Gettier Problem Sara Smollett
December 14, 2001

The Gettier Problem

What is the Gettier problem? Are any of the solutions proposed in Paul Moser ed., Empirical Knowledge adequate?


In this paper I will argue that knowledge is nearly synonymous with justified true belief, although not in the ordinary everyday sense in which we typically mean ``justified''. Because of this confusion surrounding the word ``justified'', an alternative definition of knowledge may be needed. I propose something along the lines of objective, correctly justified true belief. To demonstrate the adequacy of this definition, I present several Gettier examples and assess them in light of various conditions for knowledge. I give serious consideration to Pollock's position, which I think is quite similar to my own. Additionally, I address what I consider to be the real problem with the Gettier examples and explore how it can be resolved.

What is knowledge? Throughout the semester we have been equating knowledge with justified true belief. Something is knowledge only if it is believed, it is true, and the belief is justified. But is this enough for knowledge? Not according to Edmund Gettier. In his essay Is Justified True Belief Knowledge? Gettier provides two examples to illustrate his objection. A common modified example follows.

The Ford example: You have good evidence that Jones, a coworker, owns a Ford (you see him driving it every day), and so you conclude that at least one of your coworkers owns a Ford. Call this proposition P. You might say that you know P. It turns out, though, that in spite of all of the evidence, Jones doesn't really own that car. But, amazingly enough, one of your other coworkers owns a Ford, so your proposition was true (Williams, 94).

P, that one of your coworkers owns a Ford, is a justified true belief. You believed P. You had evidence for believing P. And P was true. But the belief P doesn't really seem to be knowledge. One might say that you didn't know P; you merely thought P and happened to be correct. This objection leads Gettier to conclude that knowledge is not merely justified true belief. Yes, something is knowledge only if it is a justified true belief, but there is something more. Gettier suggests that we need to add an additional requirement (or several) for knowledge (Gettier, 237).

Before considering what is missing from this definition of knowledge, I wish to present several similar Gettier (or Gettier-like) examples to help generalize the problem.

The candle example: You claim that there is a candle in front of you. This belief is based on the fact that you see a candle in front of you. There is a candle in front of you, but it is hidden from your view by a mirror. What you are seeing is the reflection of a second candle that is behind you (Williams, 94).

The affair example: Because your wife is acting secretively, you conclude that she is having an affair. Your wife is having an affair; however, the reason she is hiding something from you is because she is planning a surprise party. Being secretive has nothing to do with her affair (Pryor).

The barn example (Goldman, 1976): You are driving along the highway and see what looks like a barn out the window. Having no good reason to suspect it is anything besides a barn, you may say that you have a justified true belief you just passed a barn. Later, you tell this to a friend who explains that there are a lot of fake barns (barn look-alikes) in that area, and that most of the structures that look like barns really aren't. The structure you were looking at was, in fact, one of the real barns (Pollock, 246).

Let's return now to the car example and study it in a bit more detail. There are a number of reasons why P fails to be knowledge. The evidence for your belief P is not directly linked to P. In other words, it is only a lucky accident that your claim P is right. You might even exclaim after finding out that one of your coworkers owns a Ford, ``Wow, I was right after all.'' Surely this statement would not be uttered by someone who knew P.

Then what is knowledge? Where is the problem and what do we need to add to fix the definition? One problem, as suggested by Harman, seems to be that the conclusions in the above cases rely on false intermediate conclusions. The belief that one of your coworkers owns a Ford is based on the false belief that Jones owns a Ford. The belief that there is a candle in front of you stems from the false belief that you are seeing a candle in front of you. The belief that you are seeing a barn stems from the false belief that objects that look like barns really are barns. Harman suggests that for a belief to be justified, the inferences in the justifying argument cannot rely on false intermediate conclusions (Williams, 94). Armstrong suggests not only that intermediate conclusions must be true, but that they must be known to be true (Feldman, 241).

Similarly, knowledge might be said to require certain evidence. An additional requirement might be that the reasoning for the justification must be valid and sound. Another possibility is that upon further examination you wouldn't find more evidence that would contradict your justification. And, as was shown with the car example, an accidental belief does not qualify as knowledge.

All of these possibilities lead to modifications of the requirements for knowledge. But was there really a problem with the definition of knowledge in the first place? That is, are the above examples really examples of justified true belief? They are beliefs, yes; they are true, yes; but are they justified? Isn't it merely the case that you thought you were justified, but it turned out that you weren't?

That depends on what is meant by justification. For Chisholm, knowledge requires ``adequate evidence'' (Gettier, 237). That Jones drove a Ford to work every day was adequate evidence to you that he owned it. Seeing a candle in front of you or a barn sounds like adequate evidence that there is a candle or a barn. Knowledge, for Ayer, has to do with a ``right to be sure'' (Gettier, 237). Didn't you have a right to be sure that you were seeing a barn, a candle, or Jones' Ford?

Gettier thinks that these beliefs are justified, but I am not so easily convinced. Were you really justified in believing that Jones owned a Ford just because you saw him driving one? No. It was a good theory, a probable belief, but you had no right to be sure. While you may have thought you had good evidence for the belief, you didn't. You accepted this as fact without questioning it adequately. You might have thought you knew your wife was having an affair, but you were wrong. ``Are you sure?'' you might be asked on a witness stand. The answer, I believe, is no.

If, however, the answer is yes, it seems like what we need is more than justified true belief; we need correctly justified true belief. Not only the belief, but the justification as well, must be considered. The justification must be logically sound. False or inaccurate justification does not seem to satisfy these requirements for knowledge.

But do we need to know that our justification is correct? In other words, must we know that we know P in order to know P? If so, it seems that we need to know that we know that we know in order to know that we know? And so on ad infinitum. Every inference and any evidence would have to be called into question. This sounds like an argument that we can never know anything. How can we ever know anything? Do we ever have a right to be sure?

Fortunately, Gettier is immune from this argument from infinite regress because he does not believe that we need to know that we know. Furthermore, Gettier says that we can be justified in believing things that are false (Gettier, 238). Then perhaps I was too quick to claim that we don't have justification in the Gettier examples. Maybe we just have false justification.

While this contradicts my intuitive notion of justification, at the same time, it seems to be true. We do have justification for believing that a coworker owns a Ford. It is not an arbitrary belief, but rather a theory that was seriously considered and has an evidential basis. We do have justification for the belief that we are seeing a candle in front of us. We know what a candle in front of us looks like, and we rely on our sense. In both examples we have justification. It just turns out that our justification, while pretty good justification, was not the correct justification. Perhaps we never (or hardly ever) have or need absolutely certain justification, and instead what we need for justification is this approximately certain justification. It seems like this ``weaker'' justification is really what we have in mind most of the time, and what is referred to in justified true belief. Not that we are absolutely justified beyond even the tiniest shadow of a doubt, but that we, as rational beings, are reasonably certain.

This ``weaker'' justification might be called subjective justification. A belief is subjectively justified (for a subject) if given what he/she knows and can know, he/she is justified in holding that belief. It is not objective justification based on all possible knowledge that is available in the world.

I accept then, that this is what is meant by justified in justified true belief and that the Gettier examples are examples of justified true belief. Then I agree with Gettier that something more than justified true belief is needed for knowledge. I believe that something will have to do with correct, sound justification and justification that will hold up in light of future evidence.

In his ``The Gettier Problem'', John Pollock describes something similar to my distinction between kinds of justification. Pollock contrasts the subjective and objective senses of ``should believe'' (Pollock, 248). Support for subjective belief involves what we actually believe whereas support for objective belief concerns what we should believe given what is in fact true. Pollock claims that knowledge requires this objective ``should believe'' (Pollock, 251).

Rather than consider justified true belief, Pollock defines objective epistemic justification, which encompasses justified true belief. For a subject S to have objective epistemic justification for P, not only must S be subjectively justified in believing P, but S must also be subjectively justified in believing P for the same reasons when new truths are added (Pollock, 254).

The Ford example does not satisfy Pollock's requirement for this objective justification because there is a truth (that Jones does not own a Ford) that would negate the justification for believing that a coworker owns a Ford. The affair example also does not satisfy objective epistemic justification because the addition of the belief that the wife is planning a surprise party removes the belief that the wife is being secretive because she is having an affair. None of the examples discussed thus far are examples of objective justification.

Pollock claims that knowledge and objective epistemic justification are very closely related (Pollock, 251). He offers the following slightly modified formal definition of objective justification:

``S is objectively justified in believing P if and only if S instantiates some argument A supporting P which is ultimately undefeated relative to the set of all truths'' (Pollock, 255).

Does this definition satisfy our requirements for knowledge? The instantiated argument A is our justification for P. In the Gettier examples, the justification was defeated by the addition of further information. What Pollock says is a bit more subtle; objective epistemic justification may be defeatable, but it may not be ultimately defeated. To understand why this distinction is important, consider an example that satisfies Pollock's definition.

The theft example (Lehrer and Paxson, 1969): You see a boy hide a book and sneak out of a library. On the basis of this observation, you justifiably believe that he stole the book. The boy's mother tells several people that he was out of town at the time and his twin was the one who stole the book. This is a complete lie; there is no twin brother, and the boy did steal the book (Pollock, 247).

In this example, there is additional information (what the mother says) that contradicts your original justification. Your justification is defeated. However, there is even further evidence that negates the contradictory evidence and restores your justification. There is no further evidence to re-unjustify your justification. Then, according to Pollock, you are objectively justified in believing that the boy stole the book; you can be said to know that he stole the book (Pollock, 252).

It should be clear that we do not want to say that any additional inconsistent evidence can defeat our justification. This is why the ``ultimately'' clause is needed. What is needed is not for the justification to be consistent with some additional truths, but for it to be ultimately consistent with the set of all truths, or at least with the set of all truths that you can know.

This amendment is significant. What is really meant by ``the set of all truths''? It does not actually mean all truths, but only all truths that can possibly be subjectively known. If there is no way that a truth can be known, it is not relevant to objective justification because our notion of objective justification is based on potential subjective justification. Harman presents an example that he believes satisfies objective justification but not knowledge.

The assassination example (Harman, 1968): You see on one television station that the president has been assassinated and then turn off the television. In an attempt to keep people from panicking, all of the stations report that the assassination attempt failed for the rest of the day. Thus, while everyone else believes the president is alive, you believe he is dead (Pollock, 256).

In the assassination example, you are the only person who believes that the president is dead. You might say that you know the president is dead, but there is information you could have, indeed that you should have, that would defeat your justification for believing the president is dead. Unlike the library example where you have seen the boy with the book, you have less, rather than more, knowledge than everyone else. You are missing a piece of ``common knowledge'' that you are reasonably expected to have (Pollock, 257). Because your justification stems from ignorance, you do not have knowledge that the president is dead.

It still seems to me that you could have knowledge that the president is dead because there might be a defeater-defeater (namely, that all of the TV stations are lying). Still, the idea that knowledge can come from ignorance of truths is unsettling.

Knowledge, I believe, involves this objective justification. To have knowledge of a proposition one must have a correct, ultimately undefeatable argument. But before we conclude to know anything, we need to test the validity of our judgment. We do not want to say that we have knowledge until we have searched for additional truths, defeaters, and defeater-defeaters.

I am sufficiently satisfied with the definition of knowledge that I have outlined, that knowledge is objective, correctly justified true belief. I do not see how any of the Gettier examples discussed point to a problem with the definition of knowledge, because I do not think any of the examples illustrate knowledge as just defined. It seems that the real problem in the Gettier problem is not a question about whether the scenarios present knowledge or not, but whether knowledge is ever obtainable.

At a first glance, it appears that the sceptic may be right; there is no evidence that we can ever know something for sure. But although this is a central question in epistemology, perhaps it is the wrong question to be asking. Instead, it might be worth considering not this ideal knowledge, but a more practical notion of knowledge, one in which our justification is certain enough for us to depend on in everyday matters; that is, an approximation of ideal knowledge based on coherence with all truths that we can possibly know. This, I believe, is the kind of justification with which we should be concerned. With this clarification of what is typically meant by justification and knowledge, it seems that we are in a position to know, and that truth and correct justification are exactly what we need for knowledge.


Feldman, Richard. ``An Alleged Defect in Gettier Counterexamples''. Paul Moser ed. Empirical Knowledge. Lanham, Maryland: Rowman & Littlefield, 1996.

Gettier, Edmund. ``Is Justified True Belief Knowledge?''. Paul Moser ed. Empirical Knowledge. Lanham, Maryland: Rowman & Littlefield, 1996.

Pollock, John. ``The Gettier Problem''. Paul Moser ed. Empirical Knowledge. Lanham, Maryland: Rowman & Littlefield, 1996.

Pryor, James. ``The Gettier Problem''.

Williams, Michael. Unnatural Doubts: Epistemological Realism and the Basis of Scepticism. Princeton, New Jersey: Princeton University Press, 1996.

File translated from TEX by TTH, version 2.51.
On 14 Dec 2001, 15:08.