Predicting transfer from training performance

The research in this paper was designed to examine the extent to which impr ovement on a training task can be used to predict performance on a transfer task. This aim involved evaluating the proposition that when old skills ar e executed in the context of new tasks, they continue to improve as if stim ulus conditions have not changed. That is, power functions that describe im provement on old skills during their initial acquisition should predict fur ther improvement on these skills during their execution in new tasks. Three experiments were performed to achieve the aim of testing this proposition. Experiment I revealed that old skills were executed slower in the context of a new task than was predicted on the basis of training performance. Henc e improvement in the old skills appeared to be disrupted by performance of the new task. Experiment 2 was designed to examine whether this disruption was due to an increase in complexity in the task from training to transfer, or simply due to any change in task. The results suggested that any change may cause some disruption, but this disruption was greatest with an increa se in task complexity. Experiment 3 was designed to examine two variables t hat may affect the magnitude of this effect: the relative change in task co mplexity from training to transfer, and the amount of practice on a task pr ior to a change in task. The results indicated that only the former variabl e had any effect. In all three experiments no effects on performance accura cy were noted, and response times in the transfer tasks eventually returned to levels predicted by training learning functions. These results were int erpreted as indicating that old skills do continue to improve in new tasks as if conditions are not altered, but that disruptions caused by transfer a re related to performance overheads associated with reconceptualising the t ask. (C) 2001 Elsevier Science BN. All rights reserved.