Sensory prediction (SP) is the ability to anticipate future stimulations on the basis of previous sensory inputs. It is related to another basic skill named repetition suppression (RS) which is the reduction of activity in the brain when a stimulation is repeated or becomes irrelevant. Oddball protocols, in which a rare deviant stimulus (target) appears randomly in a sequence of frequent repeated stimuli, and stimulus omission protocols, in which an expected (based on repetition) stimulus is omitted, are usually used to highlight these skills. For example, 2-3 months old infants present a decreasing neuronal response to repeated stimuli in the sensory cortices, and an increase of activity in sensory and frontal cortices during a deviant stimulation (Dehaene-Lambertz and Dehaene, 1994). Recent studies indicate that children born preterm (born before 37 weeks of gestational age, GA) and children with neurodevelopmental disorders (such as autism spectrum disorder or attention-deficit disorder) may have altered sensory prediction and repetition suppression. Infants born preterm have lower brain activation during sensory prediction tasks compared to full-term infants (Emberson et al, 2017; Boldin et al, 2018). Gonzalez-Gadea et al. (2015) demonstrated that children with neurodevelopmental disorders (NDD) could have different brain responses to both expected and unexpected deviant stimuli compared to controls in an auditory oddball protocol.The prevalence of neurodevelopmental disorders (NDD) is higher in the premature population (Johnson et al, 2011) and both prematurity and NDD are associated with sensory deficits, especially tactile hypo- or hypersensitivity, suggesting a common mechanism for altered RS and SP in prematurity and NDD. The aim of this work is to describe somatosensory repetition suppression and prediction from birth to 6 years of age in both neurotypical and neuroatypical children. We hypothesize that neurotypical children will have a larger RS and SP in somatosensory and frontal cortices compared to children born preterm or children with NDD. To test these hypotheses, we built a 17 minutes tactile omission protocol designed to generate RS and SP responses. Brain activity will be measured with electroencephalography (EEG) and functional Near Infrared Spectroscopy (fNIRS). Data will be analyzed using a mixed model analysis between conditions (first 50 stimuli, last 50 stimuli, standard, omission and postomission), age group (0, 2, 4, 6 years old) and neurodevelopmental status (neurotypical, degree of prematurity, neurodevelopmental disorder). We aim at including 160 children from birth to 6 years old, divided in 4 age groups (N=40 per group): premature newborns, 2 years old, 4 years old and 6 years old. Each group will be divided in 2 subgroups (N=20) sketching the neuroatypical and neurotypical developmental trajectories: at birth, half of the participants will be extremely premature (born before 32 weeks GA) and the other late premature (born after 34 weeks GA), at 2 and 4 years old, half of the participants will be born premature and the other half at term, and in the 6 years old group, half will be diagnosed with a NDD and the other half will be neurotypical. The protocol contains 300 stimuli divided in 3 parts. The first and the last 50 stimuli are standard and used to measure RS. In between, 40 blocks of four stimuli and one omission each, are used to evaluate SP. Each stimulus is a 200ms vibration that feels like moving on the skin on the anterior part of the forearm. Interstimulus interval is jittered between 2 and 4s. Omissions generate a 7s interval instead. Stimulation is delivered by custom-made vibratory matrices (Caylar SAS, Villebon-sur-Yvette, France). The matrix for newborns contains a column of 4 vibrators, the one for children contains 3 columns of 6 vibrators. Newborn brain activity will be recorded at 35 weeks of corrected GA during natural sleep in the neonatal intensive care unit of the University hospital of Caen, France. Children brain activity will be recorded in the laboratory while they watch moving circles on a screen with classical music. Children of 4 and 6 years old will also undergo behavioral, motor and cognitive evaluations in order to compare brain measures with development. Neuronal activity is measured using a 128 channels EEG (Magstim EGI, Eugene OR, USA) at 1000 Hz sampling rate. Impedance is kept under 50kΩ. The data are processed using Magstim EGI software Netstation: bandpass filtered (1-20Hz) then segmented (-100-900ms for the standard, and 7000ms for the omissions). Segments are visually inspected to remove artifacts before averaging. We apply baseline correction to all the segments at 100ms before stimulus onset. Using Matlab (The Mathworks, Inc. Natick MA, USA), we calculate amplitude and latencies of evoked potentials in three electrodes placed over the contralateral somatosensory cortex, and two electrodes placed over the posterior superior frontal gyrus. We calculate the difference between first and last 50 stimuli response averages to quantify RS in both regions of interest. The amplitude of potentials evoked by omissions is used to quantify SP. Newborn brain activity will also be recorded with simultaneous fNIRS (Imagent, ISS, Champaign IL, USA) at 690nm and 830nm, with 2 detectors and 8 sources placed over the two regions of interest (somatosensory and posterior frontal). Data will be processed using Homer3 software (www.bu.edu/neurophotonics/research/fnirs) (Conversion of light intensity to Optical Density (OD), bandpass filtering, Δ OD to hemoglobin concentrations, motion artifact removal and block average). In Matlab we will compare the difference between first and last 50 stimuli response (oxygenated hemoglobin concentration changes during the 3s following stimulus onset, relative to baseline: ∆HbO) averages to quantify RS. The amplitude of the hemodynamic response to omissions (both event-based and block-based) will be used to quantify SP. For EEG, we will compare values of RS and SP in both areas of interest using a mixed model analysis between conditions, age group and neurodevelopmental status. For fNIRS, we will compare values of RS and SP in both areas of interest using a mixed model analysis between conditions and neurodevelopmental status, for event-based and block-based responses. Finally, we will compare fNIRS and EEG responses in each condition to evaluate consistency between techniques, and determine which is most sensitive to neurodevelopment.