In the introduction to their collection of papers, the new realists wrote that "perhaps the most notable feature" of realism is "the emancipation of metaphysics from epistemology" and added that by this they meant that "the nature of things is not to be sought primarily in the nature of knowledge."1  This is a grand slogan, but it is also a dark one. If we take metaphysics as dealing with questions of existence or, to use older terminology, being, and epistemology with knowing, we can interpret the new realists as claiming that existence does not entail being known. That is, the slogan can be taken as:

        (1)   X exists does not entail X is known

since the cardinal principle of idealism is "To be is to be perceived" and, for an empiricist, all knowledge stems from perception, the idealist is committed to the view that nothing can exist without being known; i.e. to the view that X exists entails X is known. Clearly this is to reduce metaphysical questions to those of epistemology. As opponents of idealism, however, the new realists could not accept this; as a result, they accepted (1).

        There is another way of interpreting the slogan, however. We can take it to mean that existence is not only independent of being known, but is also independent of being knowable; that is, we can interpret the motto as the claim that things can exist even though they are in principle unknowable. Taken in this way, the slogan is:

        (2)   X exists does not entail X is knowable

To hold this is also to emancipate metaphysics from epistemology, for the claim that the existence implies knowability, like to be is to be known, involves a reduction of metaphysical issues to issues in the theory of knowledge.

        Although they accepted (1), the new realists did not accept (2); for them, emancipation meant only partial freedom. Perry, for example, held that "the thing transcends the thought, but it remains perceivable, or in some such manner immediately accessible"2 while Marvin, who apparently thought up the phrase "the emancipation of metaphysics from epistemology," held that agnosticism is "an absurdity" if it means that "any sensible question can be asked of reality and not admit of a conceivable answer."3 It is clear from these passages that both Perry and Marvin reject (2); or, to put their view positively, they both hold the following principle:

        (3)   X exists entails X is knowable

which, of course, is the negation of (2). I think this was a mistake. Specifically, I will argue that there is no reason to accept (3) as a principle that will allow us to justify metaphysical claims by appealing to what we can know.

        By way of clarification, three points should be noted about (3). First, 'knowable' is taken to mean knowable in principle by humans. Locke and Kant, who are the prime targets of the principle, never denied that God or the angels could know insensible particles or things in themselves; rather their claim was that these entities are beyond all possible human knowledge. Second, as I have already suggested, the exponent of (3) assumes a broadly empiricist account of knowledge that refuses to countenance entities that are unobservable by humans. This further restricts the knowable to what we can directly observe or perceive. As a result, (3) when rephrased becomes the view that everything that exists can be observed in one way or another by humans.

        Nor is this all. On the face of it, (3) seems quite plausible even with these restrictions. With suitable beneficial mutations, humans could, it seems acquire capacities to observe all sorts of things they cannot now observe. If this is right, (3) is quite innocuous, but this is not what the usual defender of (3) believes. This brings me to the third point about (3). One who accepts (3) wants to argue that some alleged entities are unobservable in principle and not just unobservable given our present condition. That is, he holds that:

        (4)   X is unobservable in principle by humans

is true for some substitutions of 'X'. The reason for this is that he wants (3) to do some metaphysical work. Thus, if he can show (4) to be true for some substitutions of 'X', he can conclude that there are no such entities, or, perhaps even that talk of such entities is meaningless. However, there seems to be only one way for him to carry out this program:  He must hold that our capacity of knowledge is of necessity limited to  a certain range. And this commits him to the view that humans have essential cognitive capacities.

        I think this sort of essentialism is indispensable for the use of (3) in metaphysics. For, without it, there is no way to deny the consequent of (3) and draw the conclusion that a certain alleged entity cannot exist. I also think essentialism is indefensible, but I will not argue that here. Rather I wish to make the weaker point that the conjunction of (3) and essentialism is implausible. Thus, so far as the present paper is concerned, I have no objection to one who accepts essentialism without (3) or, on the other hand, (3) without essentialism. My only argument is with those who accept both (3) and essentialism about human cognitive capacity. If I am right, although (3) is quite innocuous in itself, it cannot do any metaphysical work; hence, yet another supposed link between metaphysics and epistemology ought to be severed. For convenience, I will call those who accept the conjunction of (3) and essentialism 'para-idealists' on the ground that, although they need not be idealists, they still accept part of idealism's program. Among the para-idealists, I would include Peirce and most pragmatists as well as Carnap and most logical positivists.4

        One's first reaction to the conjunction of (3) and cognitive essentialism, I think, is surprise. The para-idealist takes humans to be in a privileged position with respect to what exists, to be at the ontological center of the universe, so to speak, for he holds that the cognitive capacity of humans limits reality itself. This, however, seems thoroughly anthropocentric and parochial. One wants to say that the impossiblity of humans observing gods or neutrinos affects their existence (if they exist) no more than does the fact that spiders cannot observe them. Why should we be the ontological pivot of reality and not, say, dogs or mosquitoes? The view seems to favor humans in the most unecumenical way. Perhaps this is part of the doctrine's appeal, but it seems to me more a clue of its falsity than of anything else. These remarks are, of course, more rhetorical than anything else. Surprise that something is true is no argument that it is false. If it were, people who are surprised that they exist would have a reason to think they don't. Nevertheless, I think the initial implausibility of para-idealism warns us that some weighty reasons must be given for the doctrine before we can accept it.

        Let us look then at some of the reasons that have been given for para-idealism. Ignoring the religious arguments, I think three arguments have been thought to be persuasive. One stems from idealism itself. According to this argument, nothing can be attributed to what is unknowable since, as Perry put it, the unknowable "to which knowledge points or refers" is "always 'other' than the content of knowledge" and all qualities are annexed by knowledge. As a result, the unknowable reduces "to a bare X, entirely devoid of qualities and characters."5

        This argument apparently had wide appeal in the 19th Century. Despite its heritage, however, it is either a non sequitur or beside the point, depending on how it is interpreted. The most natural paraphrase of it is this:  an entity must have qualities to exist, an unknowable cannot have any qualities, thus an unknowable cannot exist. The second premiss is false, however. We attribute qualities to objects through our theories, but this does not mean that they have no qualities in themselves independent of our theories. To claim that something has a certain quality does not give it that quality; rather the object has its qualities in itself and, if we are lucky, we can discover what they are by hitting on the true theory of the object. Thus, the argument fails to show that an unknowable would have no qualities. The most it shows is that such objects, if there are any, have no knowable qualities, but this, of course, is granted. Indeed, that is what it means for them to be unknowable. Another reading of the argument would interpret it as holding that we cannot describe the qualities of the unknowable; hence, the question of its existence is otiose. This may be true, but it is irrelevant. To claim that the assertion of an unknowable is idle does not show that something unknowable cannot exist and this is what the para-idealist claims. As a result, the argument is either unsound or beside the point.

        A second influential argument stems from Wittgenstein who held in the Tractatus that "to understand a proposition means to know what would be the case if it is true."6 This is a very plausible view. As it is usually interpreted, however, the claim is that to understand a proposition is to know how to verify it. Thus, since by definition an unknowable cannot be verified, we cannot understand what it means to say that one exists. Schlick, for instance, takes this line. In commenting on the hypothesis that every electron has a nucleus that never manifests itself in observation, he asks "What would be the case if it didn't exist?" and answers that "everything would be the same as before."7 But this is a mistake. If such entities do exist, the world will be different from what it would be if they do not, even though by hypothesis we could never tell that they do. In other words, their existence would not affect our theory of the world, but this does not mean that the world our theory is about would be the same as it would be if they did not exist; it would only mean that our theory is false, and, if we accept further assumptions, e.g. that we cannot infer their existence in some way, could never be revised to make it true.

        The problem, I think, is that Wittgenstein's thesis can be interpreted in two ways:  as identifying meaning with truth conditions or as identifying it with observation conditions. The former is a highly plausible view, but it is quite different from Schlick's claim that every difference in meaning must manifest itself in observation. There is no reason to think, as Schlick does, that the conditions under which we can come to know a sentence to be true are the same as the conditions under which it is true unless, of course, we assume para-idealism to begin with. As a result, Wittgenstein's view does not provide a ground for the use of (3) against metaphysics.

        The third argument is less often articulated. According to this argument, language is a system of conventions created by humans for humans and has whatever limitations humans have. As a result, meaningful speech must be restricted to what is knowable since there would be no point in having possible sentences whose truth values are undiscoverable. As Locke put it:  "Words being voluntary signs, they cannot be voluntary signs imposed by him on things he knows not. That would be to make them signs of nothing, sounds without signification."8  A more modern form of the argument is this:  To learn a language is to learn the criteria for using signs correctly. Since there can be no criteria for what is unknowable, talk of an unknowable could not be learned and hence must be nonsense.

        This argument raises interesting problems that cannot be pursued here. Even if we accept its assumptions about language, I think it does not support para-idealism. The conclusion is that no meaningful sentence can express an unknowable fact, but, on the face of it, this is doubtful. The sentence 'There are beings who disappear without a trace when humans turn their way' is grammatical and seems to have a truth value even though, we may assume, such beings cannot be observed by humans. This, however, is not the main defect of the argument. Suppose that no sentence can express an unknowable. This will show that everything is knowable only if we assume that, if X exists, some possible sentence will depict X, and there seems no reason to believe this. Surely X, whatever it is, need not depend for its existence on whether we can have sentences about it. The standard positivist answer to this is that talk of an inexpressible fact seems to make sense only so long as we keep to the material mode. Once we put it into the formal mode, however, we can see  it to be meaningless. Unfortunately, this line begs the question, since the material-formal distinction is itself built on the rejection of unverifiable metaphysics as meaningless and this is just para-idealism in a different guise. The result, I think, is that the argument fails to support para-idealism.

        Each of these arguments for para-idealism thus seems defective. Where does this leave us? My suggestion is that we drop the use of (3) as a weapon in metaphysics and go a step beyond the new realists in emancipating metaphysics from epistemology. This does not mean that there is no unknowable. It only means that there is no reason to think that we can draw conclusions about what does not exist by examining the limits of our knowledge, if there are any. In short, I think we should follow Hume's advice and make "a fair confession of ignorance in subjects that exceed all human capacity"9 rather than conclude that they are meaningless or involve objects that cannot possibly exist.

                            FOOTNOTES

        1.The New Realism, E. B. Holt, et. al. (New York, 1912), p. 32.

        2.Ralph B. Perry, Present Philosophical Tendencies (New York, 1912), p. 312.

        3.Walter T. Marvin, An Introduction to Systematic Philosophy  (New York 1912), p. 419n.

        4.Carnap:  "In the thesis of the decidability of all questions, we agree with positivism as well as idealism," The Logical Structure of the World (Berkeley, California, 1969), p. 292. On Peirce, see his Collected Papers (Cambridge, Massachusetts, 1931-1958), 5.255; 5.257; 6.493.

        5.Perry, op. cit., pp. 311-312. See also Josiah Royce, The Spirit of Modern Philosophy (Boston, 1892), pp. 363-368.

        6.Ludwig Wittgenstein, Tractatus Logico-Philosophicus (London, 1961), 4.024.

        7.Moritz Schlick, "Positivism and Realism," in Logical Positivism, ed. A. J. Ayer (Glencoe, Illinois, 1959), pp. 88-89.

        8.An Essay Concerning Human Understanding, III, 2, 2.

        9.A Treatise of Human Nature, bk. I, pt. II, sect. V. I wish to thank Kenneth Stern for suggestions on an early version of this paper.