Which event-related potentials reflect memory processing in a digit-probe identification task?

Auditory and visual event-related potentials (ERPs) generated by digit-probe identification and matching in a modified Sternberg paradigm have been recorded in 37 healthy subjects with the aim of identifying the potentials which best reflect the memory processes associated with this task. We analysed the effects of memory load (one, three or five digits to memorise), probe type (probe digit present or absent from the preceding memory set) and recording site, on the ERPs. With conventional methods of determining component amplitudes and latencies, the main effects of increasing memory load on the major positive wave varied according to stimulus modality-there was an amplitude decrease for the auditory ERPs and a latency increase for the visual ERPs. However, subjective component identification methods may be prone to errors when comparing responses recorded under different stimulus conditions. Waveform changes with increasing memory load may be misinterpreted as latency (or amplitude) effects if non-analogous potentials are compared. Further, component analysis may provide only partial information, because of its relative insensitivity to sustained amplitude shifts. For these reasons, an objective computer method was used to determine the mean amplitudes for multiple '50 ms' epochs. This showed, for both auditory and visual stimuli, that the main effect of increasing memory load was a 'negative amplitude shift'. It was seen between 315 and 525 ms for auditory stimuli and between 210 and 472 ms for visual stimuli and could be distinguished from other ERP features that were sensitive to stimulus modality. These changes are either specific to the memory processes involved in carrying out this task or reflect other parallel processing which covaries with memory processing.