LONG‐TERM RETENTION AND TRANSFER OF AN INDUSTRIAL SEARCH SKILL

Abstract

The classical finding that transfer, compared with retention, is relatively stable over long periods of time was investigated using a fault‐location task. Following initial, unaided searching, the subject practised fault location with the help of a decision tree specifying the most efficient search sequences. Instructions explained that the subject might subsequently be asked to find faults in a system where the failure probabilities and test costs would be different (the transfer task), and that, besides learning the specific decision tree test sequences, he should attempt to learn more general search strategy. Subjects' performance was measured on one subsequent occasion, either on the training task or on the transfer task, after an interval of either 6, 58 or 182 days. There was a significant decrement in retention but not in transfer. All retention and transfer means were significantly superior to untrained controls. When subjects were divided into high‐ and low‐ability groups (where ability is measured by the efficiency of unaided fault location before training), significant decrement was found only in retention of low‐ability groups. No significant decrements were found in retention of high‐ability groups, or in transfer at either ability level. It is argued that stability of training effects over time depends not so much on whether the criterion is retention or transfer as on the extent to which the subject learns general search strategy, which is well remembered, as distinct from specific search routines, which are more quickly forgotten. A simple law of exercise may explain the superior retention of general strategies over specific response sequences.