Perception is widely regarded as the most basic source of empirical knowledge, yet its epistemic role is deeply contested. Direct (naive) realists hold that perception gives us immediate, unmediated awareness of mind-independent objects. Indirect realists argue that we perceive only mental intermediaries — sense data, appearances, or representations — from which we infer the external world. The sense-data theory, championed by Russell and Ayer, makes perceptual justification explicit but opens a gap between experience and reality that skeptics exploit. Contemporary disjunctivism attempts a middle path: in veridical perception, the subject is directly related to the object, while in hallucination, the mental state is of a fundamentally different kind.
Consider the argument from illusion: a stick looks bent in water, yet the stick is straight. What, exactly, are you perceiving? Each theory of perception gives a different answer, and each answer has consequences for how much of our empirical knowledge is secure.
From your prerequisite on sources of knowledge, you know that perception is the primary source of a posteriori knowledge — the channel through which we learn about the world from experience, as opposed to the a priori knowledge we can have independently of experience. But saying "perception is a source of knowledge" leaves a deep question unanswered: *how* does it work? When you look at a cup of coffee and come to know that there is a cup of coffee in front of you, what exactly is happening between your eyes and your knowledge claim? Theories of perception are competing answers to this question, and each answer has major consequences for how secure empirical knowledge is.
The most intuitive answer is direct (naive) realism: perception gives you immediate, unmediated contact with mind-independent physical objects. When you see the cup, you are directly in relation with the cup itself — a real external object. The mind is not an intermediary; it is a transparent window onto the world. This view matches how perception *feels* from the inside, and it has the philosophical advantage of explaining why perceptual knowledge is straightforwardly about the world. But it faces a well-known challenge: the argument from illusion. A stick partially submerged in water looks bent, though it is straight. A tower in the far distance looks small, though it is large. If you are directly perceiving the stick, what are you perceiving when it looks bent? If the object of perception just is the physical stick, perception seems to be misrepresenting it — yet what you are directly aware of is something with a bent appearance.
The sense-data theory (associated with Bertrand Russell, G. E. Moore, and A. J. Ayer) responds by interposing a mental intermediary. What you are immediately and directly aware of in perception is a sense datum — a mind-dependent representation with its own intrinsic properties. The bent-stick sense datum and the straight-stick sense datum are both real, as mental objects; they just differ in their properties. This makes perceptual error easily intelligible: you have a sense datum with a bent appearance, and the physical stick doesn't match. But the theory creates a new problem in place of the old one: if what you directly perceive is always a sense datum — never the physical object itself — how do you know your sense data track external reality? You cannot step outside your perceptual representations to compare them against the world. This opens exactly the gap the skeptic exploits: your experience could be exactly as it is whether or not the external world exists.
Disjunctivism, a more recent position, attempts to preserve the intuitive advantages of direct realism without making sense-data theory's concessions to error cases. Its key move is to deny that veridical perception (genuinely seeing the cup) and hallucination (vividly seeming to see a cup that isn't there) are the same kind of mental state with different accuracy. Instead, they are fundamentally different states: in veridical perception, the subject is directly in a relation with the physical object — this is a factive, world-involving state. In hallucination, the subject has a phenomenologically similar but metaphysically different experience with no such external relation. The implication is that illusion and hallucination cases do not generalize to ordinary perception: they show only that non-veridical experiences exist, not that all perception is indirect. This preserves direct realism for the good case while accommodating error cases without retreating to the sense-data intermediary.
Topics in reflective domains aren't scored by quiz answers. Read, reflect, and mark when you've thought it through.