Retroactive inhibition and natural language mediation

Abstract

Natural language mediators (NLMs) or idiosyncratic associations, were studied for their effect on A-B recall in an A-B, A-C paradigm. A list of eight paired adjectives was used, and NLMs were obtained by questionnaire after original learning, interpolated learning, and recall. The level of NLM usage was high, being 66% in original learning and 70% in interpolated learning. No loss of NLMs was found from original learning to recall. Recall was positively related to the presence of NLMs in learning A-B, but independent of whether A-C items were mediated or rote. The results were discussed in terms of NLMs being either a correlate of the associative process or a direct determinant of it, but no basis was found for a choice between these alternative interpretations. In both cases it was considered likely that items with NLMs in original learning had reduced retroactive inhibition because NLMs were an index of greater associative strength. Another possibility was that an NLM decreased retroactive inhibition by decreasing the similarity of pairs.