Event-Related Variability is Modulated by Task and Development

Abstract

Patterns of trial-averaged post-stimulus neural activity, e.g. event related potentials (ERPs) are traditionally interpreted as the correlates of cognitive operations. However, single-trial trajectories of neural responses approach these ERP components only in a loose and stochastic manner, questioning legitimacy of these components, proposed roles. Deconstructing the conventional ERP analyses, here we studied patterns of event related variability (ERV) in 2-6 month-old infants and adults using electroencephalography. Our analyses reveal that the ERP components are analogous to the biasing forces on the ongoing dynamics, which instrument the variability quenching and boosting events along neural trajectories. Moreover, the observed ERV possesses a rich temporal structure, modulated by both age and task, rivaling in complexity with that of the classical ERPs. Our findings suggest that since the early infancy, neural trajectories are actively controlled. This is compliant with the hypothesis that neural variability is a resource for cognitive processing, rather than unwanted noise.

Automated Pipeline for Infants Continuous EEG (APICE): a flexible pipeline for developmental studies

“…One possible explanation is that these methods are not optimal for infant data due to the intrinsic properties of the signal. For example, the high inter-trial variability observed in infants (Naik et al, 2021) might complicate the separation of variance related to evoked and non-evoked activity, as the method is based on the presence of highly reproducible activity across trials. The developmental changes and variability in infant EEG neural responses and physiological artifacts might also result in a poor decomposition during ICA for infant data.…”

“…The poor performances of the method are partially due to the nature of infant EEG data. Young infants' recordings contain more slow waves than adult data (Eisermann et al, 2013;Marshall et al, 2002), ERPs are less precise in time (Kushnerenko et al 2002) and more variable (Naik et al, 2021). The nature of the artifacts is also different in infants and adults.…”

“…Second, the power spectrum of the EEG signal (Eisermann et al, 2013;Marshall et al, 2002) and the properties of the evoked responses (Kushnerenko et al, 2002;Nelson & Monk, 2001) evolve throughout development due to maturational changes. More specifically, the background activity is rich and ample in low frequencies, and the signal variability between trials is higher in infants than in adults (Naik et al, 2021). Third, exogenous artifacts vary according to infants' age (e.g., fewer blinks and less motion in younger infants).…”

Interdependence patterns of multi-frequency oscillations predict visuomotor behavior

“…Typically, the slow fluctuations (ERD/ERS) visible in trial-averaged power profiles cannot be detected in individual trials (e.g. Tallon-Baudry & Bertrand, 1996;Naik et al, 2021); they do emerge as new and abstract features from smoothing out inter-trial fluctuations, treated as uninformative noise. But, concretely, behavior is performed in each individual trial, and also varies from trial to trial.…”

“…Such coordination arises in a self-organized manner, as the system is constrained by global structural and dynamical determinants to sample lower-dimensional manifolds within the high-dimensional space of configurations it could theoretically access if its parts were separately controlled (Bressler & Kelso, 2001;Pillai & Jirsa, 2017). Structured behavior along a task would thus just be associated to transient adaptations of system's ongoing trajectories on these manifolds, rather than to abrupt reconfigurations (Shine et al, 2019;Naik et al, 2021). These dynamical system views are akin to early proposals that evoked activity and networks are very similar to spontaneous fluctuation patterns (Kenet et al, 2003;Cole et al, 2014, Naik et al, 2021, possibly reflecting capability for probabilistic computations (Orban et al, 2016).…”