Properties of taking-decision function in psychophysical tests

Fundamental features of taking-decision functions have been established from the expert evaluations of perceptual image quality. (1) The functions of different stimuli consist of high-reliability tolerance plateau and one or two semi-logarithmic portions. (2) A multistimulus decision is a product of single-stimulus functions interacting like the probabilities of independent events. (3) The highest-reliability plateau of a stimulus optical property is recorded in the human memory and may be changed in a statistical manner by learning or experience. (4) A simple adaptive process of feedback comparison of neuronal input with the memorized data may underlie the uniform semi-logarithmic performance out of the high-reliability tolerances. A quantitative method was elaborated to extract the complete set of independent stimuli exhausting the experimental spread of decision response and to detect their relations to the brain response. The data are suggestive of some general principles of fuzzy-logic performance in neural networks