True by luck
On the axis: Justification of belief · a belief true by luck
The classic definition holds knowledge to be a belief that is both true and justified. Yet one can believe something for good reasons, and have it be true, without thereby knowing it: the stopped clock we glance at just at the hour it still shows, the rock we take from afar for a sheep while a real one grazes behind it. The belief is true and justified, but it hits the mark by accident. Must we then redefine what knowledge is, or tighten what we are entitled to demand of a justification, until it links belief to truth otherwise than by luck?
Justification
If all knowledge of the world comes from experience, there must be that are not inferred: Russell calls them hard data, what resists the utmost effort of doubt, and out of them he builds everything else by : objects, other minds, the past. The difficulty is that these data are private and momentary, so that the public world has to be rebuilt at great cost. Russell admits this and retreats: in 1912 he held the data to be indubitable, whereas (1948) grants them only a probable base, weakening the foundation by that much.
Our Knowledge of the External World (1914), lectures III-IV · Human Knowledge (1948), part II