The present study addressed the pre-attentive processing of sound order. Event-related potentials were recorded from reading subjects while they were presented with pairs of two tones differing from each other in frequency (1000 vs. 1500 Hz). The within-pair (silent) interstimulus interval (ISI) was, in separate blocks, varied between 0 and 245 ms to determine the minimum separation in time needed for detecting the reversed order of the two frequencies. In standard tone pairs (p = 0.9), the frequencies were in an ascending order, whereas in the deviant pairs (p = 0.1), their order was reversed. Tone durations of 5 and 20 ms were employed in separate experiments. With the 20-ms stimulus duration, the change-specific mismatch negativity (MMN) component was elicited with all within-pair ISIs employed (0, 10, 30, 90 ms). With the 5-ms stimulus duration, however, MMN was elicited only with the 245-ms ISI but not with 95-ms or shorter ISIs. These results show that increased stimulus duration considerably improves perceiving the order of two tones at the pre-attentive level. They also indicate that the accuracy of the processing of temporal information can be probed with MMN. This finding, together with the fact that MMN elicitation does not require the subject's voluntary attention, suggests that MMN might be used in the assessment of temporal processing deficits in clinical disorders in which patients are not motivated or able to give their verbal or motor response.