Assembling long-term working memory: Retrieval structures and integrative encoding as critical components of a simple memory skill

The Long-term Working Memory (LT-WM) theory (Ericsson and Kintsch, 1995) proposes that under special circumstances skilled performers acquire the ability to use LTM as the storage medium for working memory processes. Extensive background knowledge and experience with specialized tasks allow skilled performers to efficiently organize and store information in LTM. Two experiments are reported that were designed to identify and isolate processes that may be responsible for improvements in memory performance observed in skilled tasks. In the first experiment, subjects were instructed to assemble a retrieval structure for integrative encoding in an attempt to reconstruct skilled text comprehension processes in a simple serial word recall task. The task involved learning lists of word triplets. The experimental subjects were taught a set of four scenes that they used for learning the triplets. Consistent with the use of LT-WM, it was found that compared to controls, experimental subjects' recall was considerably superior after brief presentations and disruptions to STM prior to recall. Additionally, subjects who spontaneously used this encoding technique showed similar benefits in memory retention. The second experiment compared the encoding and retrieval processes seen in this task to the Method of Loci (ML). It was found that the encodings generated by the experimental subjects using the scenes showed patterns of recall consistent with the idea that the triplets were encoded as integrated elements in a semantic structure. This is contrasted to the method of loci where words within a triplet were encoded independently by location. Further, it was found that recall accuracy for the experimental subjects using the scenes was not reliably influenced by presentation time or order of the words in a triplet. This provides strong evidence that having a set of semantic cues available that the retrieval structure provided, allowed the subjects to rapidly integrate the words into a LTM semantic structure. Both experiments demonstrate that people can be instructed to use and may spontaneously adopt the use of encoding and retrieval processes that are consistent with the use of LT-WM and resemble working memory processes that have been observed in text comprehension.