2002 Lecture Notes, Lehrer's Theory of Knowledge, second edition, Chapter 1

The Analysis of Knowledge

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

Theory of Knowledge and Theory of Reality

Philosophers in the West at least since Socrates and Plato in the 4th century BCE have investigated the nature of knowledge. Since then, all of the great philosophers of the Western tradition have had a great deal to say about knowledge. Until Descartes in the seventeeth century, however, questions about knowledge did not occupy center stage in most philosophical work. It was only until the nineteenth century that a separate sub-discipline called "theory of knowledge" or "epistemology" emerged. (See Richard Rorty, Philosophy and the Mirror of Nature, Chapter III for an unsympathetic account of how this took place. Rorty argues in that chapter that "theory of knowledge" is a project best abandoned. A summary of that criticism can be found in an on-line paper, A Perennial Philosophy Perspective on Richard Rorty's Neo-Pragmatism, in the first section, "Richard Rorty's Story about Modern Philosophy.") Since then, theory of knowledge has become a specialized "core" sub-discipline of analytic philosophy.

Another "core" sub-discipline of analytic philosophy is metaphysics, which Lehrer calls "theory of reality." This was the dominant part of ancient and medieval philosophy. Lehrer mentions Plato and Aristotle as practitioners of "metaphysical epistemology", which establishes a theory of reality on some grounds and then "appends" a theory of knowledge to it. He regards this procedure as "uncritically" assuming knowledge of the nature of reality. This, he goes on, begs the question against someone (a skeptic) who challenges whether the theory of reality is correct (pp 1-2). In the introductory notes, it can be seen that Plato at least is not guilty of this kind of practice. In fact, it is probably much more common nowadays, given the high degree of specialization within the sub-disciplines of analytic philosophy.

Theory of knowledge as an autonomous sub-discipline really began with Descartes in the seventeenth century. Concerned to avoid error at all costs (and bent on overthrowing the prevailing theory of reality) Descartes refused to admit the existence of anything real until he could establish that it is known and not merely believed to exist. The theory of reality is firmly subordinated to the theory of knowledge. The peculiarity of Descartes's approach was his "method of doubt," whereby began his investigation of knowledge by considering as false whatever was subject to the slightest doubt. This way, he could be sure to avoid error. Descartes claimed to have (more or less successfully) suspended his belief in the existence of his own body and the entire physical world until he could dispel the slight doubt attached to their existence. Thus he was what Hume called an "antecedent skeptic" (An Enquiry Concerning Human Understanding, Section XII, Part I), which yields what Lehrer calls "skeptical epistemology" (p. 2).

As Lehrer notes, the only thing that Descartes could not doubt was the existence of ideas or thoughts. Even if he was a victim of a powerful evil being, bent on deceiving him, there still must be the thoughts that are the targets of deception. Whether or not they were true, the thoughts existed. Lehrer notes that this does not make for an adequate starting point for the rest of knowledge. He points out that Hume argued convincingly that one cannot establish the existence of things outside our ideas unless one already knows that the ideas are correlated with the things, which begs the question against the skeptic. (See Chapter 3 for more on this kind of argument.)

Lehrer is correct in claiming that this "skeptical epistemology" makes it very difficult to show that we have any knowledge at all. Most scholars think that Descartes, despite his best efforts, could not find a way out of his skeptical starting point. Here is what the eighteenth century "common-sense" philosopher Thomas Reid, whom Lehrer admires, said about the matter.

But is this [Cartesian skepticism] to be despaired of, because Descartes and his followers have failed? By no means. This pusillanimity would be injurious to ourselves and injurious to truth. Useful discoversies are sometimes indeed the effect of superior genius, but more frequently they are the birth of time and of accidents. A traveller of good judgment may mistake his way, and be unawares led into a wrong track; and, while the road is fair before him, he may go on without suspicion and be followed by others; but when it ends in a coal-pit, it requires no great judgment to know that he hath gone wrong, nor perhaps to find out what has misled him.

Reid himself faulted Descartes's "way of ideas" as leading to skepticism about material things. But Lehrer adds that the reverse holds as well: we cannot avoid skepticism about ideas if we begin exclusively with claims to knowledge about material things.

The way to avoid the dogmatism of metaphysical epistemology and the skepticism of skeptical epistemology, according to Lehrer, is "critical epistemology" (p. 3) The idea here is to begin with assumptions about what is real and what we can know, keeping open the possibility that those assumptions can be overthrown as the result of investigation. Lehrer assumes that common sense and science are correct in their descriptions of reality and of what we know. Common sense holds that real, knowable things are the ordinary objects we encounter: the earth beneath our feet, the bodies of which our feet are a part, the mind which directs the body, etc. Science has a much richer field of objects, and the knowledge claims it makes are often rather far removed from what common sense acknowledges.


We leave the track of Lehrer's text for a discussion of how to begin the study of the theory of knowledge. Socrates and Plato began their investigations of various subject-matters by asking what they are. What is piety? virtue? justice? knowledge? If the theory of knowledge is an investigation of what knowledge is, then we should look for the best way of answering the Socratic question "What is knowledge?"

We could try a "bottom-up" approach: to begin by giving instances of knowledge as the first step in answering the question. A list of instances would not satisfy Socrates and Plato, since they demanded that we provide some common element that makes the instances cases of knowledge. The bottom-up strategy would provide this by a process of abstraction: discovering what is common between the cases.

But the bottom-up strategy engenders a paradox. The goal is to discover what knowledge is through cases, but what are the cases? It would seem that we need to know already what knowledge is in order to determine whether these cases are really cases of knowledge. This paradox is a variant of the "learner's paradox" found at the beginning of Plato's Meno. Chisholm calls it "the problem of the criterion" (though this is not the skeptical problem that ordinarily passes under that name).

One solution might be to appeal to general agreement about what the cases of knowledge are. It might be admitted that there is no universal agreement, but that there is a core of cases that nearly everyone agrees upon, so that there is no need for a prior account of knowledge. Then we could ask what all these cases have in common. It should be noted, though, that skeptics will disagree with any list of cases, and they can complain that general agreement is of no value in the light of their skeptical arguments. Even if nearly everyone agrees that they know certain things, they may be wrong.

A barrier to general agreement is that there appears to be no univocal conception of knowledge. As we will see below, Lehrer picks out three different broad conceptions. But even if we restrict ourselves to one of them, there may still be disagreement. If we sub-divided the broad conception of knowledge even further, would we ever reach a point of correct agreement about cases? And if we do, can we discover some element or elements in that conception which captures the cases on which there is general agreement?

It appears that Lehrer is advocating the bottom-up approach when he says that we should settle on the beliefs of common sense and science as counting as knowledge, but there is evidence to the contrary. The bottom-up approach is rarely taken by philosophers. Instead, they go from the top down, as did Socrates and Plato. The method works by proposing a common element and then asking questions about how it relates to particular cases.

There are two variants of this strategy. What Chisholm calls the "methodist" approach holds that the common-element of knowledge is independently justified. That is, the methodist thinks that the common element fits with what the ways in which we think about knowledge itself. Suppose I proposed this account of knowledge:
(LG) S knows that P if and only if S has guessed that P and one's guess turns out to be correct
Most people would object that this is just wrong conceptually. We do not think of knowledge as being the product of luck. The burden on the methodist is to find an account that that conforms to the ways in which we understand what knowledge is.

According to the methodist, if a particular case fails to conform to that common element, it is thereby not to be counted as a case of knowledge. As Chisholm notes, this approach runs the risk of ruling out many cases that are generally agreed upon as cases of knowledge. Also, the "problem of the criterion" arises again. How can we tell what our conception of knowledge is without examining cases of knowledge, and how can we tell what are cases of knowledge without having an account of knowledge?

The second variant of the top-down approach is what Chisholm calls the "particularist" strategy. Here, any proposed account of knowledge must conform to a pre-established list of cases of knowledge. If it does not, it is inadequate and must be replaced by another account, until a satisfactory account is reached. So, if (LG) is the account in question, it might be defeated by the following counter-example. If someone who won the lottery (and who was not in on a rigged draw) and guessed beforehand that he would win were to say, "I knew I would win," most of would likely reply that he did not know it. But on (LG), we would have to say that the person had knowledge.

It seems that most approaches to developing accounts of knowledge are neither purely particularist nor purely methodist. Philosophers have a rough idea of what the cases of knowledge are and of what the general features of knowledge are. Then they try to bring the two into balance with each other as much as possible. A final state in which they are in complete conformity has been called "reflective equilibrium." Lehrer holds out the possibility that on his approach ("critical epistemology"), we may be forced to admit that we do not know some of the things we initially thought we knew (p. 4).

Where do we begin in looking for proposals for a common element in all cases of knowledge? The standard approach in contemporary epistemology is to look to previous proposals. Basically, this is the method that Lehrer employs. Aristotle thought it was important to capture as much of traditional views on a topic as possible, and Lehrer, it will be seen, also tries to accommodate the insights of the main rival accounts of knowledge.

The appeal to past accounts of knowledge is subject to to the problem that people flat-out disagree over where to start. Here are two examples.

1. Classically, philosophers have insisted that knowledge be ultra-stable. That is, a person who knows that P cannot be mistaken about the truth of P. Mathematical knolwedge is the model to which all knowledge is supposed to conform. These philosophers can be called "infallibilists," and their view "infallibilism." But such a demanding requirement means that much of what we ordinarily think we know is not knowledge: it is mere opinion. Perhaps it is excellent opinion, when the information is highly probable, but it is opinion all the same. This is where particularism comes into play. We can say that any infallibilist account is inadequate because it does not cover vast numbers of established items of knowledge. (Of course, the infallibilist would never accept as knowledge most of the beliefs held on the basis of knowledge and common sense.)

Starting at around the beginning of the 20th century (following the American philosopher C. S. Peirce), we see the emergence of "fallibilism." Philosophers such as Lehrer become fallibilists because they want a more comprehensive account of knowledge, one that will accommodate the general view that people have knowledge even when they fall short of certainty. Nearly everyone these days is a fallibilist with respect to human knowledge.

2. During the last half of the 20th century, philosophers (and many non-philosophers, especially in the social sciences) have become even more liberal, wishing to count as knowledge what even fallibilists have rejected as mere opinion. Basically, the view is that knowledge is relative to some social structure: it is "knowledge-for-a-social-group." This allows for the possibility that members of one social group could have "knowledge" that completely contradicts the "knowledge" of another group.

Perhaps the best way of dealing with these disagreements is to hold that they represent distinct conceptions of knowledge, and so are not opposed to one another.

There is still plenty of room for disagreement within each of these broad approaches to knowledge. Most importantly, there is a great deal of controversy over what is the correct fallibilist account of knowledge. One such disageement is discussed below, and other disagreements will be uncovered in subsequent chapters. To resolve these disputes, we must try to determine whether they are conceptually correct and whether they conform to accepted cases of knowledge, as outlined above.

What is a Theory of Knowledge?

After having settled on a stock of real and known items, Lehrer proceeds with the production of a theory of knowledge. Following Carnap, Lehrer conceives of the task of a theory of knowledge as providing an "explication" of the concept of knowledge. Explication consists of two parts:

These two elements of an explication of knowledge will be the topic of the rest of our discussion.

Types of Knowledge

We begin with the task of clarification. There are several ways in which we describe cases of knowledge. To each of these ways there might correspond a theory of that kind of knowledge.

Lehrer's theory is limited to the third type of knowledge.

When an individual displays competence, we say that it knows how. This kind of knowledge is attributed very liberally, e.g. to animals with small brains ("The spider knows how to catch flies") and to inanimate objects ("The computer knows how to find a file"). This liberality results from the fact that competence can be the result of instinct or merely the ability to follow instructions. Many recent philosophers have taken competence as the central form of knowledge, but Lehrer does not. (This theme is developed in Chapter 8.)

An individual may be said to know that with which it is acquainted. I know my mother-in-law, for example. To say that one knows something in this sense is to say that it has had some experience with what it knows. There is an association with competence here, since we might want to deny that there is knowledge in cases where the individual does not have the ability to recognize the thing.

Finally, there is knowledge in the (correct) "information" sense. To know is to recognize correct information as being correct. I know that 2 + 2 = 4 because I possess the information that 2 + 2 = 4, the information is correct, I consider it to be correct, and I have a good idea why I think it is correct. Competence often demands knowledge in the information sense, though in the cases of competence mentioned above, the spider and the computer lack knowledge in the information sense.

Knowledge in the information sense is, according to Lehrer, the most important kind of human knowledge. It is essential to theoretical and practical reasoning that pervades all human endeavors. When we reason, we begin with information we take to be correct and apply it to other information. Mere possession of correct information is not sufficient for knowledge, however. In science, as well as in more practical endeavors in life, our knowledge arises from investigation into the correctness of the information we are given.

The Conditions for Knowledge

Now that we have narrowed down the sense of knowledge with which we will be concerned, we can give an analysis. In general, an analysis breaks something down into its constituents, e.g. a drop of blood is analyzed to determine the characteristics of its DNA. In philosophical analysis, we try to find a set of necessary and sufficient conditions for something's falling under the concept which we are concerned. That is, we are looking for conditions which must be satisfied for one to have knowledge and which, if met, enough for one to have knowledge. Here, it is one in which an individual S knows the information that p. Here 'S' and 'p' are variables. 'S" stands for any individual which can know in the information sense, and 'p' stands for any information that can be known.

The account of the information sense of knowing yields three natural conditions:

This kind of account is often called the "traditional analysis" of knowledge, and it is often attributed to Plato (Theaetetus, 202c. The list is supposed to constitute a set of necessary and sufficient conditions for something's falling under that concept. A necessary condition is of the form: If S knows that p then ... Once all the necessary conditions have been found, they should be jointly sufficient, that is, any S and p satisfying all the necessary conditions will be a case of a knower and something known. Just giving the conditions does not complete the theory of knowledge. One must also give an account of how those conditions are fulfilled. Most of the course will be devoted to this task.

The Truth Condition of Knowledge

The first condition is that the information that p be true. It is not as easy as it might seem to explain how the truth condition is satisfied. On the one hand, a minimal explanation like this:

The information that p is true if and only if p

borders on the trivial. But if we try to give a more substantive account, serious problems arise, as will be discussed by Lehrer in Chapter 2.

The Acceptance Condition of Knowledge

To recognize information as correct is to have an attitude toward it. The knower S endorses the information in the sense that S stands behind it or endorses it as being correct. (Note that this does not mean that S approves of the information that p. I may accept that my best friend has cheated me but not endorse the activity of cheating.) Another way to describe the endorsement is to say that S thinks that p is correct or true information. Lehrer thinks that what we accept functions as a kind of assumption. We accept information as being true because we have as our goal the acceptance of all and only what is true.

By contrast, we accept many things for other reasons, usually having to do with some practical consequences. I might assume that my wife is faithful to me despite a lot of evidence to contrary because I could not stand to admit that she is unfaithful. What we assume with the goal of attaining truth plays a "functional role" in our thinking and acting (Chapter 2). It is what we rely upon to guide us when deciding what else to accept and what to do.

Acceptance, for Lehrer, is a technical concept which he distinguishes from belief. The key difference between the two is that acceptance is goal-directed, hence relative, while belief is absolute. To simplify his exposition, Lehrer is willing to consider acceptance to be a kind of belief, in which case believing that p is a necessary condition for knowledge. (In a later work, however, he claims that one can accept that p without believing that p.

The Justification Condition of Knowledge

Justification is the heart of the analysis. Keeping in mind that our conception of knowledge is that of the recognition of correct information as being correct, we can see that knowledge demands some means for sorting out the good from the bad information. To recognize information as being correct, S must not be correct as a matter of luck. For example, if his secretary entered and left her office at random times, and Lehrer accepted that she was in the office now, he might be right by sheer luck. Now change the case so that the secretary is ordinarily in the office at this time. If he accepts that she is there now, it is more than mere luck that his information is correct. It is very reasonable for him to accept what he does, but he is not justified in accepting it. He cannot exclude the alternative that she has left the office for some reason.

Were Lehrer to improve his position, by looking into the office, he would be able to exclude the possibility that she had stepped out. Would this be enough for him to fulfill the justification condition? Lehrer has little to say yet on this score, but we can follow up on some hints that he gives in the discussion of the justification condition. First, we learn that justification lies between reasonableness and complete certainty. (Note that 'certainty' here means nothing more than the strongest possible form of justification. It does not mean 'unyielding conviction,' since strength of conviction could be dogmatic and not justified at all.) To be certain that his secretary is in her office, Lehrer might have to go to extreme measures, such as analyzing DNA samples of the the person he sees sitting there. And this might not be enough either, as an identical twin has the same DNA. There is a lot of ground between reasonableness and complete certainty. So if he looks in and sees someone who looks like his secretary sitting at the desk, is his justification sufficient for knowledge? No answer to this question is forthcoming untilChapter 6.

One other comment Lehrer makes about justification is worth noting. He says, "the person must be justified in a way that would justify him in accepting that he knows, if he considers whether he does" (p. 14). Suppose Lehrer has the power to detect his secretary's presence in some extra-sensory way, but he is not aware of this power. Then successfully exercising his ESP does not give justification sufficient for knowledge. If he considers whether he knows that his secretary is in the office, he will not take this information into account, and so it would not make it any more reasonable for him to accept that she is there. This theme is played out in detail in Chapter 8.

Much of the text is devoted to the question of how to understand justification as such. Historically there have been many accounts of justification proposed by theorists of knowledge. We will here sketch the most prominent contenders. They will be treated in detail when we discuss later chapters of the book.

Theories of Justification

Most broadly, accounts of justification can be divided into two categories. The internalist accounts make justification a matter of having reasons, where reasons are information that we accept, existing, so to speak, in our heads or "internally" to the knower. Lehrer's embrace of internalism is already clear from his proviso that justification sufficient for knowledge must be a factor when we conside whether we know. Two ways of construing what constitutes reasons leads to a distinction between two main kinds of internalism.

For the foundationalist, there are "first premises of justification" (p. 15). The need for first premises found when we think of justification as having the structure of an argument. (Foundationalists need not consider justification "in terms of an argument for a conclusion." See especially John L. Pollock and Joseph Cruz, Contemporary Theories of Knowledge (second edition), Chapter 2.) In order for a premise to be a good reason, the foundationalist continues, it must itself be supported by a good reason. As this evidential support cannot continue without end, there must be a stopping point, a set of "first premises" which are not the conclusions of any argument. These first premises comprise the foundation on which all justification is built.

The coherentist rejects the notion that justifying reasons are premises of arguments. Instead, the coherentist proposes that one has good reasons to think the information that p is correct if that information "fits in" with all the other things the person accepts. No information is privileged like the first premises of the foundationalist. Roderick Chisholm likens coherence to a house of cards, where each card supports all the others but no card is self-supporting. As we will see in subsequent chapters, Lehrer is very sensitive to the foundationalist criticism that this relation of mutual support is in reality a kind of circularity that undermines justification.

Externalists reject the view that justification is a matter of having good reasons (whether in argument form or not) in favor of the view that it is the result of the right relation between the subject and the world "outside the head." (Sometimes they reject the notion of justification altogether, in favor of an analysis of knowledge which contains a replacement for the justification condition.) An example of an external theory is a causal theory, according to which S knows that p when the fact that p (not the information that represents the fact) causes S to accept the information that p.

The Adequacy of the Justified True Acceptance Analysis of Knowledge

Having discussed what Lehrer takes to be three necessary conditions for knowledge, we may now ask whether together they are sufficient. A number of examples indicate that they are not. A case noted by Lehrer in another context was that of a person who sees a broken clock showing the correct time, which it does once every twelve hours. If we assume that the person had good reasons to accept that the clock was working properly, we could say that the person had a justified acceptance of true information which, nonetheless, is not knowledge.

Lehrer and others have presented counterexamples to the tri-partite analysis of knowledge as justified true acceptance. A counterexample is a scenario or case which is entirely under the control of the person devising it. We are asked to consider a situation in which specific circumstances hold. What is true or false in the counterexample is stipulated by its author. Thus the counterexample is a description of a possible state of affairs which may never hold in reality. But as Lehrer states, an analysis is supposed to cover all possible cases. So the "armchair philosopher" is free to let his or her imagination roam. Since the publication of Gettier's paper, numerous imaginative counterexamples have appeared.

A case from an old movie, "The Shadow of the Thin Man," has a detective talking to a two-time loser at a race track. The guy mentions that he just got out of jail, commenting, "I was a victim of circumstances. The DA framed me, not knowing that I was guilty. Ain't that a coincidence?" A member of the jury, convinced by the arguments of the DA, would have had justified true acceptance without knowledge.

Another case from film is from the Orson Wells 1958 noir classic "Touch of Evil". A policeman has a perfect record of "solving" cases by "finding" incriminating evidence. He is investigating a case where a man's car was blown up, killing him. He has a suspect with a motive: the victim was his girlfriend's father and was opposed to his relationship with her. But the policeman lacks any physical evidence linking the man with the crime. When searching the suspect's home, he enters the bathroom, leaving soon thereafter and asking another policeman to continue the search. Predictably, the second search turns up a couple of sticks of dynamite in a box. This presumably would have gotten a conviction from a jury, who would have found the suspect guilty beyond reasonable doubt. It turns out that the suspect was guilty. (His partner asks him in the climactic scene, "How many did you frame?" He answers, "Nobody that wasn't guilty. Guilty...") But the jury would not have known that, since the physical evidence was misleading and the motive only made it reasonable to believe that he was. For a detailed account of the plot, click here.

The case in the text is a modification of some cases published in 1963 by Edmund Gettier (and now referred to as "Gettier cases."). A teacher has two students, Mr. Nogot and Mr. Havit, in her class. Mr. Nogot seems to be the proud owner of a Ferrari (a rare and expensive car). He says he owns one, drives one around, and has papers which state that the car he drives is his. The teacher, on the basis of this evidence, concludes that someone in her class owns a Ferrari. This is true enough, but only because Mr. Havit, who shows no signs of Ferrari ownership, secretly owns one. Again, it seems that the three conditions of knowledge have been met, but that there is no knowledge.

Another example of a Gettier case can be developed from Lehrer's example concerning whether his secretary is in her office. Suppose that he looked into the office and saw sitting behind the desk a figure who looked to him exactly like his secretary. We may suppose that he would be justified in accepting that his secretary is in her office. However, it may be that a hoax is being played on Lehrer, and that the person sitting at the desk is his secretary's identical twin sister. The real secretary is hiding behind the desk, waiting to leap up and surprise him. So it is true that the secretary is in the office, Lehrer accepts that it is true, and he is justified in so accepting that she is.

A Fourth Condition of Knowledge

The problem in all these cases is diagnosed in this way: the justification depends on a falsehood. In the clock case, it depended on the false information that the clock was working properly. In the movie case, the false information was what the DA presented to jury in framing his victim. In the Ferrari case, the false information was that Mr. Nogot owns a Ferrari. In the secretary case, the false information was that the person sitting behind the desk is his secretary. In each case, the justification depends on the falsehood in the sense that dropping the assumption that the information is true destroys the justification. Condition (iD) is supposed to formalize this solution by requiring that to know that p, S's justification may not depend on any false statement. It is no trivial task to explain how it is that a justification depends on false information. As noted, the false statement in the teacher case is that Mr. Nogot owns a Ferrari. It is false because Lehrer stipulated that it is when concocting his scenario. The teacher's evidence depends on this false statement because she has no other evidence that someone in her class owns a Ferrari. Ferraris are rare and expensive, and there was no reason to think that anyone else in her class owns such an exotic car. Since her justification depends on a false statement, she fails to satisfy the fourth condition of knowledge, so according to the analysis she does not know.

This is the right result given that we should not attribute knowledge to the teacher. Lehrer's "intuition" (and those of almost everyone else) is that the teacher lacks knowledge because she is correct by accident. It is possible that some people do not share this intuition. (An intuition, as philosophers understand the term these days, is an untutored opinion which expresses the way a person normally views a situation. Many philosophical disputes come to a halt because of irreconcilably different intuitions. We will run up against such disputes later in this course.)

Some people might dispute whether condition (iD) is required to block the teacher counterexample. It is somewhat plausible to argue that the teacher's justification simply was not complete, and hence that her situation failed to satisfy condition (iJ). The incompleteness of her justification is due to the fact that she did not rule out the possibility that Nogot was a fraud.

Lehrer believes that his condition (iD) is the solution to the Gettier problem. So confident is he that he states that adding condition (iD) renders the analysis "impervious" to counterexamples (p. 20). But experience shows that one ought to be more cautious in providing solutions. There is a huge body of literature in which more and more clever examples are found to defeat each "solution." to the problem. Lehrer's approach must stand up to some very sophisticated challenges, as we will see when discussing Chapter 7.

[Introductory Lecture | Next Chapter | Lecture Notes Menu ]

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