Southwell, R.; Tufo, C.; Chait, M.

In classical studies of auditory pattern learning, neuronal responses exhibit repetition suppression, where sequences of repeated tones show a reduced evoked response. This may be due in part to adaptation, but is also hypothesized to indicate suppression of expected stimuli. Repetition suppression is thought to form a building block of regularity learning, and is the paradigmatic example of predictive coding in humans and other animals. However, stimuli with more complex patterns appear to show the opposite effect. Predictable regular (REG) patterns distributed over a range of frequencies show a strongly enhanced sustained brain response compared to frequency-matched random (RAND) sequences, sitting at odds with the usual reduced evoked responses to predictable stimuli. A limiting factor in reconciling these findings is that they are obtained using different stimuli and analysis methods. This human EEG study (N=20) brings together auditory sequence predictability and repetition in a single paradigm, based on rapidly unfolding tone pip patterns, incorporating sequences consisting of exactly repeating tones at a single frequency, alongside REG and RAND of varying complexity. We demonstrate that regularity is associated with increased sustained responses, offset responses, tone locked responses and cycle-locked responses. We further show that both repetition suppression and repetition enhancement occur over different timescales, and that sustained brain responses to simple repetition show qualitatively different effects than to more complex regularities during automatic tracking of stimulus statistics. Our results indicate a system for automatic monitoring of predictability in the auditory environment, which is distinct from, but concurrent with, repetition suppression.