Reading 3 min
Perceptual Constancy: Unconscious Inference or Perceptual Organization?
A sheet of white paper lying half in shadow and half in sunlight reflects very different amounts of light from its two halves. Yet you see one white sheet, partly shadowed, not a gray sheet stuck to a white one. A friend walking away from you projects a smaller and smaller image on your retina. Yet she doesn't seem to shrink. She seems to recede. Colors, sizes, and shapes stay stable across enormous changes in what reaches the eyes.
Vision science calls this perceptual constancy, and it's one of the field's central problems. Any account of vision has to explain how a changing input yields a stable world. The problem isn't marginal. Without constancy, every step you took would change the size of everything around you, and every passing cloud would recolor it.
The oldest and most influential answer is that the visual system infers. It takes the changing input, combines it with information about distance, orientation, and lighting, and calculates the object's real size, shape, or color. Because we're not aware of this calculation, it's often called unconscious inference. The idea has a long history, running from medieval optics through Hermann von Helmholtz in the 19th century to much of current computational vision (Helmholtz [1867] 1925).
Gary Hatfield traces that history and asks what the word "inference" is doing in it (Hatfield 2002). Is the visual system literally reasoning, drawing conclusions from premises? Or is "inference" a metaphor for processes that don't resemble reasoning at all? Hatfield argues that the literal reading faces serious difficulties, and that the question of what kind of process underlies constancy remains open.
The alternative to inference is to treat constancy as a feature of how perception is organized. On this view, the stability of things isn't calculated after the fact. It belongs to the structure of the perceptual field from the start. The paper is seen as white under a shadow because shadow and lighting are perceived as such, as conditions of seeing, not as properties of the paper.
Maurice Merleau-Ponty developed the second view in detail (Merleau-Ponty [1945] 2012, pt. 2, ch. 3), and the distinction matters for research. If constancy is inference, the task is to find the hidden computation. If constancy is organization, the task begins with describing the structure the computation is meant to produce. Learning to describe perception's stability without assuming hidden calculations lets you ask of any model which of these tasks it has actually taken on.
References
- Hatfield, Gary. 2002. "Perception as Unconscious Inference." In Perception and the Physical World: Psychological and Philosophical Issues in Perception, edited by Dieter Heyer and Rainer Mausfeld, 113–43. Chichester: Wiley.
- Free: Author's PDF
- Publisher: doi:10.1002/0470013427.ch5
- Helmholtz, Hermann von. (1867) 1925. Treatise on Physiological Optics. Vol. 3. Translated from the 3rd German edition and edited by James P. C. Southall. Rochester, NY: Optical Society of America.
- Merleau-Ponty, Maurice. (1945) 2012. Phenomenology of Perception. Translated by Donald A. Landes. London: Routledge.
- Free: None. Landes prints the French pagination in the margin; cite by part and chapter so learners using Smith can follow along.
- Publisher: Routledge
- Free companion: SEP, "Maurice Merleau-Ponty" (Toadvine), §3 · IEP, "Merleau-Ponty, Maurice" (Reynolds)