An optimal strategy for the presentation of paired-associate items

Abstract

This study was concerned with the implementation of certain mathematical results that give optimal ways to present stimulus items in learning experiments. This optimization theory is based on a stimulus‐sampling model of a learning experiment. We assumed that this particular model of learning would describe paired‐associate learning and then compared the effects of training subjects in several paired‐associate experiments by the optimal strategy and by another simple presentation strategy. Our empirical results did not support the adequacy of the assumed model of learning for representing the learning processes in these experiments. In particular, it appears that suitable models of these learning situations must incorporate some mode of representation of short‐term retention.

Optimal Static and Dynamic Training Schedules: State Models of Skill Acquisition

“…Clearly, such an assumption is unrealistic. It is the failure of this same assumption which may well explain why Dear et al (1967) failed to obtain an advantage for the Karush Dear algorithm over the standard procedure. In fact, when the conditioning probability is allowed to vary across items and subjects, the locally optimal strategy described by Atkinson 6 Paulson (1972) shows an improvement of almost 500 over the Karush Dear algorithm.…”

“…But, strictly speaking, this is not optimization (Fisher, 1992(Fisher, , 1993Sheridan, 1988), at least not optimization as more formally defined. The only three studies (Atkinson 6 Paulson, 1972: Dear, Silberman, Estavan, 6 Atkinson, 1967Karush 6 Dear, 1966) that explicitly addressed optimization have done so for just one of the six different situations considered below.…”

Models of the Learner in Computer-Assisted Instruction

“…Dear, Silberman, Estavan, and Atkinson (1967) reported the first atplication of a quantitative memory model to CAI.…”

An approach to the psychology of instruction.

“…The standard cyclic presentation procedure used in most learning experiments may mask certain deficiencies in the all-or-none or random-trial increments models which would manifest themselves when the optimal presentation strategy specified by one or the other of these models was employed. 2 2 This type of result was obtained by Dear, Silberman, Estavan, and Atkinson (1967). They used the all-ornone model to generate optimal presentation schedules where there were no constraints on the number of times a given item could be presented for test and study within an instructional period.…”

“…The problem was that the all-ornone model provides an accurate account of learning when the items are well spaced, but fails badly under highly massed conditions. Laboratory experiments prior to the Dear et al (1967) study had not employed a massing procedure, and this particular deficiency of the all-or-none model had not been made apparent. The important remark here is that the analysis of instructional problems can provide important information in the development of learning models.…”