Edward N. Martin
«Am I crazy, or are there now only two beds in this room?»
Constable to Otis Driftwood
(A Night At the Opera, 1936)
Each of us makes assertions, frequently and without hesitation, which amount to truth-claims or knowledge-claims about our world. I am prepared to assert that I know, for example, that Bach is my favorite composer, that the mailman has delivered mail today, and that bulls have horns. As well as these, there exist a myriad of other propositions about which I am in a good position to say I know are true.
Sometimes my assertions claim that a certain state of affairs obtains or has obtained (for example, that Bach is someone's favorite composer). However, sometimes my assertions concern the contents of particular states of affairs (for example, that I received a long-awaited book in the mail today), and at other times the contents of general states of affairs (that some animals have horns). Perhaps we could distinguish these two types of states of affairs by defining them in the following way. Say that a particular state of affairs is one which may be described by a proposition which designates rigidly an event or happening contained in that state of affairs. A general state of affairs is one which cannot be rigidly designated in the way a particular state of affairs can.<63>Foot note 4_1
Now it is certainly the case that we sometimes err in what we say about the contents of particular states of affairs. But it is a much more tenable claim, it seems, that there must exist at least some things which form the substance, the very stuff of, general states of affairs. So it seems we are all in a good position to assert that
(1) Material objects exist.
G. E. Moore, in his book Some Main Problems of Philosophy, has recommended a version of anti-skepticism which attempts to show that we are in our so-called epistemic rights to say, with conviction and certainty, that we know that (1) is the case. In this paper I shall do two things. First, I shall examine Moore's main argument in favor of anti-skepticism. Second, I shall argue that there is an epistemic defeater which Moore's system is not equipped to defeat, and which either calls for Moore to relinquish the certainty of his knowledge of objects, to probabilify his knowledge, or to bolster his system so as to defeat the proposed epistemic defeater.
Sect.1. Moore the Anti-Skeptic
In his arguments against skepticism, Moore employs the following logical truth. Any two opposing arguments patterned after these forms are logically equivalent:
His employment of this logical principle can be seen in the following passage:
My argument is this: I do know that this pencil exists; therefore Hume's principles are false. My opponent's argument on the contrary is: Hume's principles are true; therefore you do not know that this pencil exists. And obviously in respect of the certainty with which the conclusion follows from the premiss, these two arguments are equally good.<64>Foot note 4_2
The missing premiss of these two enthymemes is: «If Hume's principles are true then I do not know that this pencil exists» [q]. Moore's claim is that he knows the pencil exists; the skeptic's claim goes the other way, as it were. He claims to know that Hume's principles are true and thus that whatever follows from such principles are true. This leads the skeptic to deny that Moore's pencil is known by anyone truly to exist. Moore, that is, has not met the grounds for knowledge which Hume's principles lay down. Clearly, both the Humean skeptic and Moore are using a valid argument form. Moore opts for the following form:
p. I do know that this pencil exists.
q. If Hume's principles are true, then I cannot know that this pencil exists.
r. It is not the case that Hume's principles are true.
The skeptic's argument takes the form: ~r; q; therefore, ~p, concluding that Moore cannot not know that this pencil exists. Deciding which argument is the best will come down to deciding which person has better evidence, grounds, or justification for his assertion of the first premise of his argument. A bit more on this point might be gained by making the following observations.
It is true that if Moore knows p, then Moore has grounds for his assertion that he knows that p. On the other hand, it seems reasonable to say that the skeptic knows that he has principles which, if not measured up to by any of Moore's knowledge-claims, entail that Moore does not really know what he asserts. So, the skeptic seems to be able to know something; Moore's skeptic is not a complete skeptic. He is merely one who is willing to doubt that Moore can know that some singled-out object really exists. We might say that the skeptic must accept this principle:
[AIK] (Assertion Implies Knowledge): If an individual A makes a propositional assertion p, then A claims to know that p.
If AIK is accepted, as it seems it should be, the skeptic also is making a knowledge-claim when he asserts that there is no pencil to which Moore points. Apparently, then, both Moore and the skeptic must have sufficient grounds for making their claims. It is at this point that Moore sees his way out. If the argument forms pitted one against the other are equally logically valid, then we must ask which one of (p) and (~r) is more probably true.
Which of (p) and (~r) look more likely to be the case? In behalf of Moore, (p) seems to have a bigger draw on its side. Two initial reasons could be adduced for (p)'s doxastically superior position. First, it seems self-evidently or nearly self-evidently true that a particular statement is epistemically easier to form, maintain, revise, and support with sufficient grounds than a general statement. Moore readily concurs on this point.
In fact any general principle to the effect that we can never know a particular kind of proposition, except under certain conditions, is and must be based upon an empirical induction .... it follows that no such general principle can have greater certainty than the particular instances upon the observation of which it is based.<65>Foot note 4_3
It's much easier to believe or find more probable that a particular pit bull, Bowser, is tenacious, than that all pit bulls are tenacious. (Though even here there may be fairly common exceptions. It's not necessarily the case when looking out over Wrigley Field in the top of the second inning that «the man on the mound exists» is more probably true than that «all the players on the field exist» or even «every one presently in my field of vision exists» are true. But I take it that these points subvert neither Moore's point nor the doxastically superior position of (p) over (~r).) Second, there is the indirect point that even the skeptic must not be committed to an all-out abandonment of knowledge. For he relies with certainty on his knowledge that his principles are what they are, and that they can properly be applied this way and that.
Now, there is one obvious rejoinder that the skeptic might make against our assertion that (p) is a more credible assertion than (~r). Namely, he could point out that when one compares what is said in (p) and (~r), one finds that different sorts of objects are referred to, and different claims about those objects are being made -- claims which are conceptually more complex in Moore's premiss (p) than in the skeptic's premiss (~r). And perhaps it is the complexity of (p) which ultimately makes the skeptic's claim that (~r) more tenable. For, in (p), the claim is made (by Moore) that
(p) I know that this pencil exists.
A pencil is an extended object, taking up space, whose existence entails that there exists external objects. To speak conclusively on the doxastic hopefulness of (p), one thus needs to have some account of the terms «exist» and «external world». What about (~r)? It claims that
(~r) Hume's principles are true.
(~r)'s commitments to the existence of objects is prima facie slimmer. This statement presupposes there being the (coherent) concepts of «principles» and «true». Perhaps Moore, then, has a more difficult position, thus making (p) less doxastically attractive? Of course, the point here is that it may be that Moore's having to explain the concepts he makes use of -- «exist» and «external world" -- will land him in more difficulty than the skeptic's having to explain the concepts of a principle and of the property of being «true» which some principles enjoy. But I would submit two considerations which pull back the doxastic edge to Moore's side. First, it seems to me that speaking of «exist» and «external world» are not at all any more conceptually problematic than is speaking of «true» (as the skeptic does). For, many propositions which, when thought of together lead one to form principles, are made true by the external world (or what we take to be an external world of which we are a part). Again, secondly, Hume's «principles» spoken of by the skeptic are forged in the same furnace. For principles, at least the Humean ones under consideration, are most definitely intended to say something true of the external world, viz. that no one is in his proper rights to claim knowledge about the world or any of its particulars outside himself.
Another important argument against skepticism of some merit is the argument from prior probability. Laurence BonJour among others has given an argument of this kind.<66>Foot note 4_4 Suppose we have two hypotheses which explain Moore's being appeared to in the pencil-like way: (1) Moore does see a pencil in front of him (`S'); (2) There is a malevolent demon named Rene who is massively deceiving Moore so that he seems to see a pencil in front of him (`D'). Since both hypotheses entail the seeing of the pencil (`P'), Bayesian analysis tells us that we are left to consider which is greater, the prior probability of S or the prior probability of D. The higher prior probability in this instance clearly falls, one might argue, on the side of S. Thus, Moore can overcome the skeptic's argument in this way.<67>Foot note 4_5
The few items I have considered tend to confirm Moore's assertion (p) over against the skeptic's assertion (~r). Up to this point we have undertaken to judge the merits of Moore's claim to knowledge when compared to the merits of the claims of the Humean skeptic. We have only examined Moore and knowledge largely derived from empirical observation. Of course, I think it's clear enough (from his Principia Ethica) that Moore holds that we also have a (largely) non-empirical type of knowledge, moral knowledge, to which we have access through our moral intuitions. An empirical state of affairs (say, Bowser's having bitten the mailman) may be the ground for making the judgment that being in pain on this occasion is intrinsically evil; but arguably the intuition itself which allows such a judgment or knowledge-claim to be formed and entertained is itself non-empirical. What I propose in the next section is primarily designed to disrupt Moore's theory of our knowledge of the external world. But it shouldn't be thought that my argument against Moore cannot also be applied to his claim that we can have moral knowledge as well. If my claim is right, in saying that there may be a defeater to Moore's certainty, which is rooted somewhat like a virus in one's belief-forming mechanisms, it will most definitely follow that all types of knowledge will be infected (for all knowledge has belief as a necessary requirement). But, it is good to remember that my main intent is to cast doubt upon Moore and his knowledge of the external world. That primarily, then, concerns empirical knowledge.
Sect.2. Sense-Data Data
There are a number of philosophical commitments to which Moore seems fondly attached but which weaken, so I shall argue, his anti-skepticism. The two most relevant of these are Moore's naturalism, and, his sense-data theory. I wish to show that it is the former which weakens Moore's anti-skeptical argument. However, we should dwell on his sense-data theory here to prepare for the last section. By so dwelling I want to make it clear that the epistemic defeater of Moore's system which I shall propose later in this section is even tougher to overcome by Moore given his committment to a sense-data theory. For a number of the knowledge claims, and the content of those claims, that Moore thinks he is entitled to make go far beyond (by way of content) what any «directly apprehended» sense-datum might tell a person.<68>Foot note 4_6
When a human agent is in the right circumstances, Moore holds that that person will come to certain truths about her world by inferring truths from her sense-data. In his essay «Material Things» Moore presents this theory about inferring certain truths from sense-data.<69>Foot note 4_7 We might call this theory the causal implication thesis. To illustrate the thesis, Moore employs an example of a group of people riding a train. The people in the train know that they are moving along the ground at a certain speed, that they are shaking back and forth during the ride, that the train is extended in space because it is a material object, and so forth. But they also know certain things because of the causal activity of the train. We infer certain things about the world from our sense-data. Moore says that it is reasonable to claim that we know that there is something, in the world, which causes our sense-data to be what they are. In his 1910 paper «Hume's Philosophy», Moore was not yet convinced that Hume's principles were very telling in regard to the limits of human knowledge. So Moore:
We may quite well know many things which do not logically follow from anything else which we know. And so ... we may know that two things are causally connected, although this does not logically follow from our past experience, nor yet from anything else that we know .... And as for Hume's argument to prove that we can never know any external object to be causally connected with anything which we actually observe, it is, I think, obviously fallacious.<70>Foot note 4_8 [his emphasis]
Moore's attitude toward Hume's argument changed during that year, for in his lectures (Winter 1910-1911) which make up SMPP, Moore says that Hume's argument is valid. We saw this in section one of this paper. What I add here, then, is the caveat that Moore only recently had adopted a new respect for Hume's argument. But, also, Moore for the first time in his Morley College lectures speaks of sense-data. This is where I think the causal implication thesis plays a significant role in Moore's new ideas of this time. For even if we grant that there is a material object which is not a mind and is extended in space and is not a sense-datum or a collection of sense-data, there remains the ultimate question: how do you know, Moore, that the thing causing your sense-datum is correctly (re)presented by the sense-datum? Moore writes:
I, for instance, claim to know that there does exist now, or did a moment ago, not only these sense-data which I am directly apprehending -- seeing or feeling -- but also something else which I am not directly apprehending. And I claim to know not merely that this something else is the cause of the sense-data which I am seeing or feeling: I claim to know that this cause is situated here; and though by here I do not necessarily mean in the space which I directly apprehend, yet I do mean in space -- somewhere in some space. And moreover I claim to know, not merely that the cause of my sensations is situated here in space, and has therefore some shape, but also roughly what its shape is.... It is, I think, plainly things like these that we all of us believe, when we believe in the existence of material objects.<71>Foot note 4_9 [his emphasis]
In all of these knowledge-claims, of course, there is really no doubt (in Moore's mind) that there is something in space which causes Moore to have the sense-data that he does. Going back to an earlier distinction I made between general and particular states of affairs, we need to grant to Moore that there must exist something in order for us to think that there are states of affairs which obtain and which contain the objects (or at least some of the objects) which we claim to know that they do. But this in effect turns the table on Moore.
We have seen that Moore favors moving from particular sense-data (cylindrical, yellow, solid, hard, etc.) to the existence of objects. But it seems much more probably true that
(1) Material objects exist
(2) This particular object exists.
Certainly the probability of the truth of a disjunctive set of knowledge-claims is much more probable than the truth of any particular knowledge-claim. It's much more probable, for example, that our school team win one of their games during the season than that they win any one particular game. And so also with our knowledge of material objects: we seem to have knowledge that they exist, but it's much less probable (pace Moore) that any particular one exists than that in general at least some material objects exist. This point is of no special consequence unless we judge that the probability of the particular claim that Moore makes (say, that the pencil exists) is far from being near 1 (where «1» means certainty). In other words, Moore could grant that it's more probably true that there exist many material things than that one particular ostensively referenced thing exists. But why think that that putatively referred to thing is not known to exist? The best way to show that Moore is not entitled to say that he knows that, say, his pencil exists is to show that a human belief-producing mechanism is not itself known to be reliable.<72>Foot note 4_10 Perhaps we can find a defeater to Moore's claim that he knows that the pencil exists by looking in this direction.
Sect.3. An Epistemic Defeater?
There is a certain defeater to Moore's claims to knowledge which his system is not able to defeat. Moore was himself a naturalist.<73>Foot note 4_11 This implies that humankind, according to the best going naturalistic theories of Moore's day and ours, arrived on the scene after a period of millions of years of evolutionary development. Humans, as we know, have belief-forming mechanisms which have as one purpose or function that of producing mostly true beliefs. But what of these mechanisms? How reliable can we take them to be if they have been produced over a long period of time by blind forces of chance, time, and natural selection? Darwin himself expressed doubt at this point.
With me the horrid doubt always arises whether the convictions of man's mind, which have been developed from the mind of the lower animals, are of any value or at all trustworthy. Would any one trust in the convictions of a monkey's mind, if there are any convictions in such a mind?<74>Foot note 4_12
Patricia Churchland and Alvin Plantinga have recently expressed similar doubts.<75>Foot note 4_13 Quine demurs from these doubts. He writes:
There is some encouragement in Darwin. If people's innate spacing of qualities is a gene-linked trait, then the spacing that has made for the most successful inductions will have tended to predominate through natural selection. Creatures inveterately wrong in their inductions have a pathetic but praiseworthy tendency to die before reproducing their kind.<76>Foot note 4_14
These quotations seem to pose two questions. First, which one of these sentiments is more nearly correct? And second, what value is this discussion for Moore's project? Let me address each of these in turn.
First, Quine's sentiment does not sit well with all of modern evolutionary theory. A bad trait (say, inferior hearing) may not be eliminated and may be indefinitely perpetuated by its being linked with a good trait (say, overall sense of balance). A gene can carry the code of more than one trait by pleiotropy, «where one gene codes for more than one trait or system».<77>Foot note 4_15 This means that the organism may never achieve optimal genetic coding, and this may or may not be telling for the reliability of the organism's belief-producing mechanisms.
Our belief-producing mechanisms are very complicated configurations many aspects of which do weigh considerably on the issue of human knowledge. Not only perception and the propositional content which comes from perception, but also memory and reason and its contents, play important roles in perceptual knowledge. But why think that the naturalistic hypothesis is ill-equipped to produce reliable belief mechanisms in humans?
It's certainly the case that an animal species's survival is the first concern of natural selection; thus, whether an animal's capability to picturing the world aright is reliable is not completely at center stage here. If it were reasonable to suppose that an organism best equipped for survival would be very much like an organism equipped with mechanisms which produced true beliefs about the world in which it lived, then Moore would be vindicated from the charge that I've leveled against him. But I don't see that an organism set on survival must necessarily have, as a sort of concomitant property, a reliable belief-forming mechanism (one which was able to reliably determine which things existed and which ones did not in one's perceptual field). For example, an animal's belief-producing mechanisms may quite often alert it to «dangers» which are not really dangers of any kind. Many times for us a «sensed» danger amounts to an imagined presence or an «I thought I saw something». As the Proverb says, «the wicked man flees though no one pursues».<78>Foot note 4_16 Or, if an animal is to survive, it must display some sort of danger-avoidance behavior. Why think that the animal's beliefs about the most appropriate danger-avoidance behavior in some circumstance are true beliefs about the world? If bad cognitively-related traits are passed on by pleiotropy, our belief-producing mechanisms, if belief is causally related to our behavior, would be maladaptive and would tend to work unreliably in some set of circumstances. Unfortunately, this means that we would not know which beliefs we entertained were in fact false beliefs. In favor of avoiding tooth and fang, man's cognitive abilities would be no guarantee for anything but survival. And survival is different from truth. We may trade on Thrasymachus's view of justice in this context: might may make right, but why truth and proper belief?
This, then, is where Moore's project is pertinent. We can construct an argument which is similar in logical form to the Humean skeptic's argument we examined in section I. It can be cast in this way: (a) Naturalism & Evolution are not sufficient to establish or guarantee that our belief-producing mechanisms are reliable. (b) Therefore, we don't know that the pencil which seems to be before us really exists. And, since A's believing that p is necessary for A's knowing that p, A would not know that p if A arrived on the scene by merely naturalistic means. There would be a defeater, D, which could not itself be defeated by Moore's ostensive demonstration or appeal to his senses.
[D] Moore cannot be certain that his belief-forming mechanisms are currently producing true beliefs.
So, Moore cannot say of any particular object that he knows that it exists. Of course, this is using «know» in a certain strict way, which I cannot develop here, but which I submit meets Moore's criteria for perceptual knowledge of the type with which he is concerned. (See Appendix)
Moore might make the following initial response. If his system had an undefeated defeater, then the defeater D itself will have arrived on the scene through the same process as the other beliefs that Moore entertains; perhaps it itself, then, is unsubstantiated. But that doubt, which defeats the defeater, is itself defeated, on account of premise (a), that our beliefs are unreliable because of our evolutionary ascent. Moore seems plagued by the difficulty of having a defeater which, though challenged, is never undercut and thus never quelled in his system.
Perhaps a better response open to Moore is the following. William Alston has in a series of important papers made use of the concept of level confusions.<79>Foot note 4_17 One would be committing a «level confusion» if he thought that being justified in accepting some claim C amounted to the same thing as showing that one was justified in accepting C. Cannot Moore just rest content that he is justified in his assertion that the pencil before him exists, without worrying over whether he can show that he is in fact justified (or that he knows)? I don't think so. What I am questioning is not simply whether Moore is justified in his assertion. I think that he is justified. I am rather questioning his theory's explanation of the source or origin of his (our) epistemic equipment, and asking whether the mode of construction of that equipment is sufficient for epistemic and doxastic success. In one sense we are asking Moore to show us something. But that which we want to be shown or addressed is not whether he is justified in his claim per se, but whether he qua naturalist is justified in his claim. For it follows that if he were justified in making his claim, but was not justified qua naturalist, then his theory would be incomplete or lacking something given its commitment to naturalism as an explanation of the origin and source of our doxastic mechanisms.
It is difficult to see what Moore would say at this juncture. What my argument points out, I think, is that there is a way to turn Moore's project on its head. He has claimed that the probability of the existence of a particular object (a pencil) is higher than any generalization from particular observations (Hume's principles). It appears that the naturalistic hypothesis supports the idea that we can know that material objects do exist, but that any particular knowledge-claim is subject to doubt because of its being produced by a perhaps unreliable belief-producing mechanism. That we doubt is enough to establish that we exist; however, the exact status and real existence of objects of perception can only be probabilistically known. No certainty concerning the existence of material objects besides a subjective kind is produced by immediate perception of the objects. Moore, I claim, is reduced to probable knowledge or to a call to bolster his system. If he were to claim that his belief-producing mechanism is reliable (and it probably is), then we either have to say that the naturalistic hypothesis is sufficient to guarantee that humans have reliable epistemic systems, or, that there are some other grounds for why our epistemic systems are reliable. I believe Moore must take the second option, if he is to get out of the undefeated-defeater circle. Moore must call upon some other grounds to fortify his claims to knowledge. I leave it to the reader to investigate what these grounds might consist of. Like Gaunilo of old, who agreed with Anselm's belief but rejected his proof, I accept Moore's conclusion: he does know, when he waves, that his hand exists. But I've argued that if survival is the only force that drives the ascent of the human organism, the reliability of the human belief mechanisms is not at all probable to be present. But if we find that they are reliable, something must account for that reliability. I conclude that Moore's system is inadequate and in need of repair or additional explanatory elements.
SMPP contains a paper entitled «Ways of Knowing» in which he explores four different types of knowledge. One type of knowledge which Moore calls «knowledge proper» sounds very much like our concept of knowledge after Gettier and his suspicion's about knowledge as `justified true belief'. Moore's treatment of knowledge proper is more or less captured in the following definition of knowledge which I want to adopt for my purposes in section II:
[K] A human person knows p iff:
(i) A believes p;
(ii) A has grounds sufficient to justify his claim to know p;
(iii) No defeater d consistent with A's other beliefs defeats his grounds for p;
(iv) p is true.
In «Ways of Knowing», Moore says that in regard to material objects, we cannot say that we can know them with knowledge proper. But the reason for this is slim: he says that «knowledge proper is a relation which you can only have to a proposition; and a material object is certainly not a proposition».<80>Foot note 4_18 What is the relationship between a material object of which some human agent is conscious and a proposition? Can we not see a connection between a perceived object and a proposition in this way: the proposition only goes to linguistically depict what is true of the vision or the «seeing» of the object? If Moore wants an ostensive reference to a pencil to count as proof of the real and true existence of material objects outside of a perceiving subject, then it seems reasonable to adopt his criteria for knowledge proper in relation to some subject S's knowing the existence of an object O. All that is needed is for S to have some belief mechanism M such that M instills in S the belief that some proposition is true, namely, that O, where O is the propositional counterpart to a state of affairs which is present to S and in which S perceives that there exists a material object. I see no reason not to adopt this type of Moorean knowledge with regard to material objects, since perceptions of material objects are subject to direct translation (by M, say) into S's propositional beliefs.
Key here, of course, is that a subject S has or has had a conscious awareness of an object O, and that at that time M must have a propositional attitude in order to say that he knows that O exists, whether he says it to himself or to another. So, for my adoption of Moore's `knowledge proper' to fail, one of two things must fail: (1) in S's inner dialogue, M forms the proposition that S knows p iff the contents of p seem to S to be part of his perceptual field; and (2) M produces a propositional attitude in S simultaneous to S's knowing that p. It appears to me that both (1) and (2) are true. [K], my tentative definition of knowledge, is amenable to Moore's use of «know» in his statement «I know that this pencil exists» in the ways that I have shown. I adopt, then, [K] in the last section of the paper.
I would like to thank Jan Cover and William Rowe for helpful commments on an earlier draft of this paper. Edward N. Martin
4721 Timber Lane, #1110
Evansville, IN 47715