We used behavioral and event-related potential (ERP) measures to study the neural mechanisms of involuntary attention switching to changes in unattended sounds. Our subjects discriminated two equiprobable sounds differing in frequency (fundamental frequency 186 or 196 Hz) while task-irrelevant intensity decrements or increments (-3, -6, -9, +3, +6, or +9 dB, standard intensity 60 dB HL) infrequently occurred in the same sounds. In line with the results of previous studies, discrimination performance deteriorated with increasing magnitude of the task-irrelevant intensity change. However, these distraction effects were dissimilar for intensity increments and decrements: while there were no differences in reaction time (RT) between intensity decrements and increments, hit rates (HR) were lower for large intensity increments than for large decrements. ERPs to task-irrelevant intensity increments and decrements were also distinctly different: the response to intensity increments consisted of an N1 enhancement, mismatch negativity (MMN), and P3a, while the response to intensity decrements consisted only of MMN. These results are consistent with the assumption that two separate mechanisms (indexed by N1 and MMN) underlie auditory change detection. However, the finding that distinct distraction effects were obtained for both intensity decrements and increments but that the P3a is elicited only by the intensity increments seems to suggest that P3a may not be regarded as a general index of attentional shift but rather it is only generated in conditions in which an enhanced N1 is elicited, too.