John A. Hebert

Training and testing effects in the generalization of a voluntary response

The effects of amount of training on the generalization of a voluntary response were investigated under two testing procedures—one in which the original training stimulus, a 100-g weight, was presented five times more frequently than the other test stimuli, and another in which the original training stimuluswas presented with the same frequency as the other test stimuli. No effects of amount of training were observed, in support of previous findings. However, the two procedures produced different-shaped gradients of stimulus generalization, also in support of previous findings.

Generalization of a voluntary response as a function of presentation frequency of the training stimulus in testing

Frequency of presentation of the original training stimulus in the test phase of a study of human voluntary generalization was found to affect the form of the generalization gradient on a weight dimension. That is, in a situation in which all of the additional test stimuli are larger than the original training stimulus, and all stimuli are presented with equal frequency, a tendency to choose stimuli toward the middle of the test range as the original was observed However, if the original stimulus was presented more frequently than any of the other stimuli in testing, Ss tended to choose the original “correctly,” yielding a “typical” unidirectional generalization gradient.

Latency patterns in category Judgments

Adult human Ss were required to learn a conditional discrimination in which there were three distinct visual stimuli (squares) to be matched with three corresponding response buttons. When Ss had learned the task to criterion, they were presented with a test series of 11 stimuli, including, and spaced symmetrically around, the three training stimuli. Ss were told to place each successive stimulus in one of the three categories represented by the choice of three response buttons. Response latency and category choice were recorded. The prediction that Ss

Effects of field-dependency on weight comparisons

20 field-independent and 20 field-dependent Ss were selected on the basis of performance on the Rod-and-frame Test and were asked to estimate a comparison weight in the context of two different original weights. Visual cues were controlled. 10 Ss in each group estimated a heavier weight which was paired on 2 consecutive trials with 1 of 2 lighter weights and the remaining Ss estimated a lighter weight in the context of two heavier weights. The difference in estimates of the same comparison weights served as an indication of contextual influence. The field-dependent Ss, estimating lighter weights, were influenced by context to a significantly greater extent than the other groups. The field-dependency dimension may affect performance on a variety of different perceptual tasks.

Adaptation-level and theory of signal detection: an examination and integration of two judgment models for voluntary stimulus generalization.

Abstract An examination of two judgement models, adaptation-level (AL) theory and the theory of signal detection (TSD), was made in terms of their ability to handle data from several voluntary generalization studies. Both models were found to be somewhat inadequate, though in different ways. AL theory was found to be unable to predict amount or slope of the generalization gradient, while TSD was seen to be unable to account for shifts in judgment due to context and frequency effects. A combined AL-TSD model was presented which combines the advantages of each model and offers a good account of both choice and latency data in voluntary generalization

Spence’s Theory of Discrimination Learning: A Re-Evaluation

Spence’s theory of discrimination learning (1936, 1937) which was orginally proposed to account for the phenomenon of transposition, is, when combined with certain postulates of the Hullian system, capable of dealing with an impressive range of empirical data from the related areas of stimulus generalization, transposition, and discrimination learning. A new class of study introduced by Hanson, (1957, 1959) based upon the comparison of generalization gradients obtained with and without prior discrimination training has, in the authors’ view, been a recent source of embarrassment for the Hull-Spence approach. The present paper reviews a number of these and similar studies, and concludes that the Hull-Spence formulation is now confronted with such grave negative evidence that it requires extensive modification in order to remain a viable theory.

Voluntary stimulus generalization as a three-category judgment process

Human Ss were asked to heft a 150-g weight and were required to identify it in a test series of weights, ranging in value from 100 to 200 g, with 25-g steps. Following each test response, each S was asked to rate the weight he had just lifted on a 7-point scale of heaviness, with 1 being very light, 4 neither light nor heavy, and 7 being very heavy. Results showed that Ss chose the stimulus weight closest to adaptation level (125 g) as the original training stimulus more often than the 150-g weight itself. Support was noted for adaptation-level theory. Further, analysis of rating data yielded information as to why Ss rejected a given stimulus as not the original. Three gradients of response, “too light,” “same,” and “too heavy,” were drawn, showing this voluntary generalization task to be a three-category judgment process.