Error signals are the cornerstone of predictive coding and are widely considered essential to sensory perception and beyond. The mismatch negativity (MMN) is arguably the most emblematic and most studied brain error signal. It is affected in many brain disorders. However, its precise algorithmic function and the underlying physiology remain mysterious. Over the past decade, theoretical and computational explanations have been put forward. They highlight a paradox: the MMN is considered a signature of context-dependent perceptual learning, although it is defined as an evoked response averaged across trials, thus neglecting the information carried by error signal fluctuations over time. We propose to deconstruct the MMN, by virtue of hypothesis driven computational approaches whose aim it to account for these fluctuations.
