One drawback of fMRI is that subjects must endure intense noise during testing. This may be annoying to some people and acceptable to others. The aim of this study was to examine, by means of event-related potentials (ERPs), the possible influence of this noise on brain activity while performing a mental reasoning task. Subjects carrying out tasks in a silent environment were compared with two groups executing the same tasks in an "fMRI-like" noisy environment, one of which consisted of subjects who were annoyed by the noise and the other of subjects who tolerated it easily. Subjects who were annoyed performed less well (i.e., produced more errors compared to the "no noise" group) and "not annoyed" subjects showed a speed-accuracy trade-off (i.e., reacted faster but made more errors compared to "no noise" subjects). Noise led to more pronounced N1 and P2 peaks but attenuated N2. As early ERP components are influenced by attention, this observation most likely reflects different attentional requirements. The slow cortical negative shift during task processing was significantly attenuated with "annoyed" subjects compared to "not annoyed" subjects. Emotion-related subcortical structures may be responsible for the observed difference. These findings suggest that individual reactions to fMRI scanner noise should be taken into account when designing fMRI studies and interpreting results.