The magnetoencephalogram (MEG) was used to detect regional changes in spontaneous cortical activity accompanying short-term memory search. This method was chosen because magnetic fields are detectable only within a few centimeters of the projections of their sources onto the scalp. The specific hypothesis that auditory cortex is involved in scanning memory for tones was tested by sensing the field of the magnetic counterpart to N100 (N100m) which is known to originate in auditory cortex. N100m was measured at many different positions and the spontaneous cortical rhythms in the alpha bandwidth (8–12 Hz) were measured at the same places. These rhythms were found to be suppressed while subjects scanned memory for musical tones in a Sternberg paradigm. For 3 subjects, both the MEG suppression time (ST) and reaction time (RT) increased linearly with memory set size. The correlation between ST measured over the left hemisphere and set size was significant for two subjects but not significant for the third, and the slopes of the regression lines relating ST to set size were too shallow to be related to the time required to scan memory. However, the correlation between ST of the right hemisphere and set size was highly significant for all subjects, and the slopes of the regression lines were comparable to those relating RT to set size. The electroencephalogram (EEG) recorded with midline electrodes failed to reveal a significant relationship between suppression time and set size for 2 of the subjects, thus ruling out global alpha blockage and generalized arousal as the basis for the task-related suppression duration. The electric N100, measured at Cz, decreased significantly in amplitude with set size for 2 subjects, but in increased significantly in amplitude for the third subject. In contrast, RT increased with set size for all subjects. N100m measured over the right hemisphere was similar to the behavior of N100, while N100m measured over the left hemisphere showed little change in amplitude with set size, thus establishing an asymmetry in N100 between the hemispheres. Since N100 amplitude is normally larger when attention is paid to auditory stimuli, differential attention alone cannot account for the relation between ST and set size. Furthermore, the processing negativity, which may be superimposed on N100 in selective attention tasks, was not discernible for any set size. It was also found that ST prior to the button press was not correlated with RT. Hence, the covariation of set size with ST is not attributable to preparation for a motor response. In this experiment the sensors were placed where the field of N100m was approximately at its maximum value. Since many measures were made from somewhat different positions in these areas, it was possible to determine how N100m amplitude varied with sensor location over both hemispheres. These position-dependent changes in N100m amplitude were significantly correlated with corresponding position-dependent changes in percent suppression of the spontaneous MEG. Thus, auditory cortex contributes significantly to the spontaneous alpha activity suppressed during memory search. It is concluded that regional changes in cortical spontaneous activity are meaningfully related to ongoing memory scanning. Earlier work on the relation between short-term memory and the N100 and P300 ERP components is reviewed in an Appendix. It is concluded that published conjectures on the relation between N100 and short-term memory are inconclusive, and neither P300 latency nor amplitude have been convincingly related to short-term memory scanning per se. These ERP components are complementary to the regional changes in spontaneous activity studied here, and reflect different facets of the processes involved in memory scanning.