Gettier's Is Justified True Belief Knowledge?: Why Justified True Belief Is Not Enough

Imagine checking a meeting time on a normally dependable office display. It says 10:30, so you believe it is 10:30. Unknown to you, the display froze yesterday. By coincidence, you happen to look at it at exactly 10:30 today. Your belief is true, and consulting the display seemed reasonable. Whether you knew the time is a further question.

This invented version of a familiar epistemological puzzle introduces the distinction Edmund Gettier examines in "Is Justified True Belief Knowledge?" A person can have good grounds, reach a true conclusion, and still arrive at that truth through an accidental match. Gettier's two examples put pressure on an analysis that identifies knowledge with justified true belief. Their conclusion does not imply that evidence is pointless or that knowledge is impossible.

The analysis concerns propositional knowledge: knowing that something is the case. It does not directly analyze knowing a person or knowing how to ride a bicycle. Let P stand for a proposition and S for a person. The proposal says that S knows P if and only if P is true, S believes P, and S is justified in believing P.

The phrase "if and only if" makes a demanding claim. Each condition is presented as necessary: knowledge requires it. The three conditions together are presented as sufficient: satisfying them guarantees knowledge. A counterexample to sufficiency needs a case in which all three conditions hold and knowledge is absent. Showing that somebody has a justified false belief would not accomplish that, because the truth condition would already fail.

Belief and truth do different work. Someone can believe a meeting begins at 10:30 when it begins at eleven. Equally, the meeting's actual start time does not ensure that someone who has never heard of it believes anything about it. Justification introduces a third question: what standing does the person's belief have, given the evidence or reasons available?

The initials JTB abbreviate justified true belief. The brevity of the formula can conceal how much turns on its last component. What counts as justification is itself a philosophical dispute. The JTB formula alone leaves justification undefined; Gettier supplies two preliminary points about it. The formula does not by itself settle whether justification depends entirely on what the believer can reflectively access.

Gettier's paper appeared in Analysis 23, issue 6, in June 1963, on pages 121–123. It begins with recent attempts to state conditions for knowledge and discusses formulations by Roderick Chisholm and A. J. Ayer alongside the familiar three-part proposal. That is a more accurate account of its immediate target than a story in which one uncontested definition ruled philosophy unchanged for two thousand years. Gettier, p.121.

Gettier first allows that a person can be justified in believing something false. This is a form of fallibilism about justification. Evidence can be good enough to justify belief without making error logically impossible. If justification already guaranteed truth, the false starting beliefs in his examples would not meet the stipulated condition.

His second assumption concerns a person's actual reasoning. If S is justified in believing P, P entails Q, and S deduces Q from P and accepts Q on that basis, then S is justified in believing Q. Here entails means that P cannot be true while Q is false. The principle connects justified belief with a recognized and performed deduction. Gettier, p.121.

This qualification prevents an exaggerated reading. Gettier does not need the claim that every person is automatically justified in believing every remote consequence of everything they believe. A consequence that nobody has noticed, or one reached through confused reasoning, is outside the stated setup. His examples specify that the subject sees the entailment and forms the resulting belief.

The principle also distinguishes preservation of truth from preservation of justification. A valid deduction from a true premise cannot yield a false conclusion. A valid deduction from a false premise can yield either a true or a false conclusion. Gettier uses a false premise for which the subject nevertheless has justification, followed by a conclusion that happens to be true.

This combination creates the problem. The justification travels along the subject's reasoning, while the conclusion's truth is secured by something the subject does not know. Calling that gap "luck" summarizes a feature of the cases. It does not yet provide an account of every kind of luck incompatible with knowledge.

In Gettier's first case, Smith and Jones have applied for the same job. Smith has strong evidence that Jones will get it and that Jones has ten coins in his pocket. Gettier suggests evidence for these beliefs: the company president has indicated Jones will be selected, and Smith has recently counted Jones's coins.

Smith reasons from that conjunction to a less specific claim: whoever gets the job has ten coins in his pocket. The inference is valid. If Jones gets the job and Jones has ten coins, then the successful applicant has ten coins. Smith recognizes the entailment and accepts the conclusion on that basis.

Now change the facts while leaving Smith's evidence and reasoning as described. Smith will get the job, and, unknown to him, Smith also has ten coins in his own pocket. His starting conjunction is false because Jones will not get the job. His conclusion about the successful applicant's coins is true because Smith will get it and has the specified number. Gettier, pp.121–122, Case I.

The proposition we evaluate must remain fixed. We are not asking whether Smith knows that Jones will get the job. That proposition is false. Nor are we asking whether Smith knows how many coins Jones has. The target is his more general belief about the person who will be hired. That belief is the one which combines truth, belief, and the justification inherited from his inference.

Gettier judges that Smith does not know the target proposition. Smith's grounds concern Jones, while the truth of the conclusion depends on Smith's own circumstances, of which he is unaware. The successful deduction has removed the name that would have made the error obvious without repairing the connection between Smith's evidence and the actual successful applicant.

This last point explains why counting the three conditions does not settle the question. The analysis was supposed to identify the circumstances under which a justified belief reaches the truth as knowledge. In this case the conditions coexist, yet the evidence concerns the wrong candidate. Accepting Gettier's knowledge judgment therefore gives a counterexample to the proposed sufficiency claim.

The second case makes the logical structure more visible. Smith has strong evidence that Jones owns a Ford. Jones's history of owning Fords and his recent offer of a ride while driving one support the belief. Smith also has a friend, Brown, whose location is unknown to him.

Smith takes the claim about Jones and constructs several alternatives by adding an "or" clause about Brown. The relevant one says that Jones owns a Ford or Brown is in Barcelona. The "or" is inclusive: at least one side must be true, and both may be. If Jones owns a Ford, the disjunction follows regardless of Brown's location.

Smith recognizes the entailment and accepts the disjunction. In Gettier's setup, however, Jones is driving a rented car and does not own a Ford. By coincidence Brown is in Barcelona. The first side is false, the second true, so the disjunction is true. Smith's evidence supported the first side, and he has no information establishing the second. Gettier, pp.122–123, Case II.

There is no requirement that Smith have evidence for both sides before a disjunction can be justified. Good evidence that a parcel is in the hall can justify believing that it is in the hall or in the kitchen. That is why the inference principle matters. Rejecting Smith's reasoning simply because he knows nothing about Barcelona would miss the logical move Gettier is using.

Gettier again judges that Smith lacks knowledge of the true disjunction. The fact that saves its truth plays no role in his evidence. The example separates the subject's justified route to belief from the actual state of affairs that makes the resulting proposition true. It reaches the same philosophical pressure point as the job case with a different logical form.

These are proposed counterexamples, not survey results about how everyone uses "know." The argument combines formal reasoning with judgments about justification and knowledge in carefully described cases. Someone who rejects a judgment or an assumption can dispute the argument, but needs to identify which part they reject and why.

The formal component is straightforward: a conjunction supports the generalized claim in Case I, and a proposition entails a disjunction containing it in Case II. The philosophical component includes whether the subject is justified and whether the resulting true belief qualifies as knowledge. Keeping these components apart makes disagreement more useful than simply announcing that the cases feel convincing or unconvincing.

The conclusion targets sufficiency. It does not by itself disprove that truth, belief, or some form of justification is necessary for knowledge. Nor does it supply a finished replacement analysis. Finding an exception to a proposed definition and producing a better definition are different tasks. Gettier's paper concentrates on the former.

The historical claim also needs restraint. In a footnote, Gettier refers cautiously to Plato's Theaetetus and Meno; he does not establish an uninterrupted consensus beginning with Plato. Describing his paper as an objection to particular analyses avoids turning a precise argument into an unsupported story about the whole history of philosophy. Gettier, p.121, note 1.

One response denies that Smith was justified in the first place. Perhaps a president's assurance is not sufficient, or a history of seeing somebody drive a Ford does not justify believing they own one. Particular details can certainly be questioned. The issue for this response is whether it preserves the intended fallible conception of justification while excluding the cases for a principled reason.

Demanding absolute certainty would change the task substantially. If any possibility of a misleading assurance prevents justification, many ordinary beliefs supported by testimony would fail the condition. A defender is free to propose such a stringent account, but must explain its scope and consequences. Simply pointing out that Smith's evidence can mislead does not answer an argument that explicitly allows justified error.

A different response questions the transmission of justification through deduction. Perhaps the oddly constructed disjunction should not inherit the status of Smith's starting belief. But refusing that move also requires explanation: the inference from P to "P or Q" is valid, and the case says Smith understands it. We would need an account of justification that explains why this use of a valid inference fails.

These responses show why a counterexample can generate further theory without settling everything. The case presents a combination of commitments under pressure. Revising any one commitment may block it, but a revision should also make sense in ordinary cases beyond the puzzle. Otherwise the repair risks becoming a rule invented solely to exclude the troublesome example.

An appealing repair forbids reliance on false premises. Both original cases involve an inferential starting point that is false, so such a restriction addresses their visible mechanism. That observation supports investigating the proposal. It does not prove that every possible case of accidental truth must contain an explicitly false premise.

Return to the imagined frozen display. A person might read it without consciously inferring anything from the proposition that it is functioning. Whether an implicit false assumption is involved depends on how belief formation is analyzed. Merely instructing people to avoid false steps leaves that question unanswered. This example illustrates a problem a repair must address; it is not a third case from Gettier's paper.

Alvin Goldman offers a different direction in "A Causal Theory of Knowing." He asks for an appropriate causal connection between an empirical fact and the person's belief. In his discussion of the Ford case, Brown's actual presence in Barcelona is disconnected from the formation of Smith's belief. The contrast is between an appropriate connection and coincidental correctness. Goldman, 1967, pp.357–358.

Goldman's proposal needs qualifications. His paper explicitly concerns empirical propositions, so treating mathematical knowledge as an overlooked problem for that exact stated scope would misrepresent it. He also considers causal connections through a common cause and argues that an inferential reconstruction must get the important links right. The proposal is more specific than the slogan that a true fact must directly cause a belief. Goldman, pp.357, 363–365.

The broader question is how to explain a belief's connection with its truth. Evidence available to the subject, the process producing the belief, and the surrounding circumstances can all become candidates for analysis. Listing these possibilities is a way of locating the work still required. Gettier's two counterexamples alone do not demonstrate that any one later theory succeeds.

Consider an invented analyst who forecasts a delivery correctly using a report secretly copied from last week's schedule. The delivery arrives at the predicted time because of an independent change nobody on the team anticipated. If the analyst had reasonable grounds to trust the report, the case invites the same questions about justified belief and accidental truth. The stipulations matter: a guess made without adequate grounds would fail the justification condition from the beginning.

Evaluating the prediction requires more than checking its final accuracy. We can ask whether the report actually concerned this delivery and whether the evidence would have supported the answer under slightly different circumstances. Those questions investigate how the result was reached. They do not imply that every forecast based on a fallible source lacks knowledge.

The same caution applies to a hypothetical generated answer with an invented citation. The answer's central claim might happen to be true. If a reader reasonably relied on apparently credible evidence that did not actually support it, accidental agreement would remain possible. We must separately assess whether the reader was justified; an impressive-looking citation does not automatically satisfy that condition.

Neither illustration is empirical evidence about forecasting or artificial intelligence. They are cases constructed to clarify the distinction at issue. A true answer can be useful even when its origin is poor, but that usefulness does not establish that its recipient knows the answer on that basis. Evaluating a result and evaluating an epistemic achievement can lead to different judgments.

Gettier's challenge is demanding without requiring despair about ordinary knowledge. It asks for an explanation of why the subject's justified belief and the actual truth belong together in the way knowledge requires. In the original cases, inspecting the target proposition carefully reveals how they come apart. That inspection is the argument's lasting methodological lesson: before accepting a definition, test what its conditions permit.

• Edmund L. Gettier, "Is Justified True Belief Knowledge?", Analysis 23.6 (1963), pp.121–123. Read the two preliminary assumptions and both cases together; the journal-page scan allows comparison with the transcription.
• Alvin I. Goldman, "A Causal Theory of Knowing", The Journal of Philosophy 64.12 (1967), pp.357–372. An early primary response, explicitly scoped to empirical knowledge.
• Linda Zagzebski, "The Inescapability of Gettier Problems", The Philosophical Quarterly 44.174 (1994), pp.65–73. A later paper on Gettier problems; the publisher page supplies bibliographic information, with full access subject to availability.