Abstract
The editing of an episodic memory record in order to remove incorrect information embedded within naturalistic communications is an important though underinvestigated phenomenon. Experiment 1 deals with the recall and comprehension of a sequence of messages following the delayed identification of one of the messages as being incorrect. Two styles of correction were employed, and it was found that in neither case was the memory record edited effectively. Inferences based upon the old information continued to be drawn although subjects had clearly recalled that it had been subsequently corrected.
Experiment 2 showed that editing could be effective if the old information did not play a central role in the message sequence. It is concluded that the observed difficulties in editing arise when old information has to be excised from the episodic record; the uncontested insertion of new information retrospectively did not present the same difficulty.
Reading span was used to monitor subjects' editing strategies, and from its association with performance measures it is concluded that contradictions in the memory record are not dealt with immediately but are resolved locally when comprehension is questioned. At this time inferences are drawn based upon the most recent version of the contradictory messages. This recency strategy breaks down when the old information provides a better fit to the question posed. Some implications of these findings for models of memory storage are discussed.
