According to the model hypothesized by Näätänen and Michie, the generation of the mismatch negativity (MMN) requires a mismatch detection, taking place in temporal areas, followed by the activation of frontal generators, underlying attention switching toward the deviant stimulus. We aimed at verifying whether the activation of temporal and frontal regions is dependent on the amount of attentional resources allocable toward the deviant stimulus. We recorded event-related potentials (ERPs) in nine healthy subjects while reading and during a demanding visual task (Multiple Features Target Cancellation, MFTC). Raw data were further evaluated by Brain Electrical Source Analysis (BESA). During the Reading condition, distraction toward the unattended auditory stimuli was reflected by the enhancement of the N1 response to frequent stimuli and by the elicitation of a P3a response to deviant ones. The MMN distribution was explained by bilateral temporal dipoles. During the MFTC condition, no P3a was detected, while source analysis showed the activation of a right frontal generator. Temporal dipoles showed no change between the two conditions: we thus conclude that the earlier mismatch detection is independent on the attentional load. By contrast, the activation of a right frontal subcomponent occurred only during the high-load task, independently on any actual attention shift reflected by the P3a component. We thus discuss the hypothesis whether the right frontal MMN generator, rather than subserving a simple attention switching toward the deviant stimulus, plays a role in modulating the auditory change detection system ("contrast enhancement" model).