Perception refers to the experience of obtaining sensory information about the world of people, things, and events and the underlying processes.
Successive British philosophers from Hobbes to Berkeley to Mill, being empiricist epistemologists, considered perceptions as learned assemblies of more elementary experiences. Scientific study began in 1838, when J. Müller identified the separable sensory modalities; to H. von Helmholtz (Müller’s student), a modality provides elementary sensations from specific receptors, each sensitive to some physical stimulation. Thus, the eye’s retina contains cones sensitive to long, middle, or short wavelengths (L, M, S), providing red, green, or blue points of color sensation, respectively. (Graphic artists had long used arrays of a few kinds of dots to portray all colors.)
Since light at the eye is two-dimensional (Figure 1A), depth cannot be directly sensed in this approach. Fortunately, normal environments offer depth cues (Figures 1C and 1D are simple examples), well studied by fifteenth-century artists. To classical theory, depth perception derives from memory associations between such cues and accompanying actions (reaching, eye movements, etc.). Cues, as dissected by receptors (symbolized in Figure 1E), plus learned unconscious rules relating objects to their retinal images, let perceivers unconsciously infer various object attributes. Thus, although the size, shape, and brightness of an object’s retinal image change with distance, viewpoint, and illumination, and although saccades (glances) fracture what lies before us, the correct use of the cues (including eye-movement commands) provides size constancy, shape constancy, color constancy, transsaccadic constancy, and so forth. Misperceptions, or illusions, are presumably erroneous unconscious inferences (e.g., perceiving something as larger because of misperceiving it as farther).
Perceptions should therefore be predictable from a pointwise analysis of the sensory pattern, plus from a list of the probabilities of that pattern’s alternative sources in the perceiver’s past environmental encounters. (Most displays do indeed reflect such probabilistic or Bayesian principles; see Geisler [(2008)] for a recent review.)
However, the classical theory has since expired, epistemologically because some animal species respond to depth without prior visuomotor experience, and scientifically because the supposedly independent receptors are actually interconnected, not independent. Before facing that critical issue, however, the theory faced recent consequential competitors.
To Gestalt theorists (notably M. Wertheimer, W. Köhler, and K. Koffka), visual systems respond directly to an overall configuration (or Gestalt) of sensory stimulation, not by pointwise analysis. The primary visual units are shaped regions, as in Figure 2A (Rubin’s figure-ground phenomenon): there, only one region (urn or pair of faces) is shaped figure at any time, the other being shape-free ground, so laws of figure-ground organization determine what we perceive. By the law of good continuation, for example, a familiar number is concealed and revealed in Figures 2B and 2C, respectively, and the apparent cube in Figure 2E is compellingly flat in Figure 2D. Such phenomena were offered as evidence for isomorphic (same-shaped) configured current-flow cortical processes, as opposed to associative learning.
However, Gestalt theory has not made its case on isomorphism; its belief that figural organization must precede any effects of familiarity is contradicted by Figures 2I and J, which are examples of denotative shapes established by M. Peterson and B. Gibson; moreover, so-called organizational laws themselves may simply reflect learned environmental likelihoods, as E. Brunswik proposed, and as animals’ protective coloration (Figure 2F) suggests. The Gestalt phenomena themselves remain important in making visual displays (Z. Pylyshyn), in improving search procedures in airports and x-ray clinics (J. Wolfe), and in learning how our perceptual systems work, so quantifying organizational laws continues (e.g., J. Pomerantz, M. Kubovy).
More broadly, different Gestaltist-like efforts toward a quantifiable principle, assuming that we perceive the simplest structure that fits each stimulus pattern (cf. Figures 2D and E), were attempted separately by F. Attneave, by the present author (with E. McAlister), by E. Leeuwenberg (1960) and colleagues, and most recently by Z. Pylyshyn (2008).
However, a simplicity principle cannot be simple, for reasons discussed here after noting the classical theory’s other major opponent.
Opposing both Helmholtzian and Gestaltist approaches, J. J. Gibson argued that natural environments provide invariants of stimulus information (notably, the optical expansion pattern of an approaching surface varies with its slant, approach rate, etc.). Such invariants could enable direct perception of objects and layouts, making the classical inferences from depth cues and the Gestaltists’ organizational isomorphism unnecessary. This approach has generated sophisticated analyses of visual information (e.g., J. Todd) but ignores the Helmholtzian and Gestalt phenomena: Depth cues and organizational factors certainly do appear to work, at least in pictures and movies (for which Gibson’s proposed invariants simply fail to apply; see Cutting, 1987).
Moreover, like Gestalt theory, direct perception theory fails to use what we now know about eye and brain.
As noted earlier, Helmholtz’s assumptions about independent receptor neurons were elegant but wrong. D. Hubel and T. Wiesel, using microelectrodes in 1960, found individual cells with receptive fields selectively sensitive to local shapes and motions. Moreover, the retina and several levels of its cortical projections are mutually connected in both directions, providing top-down effects on sensory input (see Wandell Dumoulin & Brewer, 2007 for recent review).
Well before such research actually changed our view of the nervous system, D. Hebb had argued persuasively that neurons that are coactive produce cell assemblies and act as pattern-sensitive units, and that repeated serial activities yield phase sequences that unfold over time. Given such forward-aiming pattern-sensitive actions and the top-down interactions now amply verified, perceiving is not a passive registration process; it is a multilevel mix of purposeful attentional behaviors.
Only the retina’s tiny central region, the fovea, resolves fine detail. To see more, the viewer executes rapid intentional eye movements (saccades), with sequences of about 5 per second, following preselected routes. After such a brief moment of high detail, the region just looked at remains only as a simplified encoded memory (and/or as part of the impoverished peripheral surrounding). Detail not deliberately stored in working memory, as G. Sperling showed in 1960; it is usually lost with that glance. And it is the viewer’s brain that elects what to look or listen for next, and then what to encode first in the next glance: So perception starts with motivated attentional processes, not with the retinal mosaic of Figure 1. And limited central selection, peripheral paucity, and sparse transfer across glances may provide a multistage simplicity principle that accommodates the likes of Figure 2G.
Research now explores how attentional and saccadic routes are each programmed and with what consequences (see Figure 3A); it explores how looking behaviors are shaped both by the viewer’s anticipations about the events being attended and about the looking behavior itself (as in Figure 3B) and examines how fovea and periphery contribute to successive integration (Figure 3C). Perception is a purposeful sequential activity, and such research should accompany brain imaging in order to help the latter tell us how our systems work.
*NOTE: Authors who are cited in the foregoing text and are not listed here are given in Hochberg, 1998, Chs. 1, 11.
Related Credo Articles
Perception refers to the experience of obtaining sensory information about the world of people, things, and events, and to the underlying...
From the Latin percipi o (both “gaining knowledge through the senses” and “apprehension with the mind”), this term denotes the act or ability of...
“Perception” was understood by British empiricists to signify either (1) consciousness generally, (2) its most elementary and passive species, or...