Lecture Notes, Lehrer's Theory of Knowledge, Chapter 7, Coherence, Truth, and Undefeated Justification

2002 Lecture Notes, Lehrer's Theory of Knowledge, second edition

Chapter 7, Coherence, Truth, and Justification

By G. J. Mattey, Senior Lecturer, UC Davis Philosophy Department

The account of personal justification given in Chapter 6 is fallibilist. Coherence with an evaluation system is not a guarantee of truth because the evaluation system itself may be in error. In the present chapter, Lehrer gives an account of undefeated justification, which he regards as a necessary condition for knowledge. In general, personal justification for accepting that p is undefeated when it holds up when the evaluation system is corrected for error.

Massive Error

Lehrer first takes up the possibility that a person's acceptance system is massively in error. It may be that we should never attribute massive error to a person. If so, it might be a consequence that personal justification is automatically undefeated justification. But, "Personal justification does not automatically convert to undefeated justification as a result of the necessity of interpreting most of the acceptance system a person has is true, for there is no such necessity" (p. 152). Lehrer argues against the necessity of denying that one's acceptance system is massively in error, but it seems that he need not have taken the trouble to do so. For even if the evaluation system on which it is based it is only moderately in error, there is still plenty of opportunity for a given justification to be defeated.

As Lehrer points out, taking our own acceptance system to be mostly correct is "congenial to the account of justification" he offers (p. 152). This is tantamount to accepting our own trustworthiness, which is a central feature of personal justification as he understands it. But the basis for this acceptance is to be found in what one accepts about one's "track record," or perhaps in the principle of trustworthiness itself.

The standard argument given to support the denial of the possibility of massive error has nothing to do with a previous record of success or trust in one's self. The idea is that when we try to understand what someone believes, be it our own self or another person, we should attribute beliefs in a charitable way. One is most charitable when one describes most of those beliefs as being true. Donald Davidson urged adoption of the "principle of charity" when we attribute beliefs to others who speak a language entirely new to us. We try to interpret their use of language from their behavior, and the only way we can get an acceptable translation is if we assume that the behavior reflects beliefs that we share with them.

Lehrer claims that the principle of charity "does not itself appear to be true" (p. 152). The reason is that charity does not require the attribution of beliefs that we think are true. It may be "doxastically imperialistic" to do so, as Lehrer had argued in earlier chapters. In effect, we are not permitting the attibutee to have beliefs that deviate from our own, which deprives them of their autonomy in believing that they will. Moreover, Lehrer gives an example where there are differences in linguistic practice (for two speakers of the same language) that are so great that it would be "absurd" to attribute one's own beliefs to the other. This is a case where radically different metaphysical views, nominalism and realism, are reflected in the two people's use of a language they both understand.

Irrefutable Justification

In preparation for the discussion of undefeated justification, Lehrer says that the "intuitive idea" is that one's personal justification "cannot be refuted by any error on the part of the subject who has the justification" (p. 152). For the acceptance that p to be irrefutably justified, there must be no counter-argument to p that appeals to error in the acceptance system. Personal justification need not be irrefutable, Lehrer says. One might be personally justified in accepting that the Alps were created a few thousand years ago. The justification might be based on extensive observation in conjunction with a belief that God created the universe a few thousand years ago, though in a way that would make the time of its creation seem much earlier. If this belief about God's creation is false, then the person's personal justification is refuted. Refuted justification is equivalent to defeated justification.

Undefeated Justification

Since refutation is the result of error, irrefutable justification should be based on a system free from error. This includes not only the acceptance system but the other parts of the evaluation system as well. The ultrasystem is the result of removing the errors from the acceptance system, the preference system, and the reasoning system. It also includes a listing or "log" of what was removed from the system. So the ultrasystem has these four features:

The first three features make up what Lehrer later in the chapter calls the "t-system," which is just the evaluation system minus all error. The fourth feature, which preserves remnants of the erroneous parts of the evaluation system, is exploited later. Irrefutable justification is justification based on the ultrasystem, though for the time being we can take it to be justification in the t-system, which is a necessary condition for irrefutable justification.

As with personal justification, irrefutable justification can be illustrated by a game involving a claimant and a critic. This time, the critic is an "ultracritic," who is allowed a move not permitted the critic in the original justification game. The ultracritic can demand the removal of instances of any of the three kinds of error from the original evaluation system. Then the claimant has to answer objections based on the reduced system.

To illustrate how the game works, let us consider one of the Gettier cases discussed in Chapter 1. (Gettier cases generally purport to show that a person can have a justified true acceptance which falls short of knowledge.) There is a teacher in whose class sit Mr. Nogot and Mr. Havit, the former appearing to own a Ferrari but not owning one, the latter owning a Ferrari but not appearing to own one. The teacher fails to know that someone in her class owns a Ferrari, because (in this first version of the case) her justification is the result of reasoning from a false premise and thus depends on a false statement, i.e., that Mr. Nogot owns a Ferrari. This dependence on a false statement makes it in a way accidental that the teacher got it right: she stumbled onto the truth because her acceptance system was tied to the truth in a defective way. We could say that her justification is refutable.

Now let us look at the teacher's ultrasystem. Suppose that she accepts that Mr. Nogot owns a Ferrari, and her acceptance of this statement is required for the justification of her acceptance that someone in her class owns a Ferrari. (She has no evidence that nobody else in the class owns a Ferrari.) If the acceptance of Nogot's Ferrari ownership were to be removed from her acceptance system because of its falsehood, the teacher would lose the original justification. When asked, "How do you know that someone in your class owns a Ferrari?" she would be left without her original answer that she accepts that Mr. Nogot owns one, leaving her with no answer at all.

In terms of the ultra justification game, the teacher would bring forward her acceptance that Nogot owns a Ferrari as part of an answer to an objection that "None of the students in your class who appear to own Ferraris actually own Ferraris" (p. 155). The ultracritic can then reply, "Eliminate your claim that Mr. Nogot owns a Ferrari! This is false!" (p. 155). So, Lehrer says, the claimant loses the round, and because she must win all rounds of the ultra justification game to be irrefutably justified, her justification is refuted.

Let us look at this example, as presented by Lehrer, in more detail. The original claim is that "Someone in my class owns a Ferrari." The objection, as we have seen, is that "None of the students in your class who appear to own Ferraris actually own Ferraris." So the teacher should respond as follows: It is more reasonable for me to accept that someone in my class owns a Ferrari than for me to accept that none of the students in my class who appear to own Ferraris actually own Ferraris. (The text gives a different response.) Among the members of the teacher's acceptance system that might back this up are the pieces of evidence that give the strong appearance that Nogot owns a Ferrari. Lehrer adds to this the further claim that Nogot owns a Ferrari.

Now the critic commands the teacher to eliminate her acceptance that Nogot owns a Ferrari. Does this really defeat the teacher's justification? Even with this particular acceptance eliminated, the teacher has all the evidence that give the strong appearance that Nogot owns a Ferrari. This is what was needed, we saw, to handle the specific objection raised by the ultracritic. On the other hand, if the teacher were to accept that Nogot does not own a Ferrari, then the objection could not be answered. In the earlier versions of the text, Lehrer allowed this conversion, but he dropped it because it led to certain untoward consequences. We are pulled toward thinking that the objection could not be answered either because we implicitly change the false acceptance to a true one, or else we think that the justification would collapse without the false acceptance. But the mechanism here does not force the justification to collapse, at least in this round of the ultra justification game.

One reason to think the justification would collapse is that in the original example, the teacher uses the acceptance that Nogot owns a Ferrari as a premise in her reasoning to the conclusion that someone in her class owns a Ferrari. So she would not have accepted the conclusion had she not accepted the premise. But the test posed by the justification game is whether she can answer objections. Now suppose that the ultracritic could find an objection the teacher could not answer. This does not block another version of the case, in which the teacher reasons directly from her evidence about Nogot's Ferrari ownership to the conclusion that someone in her class owns a Ferrari, an observation Lehrer attributes to Gettier. Lehrer does not mention why the teacher might bypass the intermediate conclusion that Nogot owns a Ferrari. In his earlier book, Knowledge, a motivation is given.

If the teacher is a clever man and is only interested in the question of whether there is at least one Ferrari owner among his students, he might reason that, though his only evidence of a Ferrari owner among his students is what he knows about Mr. Nogot and a certain car, there is at least the possibility that someone else owns one, and, hence, it is safer to accept the more general statement that at least one person in his class owns a Ferrari than the quite specific claim that Mr. Nogot owns one. (pp. 19-20)
Such clever reasoning would, it appears, enable the claimant to win the ultra justification game, because her evaluation system contains no error. "This confirms Gettier's basic contention that justified true belief not based on any false acceptance may still fall short of knowledge" (p. 156).

This teacher has the resources to answer the relevant objection: "None of the students in your class who appear to own Ferraris actually own Ferraris" (p. 155). As before, the objection can be answered because, from the standpoint of the teacher's acceptance system, "It is more reasonable for me to accept that at least one student in my class who appears to own a Ferrari does own a Ferrari than to accept that none of the students in my class who appear to own Ferraris actually own Ferraris" (p. 156). (Note that this is not quite in good form, since it is p, "Someone in my class owns a Ferrari," which must be more reasonable than the objection.) The objection would be answered by S with the same reasons as before, i.e., "Mr. Nogot is a student in my class. He has told me that he owns a Ferrari, has shown me papers stating that he owns a Ferrari, and he drives a Ferrari" (p. 156). Lehrer's solution for this case (and the original case, which was not treated correctly) will be given later, after a discussion of other proposals for handling the Gettier problem.

Another case that does not require reasoning through a false statement is Chisholm's sheep case, which has already been discussed. A person may accept that p on the basis of perception alone, that is, on the basis of how something looks to him. So it will not do simply to ban error from one's reasonings, any more than it will do to ban them only from one's acceptances.

Strong Irrefutability

If we want to say that the teacher's justification is not based on any error, there are many ways to do so. The most extreme demand would be that the evaluation system contain no error at all, that it meet Descartes's ideal. But then hardly anything would be justified. Nor should we say that the evaluation system may not allow the justification of anything false. "But it may well be that whatever justifies us in accepting anything incidentally justifies us in accepting at least some false statements" (p. 157). Another proposal is that one's justification is based on error when the justification contains a false statement. We call any of these conditions "strong irrefutability," since they are intolerant of falsehood in a very general way, making no reference to how p is actually supported by the evidence.

The last proposal is the most liberal, as it allows for a great deal of justification, but it is still too strong a condition on knowledge, because sometimes the error in our evaluation system is "harmless." To see this, expand the case of the teacher so that she also has evidence to accept that Mr. Knewit in her class owns a Ferrari: she sold him a Ferrari and he still owns it. Then even though she is justified in accepting the false statement that Mr. Nogot owns a Ferrari, she still knows that someone in her class owns a Ferrari. This is because she has a line of justification running through her information about Mr. Knewit in addition to that running through her information about Mr. Nogot. Her justification has redundancy built into it, and the fact that one channel is blocked is not sufficient to close off all lines of justification. The justification contained error, but the error was not essential to the justification.

External Truths and Falsehoods: The Grabit Case

A potentially simpler approach is to consider not just how the evaluation system reflects the world, but all the facts about the world, as relevant to whether justification is defeated. There are errors of omission as well as commission. In the second, more difficult, case of the teacher, it is false that Nogot owns a Ferrari. This is an "external" falsehood, in the sense that the teacher does not accept it, and so it has no place in her acceptance system. On the other hand, the falsehood of the statement that Nogot owns a Ferrari is an objection to the teacher's acceptance that someone in her class owns a Ferrari. It would be less reasonable for her to accept the latter if she were to accept that Nogot does not own a Ferrari than if she were not to accept that Nogot does not own a Ferrari. However, this objection is answered in her evaluation system, based on the evidence she actually accepts.

So we might be inclined to say that if an objection answered in the original evaluation system is nonetheless a truth about the world, it is relevant to whether one knows. More precisely, for S to know that p, there can be no true statement q such that if S were to accept that q, S would not be justified in accepting that p. (This is the proposal of Klein and Hilpinen, stated on p. 158, slightly modified here fore ease of understanding. I will speak of accepting the truth of true statements rather than accepting the falsehood of false statements, as I will take these two to be equivalent.) Let us call this an external dependence on a false statement. The truth that Nogot does not own a Ferrari is equivalent to the falsehood of the statement that Nogot owns a Ferrari, but neither is something the teacher accepts. The teacher needs the falsehood for her justification, but it is not available in the real world, and if this external fact were added to her acceptance system, her justification would collapse.

Lehrer rejects this external account of dependence, on the grounds that it is too strong. That is, if the condition holds, then sometimes it will block the attribution of knowledge in cases where a person clearly knows. He contrives the case of Tom Grabit to illustrate his point. As the story goes, I am in the library and observe Tom Grabit taking a book. I accept that Tom Grabit took a book from the library and am personally justified in my acceptance, since my acceptance system contains the acceptance that I saw the event at close range, that I am acquainted with Tom Grabit, that I can distinguish cases of stealing from other behavior, etc. A competitor would be the proposition that Tom Grabit has an identical twin brother who is in town on the day I observe the event. It competes because accepting it makes it less reasonable to accept that I saw Tom steal the book (since I might have seen his twin instead) than it would be if I did not accept that Tom has an identical twin brother in town today.

The competitor is beaten because there is no reason for me to think that Tom has an identical twin: I have never seen one or heard of one. Now consider the statement that Tom's father today has spoken the sentence, "Tom's identical twin John is in town today." This, too, is a competitor, since it is good reason to think that Tom has a twin in town today. Again the competitor is beaten in my evaluation system, since I have no reason to believe that Tom's father said such a thing.

Additional facts of the case are these: Tom has no identical twin, but Tom's father uttered the sentence just mentioned. He has done so because he has a mental condition which causes him to fantasize the existence of John Grabit. Lehrer would say that under these conditions, I know that Tom Grabit took the book from the library today.

On the other hand, he thinks that the external account of dependence on a false statement does not allow me knowledge. The reason is that one of the objections to p, that Tom's father uttered the sentence he did, is true. If I were to accept this truth, my justification would be lost. But the truth is misleading, and its existence should not block my being said to know. So, Lehrer concludes, the external condition is too strong.

But Lehrer has not given the external condition its due. For if I were to accept not only that Tom's father said what he did, but also that he was uttering a falsehood, then my justification would not be touched. What has happened is that there is an external fact that helps "answer" the true objection in the extended domain of all truths. A systematic way of developing this response has been given by Pollock, in the first edition of Contemporary Theories of Knowledge. It is discussed below, after we have finished examining Lehrer's account of irrefutable justification.

Evidence One Does Not Possess: the Newspaper Case

In the meantime, Lehrer wants to distinguish his case from another one, due to Gilbert Harman. Harman's "newspaper" case is one in which it seems that external information can successfully prevent the attribution of knowledge. This case concerns Mrs. Readlucky, who reads in an early edition of a newspaper that a civil-rights leader has been assassinated (p). The paper and its reporter, who witnessed the death, are credible enough that the reader is justified in accepting that p. She does accept it, and it is, alas, true.

If this were all the information we had about the case, most of us would be willing to attribute knowledge to the reader. (Harman reports that a small percentage of people would not, even without the further wrinkles about to be added to the case. See "Reasoning and Evidence One Does Not Possess," Midwest Studies in Philosophy V, pp. 163-182, footnote 7.) But suppose that there was a conspiracy to cover up the killing, with the result that earlier editions of the newspaper are withdrawn from the stands and replaced with modified editions denying the death of the leader. All the original eye-witnesses except the reporter disavow the fact of the killing. Suppose further that people all around the reader are confused about what happened, though the reader is unaware of their confusion or of the earlier retractions.

Harman reports that about 10% of those hearing this case are still willing to attribute knowledge to the reader, while 40-50% would deny knowledge (the remaining 40-50% being unsure). He takes this 4-1 or 5-1 ratio to suggest strongly that something is blocking S's knowing. The lesson he draws is that sometimes evidence one does not possess can block knowledge, when it is either readily accessible or possessed by others in a relevant social group. In the case we have described, it is the possession of the information that the leader was not killed by the people around the reader that is relevant. Harman adds that this shows that knowing has a "social dimension."

Evidence one does not possess is the kind of consideration we have been calling "external." The denials issued by the newspaper are analogous to the claim made by Tom Grabit's father. The key difference between the two cases is that that denials cause widespread confusion, and thus are "social" in character, while nobody listens to Tom's father. But they have in common the fact that there is misleading evidence that is not taken into account in the original justification. The fact that someone has not taken into account misleading evidence is not accommodated directly by the account of irrefutable justification. In the ultra justification game, the skeptic may raise the objection that the claimant has overlooked negative relevant evidence. But this objection might be answered by appeal to one's due diligence in evaluating the evidence.

What is needed is something in the evaluation system that is false. Harman has described cases in which someone does not know because he has overlooked evidence that is very ready to hand. This is evidence that the person should have taken into account but did not. Lehrer could say that an objection to any claim to knowledge is that one has overlooked relevant negative evidence. And one would then have to answer the objection, appealing to his evaluation system to show that indeed he has taken all the relevant negative evidence into account. If the answer does not hold up in the ultrasystem, so that an ultracritic in the ultra justification game can say: "You overlooked relevant negative evidence!" a person's justification can be defeated, as in the newspaper case.

Anyone following this line would be left with the problem of defining relevant negative evidence. But this proposal has a bigger drawback, in that it blocks knowledge in the Grabit case, where relevant negative evidence has been overlooked. To fix this problem, one might wish to require only that relevant non-misleading negative evidence not be overlooked. But the overlooked evidence in the newspaper case is misleading, so knowledge would not be blocked by this kind of generic requirement. It is very tricky to specify a "no overlooked evidence" requirement in a way that blocks knowledge in the newspaper case but allows it in the Grabit case.

Lehrer does not proceed along these lines, but is more specific. "In the newspaper example, though this is unstated, part of what justifies the person in accepting that the civil rights leader has been assassinated is her belief that the newspaper is a trustworthy source of reliable eyewitness reports about the assassination" (p. 160). Since this is false, her evaluation system is in error in a way that makes her justification refutable. This solution may strike the reader as ad hoc, i.e., as a requirement placed on Ms. Readlucky's justification just to block her knowledge in this case.

It is not difficult to modify the case so that Ms. Readlucky's justification does not depend on any false acceptance. For example, suppose that she places her trust in the reporter who gave the initial, true, eyewitness account, rather than in the newspaper itself. Suppose further that the reporter refused to go along with the coverup and refused to allow his name to be placed on the false story in the later editions. In the first edition of Theory of Knowledge, Lehrer was willing to swallow the consequences and say that the reader can be said to know, so long as she has sufficient confidence in the reporter and resolves "not to let the doubts of others shake her confidence." Here I am in agreement with what Lehrer at least used to think. My view is that the fact that others around one are misled should not impugn the knowledge of someone who has trusted a trustworthy source. Thus I think the "social dimension" of knowledge allegedly revealed by this example is overstated.

Errors in Preferences and Reasonings

Having rejected several approaches to explaining how someone's justification can be refuted without his accepting a false statement, Lehrer makes a proposal of his own. Error can occur in an evaluation system not only in what one accepts, but also in one's preferences and reasonings. These errors should block justification just as should errors in acceptance. So the ultra justification game allows that the ultracritic can throw out false preferences and reasonings, just as he can throw out false acceptances.

This takes care of both problems raised in the application of the justification game to the case of the teacher and the Ferrari-owning student. Whether or not the teacher accepts the falsehood that Nogot owns a Ferrari, the teacher must at least prefer accepting that Nogot owns a Ferrari to accepting that Nogot does not own a Ferrari. Without this, she would lose her reason for accepting that someone in her class owns a Ferrari, since she has no evidence that anyone else does. But this preference can be blocked by the ultracritic: "You must eliminate your preference for accepting that Nogot owns a Ferrari to accepting that Nogot does not own a Ferrari! (It is false that Nogot owns a Ferrari!)" (p. 161).

This solution seems to work effectively to refute the teacher's justification. Lehrer proposes a second way of doing this, which does not work well at all. (Note that the last paragraph on p. 161 lacks continuity with what comes just before it.) Just as he tried to finesse Harman's example by appealing to an unstated belief, Lehrer attributes an unstated acceptance to the teacher.

Given our account of acceptance as a mental state having a certain functional role in inference, we may say that the claimant accepts the hypothetical to the effect that if the statements of evidence are true, then Nogot owns a Ferrari, even if, for some reason, she does not accept the conclusion that Nogot owns a Ferrari. The reason for ascribing acceptance of the hypothetical to the claimant is that the inference from the evidence to the conclusion that someone in her class owns a Ferrari rests on the acceptance of the hypothetical linking the evidence to that conclusion. (p. 161)
This acceptance is false, since the evidence about the ownership is true (i.e., it is true that Nogot drives a Ferrari, says he owns one, etc.) but it is false that he owns a Ferrari. The ultracritic demands that the false hypothetical be eliminated from the teacher's acceptance system. Then the teacher is not irrefutably justified, since now she has no basis for moving from the premises about Nogot to the conclusion that someone in her class owns a Ferrari.

The reason I say this does not work very well is that the teacher need not accept this hypothetical. He need only accept the hypothetical that if the statements of evidence are true, then someone in my class owns a Ferrari. This is what links the evidence to that actual "conclusion that someone in her class owns a Ferrari." It would make little sense for the teacher to accept the hypothetical whose conclusion is that Nogot owns a Ferrari without accepting that Nogot does own a Ferrari.

The first, and better, solution to the second teacher case is new in the second edition of the text. The second solution, though modified in the second edition, is the only proposal Lehrer made in the first edition. We may regard it as a harmless relic from the earlier version of the text.

Keeping Sight of the Original Evaluation System

Lehrer in the second edition of the text brings up an apparent problem with exclusive reliance on the t-system, which simply eliminates all error from the evaluation system. He says that it must be supplemented to avoid an unacceptable consequence. In the original version of the teacher case, for example, the teacher falsely accepts that Nogot owns a Ferrari, and that acceptance is purged from the t-system. Then, it seems, the teacher would not be ultra justified in accepting that she accepts that Nogot owns a Ferrari.

This apparent problem is patched by adding to the t-system a description of all the things one accepts, prefers and reasons. So a statement of the fact that the teacher accepts that Nogot owns a Ferrari is an element of the ultra system. This will allow knowledge of this acceptance. But when he later develops the formal definition of the ultra system, Lehrer sets it up so that the content of false acceptances cannot be used in answering objections. The teacher cannot take it as true that "Nogot owns a Ferrari," though she can take it as true that "I accept that Nogot owns a Ferrari."

One might wonder whether this addition is necessary. If the teacher is to know that she accepts that Nogot owns a Ferrari, she must accept that she accepts that Nogot owns a Ferrari. This higher-order acceptance is a true acceptance, so it will have a place in the t-system. But Lehrer puts the device to another use which may be more crucial to his solution to the Gettier problem.

He considers the possibility that allowing this additional information into the ultra system would result in a bias toward the claimant. This does not seem plausible on the face of it, since the information consists only in truths about acceptances whose contents cannot be detached. The fact that the teacher accepts that Nogot owns a Ferrari is just information about what she accepts, not about the truth of her acceptance. At any rate, Lehrer goes on to show how in fact false acceptances may work against the claimant, as illustrated in an elaborate variation of the "Knewit" case.

The Knewit case is that in which the teacher has evidence that Knewit, who is in her class, owns a Ferrari. Her evidence is that she sold Knewit a Ferrari. But here, she also accepts that Knewit in turn sold the car to someone else and does not own the Ferrari she sold him. In reality, though, he still owns it. The teacher accepts nontheless (and for no good reason) that Knewit owns a Ferrari. Here we have a failure of knowledge, as can easily be seen by imagining moves in the ultra justification game.

The reduction of the evaluation system to the t-system might seem to clear the way for the teacher to be able to be ultra justified, however. She accepts that Knewit owns a Ferrari and she has the evidence. But her ultrasystem contains the information that she accepts that Knewit does not own the Ferrari she sold him. This, Lehrer claims, blocks the justification.

We are not told how this is supposed to work. The ultracritic might raise as an objection, "Knewit does not own the Ferrari you sold him." We might be tempted to have the ultracritic add, "And this is something you accept!" There is no mechanism for this kind of move, however. The teacher would not use her acceptance that she does not accept that Knewit does not own the Ferrari she sold him to answer an objection, because she does not accept it in the first place.

The ultracritic would have to raise it as an objection that the teacher accepts that Knewit does not own the Ferrari she sold him. Accepting that she accepts this would, presumably, make it less reasonable for her to accept that Knewit owns a Ferrari than if she were to accept that she has not accepted that Knewit does not own the Ferrari she sold him. The teacher would not be able to answer the objection based on the ultrasystem, since the ultrasystem contains the very objection itself.

External vs. Internal Accounts of Dependence on False Statements

As mentioned above, another way of approaching the Gettier problem, taken from Pollock's Contemporary Theories of Knowledge, first edition, is by adding to rather than subtracting from the evaluation system. In terms of the justification game, the ultracritic would not be limited to denying what the claimant brings up in answering objections. Instead, he could bring in any truth whatsoever. If the addition of the truth undermines the justification, then the ultracritic wins the round.

But as Lehrer has noted, this means that misleading information could be brought forward, as in the Grabit case. To remedy this, we would require that a critic of the ultracritic be able to bring forth a rebutting objection. So to the objection that Mr. Grabit said that Tom's twin son John is in town, the critic of the ultra critic could respond, "But Mr. Grabit invented Tom's twin son!" If, after all objections and rebuttals are completed, the claimant remains justified, then her justification is undefeated.

This seems to be a simpler way of handling the Gettier cases than Lehrer's. It handles the issue of misleading evidence very straightforwardly, by countering it with the very fact that it is misleading. (This is reminiscent of Lehrer's own treatment of indirect objections: you just say they don't apply.) This seems to me better than having to attribute specific errors to a person's acceptance system in cases where we think there is no knowledge and denying the relevance of the misleading information when there is knowledge.

Ironically, Pollock does not include his solution in the second edition of Contemporary Theories of Knowledge. He notes that few philosophers any more there are any relatively simple solutions to the Gettier problem. As we have seen, the cases can be made more and more complex in order to block proposed solutions, which forces the solutions themselves to become more complex, and so on. But Pollock's main objection is against the whole idea of providing an analysis of knowledge.

The analysis of "S knows that P" is a fascinating problem, but it should be regarded as a side issue rather than as the central problem of epistemology. What the Gettier problem really shows us is what a perverse concept knowledge is. One can do everything with complete epistemic proprirety, and be right, and yet knowledge because of some accident about the way the world is. Why do we employ such a concept? Contemporary Theories of Knowledge, second edition, p. 14).

The Isolation Objection

Lehrer, on the other hand, thinks there is a payoff to solving the Gettier problem. He attempts to exploit the account of undefeated justification, as developed so far, to overcome a standard objection to coherence theories. The "isolation objection" holds that p may cohere ever so well with an evaluation system, but the evaluation system may be so far from the truth that the coherence does not confer justification. Laurence BonJour, who proposed his own coherence theory of empirical knowledge, calls this one of the "standard objections."

Coherence is purely a matter of the internal relations between the components of the belief system; it depends in no way on any sort of relation between the system of beliefs and anything external to that system. Hence if, as a coherence theory claims, cohrence is the sole basis for empirical justificaiton, it follows that a system of emprical beliefs might be adequately justified, indeed might constitute empirical knowledge, in spite of being utterly out of contact with the world that it purports to describe. Nothing about any requirement of coherence dictates that a coherent system of beliefs need receive any kind of input from the world or be in any way causally influenced by the world. But this is surely an absurd result. (The Structure of Empirical Knowledge, p. 108)

Lehrer's reponse to this objection is to allow that personal justification can be isolated, but that undefeated justification cannot be. The possibility of isolation is itself an objection that can be raised by the ultracritic. If the evaluation system of a person is indeed isolated, then the objection will not be answered in the ultrasystem, while if it is connected to the world, the objection will be answered. Either way, the result is appropriate. Lehrer calls this the "transformation argument," in that an evaluation system that is "connected to the truth" is transformed from personal justification to ultra justification. It should be noted that the conclusion of the argument is conditional: if my evaluation system is connected to the truth, then I am irrefutably justified. Since one cannot actually play the ultra justification game with an omniscient ultracritic, it remains for the antecedent of the conditional to be established, i.e., for skepticism to be answered.

Another Argument Against Foundationalism

A further use Lehrer makes of undefeated justification is to attack foundationalism anew. The foundationalist claims that we have sources of knowledge in the use of our mental faculties, perception, memory and introspection. The very fact that an acceptance is derived from such a source is sufficient to account for justification, the foundationalist claims. "We may have justified beliefs because they arises from source or faculty independently of what else we accept and, therefore, coherence is inessential to the justified beliefs and knowledge emanating from them" (p. 165).

Already we have it that in playing the ultra justification game, the claimant can only appeal to what he accepts to answer skeptical objections. There can be no direct appeal to the deliverances of the senses, etc. But we must also be able to answer objections directed against the trustworthiness of these sources. This requires acceptance about their trustworthiness

It is not sufficient for knowledge that a belief arise from a trustworthy source when one has no idea that this is so. The justification of the belief depends on acceptance of the unarticulated assumption that the source of the belief is trustworthy for becoming informed of the truth and not deceptive. Victory in the justification game depends on our accepting that perception, memory, and introspection are trustworthy sources of information of truth in general and in our special circumstances. Victory in the ultra justificaiton game depends in turn on our being correct in accepting these assumptions about the trustworthiness of our sources of information. (p. 166)
Coherence, reflected in our acceptances about the trustworthiness of our sources of information, is required to justify what we accept on the basis of those sources. "Coherence transforms sources of information into fountains of knowledge" (p. 166).

Absorbing Foundationalism and Externalism

Lehrer sums up his system by stating that it incorporates the best of both externalism and foundationalism. In both cases, the key is the principle of trustworthiness. If I know that p, my justification that p has made the appropriate truth connection. And my justification depends on what I accept about my own trustworthiness. So if I know, I must be right about my being trustworthy, that is, I must be trustworthy. Externalists emphasize the reliability of the subject, so in insisting on one's actual trustworthiness as a requirement for knowledge, Lehrer gives a nod to externalism.

The principle is also supposed to be self-justifying, at least "in the normal case" (p. 172). As such, it provides a sort of foundation for the coherence relation that is justification. Thus, Lehrer calls himself a kind of foundational coherentist. In my notes on Chapter 6, I criticized Lehrer for this move. I believe that he should remain a pure coherentist with respect to personal justification.

Can We Determine Whether We Are Justified?

The last topic of discussion in Chapter 7 is the issue of the determination of justification. Lehrer consistently maintains the view that there is no exit for the circle of what one accepts. So if I were to ask myself whether I am justified in accepting that p, my only appeal would be to my acceptance system at that time. If I were asked further whether my justification was undefeated, the appeal to my acceptance system at that time is again all I can make. So I cannot really determine at a time whether my justification is undefeated at that time, except to say that I am justified then.

On the other hand, when one's acceptance system changes, one is in a position to judge its earlier contents. I might decide now that my earlier justification was defeated. New information might move me to make this "diachronic" determination. (To be diachronic is to involve more than one time.) The reader may wish to compare these remarks on the diachronic evaluation of one's justification to those Lehrer makes in Chapter 6 about the diachronic dimension of trustworthiness.

[Previous Lecture | Next Lecture | Lecture Notes Menu

Philosophy 102 Home Page | G. J. Mattey's Home Page