Prediction of paired-associate latencies after the last error by an all-or-none learning model

Abstract

AbstraetAn all-or-none learning model is presented which makes predictions for response latencies in pairedassociate learning. The predicted latencies decrease on trials following the last error of a criterion run and accurately describe the latency data of an experiment by Peterson (1965). Problem Peterson (1965) has stated that "For an incremental view of learning, the latency of response would be expected to decrease after the S stops making errors. However, decreasing latencies after errors cease are difficult to incorporate into an all-or-none model of learning." Peterson (1965) gave 15 undergraduates 20 trials to learn 10 paired-associates. The stimuli were 10 consonant bigrams, and the responses were the digits 1 to 8. He found that the group mean latency decreased on the 10 trials following an SIS last error. He interpreted the data as indicating that "Speed of occurence (of the correct response) appears to be a measure of strength of an association which changes gradually over trials. tt The question of crucial importance to learning theory is not whether learning is strictly' 'all-or-none." Rather it is whether learning is grossly discrete or approximately continuous. A grossly discrete process may have a number of individual mechanisms, each strictly allor-none. When cast in these terms, the Peterson (1965) data are not at all at variance with a discrete, all-ornone position on learning.As a case in point, suppose that a complete stimulusresponse association is formed strictly on one triaL Behavioristically, an association is defined as the occurrence of a particular response to a presented stimulus. Assume further that the association is first stored in an imperfect, temporary memory and later transferred to a perfect, relatively permanent memory. Then response latency would be a direct function of retrieval time or memory search time, and it would not be unreasonable to expect that retrieval of the association from temporary memory would take longer than retrieval from permanent memory. These notions lead to an all-or-none learning model which may be defined by the following explicit assumptions;1. The response given on a trial to a particular stimulus item depends only on the state of conditioning of the individual stimulus-response association.

Trichotomous processes in early memory development, aging, and neurocognitive impairment: A unified theory.

“…The fine-grain structure of adults’ recall data—by which we mean the empirical distributions of various error and success statistics in standard recall paradigms—is known to conform closely to the predictions of such chains (for a review, see Brainerd, Howe, & Desrochers, 1982). Although Miller (1952) was the first to propose that finite Markov chains are applicable to memory paradigms, the earliest two-stage models of recall were Theios and Hakes’s (1962) model of paired-associates recall and Waugh and Smith’s (1962) model of free recall. Various investigators soon confirmed that two-stage absorbing Markov chains delivered excellent fits to paired-associates recall, cued-recall, free-recall, and serial-recall data (e.g., Bower & Theios, 1963; Estes & DaPolito, 1967; Greeno, 1968; Halff, 1977; Kintsch, 1963; Kintsch & Morris, 1965; Pagel, 1973).…”

Recall latencies of highly overlearned items

“…A newly learned association tends to be remembered more quickly each time it is tested within the course of an experimental session (Eimas & Zeaman, 1963;Millward, 1964;Kintsch, 1965;Peterson, 1965;Suppes et ai, 1966). The purpose of the present study was to see whether such a reduction in latency of recall is maintained over intervals of 24 h. If so, then it may be that overlearned associations are in fact transferred (Theios, 1965;Suppes et ai, 1966) to a rapid-access "permanent" memory store from some more temporary and less accessible system. If not, then the effect may simply denote a relatively temporary response set induced within the course of an experimental session (cf.…”