The recording time varied from 4 to 6 min, depending on each infant’s attention to the stimuli. The behaviour of the infants was videotaped and off-line coded for EEG artefact rejection. High-density EEG was recorded using a 128-channel Hydrocel Sensor Net (EGI

Inc.) referenced to the vertex (Tucker, 1993). The EEG signal was amplified, digitized at 500 Hz and band-pass filtered from 0.1 to 200 Hz. The signal was off-line low-pass filtered at 30 Hz and segmented into epochs starting 100 ms before and ending 1,000 ms after the AV stimulus onset. Channels contaminated by eye or motion artefacts were rejected manually, and trials with > 20 bad channels were excluded. In addition, video recordings of the infants’ behaviour were coded frame-by-frame, and trials during which the infant did not attend to the face were excluded from further analysis. Following artefact rejection, the average number of trials for an individual infant accepted DNA Synthesis inhibitor for further analysis was 37.4 for /ba/, 36.7 for /ga/, 37.6 for VgaAba and 37.8 for VbaAga. Although uncommon for adult ERP studies, this number of accepted trials has been proved to be sufficient in infant studies (Dehaene-Lambertz & Dehaene, 1994; Friederici et al., 2007; Kushnerenko et al., www.selleckchem.com/products/PD-0325901.html 2008; Bristow et al., 2009; Guiraud et al., 2011). Artefact-free segments were re-referenced to the average

reference and then averaged for each infant within each condition. A baseline correction was performed by subtracting mean amplitudes in the 260–360 ms window from the video onset (i.e. immediately before the sound onset) to minimise the effects of any ongoing processing from the preceding stimulus. According to Kushnerenko et al. (2008) the AVMMR resembled the auditory mismatch response and was observed mainly over the right frontocentral area (between F4, C4 and Cz), commencing at ~ 290 ms from the sound onset

and lasting beyond the epoch of analysis. In this report AVMMR was observed only in response to apparent AV mismatch of speech cues (visual /ba/ auditory /ga/). In order to link individual differences in electrophysiological PIK-5 mismatch response to the development of visual scanning, the mean amplitude between 290 and 390 ms after sound onset (650–750 ms from video onset) from the area between F4, C4 and Cz was entered into hierarchical linear regression as the dependent variable with looking times to articulating mouth and control demographic variables (age, gender and second-language experience) as predictors. (Second-language experience here is defined as experience of one or more languages spoken at home in addition to English.) For the comparison between age groups we also measured mean voltage between 140 and 240 ms from the sound onset, centred around the mean latency of the auditory infantile P2 (Kushnerenko et al., 2002a, 2007) over the frontal leads.