the others. When Smith says, "My car is red," he asserts nothing about his private experience of redness. He asserts only that his car is the same color as apples, cherries, lipstick, etc. Even "the same color as" is a purely structural notion, which entails no commitment about the specific manner in which color is experienced.

Case 2. Smith tells Jones that "One cannot see atoms." Since we are not bound by the Uniformity Assumption, let’s introduce a special premiss here. Suppose that Jones’s visual processes are such that, when looking at a physical object, he actually experiences what we would characterize as a presented aggregation of billions upon billions of tiny particles. Will Jones then dissent from Smith’s statement, claiming that he can indeed see atoms? No, he will not, because he really cannot see atoms. "Atoms" are by definition something that one cannot see; and "solid expanse" or "colored surface" are structural concepts – just like "atom" – which apply regardless of the specific form in which Jones perceives them. From earliest childhood, whenever he was presented with a physical object – irrespective of the content of his experience at the time – Jones has learned to call what he sees a "surface" and not an aggregation of atoms. He will never suspect that he sees things any differently from the rest of humanity, and, indeed, in the most important sense, he does not. His different form of visual perception makes no difference for his knowledge of the world.

What is the moral of these examples? The most important moral with which we are concerned here is that our judgments and concepts do not refer to anything outside of experience; they do not even refer to the content of our experience itself. They are about nothing more nor less than the structures which we find in our experience, and are cognitive instruments which we use to systematize our experience into an intelligible whole. This being the case, it is totally unnecessary to fret about the ontological status of "atoms." Only the Uniformity Assumption, by obscuring the true character of our knowledge, led us to believe that our concept of "atom" referred to something outside of experience. Once we have provisionally rejected the assumption and are able to see the actual cognitive role of our concepts and judgments more clearly, we understand that the atom is not so much a tiny particle of matter as it is an organizing principle of our knowledge. And understood in this way, the "existence" of atoms presents no problems for either idealism itself or for the coherence theory of the nature of truth. I conclude, with all respect, that Blanshard’s misgivings about coherence and idealism in the face of modern science are groundless.

The view that in a perfectly coherent system every proposition will entail every other proposition, defended by Blanshard in The Nature of Thought and by Joachim, among others, before him, has never had an easy time of it. Rescher’s attack on the doctrine is only the latest in a series which includes the sympathetic but formidable criticisms advanced by A.C. Ewing. Even Blanshard has now abandoned the view, writing that:

If any proposition of Euclidean geometry were deleted from human memory, it could be reinstated by deduction from the rest of the system, but no geometer would claim that from a single proposition the rest of the system could be reconstructed. I should myself be content with Ewing’s terms: a coherent system is "a set of propositions in which each one stands in such a relation to the rest that it is logically necessary that it should be true if all the rest are true, and such that no set of propositions within the whole set is logically independent of all propositions in the remainder of the set."

It is impossible, in a short paper, to do full justice to the criticisms which have been brought against the view of perfect coherence as "reciprocal entailment." Nonetheless, it is possible to show that the case against this view is a little less pat than it might appear, and that is what I will try to do here.

Let’s begin with a fairly uncontroversial assumption – at least, as uncontroversial as philosophical assumptions get. Let’s assume that the coherence theory of meaning, in some form, is correct. This view simply holds that the meaning or logical content of a judgment is determined by the system of thought in which it has a role, or, alternatively, by its logical relations to all the other judgments or propositions of the system. Versions of this view have been advanced by such diverse philosophical figures as Bosanquet, Blanshard, Quine, Davidson, and the later Wittgenstein.

If this view is correct, then for any proposition p, there will be not one but two significant meanings: first, the meaning of the proposition as it is actually used by a particular person at a particular time, which meaning is determined by an incomplete and less-than-fully-coherent system – we might call this a "partial" meaning; and second, the meaning of the proposition as it would be if we had complete and fully-coherent knowledge, which meaning is determined by a complete and fully-coherent system – we might call this the "complete" meaning. (Note that there can be many different partial meanings, but only one complete meaning.)

Then there are two closely related, but significantly different, questions involved here. The first question is this: Can we from a partially meaningful proposition derive its entire systematic context? The second question is this: Can we from a completely meaningful proposition derive its entire systematic context? It is, I would suggest, the confusion of these two questions – specifically, the substitution of the first for the second – which makes it seem necessary to abandon the strong concept of perfect coherence as reciprocal entailment. (It can be shown, by the way, that a perfectly coherent system must also be completely comprehensive, encompassing all truth. This proof will be given in a future article which explores the nature of coherence in more detail.)

Consider the example of the geometrical proposition Blanshard cites in the passage just quoted. As we ordinarily use the proposition, e.g., "a straight line is the shortest distance between two points," we do not have the entire systematic context in mind, but only a small piece of it – that piece which we need for our immediate purpose. For instance, we might need to call upon our knowledge of its logical relations to certain other propositions of the system in order to carry out a specific inference. The rest of the systematic context, although implicitly present, is not part of our conscious awareness, since we have focused only on what we need. Hence, as far as the meaning we "see" goes, it is only a partially meaningful proposition and we could subtract the parts of the system on which we have not focused without any apparent loss of meaning from the proposition. The loss of meaning, however, would very quickly become apparent if we attempted to do a different inference which required something from that part of the system that we had subtracted. But from this partially meaningful proposition (which is really just a system fragment), it is quite correct that the whole system of geometry could not be reconstructed.

The situation is quite different, however, if we are considering a completely meaningful proposition. In this case, the whole system will be implicitly present in the proposition, or else it will be less than completely meaningful; there is no "transaction" anywhere in the system in which the proposition could not, in principle, be involved. Starting from such a completely meaningful proposition, it is clear that we could reconstruct the rest of geometry.

We can conclude, therefore, that in a perfectly coherent system, every proposition (i.e., completely meaningful proposition) will entail every other proposition, as well as the whole system itself.

Two objections have been raised even to Blanshard’s current "minimal" conception that a coherent system is

a set of propositions in which each one stands in such a relation to the rest that it is logically necessary that it should be true if all the rest are true.

This is plainly false, the objection goes, because there are many propositions of, e.g., Euclidean geometry, which cannot be deduced from the rest of the system – the parallel postulate, for example.

How one views this objection will depend largely on how narrowly one thinks of "deduction." If what is at issue is a standard Aristotelian or Principia Mathematica-type deduction, then the objection is quite correct. But if we allow a slightly wider sense to "deduction," then the objection can be effectively parried.

To simplify matters in a way that still preserves the structure of the situation we wish to examine, let us suppose that all geometrical propositions fall into one of two classes: axioms and propositions derived from the axioms. The case of derived propositions presents no problem, since if they are removed from the system, they can simply be derived again. But suppose that the removed proposition is an axiom, call it "A." Then, there are two paths to the recovery of A. First, there will be many other propositions of the system whose derivation depends on A, and these propositions taken together will clearly call for a proposition precisely equivalent in deductive power to A. An axiom weaker than A would not license the derivation of all the propositions in the aggrieved set, while an axiom stronger than A would license more than are in fact present.

The second path to A’s recovery invokes the coherence theory of meaning once again. The meaning of A is not self-contained, but is scattered throughout the system. To remove A from the system would alter the deductive powers of many other propositions in it. If we assume that these propositions do not change, then we are committed to suppose that by unpacking the meaning of any one of these, we will eventually come back to A. Since in both of these cases, the recovery of A depends, in greater or lesser degree, on unpacking the meaning of other propositions of the system, we might call the process by which we recover A "m-deduction."

The second objection to Blanshard’s minimal conception is an attack on clause (ii) – that in a completely coherent system

no set of propositions within the whole set is logically independent of all propositions in the remainder of the set.

Suppose, this objection goes, that instead of merely removing A, we also removed all of the deductive consequences of A. Would not that set, which we had removed, be logically independent of the rest of the system, e.g., the parallel postulate and all of its consequences?

And the answer is, yes, it would be independent of the rest of the system, but there would not be anything left to compose "the rest of the system." Because any proposition of the system can, in principle, be involved in the derivation of any other proposition of the system (by either standard deduction or m-deduction, which can even derive axioms), then removing all the deductive consequences of A would mean removing every other proposition of the system!

Even if this line of response is denied, there is still the matter of the coherence theory of meaning and "m-deduction." To remove a large set of propositions from the system would do considerable violence to the meanings and logical powers of the remaining propositions. If we suppose that these are unchanged, then we can (at least in principle) employ m-deduction to pass from them to the missing set. In this sense, then, they would not be logically independent.

 Conclusion

In this paper, it has only been possible to suggest avenues of research which require a good deal more development. We have, however, seen that the "classical" idealist positions defended by Blanshard in The Nature of Thought are not nearly as vulnerable to the sophisticated views of a later time as one might initially suppose. In particular, coherence supplies both the criterion and the nature of truth; modern science falsifies neither the coherence theory of truth nor idealism; and there are reasonable grounds for accepting a concept of full coherence as reciprocal entailment.