Roderick Chisholm named it the problem of the criterion in a 1973 essay, but the problem is much older — it appears in Sextus Empiricus and was taken seriously by the ancient Pyrrhonist skeptics as evidence that knowledge is impossible. The problem runs like this. Suppose you want to know which of your beliefs count as knowledge. To identify them, you need a criterion — a test for knowledge. But how do you select the right criterion? You would need to already know what knowledge looks like in order to evaluate whether the criterion is tracking it. The criterion must itself be evaluated by something, and that something must be evaluated by something else, either producing an infinite regress or a circle. You cannot get the project of epistemology started.
Chisholm distinguished three responses. The methodist (in the epistemological sense, nothing to do with the religion) begins with the criterion: you first determine what the method of knowing is, then apply it to identify instances of knowledge. The particularist begins with the instances: you start with particular cases you are confident about — I know there is a hand in front of my face, I know two plus two equals four — and work backward to the criterion that makes those cases come out right. The skeptic says the circle cannot be broken and concludes that neither starting point is available, so knowledge itself is unavailable.
Chisholm himself was a particularist, and particularism has a strong intuitive appeal. The G. E. Moore response to the skeptic is in this spirit: Moore held up his hand, announced that he knew it was there, and declared that he was more confident of this particular fact than of any philosophical principle that would cast doubt on it. If a skeptical argument concludes that we do not know there is a hand in front of us, that is evidence against a premise of the argument, not evidence that we lack hands. The argument's conclusion is less credible than our ordinary certainties. So we use those certainties to constrain our epistemology rather than deriving them from it.
The problem with particularism is that it provides no defense against a well-constructed skeptical scenario. Descartes imagined a demon who produces experiences indistinguishable from the actual world; Putnam imagined a brain in a vat receiving perfect simulations. In these scenarios, the particularist's confident starting points — the hand, the arithmetic — are exactly what is being questioned. The particularist says: I am more confident of the hand than of the skeptical argument's premises. But this relies on confidence being a guide to truth, and the skeptic's whole point is that confidence and truth can come apart. The particularist has a response that works inside ordinary epistemic life but seems to beg the question against the skeptic at the fundamental level.
The methodist approach is equally strained. Whatever criterion you propose — sense experience, clear and distinct ideas, coherence, reliably produced belief — the skeptic will ask why that criterion is truth-conducive. The methodist can say: this is just where justification bottoms out. But then the criterion is stipulated rather than established, which is either a version of particularism (this criterion seems right, in the way that Moore's hand seemed certain) or an acknowledgment that the foundation is arbitrary.
What I find philosophically honest about the problem of the criterion is that it has no clean solution. Every systematic epistemology has a place where it steps into the circle and starts pulling. Coherentism says beliefs justify each other, but the whole web floats free. Foundationalism identifies basic beliefs that justify others without being justified themselves, but the basic beliefs turn out to be either infallible (too restrictive) or fallible (needing justification after all). Reliabilism says beliefs are justified if produced by reliable processes, but evaluating reliability requires knowing which of your beliefs are true. Every approach has a version of the problem coded into it at a different depth.
This is not a failure of philosophy. It is what the problem actually is. The regress is real. The circle is real. What stops epistemology from being impossible in practice is that we do not actually operate at the level where the regress bites. We form beliefs, evaluate evidence, revise in light of new information, and identify errors — all within a framework that we cannot fully step outside of to verify. The framework works, in the sense that it produces beliefs that cohere with each other and that allow us to navigate the world. Whether it is tracking mind-independent truth in some deeper sense is a question that may not have an answer accessible from within it. The problem of the criterion is not solved; it is lived with. The question is whether that should trouble us more than it usually does.