A late parietal positivity (P3) and behavioural measures were studied during performance of a two-item memory-scanning task. Stimuli were digits presented as memorized items in one modality (auditory or visual) while the following probe, also a digit, was presented in the same or the other modality. In a separate set of experiments, P3 and behaviour were similarly studied using only visual stimuli that were either lexical (digits) or non-lexical (novel fonts with the same contours as the digits) to which subjects assigned numerical values. Reaction times (RTs) and P3 latencies were prolonged to non-lexical compared to lexical stimuli. Although RTs were longer to auditory than to visual stimuli, P3 latencies to memorized items were prolonged in response to visually compared to auditorily presented memorized items, and were further prolonged when preceding visual probes. P3 amplitudes were smaller to auditory than to visual stimuli, and were smaller for the second memorized item when lexical/non-lexical comparisons were involved. The most striking finding was scalp distribution variations indicating changes in relative contributions of brain structures involved in processing memorized items, according to the probes that followed. These findings are compatible, in general, with a phonological memorization, but they suggest that the process is modified by memorizing the item in the same terms as the expected probe that follows.