Electrophysiological Correlates of Word Segmentation in Three-Month-Old Infants at High and Low Risk for Autism Spectrum Disorder

Background: Early Auditory Statistical Learning (ASL) – the ability to implicitly extract statistical regularities available in the input (Saffran et al., 1996) – is a precursor to later expressive language ability (Romberg & Saffran, 2010). ASL is crucial for language learning, as it allows infants to detect word boundaries by determining transitional probabilities between syllables in continuous speech (Kuhl, 2004). Successful word segmentation has been demonstrated in 6-month-old infants (Johnson & Tyler, 2010) but no studies to date have examined ASL in 3-month-old infants at high- and low-risk for autism.

Objectives: Here we examined whether EEG signatures of ASL can be quantified at 3-months-of-age, whether different patterns are observed in infants at high (HR: have an older sibling with ASD) and low-risk (LR) for autism, and whether ASL at 3-months predicts later language ability.

Methods: Three-month-old infants (HR:n=17, LR:n=22) were exposed to a continuous speech stream created by concatenating four tri-syllabic artificial “words”, constructed from a set of 12 syllables. Infants were then presented with the same tri-syllabic combinations (“words”), as well as tri-syllabic combinations not heard during the exposure phase (“non-words”). High-density electroencephalography (EEG) was recorded (128-electrode, EGI Inc.) and the event-related EEG components of interest included the frontal-auditory P100 and left-temporal Positive Slow Wave (PSW). A general linear model was used to examine within-subject effects of region and condition and between-subject effects of group with respect to P100 maximum amplitude and latency, and PSW mean amplitude. Evidence of word segmentation was operationalized as differences between conditions on PSW. Correlations between EEG markers of word segmentation and verbal ability (receptive and expressive language scales, Mullen Scales of Early Learning (MSEL), 1995), receptive vocabulary (MacArthur Bates Communication Inventory (MCDI), 2007) at 9- and 12-months were also examined.

Results: We observed a significant main effect of group (F(1,37)=4.59, p=.03) in the frontal P100 latency, with slower P100 peak in HR infants (M=242.5) compared to LR (M=219.4). With respect to PSW mean amplitude, we observed a significant region by condition interaction (F(1,37)=3.84, p=.05) and a significant main effect of region (F(1,37)=4.66, p=.03). Post-hoc tests revealed greater PSW mean amplitude within the left region (M=2.9) compared to the right (M=1.47, p=.03); however, a significant difference between conditions was only observed in the left temporal region only (t(38)=2.29, p=.03). No significant correlations emerged between P100 frontal latency and MSEL or MCDI measures but the difference in PSW mean amplitude between conditions significantly predicted MCDI words understood at 9-months (r(25)=.46, p=.03).

Conclusions: Our results indicate differential low-level auditory responses as a function of risk-group, with HR infants showing a delayed P100 compared to LR infants. Moreover, across risk groups, infants as young as 3-months show EEG evidence of ASL as evidenced by the left temporal PSW. ASL at 3-months of age predicted receptive vocabulary at 9-months. Further analyses will examine relations between low-level auditory processing, subsequent language development, and clinical outcomes in infants at high-risk for ASD as well as changes in ASL over the first year of life in both high- and low-risk infants.