The spoken word ‘kipi’ or ‘moma’ (400 msec in duration) was presented

550 msec after the onset of the visual stimulus. Infants passively saw and heard the stimuli. An attention-getter was presented in one fourth of the trials (randomly selected) to regularly reinforce the infants’ attention towards the display. The EEGs were continuously recorded from silver–silver chloride electrodes attached to an elastic electrode cap. EEG data were recorded at 11 electrode sites: F3, Fz, F4, C3, Cz, C4, P3, Pz, P4, and left and right mastoids (A1, A2). The ground electrode was placed at FPz. Electrode click here impedances were kept mostly below 10 kΩ. The EEG activity was amplified with Neuroscan SynAmps2, digitized online at a rate of 1 kHz, and filtered on-line (bandpass between .1 and 200 Hz). The EEG was re-referenced to the average of left and right mastoid channels (A1, A2). Artifact rejection was performed based on the criteria used in the ERP analyses (see section 2.5.2). There was a minimum of 21 valid epochs per condition in every infant participant (mean: 47.6 epochs in the match condition and 46.7 epochs in the mismatch condition). Epochs ranged from −2000 to 1500 msec after

the auditory onset. To estimate local brain networks, we extracted amplitude of oscillations in each frequency band (Herrmann et al., 2004 and Schneider et al., 2008). It was extracted by using the wavelet transform at the target frequency (f) ( Lachaux et al., 2000). The frequency ranged GPX6 from 2 Hz to 45 Hz in 1 Hz steps. To avoid problems due to the sample size bias, for each infant, the number of epochs was made the same for the match and mismatch conditions by randomly selecting the find more same number of epochs. EEG signal s(t) was convolved with the complex Morlet’s wavelet defined by: w(t,f)=fexp(−t2/2σt2)exp(i2πft),as a function of time (t) and frequency

(f). The Morlet wavelet is characterized solely by σt, which sets the number of cycles of the wavelet: nco = 6fσt. We chose nco to be 8 ( Lachaux et al., 2000). To detect auditory event-related changes in amplitude, we first computed the instantaneous amplitude of EEG signal from electrode n by deriving the length of the convolved signal as follows: Ant=|wt,f*snt|.Ant=|wt,f*snt|. Next, we averaged the instantaneous amplitude An(t) across all trials and obtained averaged amplitude AMPn(t). Finally, we standardized the averaged amplitude relative to the pre-stimulus baseline period (600 msec–100 msec before the visual onset) for each electrode and frequency. Standardized amplitude values for each time point t [AMPz(t)], were computed as follows: AMPz(t)=AMP(t)−AMPBmeanAMPBsdwhere AMPBmean and AMPBsd are, respectively, the mean and standard deviation of the AMPs computed from the baseline period at each frequency. The resulting index, AMPz, indicates standardized changes in the direction of increased amplitude (positive values) or decreased amplitude (negative values).