Several noninvasively measured neural signatures of predicting events in time have been proposed so far. These include the contingent negative variation – a slow build-up of the EEG potential before an expected stimulus; similar time-dependent modulation of alpha-band power (8-13Hz); and low-frequency entrainment, such as delta-band (1-3Hz) synchronisation to slow rhythmic streams. Beyond these candidate mechanisms, also beta-band activity has received considerable interest – perhaps not surprisingly, given its importance for both sensory processing and motor processing. In the sensory domain, beta-band oscillations have been proposed to carry sensory predictions within the influential predictive coding framework. Accordingly, increased beta power preceding expected stimuli have been observed in several studies (reviews here). In the motor domain, however, beta (13-30 Hz) power typically decreases before voluntary movements as well as anticipated events, and motor activity has been suggested to underpin predictive timing in sensory regions. So are differences between the neural implementation of perception and action sufficient to reconcile these conflicting findings?
In their recent paper in Neuroimage, Meijer, te Woerd and Praamstra from Radboud University Nijmegen show that both decreases (event-related desynchronisation; ERD) and increases (event-related synchronisation; ERS) in the beta band can be observed after visual stimuli in a sequence with predictable timing. However, neither the positive (ERS) nor the negative (ERD) peaks seem to be shifted in time by the predicted onset of the upcoming stimulus – which contradicts the previous findings mentioned above.
In the experiment designed by Meijer et al., participants were asked to attend to a series of rhythmically presented visual stimuli and count the number of incongruent stimuli in the series. Incidentally, the stimuli consisted of clock arrows and digits; the digits could be congruent or incongruent with the hour indicated by the clock arrows. Even though clock displays were used as stimuli, what was shown on the clock was randomised over stimuli and thus unpredictable; the only predictable information was when the clocks were shown (every 1050, 1350 or 1650 ms). While participants were viewing the stimuli, their neural activity was recorded using 128-channel EEG.
The results from different groups of electrodes (frontal and parietal/occipital) show that during the first 800 ms after each stimulus, the time-course of the beta-band is not predictive of when the subsequent stimulus will appear – in the sense that beta-band ERS peaks at the same latency for all three temporal conditions. The authors rightly interpret this lack of temporal specificity of the ERS peak as disconfirming previous hypotheses – at least in their narrow interpretation – that the latency of the beta maximum before an upcoming stimulus should linearly depend on the expected onset of that stimulus.
However, the figures reported in the paper also suggest that, following the last positive peak at ~800 ms after each stimulus, the speed (or slope) of beta-band desynchronisation might be predictive of when the next stimulus will appear. While the authors address this pattern of results only in passing and dismiss them as inconclusive, it should be noted that the relatively weak modulations in beta power towards the end of the trial are offset by prominent (and largely unexplained) modulations at the beginning of the trial. This is most likely due to the choice of baseline – it would be interesting to see how the results would change if the beta time-courses were normalised with respect to a period preceding the first stimulus in a sequence, instead of the average of the entire trial. In other words, while the exact timing of the maximum beta power does not appear to be modulated by the expected onset of an upcoming stimulus, more subtle beta-band mechanisms – such as the speed of beta desynchronisation just before an anticipated stimulus – might still be at play.
Another possibility, not tested here, would be that – rather than absolute beta power in a given region –beta-band connectivity between regions (e.g. the influence of frontal/motor activity on posterior/sensory regions) would be a more likely candidate for implementing predictive signals about upcoming events, consistent with the interpretation of beta-band synchronisation as mediating neural predictions between different cortical areas. Thus, while the results reported in this paper show that the peaks of beta-band synchronisation are not temporally specific with respect to anticipated events, to actually understand the possible role of beta oscillations in signalling temporal predictions we will need more insights about the neural sources and targets of the observed beta-band increases and decreases, and the connectivity influences explaining the beta-band dynamics.
Source article: Meijer D, te Woerd E, Praamstra P (2016). Timing of beta oscillatory synchronization and temporal prediction of upcoming stimuli. NeuroImage 138:233-241. doi: 10.1016/j.neuroimage.2016.05.071