Some things are easier done than said: The effect of post-encoding verbalization, saliency, and ability on transfer performance in artificial grammar tasks.

摘要

Schooler and Engstler-Schooler (1990) have shown that verbalization of difficult-to-verbalize information (e.g., describing a face) impairs subjects' memory performance (e.g., face recognition) relative to no-verbalization, a phenomenon called verbal overshadowing. Five studies are reported that attempted to investigate whether verbalization can similarly impair learning when the information is difficult to verbalize, as it is with artificial grammar tasks. Subjects memorized letter strings derived from an artificial grammar then either verbalized rules that they learned during memorization or not. It was expected that verbalization would impair subjects' performance relative to no-verbalization, the standard verbal overshadowing effect. In Experiment 1, the prediction was confirmed: verbalization impaired the subjects' performance. This finding significantly extends previous artificial grammar research that found better performance under implicit rather than explicit encoding conditions. In the present case, the verbal manipulation occurred following encoding rather than during encoding, meaning that all subjects presumably had the same resources available for memorization. Thus, these results indicate that verbalization may disrupt access to already encoded implicit knowledge.; Subsequent experiments tested the hypothesis that the mechanism underlying this effect is the difficulty of verbalization. For example, one possible mechanism is the saliency of the rules. In order to test this, strings were arranged according to rules in order to make them more obvious and therefore easier to verbalize. It was expected that in this case, verbalization would have no effect on performance. However, not only was the basic hypothesis not confirmed, but there was a failure to replicate Experiment 1, an ongoing difficulty in both verbal overshadowing and artificial grammar work. Despite this replication problem, some trends were evident in the studies. First, it appears that subjects do not use the information they verbalize, lending evidence that the information used in making grammaticality judgements is primarily implicit. Second, it appears that subjects' initial verbal ability may have some bearing on their performance on this task and on the effect of verbalization on their performance, though it is not yet clear exactly what the nature of that effect is. Implications for verbal overshadowing, artificial grammar tasks, and education are discussed.