In one pair of experiments, modeling an unambiguous auditory scene, a repetitive tone sequence including infrequent duration deviants was mixed with tones randomly varying in duration (intervening tones). In the Interference condition, the frequency range of the intervening tones included the frequency of the oddball sequence tones. In the Segregated condition, intervening tone frequencies were set to be distinctly different from the frequency of the oddball sequence tones. Subjects could detect the duration deviants in the Segregated but not in the Integrated condition. Correspondingly duration deviants only elicited MMN in the Segregated condition (while subjects performed a visual primary task).
In another experiment, modeling an ambiguous auditory scene, sequences consisted of a repeating tone triplet (ABA ABA ABA …, the A and B are two tones differed in frequency). Sounds were delivered at a pace at which this sequence could be perceived either as separate high and low streams of sound or as a single stream with a galloping rhythm. Subjects were instructed to maintain hearing the galloping rhythm and to keep a response button depressed when they could do so. ERP responses were analyzed for occasional omissions of the "B" tone. Perception of the galloping rhythm fluctuated within the tone sequences. The elicitation of deviation related ERP responses by the omission deviants was found to correspond to the perception of the galloping rhythm.
Results of these experiments show that stimulus driven and top down effects of sound organization converge on the memory representations reflected by the MMN response.