Philosophy of Mind Part 10: Perception

Direct Realism — The Common-Sense View

The natural starting point: when you see a tomato, you see the tomato. Not an image, not a representation — the red, round physical object itself. This view, called direct realism or naïve realism, has obvious appeal. It is what perception feels like from the inside. We seem to be in unmediated touch with the world.

Direct realism enjoyed a major revival in 20th century thanks to J.L. Austin (Sense and Sensibilia, 1962) and was developed analytically by John McDowell and Bill Brewer in the 1990s and 2000s. Its central claim: the objects of perception are mind-independent physical objects and their properties.

The Argument from Illusion

The classical challenge to direct realism. A straight stick half-submerged in water looks bent. The white wall under sodium lighting looks yellow. Press your eyeball gently and the world doubles. In each case you are aware of something bent, yellow, or doubled — but no such bent stick or doubled wall exists in the physical world. So whatever you are immediately aware of cannot be the physical object. There must be an intermediary — a "sense datum."

Sense-Data Theory

The dominant early-20th-century response (Bertrand Russell, G.E. Moore, H.H. Price, A.J. Ayer): perception always involves direct awareness of sense data — mind-dependent particulars whose properties are exactly as they appear. Physical objects are inferred or constructed from regularities among sense data.

The view crashed under accumulated objections. Sense data are awkward entities: are they physical, mental, or neither? Where are they located? Can they have properties (like "being elliptical from this angle") that no possible physical object has? Wilfrid Sellars's famous "Myth of the Given" attack (1956) charged that sense data theorists conflated causal and justificatory relations — turning the sensible into a foundational basis for knowledge in a way that doesn't work.

Adverbialism & Representationalism

Adverbialism (Roderick Chisholm, Frank Jackson in early work) avoids sense data by reanalyzing perceptual reports. "I see a red patch" is misleading; better: "I am appeared to red-patch-ly." There is no patch as object; there is only a particular way I am being affected. Linguistically clean, but most philosophers found it incomplete for explaining the rich structural content of experience.

Representationalism (Fred Dretske, Michael Tye, William Lycan) became dominant in the 1990s. Perceptual experience is a representational state with truth-evaluable content; it can be accurate or inaccurate. The phenomenal character of seeing red is exhausted by the representational content "there's something red in front of me, located thus-and-so." This integrates perception cleanly with theories of intentionality (Part 6) and sidesteps sense data — and conveniently maps onto the brain's representational architecture.

Disjunctivism — Direct Realism Strikes Back

The most interesting recent direct-realist position. J.M. Hinton introduced and John McDowell popularized disjunctivism: veridical perception and hallucination, despite being introspectively indistinguishable, are fundamentally different mental kinds. In genuine perception, the immediate object is the physical thing; in hallucination, something else is going on (a seeming, a "mere appearance"). The argument from illusion fails because it assumes a "common factor" — and disjunctivists deny one.

The cost is admitting that subjective indistinguishability does not entail sameness of mental state — counterintuitive but defensible. The benefit is preserving the common-sense thought that genuine perception puts us in touch with the world itself.

Predictive Processing — The Brain as Prediction Machine

The dominant framework in current computational neuroscience, developed by Karl Friston, Andy Clark, Jakob Hohwy, and others. The brain is not a passive receiver of sensory inputs; it is constantly generating top-down predictions about what should be sensed next, based on a learned model of the world. Perception is the brain's best guess about the causes of its current sensory data, refined by minimizing prediction error.

What we experience is, in a sense, a controlled hallucination — the brain's hypothesis, constrained but not constituted by sensory input. The predictive view explains many puzzling phenomena: the ambiguity flips of bistable images, the persistence of perceptual illusions even when we know they're illusions, the role of attention as precision-weighting of prediction error, the experience of presence vs absence.