An oscillator model better predicts cortical entrainment to music

Significance Previous work in humans has found rhythmic cortical activity while listening to rhythmic sounds such as speech or music. Whether this activity reflects oscillatory dynamics of a neural circuit or instead evoked responses to the rhythmic stimulus has been difficult to determine. Here, we devised a metric to tease apart the two hypotheses by analyzing phase lag across many stimulation rates. We test this phase concentration metric using numerical simulations and generate quantitative predictions to compare against recorded magnetoencephalography data. Both previously recorded and new data were better predicted by a model of oscillatory dynamics than evoked responses. This work, therefore, provides definitive evidence for the presence of an oscillatory dynamic in auditory cortex during processing of rhythmic stimuli. A body of research demonstrates convincingly a role for synchronization of auditory cortex to rhythmic structure in sounds including speech and music. Some studies hypothesize that an oscillator in auditory cortex could underlie important temporal processes such as segmentation and prediction. An important critique of these findings raises the plausible concern that what is measured is perhaps not an oscillator but is instead a sequence of evoked responses. The two distinct mechanisms could look very similar in the case of rhythmic input, but an oscillator might better provide the computational roles mentioned above (i.e., segmentation and prediction). We advance an approach to adjudicate between the two models: analyzing the phase lag between stimulus and neural signal across different stimulation rates. We ran numerical simulations of evoked and oscillatory computational models, showing that in the evoked case,phase lag is heavily rate-dependent, while the oscillatory model displays marked phase concentration across stimulation rates. Next, we compared these model predictions with magnetoencephalography data recorded while participants listened to music of varying note rates. Our results show that the phase concentration of the experimental data is more in line with the oscillatory model than with the evoked model. This finding supports an auditory cortical signal that (i) contains components of both bottom-up evoked responses and internal oscillatory synchronization whose strengths are weighted by their appropriateness for particular stimulus types and (ii) cannot be explained by evoked responses alone.