The sensory input from an object changes constantly as viewing conditions vary. A plate viewed from an angle projects an ellipse on the retina, yet we perceive it as circular. A friend's face in shadow reflects less light than a white wall in sunlight, yet we perceive the face as lighter than the wall. Perceptual constancy — the ability to perceive stable object properties despite varying proximal stimulation — is one of the most fundamental and impressive achievements of the perceptual system.
Size Constancy
An object at twice the distance projects a retinal image half as large, yet perceived size remains relatively stable. Size constancy depends on distance information: the visual system "takes distance into account" when interpreting retinal image size. Emmert's law formalizes this for afterimages — an afterimage projected on a far wall appears larger than one projected on a near surface, because the brain scales the constant retinal size by the perceived distance.
S = θ × d
Size constancy breaks down when distance cues are eliminated (as in the reduced-cue conditions of classic psychophysics experiments) and can produce striking illusions when distance cues are misleading (as in the Ames room, where a distorted room geometry creates the illusion of dramatically different sizes for people at different distances).
Shape Constancy
Shape constancy maintains the perceived shape of an object despite changes in the angle of view. A door opening from rectangular to a narrow sliver maintains its perceived rectangular shape. This requires the visual system to compensate for the projective transformation caused by slant — essentially performing an inverse perspective projection using information about the surface's orientation in three-dimensional space.
Lightness and Color Constancy
Perhaps the most computationally demanding form of constancy, lightness constancy requires the visual system to distinguish changes in illumination (which affect the amount of light reflected from all surfaces in a scene) from changes in surface reflectance (the intrinsic property of a surface). A piece of coal in sunlight reflects more light than a piece of chalk in shadow, yet we correctly perceive the chalk as lighter.
The visual system achieves this by comparing the luminance of a surface to its surround (ratio-based processing), by estimating the illuminant using cues such as specular highlights and the overall distribution of luminances in the scene, and by using prior knowledge about typical surface reflectances. Color constancy employs similar computational strategies to maintain stable perceived color despite changes in the spectral composition of illumination.
Constancy requires solving an underdetermined inverse problem: the retinal image is the product of illumination and reflectance, and recovering either one from the product alone is mathematically impossible without additional constraints. The visual system uses multiple constraints — spatial comparisons, temporal information, prior knowledge about illuminants and surfaces, and information from multiple cues — to approximate a solution. This Bayesian approach to constancy explains both its remarkable success and its systematic failures (illusions).
Speed Constancy
Objects at different distances moving at the same physical speed produce different retinal velocities (nearer objects produce faster retinal motion). Speed constancy — perceiving the true physical speed rather than the retinal velocity — relies on distance information and is weaker than size or shape constancy, showing significant departures from perfect compensation.
Neural Basis
Constancy mechanisms operate at multiple levels of the visual hierarchy. Retinal adaptation provides a basic form of luminance adaptation. Cortical mechanisms in V1, V4, and inferotemporal cortex show responses that correlate more closely with perceived surface properties than with the raw physical stimulus, demonstrating that constancy computations are implemented in cortical visual processing.