Mismatch brain responses to unpredicted rare stimuli are suggested to be a neural indicator of prediction error, but this has rarely been studied in the somatosensory modality. Here, we investigated how the brain responds to unpredictable and predictable rare events. Magnetoencephalography responses were measured in adults frequently presented with somatosensory stimuli that were occasionally replaced by two consecutively presented rare stimuli [unpredictable rare stimulus and predictable rare stimulus ; p = 0.1 for each]. The FRE and PR were electrical stimulations (...) administered to either the little finger or the forefinger in a counterbalanced manner between the two conditions. The UR was a simultaneous electrical stimulation to both the forefinger and the little finger. The grand-averaged responses were characterized by two main components: one at 30–100 ms and the other at 130–230 ms latency. Source-level analysis was conducted for the primary somatosensory cortex and the secondary somatosensory cortex. The M55 responses were larger for the UR and PR than for the FRE in both the SI and the SII areas and were larger for the UR than for the PR. For M150, both investigated areas showed increased activity for the UR and the PR compared to the FRE. Interestingly, although the UR was larger in stimulus energy and had a larger prediction error potential than the PR, the M150 responses to these two rare stimuli did not differ in source strength in either the SI or the SII area. The results suggest that M55, but not M150, can possibly be associated with prediction error signals. These findings highlight the need for disentangling prediction error and rareness-related effects in future studies investigating prediction error signals. (shrink)