Negative DC potential shifts appeared over the scalp during the performance of verbal and non-verbal short-term memory tasks. Three items were successively presented (presentation of memory items) and then had to be retained in memory for 3 sec (memory retention) before being compared to a probe which was either a member (in set) or not a member (out of set) of the memory set. Verbal items (the digits "1" through "9") were tested in the auditory and visual modality and non-verbal items (musical notes) were tested in the auditory modality. Stimulus modality had a significant effect on DC potential shifts during both presentation of memory items and memory retention. There was a sustained negative shift during these periods which was larger over frontal regions with auditory than with visual material whereas the negative shift was larger over posterior temporal regions with visual than with auditory material. Out of 21 subjects who participated in the study, 9 reported the use of visual images in the auditory task, 5 used subvocal auditory rehearsal in the visual task and 7 used imagery concordant with the stimulus modality being memorized. These different strategies had a significant effect on the amplitudes and distribution of the DC potential shifts. The speed of response affected the amplitude of the DC potential shifts in the frontal regions, being larger with fast RTs than with slow RTs but only when verbal items were being processed. These results indicate that stimulus modality, modality of mental imagery, and speed of scanning of the memory store affect DC potential shifts during a 3 sec period of memory retention.