For more than two thousand years, philosophers largely agreed on what knowledge is: a true belief that you have good reason to hold. The definition had the stability of something that had survived scrutiny — ancient, tested, intuitively sound. Then, in 1963, a philosopher named Edmund Gettier published a paper two and a half pages long that broke it. What he broke has not been repaired. And the wreckage, examined carefully, tells us something important about the relationship between being right and knowing what you're talking about — which turns out to be a relationship considerably more complicated than it appears.
I want to begin with a specific memory, because it captures the problem more precisely than any abstract formulation I have found.
I was working a job that required knowing train schedules — not memorising them, exactly, but being the kind of person who could be asked what time the next train left and give a confident answer. One afternoon, someone asked. I checked my watch — it was 3:40 — and I told them the 3:47 leaves in seven minutes. I had looked at the schedule that morning. I was certain. And I was right: the 3:47 did leave at 3:47, and it did leave in seven minutes, and everything I told them was true.
What I did not know, and discovered only after they had left, was that my watch had stopped. It was not 3:40. It was, by complete coincidence, 3:40. The battery had died and the watch had frozen at exactly the right moment. My confidence was grounded in a belief — it is 3:40 — that was true, that I believed, and that I had what I considered excellent reason to believe (I had just looked at my watch). But the connection between my belief and the truth was, to put it plainly, accidental. I was right for entirely the wrong reasons.
This is not, in the scheme of things, a consequential story. But its structure is the structure of Edmund Gettier's 1963 paper, and Gettier's paper is one of the most consequential things published in philosophy in the twentieth century.
The Definition Gettier Broke
To understand what Gettier broke, you need to understand what he was breaking.
The definition he targeted is not Plato's in any simple sense — Plato's treatment of knowledge in the dialogues is more complex and more ambivalent than a single definition — but it is traceable to discussions in the Meno and the Theaetetus, and it became, over the following centuries, the default philosophical account of knowledge. In contemporary philosophy it is called the Justified True Belief account, and it holds that knowledge consists of exactly three conditions, each necessary and together sufficient.
First: the belief must be true. You cannot know something false; if what you believe turns out not to be the case, you did not know it. This condition seems unassailable. Whatever else knowledge requires, it requires that what is known actually be so.
Second: you must believe it. Knowledge is not merely a true proposition floating somewhere in the universe. It is a relationship between a person and a true proposition — a relationship that requires the person to actually hold the proposition as a belief. Again, this seems obvious to the point of being trivially true.
Third: you must have justification for the belief. This is the condition that distinguishes knowledge from lucky true belief — from the student who guesses the right answer on a multiple-choice exam and happens to be correct. Knowledge requires that your true belief be supported by reasons adequate to make it credible — that your arriving at the truth was not a matter of mere chance.
These three conditions — truth, belief, justification — form the classical account of knowledge, and for over two thousand years they seemed, if not perfectly specified, at least pointed in the right direction. The history of epistemology between Plato and Gettier is, in large part, the history of working out the details: what kinds of justification are required, how certain must the belief be, what counts as adequate reason. But the basic structure — knowledge as justified true belief — was not seriously questioned.
Then Gettier questioned it.
Two and a Half Pages
Gettier's paper is titled "Is Justified True Belief Knowledge?" and it was published in the journal Analysis in 1963. At the time of publication, Gettier was a twenty-nine-year-old assistant professor at Wayne State University. The paper is, by academic philosophy's standards, startlingly brief: two and a half pages, two examples, and a conclusion that the classical definition is wrong.
The structure of both examples is the same structure as my stopped watch, generalised. Gettier constructs situations in which a person has a justified true belief — all three conditions are met — but in which the connection between the justification and the truth is mediated by a chain of coincidence that seems to rule out knowledge. The person is right. They have good reasons to believe they are right. But they are right in the wrong way.
In the first example: Smith and Jones are both applying for the same job. Smith has strong evidence — he has heard directly from the company president — that Jones will get the job. He has also counted the coins in Jones's pocket and knows there are ten. From this evidence, he infers: the man who will get the job has ten coins in his pocket. This seems like a justified inference. But as it turns out, Smith gets the job — not Jones. And Smith, coincidentally, also has ten coins in his own pocket, though he had not counted them and did not know this. His justified belief — the man who will get the job has ten coins in his pocket — is true. But it is true in a way he had no idea it was true.
In the second example: Smith has strong evidence that Jones owns a Ford — he has always owned one, he was seen driving one recently. From this, Smith infers: either Jones owns a Ford, or Brown is in Barcelona. This is a valid inference; if the first part is true, the disjunction is true regardless of whether Brown is in Barcelona. It happens that Smith has no idea where Brown is. But it also happens that Jones does not, in fact, own a Ford — the evidence misled Smith — while Brown is in Barcelona. Smith's belief — either Jones owns a Ford or Brown is in Barcelona — is justified, true, and believed. But it is justified by evidence that is false, and it is true only because of the coincidental truth of the clause Smith included arbitrarily.
The examples feel somewhat artificial on first reading. This is, I think, a feature rather than a flaw. Gettier is not claiming that Gettier cases are common. He is claiming that they are possible — that the three conditions of the classical account, even when all three are met, do not guarantee the thing that the account was supposed to capture. The possibility of Gettier cases is sufficient to refute the analysis as a conceptual account. If justified true belief were sufficient for knowledge, then Gettier's examples would not be possible. They are possible. Therefore justified true belief is not sufficient for knowledge.
The Aftermath
The philosophical response to Gettier's paper was immediate and extensive. The short paper generated an industry of attempted repairs — a literature that is now more than sixty years old and that has still not converged on a consensus.
The most natural initial response was to add a fourth condition to the classical account: a condition that would rule out Gettier cases while preserving everything else. The proposals varied.
One suggestion was the no-false-lemmas account: knowledge requires that your justification not pass through any false intermediate belief. In Smith's first example, his justification passes through the false belief that Jones will get the job; ruling this out would handle the case. But counterexamples to the no-false-lemmas account were not difficult to construct: Gettier situations in which the justification does not pass through any explicitly false intermediate belief but the connection between justification and truth is still accidental in the troubling way.
Another suggestion was the causal account: knowledge requires that your belief be caused by the fact that makes it true. Smith's belief about the man who will get the job is not caused by the fact that makes it true — the ten coins in his own pocket — but by a different fact. This handles the case. But the causal account faces difficulties of its own: mathematical and logical knowledge, for instance, does not seem to be causally connected to mathematical and logical facts in any ordinary sense. The fact that two plus two equals four does not cause anyone's belief that two plus two equals four in the way that rain causes the belief that it is raining.
Reliability-based accounts — the view that knowledge requires a belief-forming process that reliably produces true beliefs — became influential. Alvin Goldman's reliabilism holds that the problem with Gettier cases is that the belief-forming process, in each case, is unreliable in the specific circumstances: it happens to produce a true belief, but the process itself does not reliably do so. The stopped watch reliably shows the right time twice a day; when I consulted it, the belief-forming process was unreliable in the relevant sense. This handles many cases but generates new difficulties: what counts as the relevant process, how local or global must the reliability be, and how reliability is to be assessed across possible worlds rather than merely actual cases.
Each proposed repair either fails against new counterexamples, or raises further questions that require further repair. This is not a sign of insufficient philosophical cleverness. It is, I think, a sign that something more fundamental is at stake.
What Gettier Actually Showed
The popular reading of the Gettier problem is that it showed the classical definition was incomplete — that it needs a fourth condition, and the difficulty is just in specifying what that condition is. I think this reading, while natural, misses the deeper significance of what Gettier showed.
What Gettier showed is that the concept of knowledge is not fully capturable by a set of individually necessary and jointly sufficient conditions. The classical definition fails not because one condition is missing, but because the concept it was attempting to define has a structure that resists this kind of analysis. Knowledge is not just true belief with justification added on. It involves a relationship between the believer, the justification, and the truth that cannot be fully specified by listing properties — it involves something about how the justification connects to the truth, something about the right kind of epistemic contact between the knower and the known, that has proven impossible to fully articulate.
Timothy Williamson, whose work Knowledge and Its Limits is among the most important epistemological contributions since Gettier, argues that this situation is not accidental. Knowledge, he proposes, should be taken as the foundational concept in epistemology — not defined in terms of simpler concepts, but used as the basis from which other epistemic concepts are understood. On this view, trying to define knowledge in terms of belief, truth, and justification is like trying to define red in terms of more primitive concepts: the right relationship runs in the other direction.
This is a significant reversal. Two thousand years of epistemology had been devoted to the project of explaining what knowledge is by breaking it down into its components. Williamson's suggestion is that this project was pointing the wrong way, and that Gettier's paper, rather than identifying a gap in the analysis that better analysis could fill, was demonstrating the fundamental limits of the analytical approach.
The Personal Dimension
I want to return to the stopped watch, because I think the experience it produces is philosophically instructive in a way that the abstract examples are not.
When I discovered that my watch had stopped, my first reaction was not philosophical. It was something more like embarrassment — the specific embarrassment of having been confidently wrong, of having misled someone with false certainty. The embarrassment did not come from having lied; I had not lied. It did not come from having reasoned badly; I had not reasoned badly. It came from having been certain in a situation where the certainty was, it turned out, disconnected from the reality it was supposed to track.
This is the phenomenology that Gettier's paper, in its dry analytical precision, is pointing at. The trouble with Gettier cases is not merely that they produce false conclusions — in my case, the conclusion happened to be true. The trouble is that they produce a specific kind of epistemic failure: the failure of being right for the wrong reasons, of having your justification connected to the truth by a chain of coincidence rather than by the kind of relationship that genuine knowledge involves.
I was not wrong. But I did not know. And the gap between those two things — between being correct and actually knowing — is what Gettier's two and a half pages forced philosophy to take seriously.
There is a version of this that happens regularly in intellectual life, and that is worth identifying because it is less obvious than the stopped watch. We form a confident view on a topic, and the view turns out to be correct, but when we examine the reasoning that led us there, we find that the reasons we thought we had were not the actual reasons the view is true. The conclusion survives; the justification does not. We were right, and we were wrong about why we were right, and for a period of time we moved through the world with the confidence of someone who knows rather than the more modest confidence appropriate to someone who happens to be correct.
What to Do With This
The Gettier problem is sometimes presented as a purely academic puzzle with no practical implications — a problem for professional epistemologists, irrelevant to anyone living outside a philosophy department. I think this is wrong.
The practical implication of the Gettier problem is a specific kind of epistemic vigilance: the recognition that being correct is not sufficient evidence that you know what you think you know, and that the justification you have for a belief needs to be examined not just for its plausibility but for its connection to the truth.
This is harder than it sounds. From the inside, being right and knowing feel identical. The person whose stopped watch shows the right time does not experience their situation differently from the person whose working watch shows the right time. The confidence is indistinguishable. The correctness is indistinguishable. What is different is invisible to the person in the moment.
The implication is not skepticism — not the conclusion that because Gettier cases are possible, nothing can be known. Most of our beliefs, most of the time, are connected to the truth in the right kind of way; they are not Gettier cases. The implication is rather a specific kind of humility about the relationship between your justification and the truth your justification is supposed to track.
When you are confident that something is true, and you can articulate good reasons for believing it, that is significant. But it is not as significant as it feels. The reasons might connect to the truth in a way that would survive examination; but they might also connect to it through a chain of coincidences that you have not examined, and that you would not find reassuring if you did.
What Plato Got Wrong, and What He Got Right
The title of this piece promises a verdict on Plato, and I should be honest about what the verdict actually is.
Plato did not simply propose the justified true belief account of knowledge as if it were a finished theory. The Theaetetus is, in fact, a dialogue in which the account is examined and found wanting — the dialogue ends in aporia, without a satisfying definition, and Plato seems to have been aware that the problem was more difficult than any simple account could handle. Attributing the classical definition to Plato as an error he committed is, historically, somewhat unfair.
What Plato got right is that knowledge involves at least truth, belief, and justification — that these conditions are all necessary. What the Gettier problem showed is that they are not sufficient — that something more is required, something about the connection between the knower and the known that two thousand years of philosophy has not managed to fully articulate.
This is, in a sense, a monument to the difficulty of the question rather than to the failure of the philosophers who worked on it. The difficulty is real. The question — what distinguishes the person who knows from the person who is merely right? — is one of the most important questions in epistemology, and it does not have a clean answer. Knowing this, and holding it honestly, is more valuable than the false comfort of a definition that turned out to need two and a half pages to break.
A Closing Thought on Confidence
There is a specific intellectual habit that the Gettier problem recommends, and it is not the habit of constant doubt. It is the habit of distinguishing between I am right and I know why I am right — of treating the second as a separate question from the first, and as one that requires separate work.
We are right about many things. We are right about many things without having fully examined the reasons we are right. Some of those reasons, examined, would turn out to be exactly as good as we believed them to be. Others would turn out to be connected to the truth through chains of coincidence we had not noticed, held by assumptions we had not examined, dependent on conditions we had not checked.
The difference matters. Not because it changes our conclusions in the moment — being right is being right — but because it changes how we hold our conclusions, and how we respond when the conditions that happened to make them right change. The person who knows why they are right can update when the reasons shift. The person who is merely right, for reasons they have not examined, will often find that their confidence persists long after the conditions that warranted it have changed.
Gettier showed that Plato's definition was incomplete. He did not show that the search for understanding was misguided. He showed that it needed to go further than it had — which is, perhaps, the most useful thing a two-and-a-half-page paper has ever done.
The distance between being right and knowing is not always visible, and it is never comfortable to examine. But it is there, and it matters — most clearly in the cases where what you were certain of turns out to have been true by accident, and you are left holding the confidence without the grounds that would justify it. This is Gettier's gift to epistemology: not a solution, but a precise statement of the problem. Precise problems, examined honestly, are worth more than comfortable answers to questions that were never quite right.