Anna Truzzi, Rhodri Cusack Trinity College Institute of Neuroscience, School of Psychology, Trinity College Dublin, Ireland
Events in the environment have distinct characteristic timescales. Each cognitive system needs to be tuned to the appropriate timescale in order to extract from the sensory stream the information relevant for cognitive development and learning. Neuroimaging studies in adults have found that each brain region has a characteristic intrinsic timescale over which it integrates information. However, we do not yet know if the timescales in the brain emerge as a consequence of learning the statistics from the environment or whether, on the contrary, the development of brain timescales precedes the learning process and drives and shapes it. We investigated intrinsic timescales in neonates shortly after birth. We characterised the timescales using an autoregressive model from resting-state fMRI data in low-movement neonates from the developing human connectome project (dHCP, N_infant=267) and low-movement adults from the CamCAN project (N_adult=252). The timescales in the neonates significantly differed by brain area and the difference was reliable across infants. Structure is therefore already present shortly after birth. Variation within some brain networks was similar between adults and infants, but the timescale distribution differed in that infants had overall slower timescales than adults. These longer timescales, especially in sensory areas, could be linked to lack of pruning and myelination. This characteristic might lead the infant to learn slower changing statistics and favour the emergence of more holistic cognitive representations less binded to little perceptual details. During development, the brain could then differentiate its tuning properties to match a variety of events with different statistics.
