by Shruti Naik (Neurospin, Saclay & Aix-Marseille University, Marseille, France).
The neural correlates of cognition have classically been studied by averaging the neural signal over many presentations of the same stimulus. The underlying hypothesis is that the average process recovers the stereotyped response evoked by each stimulus while the ongoing background activity, of larger amplitude but uncorrelated with stimulus timing, is cancelled out. Such assumption informs the framework of Event Related Potential (ERP) analyses that has been very popular to segment different stages in cognitive processing linked to perception and cognition, as well as the build-up of these stages themselves through learning and development. However, a long-time known issue with ERP analyses is however the extreme variability between different trials in response to similar stimuli or conditions. Since complete cognitive processes take place within each individual trial – we don’t need to see a stimulus many times to perceive it! – it is important to understand whether this extreme variability is noise or, on the contrary, signal. Indeed an increasing number of studies suggest that large fluctuations of responses and background activity convey task-relevant information and are crucial for cognitive function. Here, to probe the interplay between on-going activity and cognitive dynamics, we analyzed high-density EEG data obtained in 5 to 24 week-old human infants, which clearly display the gradual development of adult-like ERP responses eventually paralleled by an analogous development of variance modulations. Adults were also tested with the same lateralized faces experimental paradigm to have a comparison point. We find that the dynamics of variability quenching relative to stimulus can have a time-course way more complex than previously reported. We describe individual trials as trajectories in the space of possible brain topographies and find that the effects of development can be described as modulating the speed and dispersion of the “swarm” of possible trajectories, which are characteristically modulated at “fly-by” moments close to known ERP components. These trajectories follow characteristic patterns of deceleration and acceleration in the surrounding of ERP components that can be interpreted as motion in presence of effective energy barriers that have to be crossed. Development does not reduce between and within-trial variability –which on the contrary increases, leading to distributions of fluctuations reminiscent of power-laws– but seems to increase the degree of structuring of variability modulations, as revealed by the crossing of growing effective energy barriers along the trial dynamics. The classic ERP approach is thus revisited and reinterpreted as tracking the shadow surviving averaging of an underlying rich Event-Related-Variability (ERV) dynamics which occur within each trial and gradually builds up along development in the first critical semester of life.
Spotlight talk presented on October 8th 2020, at the Brain Criticality Virtual Conference 2020 (Plenz D., Chialvo D., de Arcangelis L. & Battaglia D. organizers)