Neural theories of simple visual discriminations.

Simple visual discriminations such as detection of presence and recognition of small differences in the spatial and temporal patterns of objects have been explained in some detail in recent years by variants of the physical quantum theory. The present paper summarizes the status of an alternative group of theories, which is classed as neural, primarily to distinguish them from the physical quantum theory, and compares the two groups of theories. The neural theories are models of the visual system as a whole and include assumptions concerning the transformation of stimulus to sensory events, and the decision processes involved in discrimination. They are neural by inference only, and rarely specify the type of neural activity or the neural site at which the events occur. The bulk of the paper is devoted to a description of specific neural theories, attributable to the writer, concerned with (a) the detection process and the form of threshold probability data; and (b) the relation of threshold ΔI/I to I, and to the spatial and temporal characteristics of the stimulus. These theories are compared with selected related theories. The Swets–Tanner decision theory is contrasted with the present detection threshold theory and various lines of evidence are summarized. It is shown that data previously used to support the Swets–Tanner theory are inconclusive. Chief evidence against the Swets–Tanner theory comes from data obtained in a stimulus-comparison experiment, and from the contradiction implied when the relations assumed to exist between stimulus and sensory magnitudes over the threshold range are extended to a wider range. Evidence is presented which is in favor of the present theories and opposed to the neural theories of Graham and his colleagues which relate threshold ΔI/I to the spatial and temporal characteristics of the stimulus. This evidence comes primarily from detection thresholds for stimuli composed of twin spatial points or double temporal pulses, which the Graham theories cannot explain. In addition to spatial and temporal summation, double spatial and temporal stimuli exhibit probability summation whenever the sensory correlates of the stimuli exhibit twin modes. A 5–6 cps scanning mechanism is shown to exist which breaks up prolonged temporal stimuli and provides probability summation among the separate parts.