By Thomas Meindertsma & Tobias Donner.
The brain continuously tracks the statistical structure of the environment to anticipate behaviorally relevant events. Deviations from such predictions cause surprise, a violation of the agent’s expectation about an event, which might be indicative of an underling change in the environment. So, it would make sense if the brain uses surprise for dynamically modulating its sensitivity to evidence about the state of the environment. Indeed, it has been proposed that surprise transiently boosts central arousal state, mediated by phasic responses in neuromodulatory brainstem systems (such the locus coeruleus noradrenaline system), which project widely to the networks of the cerebral cortex engaged in evidence accumulation and inference (Dayan & Yu, 2006). Previous neurophysiological work in non-human primates has linked surprise, here about the timing of rewards, to phasic responses in dopaminergic midbrain nuclei (Fiorillo et al., 2008). However, little has been known about the impact of surprise on ongoing cortical network dynamics.
In this new study, we show that surprise induces a transient suppression of ongoing beta-band oscillations in the human brain – specifically within regions of prefrontal and parietal association cortex. To this end, we combined a Bayesian belief updating model with magnetoencephalography (MEG) recordings of cortical network dynamics. We developed an analysis approach that enabled us to comprehensively map out surprise-related modulations of cortical network activity across space, time, and frequency. Surprise was elicited by variations in the timing of the changes (on- or offsets) of a salient visual target stimulus, which had to be reported by participants with a simple behavioral response (Figure 1A). Our Bayesian (ideal observer) model updated a ‘belief distribution’ about the timing the upcoming sensory event based on the previously experienced timings. Specifically, we used the model to calculate how unlikely the timing of a stimulus change was at any given moment in time, given this belief distribution (Figure 1A-B).
Our model-based estimates of surprise varied substantially from trial to trial (Figure 1B). We found that these variations predicted the reaction times (RT) of the participants: when surprise was high the responses were slower. This relationship between surprise and RT existed even when we controlled for the timing of the two preceding stimulus changes; this shows that participants indeed tracked the temporal structure of the task in a similar fashion as our ideal observer model.
Trial-by-trial variations in surprise were correlated negatively with the power of cortical population activity measured with MEG. This surprise-related power suppression occurred transiently around the behavioral response, specifically in the beta frequency band (Figure 1C). It peaked in parietal and prefrontal cortices, remote from the motor cortical suppression of beta power related to the button presses used to report of stimulus changes (Figure 1D).
Our results indicate that surprise about sensory event timing transiently suppresses ongoing beta-band oscillations in association cortex. This transient suppression of frontal beta-band oscillations might reflect an active reset triggered by surprise, and is in line with the idea that beta-oscillations help maintain cognitive sets (Spitzer & Haegens, 2017). Further, the approach that we have developed here will be useful also for tracking the impact of other important computational variables on ongoing cortical network dynamics.
Reference: Meindertsma T, Kloosterman NA, Engel AK, Wagenmakers EJ, Donner TH (2018) Surprise About Sensory Event Timing Drives Cortical Transients in the Beta Frequency Band. Journal of Neuroscience xx(xx). DOI: https://doi.org/10.1523/JNEUROSCI.0307-18.2018