Why No Fourth Condition Fixes the Gettier Problem

Smith counts his evidence twice and finds it clean. The personnel file names Jones as the man who will be promoted; the manager, pressed, all but confirms it. So Smith concludes that the man who will get the job has ten coins in his pocket. He is entitled to that conclusion, and the conclusion is true — because Smith himself is hired, and Smith himself happens to be carrying ten coins. Nothing in the story is false except the reasoning that produced it.

The instinct, on first contact with a case like this, is that some small component has slipped loose and can be screwed back on. Add a clause, tighten a clause, and knowledge returns to being justified true belief plus something modest. That instinct has now been tested for six decades and has failed every time, and the pattern of failure is instructive enough to be worth walking through.

The repair everyone reaches for first

The earliest serious proposal was the no-false-lemma condition, associated with Keith Lehrer and Thomas Paxson: a belief counts as knowledge only if the justification supporting it does not rest on any falsehood. In Gettier's original cases the trick seems obvious — Smith's reasoning leans on the false premise that Jones will be hired — so closing that door should seal the leak.

It does not. The condition handles the easy cases and then stalls. Justifications can be built entirely from true premises and still terminate in a lucky true belief; and a justification that is sound at the moment of belief can be defeated later by evidence the believer has not yet encountered. The clause was also accused of vagueness: what exactly counts as resting on a falsehood, and how deep does the dependency run?

Causal chains, sensitivity, and the cost of strictness

Alvin Goldman tried a different route in 1967, arguing that the fact itself must be causally connected to the belief. When Smith believes the promoted man has ten coins, the actual cause of his belief is the file and the manager's remark, not the state of affairs that makes his sentence true. The diagnosis is sharp. The mechanism, however, is hard to specify for beliefs about mathematics, about the distant past, or about anything general — cases where no obvious causal thread runs from world to head.

Robert Nozick's sensitivity condition asks whether the belief would have survived falsity: if the proposition were false, would you still believe it? Ernest Sosa's safety condition asks a near neighbour: in nearby worlds where you hold the belief, is it true? Both capture something real about knowledge and both pay for it. Sensitivity is so demanding that it strips you of knowledge in skeptical scenarios. Safety is gentler, but gentler in a way that lets a certain amount of luck back through the door.

Zagzebski's grinding machine

Linda Zagzebski's contribution was not another patch but a proof that patching cannot work. In Virtues of the Mind she set out a recipe: take any proposed analysis of knowledge that consists of true belief plus a set of conditions that do not entail truth, construct a case where the first conditions are satisfied and the final one fails only by a stroke of luck, and you have a Gettier case on demand. The generator runs because the analysis has a seam. If the components never guarantee the conclusion, there is always a gap for luck to enter.

The deeper difficulty is not philosophical bookkeeping. It is that our reasons age. Evidence assembled yesterday can be true, well-sourced, and quietly irrelevant to the fact it is supposed to support, and nothing in the feeling of confidence announces the drift. Gettier cases are what that drift looks like when it is measured and given initials.

Abandoning the workshop

Faced with a device that regenerates its own counterexamples, several major figures left the repair business. Timothy Williamson argued that knowledge is conceptually primitive — that belief, truth, and justification are understood in terms of it, not the reverse, and that the analysis was always running backwards. Edward Craig proposed a practical turn: knowledge is whatever plays the social role of flagging reliable informants, which makes the hunt for necessary and sufficient conditions the wrong question asked with great diligence.

What remains is a genuine result, not a failure. Justification and truth are separable, and their alignment can be an accident. Any account that treats them as a package deal will keep producing Smiths: people whose paperwork is impeccable and whose understanding is nil.

If you found this valuable, consider supporting our work.