When implicit learning fails: amnesia and the problem of error elimination.

We propose that one of the major functions of explicit memory is the elimination of learning errors. The hypothesis is explored by means of a stem completion task in which subjects are presented with stems having many potential completions, and in the initial phase are either encouraged to guess, the "errorful" procedure, or are provided with the correct completion, the "errorless" condition. Learning is then tested over a sequence of nine trials. The performance of amnesic subjects who are assumed to have good implicit but bad explicit learning is compared with that of normal elderly subjects, who are assumed to have an intermediate level of explicit learning skill, and young controls who are expected to be high in both implicit and explicit learning capacity. As predicted, errorless learning is beneficial, with the effect being particularly marked for the amnesic group. A detailed analysis of the intrusion errors supports an interpretation of the results in terms of the relative contribution to the three groups of implicit learning, which is assumed to be particularly vulnerable to interference. Implications for the analysis of normal learning, and for the rehabilitation of brain damaged patients are discussed.