Thornton B. Roby

Dimensions of decision-making behavior.

Most laboratory research on decision making has been derived from normative theories, whose purpose is to advise the decision maker on what he should do. The present study, however, approached the problem of pinpointing the variables which are most relevant to unguided decision making behavior. A factor analysis was done of a battery of 70 scores which included a wide sample of decision-making behavior and certain cognitive and personality measures. The decision behavior was obtained in a variety of laboratory tasks designed to measure different traits. The cognitive and personality measures were included to help clarify the behavior represented by each factor. Eight oblique, but nearly orthogonal, factors were obtained: (1) readiness to make distinctive or informationally more certain responses, (2) intellectual passivity, (3) rational, active approach to new problems, (4) formation of sound concepts, combined, in decision making, with mimicry of the outcomes of prior events, (5) pessimism, with unstructured or global responses, (6) constrictive rigidity when required to make too many decisions in a short time, (7) consideration of each decision element independently, with active search for information and (8) contemplative appreciation of the structure of events.

SELF-ENACTING RESPONSE SEQUENCES AND REINFORCEMENT: CONJECTURE

In large measure the difficulties of the principal learning theories may stem from the use of discrete, overt, responses as the units of learned behavior. Internally chained sequences of implicit responses are suggested as alternative units. The latter are assumed to result in overt behavior by a process of neurological summation. It is then proposed that the principle governing reinforcement of such sequences is that the elicitation of any initial segment of a successful sequence increases the probability that remaining elements of the sequence will be elicited. The principle is applied to habituation, classical conditioning, instrumental learning, and avoidance. Finally, points of similarity and contrast with the major theories of learning are discussed, and several general criteria for a behavior theory are listed which seem to favor a formulation of the kind proposed.

Signal and channel load in vigilance tasks.

In a complex vigilance task, probability of error was greater with greater load of signals assigned, greater with greater load of locations (channels) assigned, and less with practice. The effect of signal load on errors was greater than that of channel load, and was more strongly influenced by practice. The methodology and results are seen as relevant to sub-task assignment in group vigilance tasks.

Use of State and Transitional Information

The problem concerned relative human ability to digest information describing the state of the environment (σ) or state transitions (τ). The task required the summing or differencing of symbolically presented σ or τ information items. In the first study, subjects were presented σ or τ information in mixed order and σ or τ queries at variable intervals. A second study entailed uniform translation from σ to τ or τ to σ modes. The chief results were: (a) acquisition across modes—τ—σ or σ-τ—is more difficult than within mode acquisition; (b) τ—σ acquisition is superior to σ-τ acquisition; (c) both number of elements and number of phases or levels within elements add to task difficulty; and (d) the relative advantage of σ over τ acquisition decreases with an increasing number of presented items. It is suggested that the results may be explained in part by the comparative economy of state information storage for normal decision making tasks.

Some task parameters in simple pattern recognition.

The task presented to individual Ss required identification of a particular set of two digit numbers that was wholly contained in a larger set of displayed elements. Five stimulus factors were experimentally manipulated. Size of the display set, number of presumptive pattern sets, and density of display elements in the incorrect presumptive pattern sets were negatively related to accuracy of identification. The size of the pattern sets and the spatial distribution of pattern elements in the display were not related clearly to performance. The relevance of these studies to pattern identification in small group situtations is discussed and illustrated.

ECONOMICS OF PSYCHOPHYSICAL SCALES.

The basic assumption of this paper is that the scalability properties of stimulus continua have functional applicability to the generalization and transfer of adjustive responses: a ratio scalable stimulus continuum, for example, lends itself to more efficient and precise extrapolation of behavior to new conditions than does a nominally ordered set of stimuli. The paper lists auxiliary assumptions necessary to translate this contention into quantitative estimates and derives the comparative performance estimates for the acquisition and retention of an S-R mapping under various scale relations. A final section assesses the assumptions from the standpoint of realism and susceptibility to experimental test.