Memory as a Constructive Process: The Parallel Distributed Processing Approach

James L. McClelland In Harold Pinter’s play Old Times, a husband and wife of many years reminisce about the early days of their relationship, while awaiting a visit from the wife’s best friend from that era. In these reminiscences, we learn just how differently two people can remember what were ostensibly the same events, and ostensibly the same people – most notably, themselves and the wife’s best friend. For each, these reminiscences have become embedded in a complex, not fully consistent, and self-serving personal history that does not survive well when juxtaposed against the reminiscences of the other. In an interview at the time of the opening of Old Times, Pinter was asked to comment on his thoughts while writing his play. He said “what fascinates me is the mistiness of memory”(1). Cloud-like, forever changing, memories are clearly not like frozen snapshots taken on a day long ago, pulled out from the back of a drawer for reinspection. Indeed, since the work of Frederick Bartlett in 1932, memory researchers have been keenly aware of the constructive nature of memory (2). Bartlett asked educated people at Cambridge to read, and then later to recall, a story from a native North American culture. Though written in English, the story had an unfamiliar structure and content. The recollections of the participants retained elements from the original story but many details were omitted or transformed in ways that seemed to Bartlett to fit better with the cultural context of the individuals who were recalling them. Repeated attempts at recall by the same individual resulted in gradual fixing of the elements, but into a story sometimes quite different from the original. Such findings led Bartlett and others to view recollection as a process not unlike the activities of an archaeologist faced with the task