Edmund Gettier's 1963 paper showed that justified true belief is insufficient for knowledge: one can have a justified true belief that is only accidentally true, and such beliefs are not knowledge. Nozick's response is to replace the justification condition with two counterfactual conditions. Sensitivity: if P were false, S would not believe P (roughly: S's belief is not detached from the facts). Adherence: if P were true, S would believe P (roughly: S's method is reliable enough to catch P when it's there). Together, these conditions ensure that S's belief tracks the truth — it is sensitive to and moves with the actual state of affairs.
The tracking analysis has an important consequence for scepticism. The sceptic argues: you cannot know you have hands, because you cannot rule out that you are a brain in a vat. Nozick's response is that knowledge does not require ruling out all possibilities of error. In the actual world, you have hands; if you didn't have hands, you would know it (your perceptual system tracks the presence and absence of hands in normal conditions). The sensitivity condition is met. That you cannot rule out the brain-in-a-vat scenario does not undermine your knowledge, because in the actual world, that scenario doesn't obtain, and your perceptual states track the actual situation.
Nozick extends the tracking account to knowledge through inference and testimony. Knowledge is method-relative: whether S knows P depends on which method M S uses to form the belief. S knows P via M if M is a reliable tracker — if, had P been false, M would not have led S to believe P. Different methods (perception, memory, testimony, inference) have different reliability profiles, and knowledge claims must specify the method. This generates a more nuanced picture than simple reliabilism: not all reliable methods confer knowledge in all circumstances, because reliability must be assessed relative to the counterfactual situations most relevant to the belief in question.
The tracking theory of knowledge is developed in chapters 3–4 of Philosophical Explanations and has generated a large critical literature. Key objections focus on the "grandmother" case (where sensitivity fails for deductive closure) and the problem of specifying the relevant counterfactual situations.
