Morton P. Friedman

Feedback and Psychophysical Variables in Signal Detection

144 observers, divided into eight groups of 18 each, were run in a two‐alternative, temporal, forced‐choice auditory‐signal‐detection task. At each of two signal intensities, four levels of information feedback were used. No feedback (NF); correct feedback on every trial (F100), on three‐fourths (F75), or half (F50) of the trials, with incorrect feedback on remaining trials. The results were that (a) NF and F100 led to higher probability of correct responding P(C) than either F75 or F50 for both signal intensities; (b) P(C) for NF was higher under the higher intensity but lower under the lower intensity than for F100; (c) on trials immediately following trials on which observers response and feedback agreed, detection rates were higher and false‐alarm rates were lower than following disagreement trials, whereas these differences were close to zero for F50. It is argued that feedback leads the observer to change his criterion following disagreements. The effect of this variability is to depress the mean d...

Mach Bands in Hearing

Everyone believes that there must be laterally inhibitory nerve nets fed by cochlear potentials, for otherwise it is difficult to explain the extraordinary capacity of the ear in discriminating pitch. Mach bands have not been demonstrated in audition as directly and simply as in vision. We have sought to show the existence of Mach bands (“edge effects”) in direct masking. Noise bands having theoretically infinite attenuation rates outside the passband were generated by computer from 56 sinusoids spaced randomly by frequency. Monaural masked audiograms were obtained for each of four subjects at each of two noise bands (480–580 Hz and 960–1160 Hz) and sensation levels of 20, 30, 40, 50, and 60 dB. Edge effects, as measured by contours of threshold shift, were revealed for all subjects, at every loudness, and are similar for both noise bands. Sharpening is greater at the low‐frequency end of a band and grows nonlinearly with loudness, as does upward spread of masking. An account of these data can be given in...

Comparisons of some learning models for response bias in signal detection

The effects of variations in signal probability and varying degrees of correct feedback on response bias were studied in a yes-no auditory signal detection task. The main finding was that the bias towards saying yes was an increasing function of the frequency of signal feedback events, but did not depend on the correctness of the feedback. Several learning models coupled with a simple psychophysical and decision model yielded predictions about overall biases and certain sequential statistics. Only one model, which can be decribed as an “informational” model, gave a good account of both observed overall biases and sequential statistics. This model assumes the observer’s response bias is strengthened for the feedback-reinforced response when the observer’s sensory information is ambiguous or is contradicted by the feedback information.

Effect of False Information on Responses in a Signal‐Detection Task

In a 2‐alternative, temporal, forced‐choice signal‐detection task, observers received degrees of correct information as to the interval in which the signal had occurred. Groups π(100), π(75), and π(50) received correct feedback following every trial in the proportions 1.00, 0.75, and 0.50, respectively. Group π(0) received no information on any trial. 18 observers were run for 400 trials under each combination of 2 E/N0s and the 4 information conditions. Results were (a) detection rate was greatest for π(100) and π(0) within both E/N0s; but (b) rate of Group π(100)>π(0) for low E/N0 and π(0)>π(100) for high E/N0, in agreement with an earlier result obtained in a Yes‐No experiment. Are S1EjAk states on trial n independent of S1EjAk states on trial n−1 (where i, j, k=1, 2 and Si=stimulus interval, Ej=experimenter feedback to observer, Ak=response of observer)? That is, are transition probabilities stationary? For both E/N0s: (1) all states for π(0) were stationary; (2) a single state (S2E2Ak) was nonstat...

Feature saliency and recognition memory for schematic faces

Memory for schematic faces was studied using a recognition memory paradigm. Ss indicated whether two faces, separated by delays of 6, 12, or 18 sec, were the same or different. Theoretical interest centered on the effects of component feature saliency on response bias and detectability. The main results were that there were fewer misses and more false alarms on the more salient features. Analysis showed that the main effect of feature saliency was on response bias, not detectability.

Perceptual vs conceptual categorization

The study considered whether Ss use the same strategies in categorizing biographical descriptions as in perceptual classifications. A biographical description consisted of a person’s age, income, number of children, and years of education. The Ss were asked to classify these descriptions as residents of two different suburbs in order to compare the results with a previous study using schematic faces. The 123 Ss were assigned to one of three alternative organizations of the descriptions: a table, a “name” organized paragraph in which each person was successively described, or an “attribute” organized paragraph in which each attribute was successively described. Essentially the same results were obtained for schematic faces and biographical descriptions, except for a few differences which were attributed to the use of more realistic categories (suburbs) in the present study. When more realistic categories are used, a S’s prior experience can influence which attributes he emphasizes and possibly his formation of an abstract prototype.

The role of cue validity in stimulus compounding

Abstract Several alternative models for cue selection and responding in stimulus compounding situations are discussed. All models assume that the previous validities of the component cues of a compound are the principal determiners of cue selection and transfer, where the validity of a cue may be defined as the accuracy of the predictions which can be made on the basis of that cue alone. In an experiment designed to test the models, fifty human subjects received orobablistic reinforcement training on a set of five component cues. They were then tested on novel two-element compounds of the training cues. Two variants of a “scanning model” were found to give a good account of cue selection on the compounding tests.

Tests of a mixed model for paired-associates learning with overlapping stimuli☆

Abstract This research was concerned with the overlap problem of discrimination-learning theory—how discriminations among stimuli with overlapping cues are learned. One hundred and twenty subjects learned a paired-associates list in which the overlap among the compound stimulus members of the items was varied to produce a wide range of component cue validities and intratask transfer effects. After learning the list, subjects were tested on the component cues alone and on a series of novel compounds which overlapped the training compounds. An analysis of these transfer tests indicated that subjects tended to respond on the basis of the validity or relevancy of the component cues, and that frequency of presentation of the cues during learning was not a critical variable. An extended version of a mixed model of stimulus-sampling theory yielded a good quantitative account of acquisition and transfer performance.

SELECTIVE ATTENTION IN CHILDREN WITH LEARNING DISABILITIES

49 children, aged 11 to 14 yr., a learning disabled group and a normal group, performed a primary, reading-like, card-sorting task. After they completed the primary task, they were tested for memory of incidental materials presented during learning. While the normal children showed better recall of incidental materials related to the primary task, 24 children with disabilities showed superior recall of material irrelevant to the primary task. The results were discussed in terms of alternative “motivational” and “developmental lag” interpretations.

Mental operations on number symbols by children.

Children 4 to 6 years of age made choices among colors that had been associated with numbers. Prior to learning four number-color associations, pretraining on counting and ordering objects was given to half of the Ss. Associations were established by either absolute (paired associate) or relative (choice discrimination) methods, and then tests involving “more or less” questions on each of the possible six color pairs were given. Numerical associations were acquired best with absolute training or counting-ordering pretraining. Using the children’s justifications for their responses, the test data were fit by Estes’s (1966) scanning model; a response strength model based on Luce’s (1959) axiom failed. Young children apparently possess an ordered numerical scale, can map this scale onto other nonquantitative symbols, and can generate, compare, and make ordinal decisions using these symbols.