Grimm, S., & Schröger, E. (2007). The processing of frequency deviations within sounds: Evidence for the predictive nature of the Mismatch Negativity (MMN) system. Restorative Neuroscience and Neurology, 25, 241-249.
Purpose: The present study aimed at further clarifying the nature of the automatic deviance detection system indexed by the mismatch negativity (MMN) component of the event-related potential (ERP). We investigated the pre-attentive detection of a transient frequency deviance occurring within a short tone at different temporal distances relative to tone onset. It was of specific interest whether these different types of deviations were categorized as distinctive events by the MMN system or not. Methods: We compared the MMN elicited by a rare frequency deviation occurring 200 ms after tone onset among frequent standard tones without frequency deviation. In three conditions, the proportions of standards and deviants were 90-10 (90 % standards, 10% frequency deviation occurring 200 ms after sound onset), 70-30 (70 % standards, 30% frequency deviation), or 70-10-10-10 (70% standard, frequency deviation 10% after 100 ms, 10% after 200 ms, 10% after 300 ms). Results: We found that the MMN in the 70-10-10-10 condition resembles the MMN in the 90-10 condition, both MMNs being larger than the MMN in the 70-30 condition. In other words, although a frequency deviation in the 70-10-10-10 condition was as likely as in the 70-30 condition (i.e. 30%), it was treated as if it were as unlikely as in the 90-10 condition (i.e. 10%). Thus, the amplitude of the MMN was related to the specific temporal occurrence of the frequency deviance rather than the global probability of the frequency deviance. Conclusions: The data support the hypothesis that the MMN system was operating on the basis of a spectrotemporal representation rather than on independent feature dimensions. Results are compatible with a new theory of the MMN system according to which MMN is based on a pointwise comparison between the current sound and a model-based concrete prediction of a forthcoming sound as it evolves over time.