November 1974
A Listing of Supposed Operations on the Data of Sense
J. J. Gibson, Cornell University
The World Wide Web distribution of James Gibson’s “Purple Perils” is for scholarly use with the understanding that Gibson did not intend them for publication. References to these essays must cite them explicitly as unpublished manuscripts. Copies may be circulated if this statement is included on each copy.
To say that theories of sense perception are “sensation-based” is to say that they are hypotheses about the operations that must occur if sensory inputs are to become perceptions of the environment. The different theories emphasize different operations and it would, therefore, be useful to attempt a sort of inventory of all the operations that have been proposed, ancient or modern, so as to compare the theories. They have been classified generally in such terms as nativism, empiricism, rationalism and gestalt theory; but perhaps an inventory would be more revealing. The operations have also been classified as “mental” or “physiological” but I shall neglect this dichotomy.
By “the data of sense” I mean both the inputs of the sensory nerves and the supposed sense impressions that immediately result, that is, the earlier results of the stimulation of receptors as distinguished from later processes. Perception is supposed to come after sensation, and sensation is usually supposed to be a necessary (but not sufficient) cause of perception. By “data” I mean the incoming data from a sense organ; I do not mean available or potential data that the organ might have sent to the brain but on this occasion did not.
Although different theories of sensation-based perception emphasize different operations on the data all of them must postulate certain operations. I can think of three. There has to be, first, some kind ofselection from the mass of simultaneous inputs, some kind of internal attention, or filtering, or exclusion. There must be, second, some sort of classifying of the inputs, or sorting, or categorizing of them. This necessity can be put in another way by saying that the same input on a later occasion must be “recognized” as the same. And there has to be, third, some kind of operation on the sensations that compensates for the changes due only to the movements of the observer himself, some operation that explains the “constancy” of the perception of an object despite the “inconstancy” of the sense impressions corresponding to it. In the case of vision this is bound up with the operation that supposedly adds depth to the retinal images.
Historically, theories of perception might be classified in three ways according as the emphasis is put on association, inference, or organization. All theories, whatever their emphasis, appeal to memory or the “influence of past experience on perception.” Association theory rests on the theory of memory most heavily, perhaps, but both the theory of unconscious inference as formulated by Helmholtz and the theory of sensory organization as formulated by the Gestalt psychologists could not get along without the notion of traces, records, images, or engrams and the notion of storage, retention, or the preservation of the past in the present.
Following is a tentative list of perceptual operations under three loose headings.
Operations Emphasizing Association with the Data
— The evoking of images of past experience.
— Supplying to the selected “core” of present sensations a “context” of memory images and feelings. This has been described as the supplementing or enriching of bare sensations (e.g. Titchener). This explains meaning.
— The assimilation of sense impressions to memories, or the fusion of sensations and images. This is supposed to explain the effect of preconception on perceptions (e.g., Kant).
— The fusion of images or memories with one another to form concepts. Also the “accommodation” of old concepts to new impressions (Müller, Bartlett, Piaget).
— The differentiating of inputs by the associating of different responses to them. Whether or not this can explain the categorizing of sense impressions, however, has been debated (“acquired distinctiveness”).
Operations Emphasizing Inference from the Data
— The rational interpretation of data considered as clues. According to Helmholtz the process is unconscious. (The data of sense are often termed “cues”).
— The “reading” of data considered as signals, or signs, or indicators, by analogy with human messages.
— The learning of assumptions about the data, or expectations. If the assumptions (expectations, preconceptions) are supposed to be unlearned rather than learned the theory is classed as “nativistic.”
— The making of computations based on the data. The rational or logical operations of a computer upon its inputs, together with its “memory”, are taken to be at least analogous to the operations of a brain on its sensory inputs.
Operations Emphasizing Organization of the Data
— The imposing of structure on sensory data.
— The “spontaneous self-distribution” of sensory processes in the cortex of the brain (Köhler). This is supposed to explain visual form perception.
— The grouping of sensory data in accordance with the laws of organization (Wertheimer).
— The completion or closing of contours and the filling-in of gaps in accordance with the figure-ground phenomenon
— The increasing precision (pragnanz) of traces during memory, and the tendency toward “good” forms.
— The accumulation of successive images during memory, as in Koffka’s theory of a “column” of traces laid down in the brain which interact with one another.
Combinations of Operations
Some theorists of perception, Piaget for example, are neither strict empiricists nor rationalists nor gestalt theorists but combine operations, and accept association, inference, and organization as compatible with one another. What should be noted, however, is that sensory inputs or sense data are presupposed as the basis of perception in all current theories. The assumption of a stimulus flux with both change of structure and concurrent nonchange is an entirely different starting point. The above listed operations become irrelevant if sense-data are irrelevant. They will serve, however, as the basis of a critique of current theories of perception.