Bias Functions and Operating Characteristics of Rats Discriminating Auditory Stimuli

Abstract

Rats were trained to discriminate between two bursts of random noise that differed in intensity. In a two-lever, discrete-trial procedure, correct responses were reinforced with brain stimulation, and incorrect responses produced timeout. Responding was studied as a function of the decibel difference between the stimuli, the probabilities of presenting the stimuli, the relative duration of timeout consequent upon the two types of incorrect responses, and the absolute duration of timeout consequent upon incorrect responses. The results showed that the distribution of responses between the two levers depended upon the stimulus probabilities, but were independent of either the absolute or relative durations of timeout. When the stimulus probabilities were varied, the response probabilities did not match the stimulus probabilities; when the relative durations of timeout were varied, the animals did not obtain the maximum rate of reinforcement per unit time. Instead, the animals distributed their responses so as to obtain the maximum number of reinforcements at each level of discrimination. In addition, the level of discrimination increased as a function of the decibel difference between the stimuli.

Towards a behavioral theory of bias in signal detection

“…In general, correct responses are reinforced with, perhaps, food or brain stimulation for animal subjects and money, points, or feedback for humans. Sometimes payoffs for correct responses are given each time a correct response is emitted (e.g., Hume, 1974aHume, , 1974b & Irwin, 1974), and sometimes it is given intermittently on VI schedules (McCarthy & Davison, 1979)' or on probabilistic variable-ratio schedules (e.g., Elsmore, 1972;McCarthy & Davison, 1979, 1980a, 1980bStubbs, 1976). Usually, incorrect responses have no consequence or are punished in some way (e.g., time-out with animals; Hume, 1974b).…”

Optimal response biases and the slope of ROC curves as a function of signal intensity, signal probability, and relative payoff

“…The corresponding decrease~in d '~ere f~~m approximately 2.0 to 0.5. At Signal tntensines exceeding 8 or 10 dB, these animals achieve a large proportion of correct trials; points lie in the upper left corner of the ROC space where matching and maximizing response strategies are not readily distinguishable (Hume & Irwin, 1974 Terman and Terman (1972) showed, on the basis of a different analysis, that their animals "deviated systematically from a strict matching strategy as the discrimination problem became more difficult." They did not, however, demonstrate to what extent their animals' response biases became more maximal with decreasing signal detectability.…”

“…The animals' sensitivity to increments in the intensity of random noise increased with the magnitude of the signal, and was independent of variations in response bias induced by varying either the signal probability or the relative payofffor correct responses (cf, Stubbs, 1968;Irwin & Terman, 1970;Terman, 1970;Clopton, 1972;Terman & Terman, 1972;Hume & Irwin, 1974;Hume, 1974).…”

“…This result directly supports previous findings (Hume, 1974) and the conclusions from that and a previous study, namely that in the present psychophysical procedure; animals will more closely optimize the probability of a correct trial, and hence the probability of a single brain stimulation, than they will optimize the overall number of brain stimulations. This, in turn, they will more closely optimize than the rate of obtaining brain stimulation in time (Hume & Irwin, 1974).…”

“…Both ROC curves and their complementary bias functions have been obtained in stud ies with' animals *This research constituted a portion of a thesis submitted in partial fulfillment of the requirements for the PhD degree at the University of Auckland. The research was supported in part by grants from the New Zealand University Grants Committee and the Scientific Distribution Committee to R. J. by varying either the signal probability (Clopton, 1972;Elsmore, 1972;Hodos & Bonbright, 1972;Terman & Terman, 1972;Hume & Irwin, 1974;Hume, 1974) or the symmetry of the reinforcement contingencies (Stubbs, 1968;Hume, 1974). To date, only Terman and Terman (1972) have considered how response biases may vary as a joint function of the signal probability and the stimulus parameters.…”