Event-related potential correlates of encoding and retrieval processes in human memory

Abstract

In electrophysiological research over the last few years there has been a growing tendency to utilise the Event-Related Potential as a tool in the study of cognitive processes, especially those which are well defined and understood. Over the last few years too, the study of encoding and retrieval processes in human memory has achieved a certain measure of consensus, in that many researchers have suggested that retrieval performance depends largely upon the type of encoding process performed upon the item at presentation, in particular it is suggested that the degree to which the physical characteristics of a stimulus item are analysed will largely determine whether the item can later be recognised and that the degree to which the semantic content of the item is analysed will largely determine whether the item can later be recalled. The present series of experiments sought to determine whether there existed ERP correlates of the two types of encoding process and of the two related retrieval processes, recognition and recall.

In the first experiment, ERPs generated by words which were thought to have been analysed at a physical level, as determined by whether they were recognised 24 hours later, were compared with ERPs generated by words thought not to have received such processing, during both initial and subsequent presentation. The ERP encoding data indicated that enhancement was seen in late positive activity generated at Fz by words which were later recognised with a high degree of confidence. This was taken to imply that the enhanced positivity was generated by elaborative encoding processes. The ERP retrieval data indicated: 1) that between 300-500 msec post stimulus, words which were correctly recognised as "old" generated potentials of greater positivity than words which were correctly recognised as "new". This was interpreted as an ERP index of a retrieval process based on familiarity only. 2) Between 500-924 msec post stimulus, items which may have been recognised due to the retrieval of encoding context generated greater positivity than items recognised on the basis of familiarity alone.

In experiment 2, the basic design was repeated with the exception that a cognitive task was interposed between trials to ensure that all processing related to subsequent memory performance was restricted to the recording epoch. ERPs were recorded only during the initial presentation of stimulus items, and those generated by words later recognised were again compared with those generated by items not recognised or recognised with low confidence. The ERP data revealed the same enhanced late positivity at Fz generated by words correctly recognised with a high degree of confidence, although, as in experiment 1, the effect was small.

In experiment 3, ERPs were again only recorded to words during the first presentation, but were this time compared on the basis of whether words had been recalled or not. It was thought that if subsequent recall of items depends on elaborative processing at presentation, this manipulation would ensure that ERPs generated by recalled words would reflect activity selective to elaborative encoding. The recording epoch was also lengthened in order that the ERPs might be sensitive to slow, long latency effects. The data indicated that words which were recalled generated ERPs of significantly greater positivity in the region 800-1400 msec at Fz than did words not recalled.

In experiment 4, ERPs generated by words during retrieval were recorded, and in this case ERPs generated by words whose experimentally learned associates were recalled from memory, were compared with ERPs generated by words whose associates were not recalled. It was suggested that since recall depends upon retrieval of encoding context, ERPs generated by words whose learned associates were recalled, should reflect such processing. The ERP data showed that words whose associates were recalled, generated activity of greater positivity than words whose associates were not recalled, from 500 msec onwards at all three midline sites. It is concluded from these experiments that at encoding, the activation of elaborative processing is reflected by an enhancement of the ERP activity at Fz from approx. 500 msec onwards, and that at retrieval, 1) the activation of the "familiarity-checking" process generates enhanced positivity at all midline sites between 300-500 msec, and 2) that the "retrieval of encoding context" process generates enhanced positivity from 500 msec onwards. These data are related further in the conclusion to both physiological and cognitive theories of human memory.