Investigating Brain Response to Speech in 6- and 7-Month-Old Infants: Evidence for Atypical Statistical Learning in Infants at High Risk for ASD

Background: Statistical learning is a mechanism of language acquisition that may be atypical in ASD (Romberg & Saffran, 2011). Previous studies have investigated statistical learning by measuring changes in brain activity across trials as infants listen to repeated stimuli. When listening to repeated patterned speech, typically developing (TD) neonates exhibit increased brain activity overtime (Gervain et al., 2008). By 3-months, TD infants exhibit suppressed brain activity overtime, but infants at high risk for ASD do not exhibit changes in brain activity overtime (Edwards et al., 2017). It is however unknown whether these high risk infants continue to exhibit atypical statistical learning as they age, and whether this is early brain activity predicts language abilities later in development.

Objectives: The present study investigated whether 6- and 7-month-old infants at low and high risk for ASD exhibit neural evidence of statistical learning when listening to patterned versus non-patterned speech, and explored whether early brain response predicts receptive language abilities at 24-months.

Methods: 26 (14M, 12F) low risk control (LRC) and 12 (6M, 6F) high risk for ASD infants (HRA-), ages 6 months;5 days to 7 months;28 days, listened to two syllable sequence types, patterned (ABB; “ba-ga-ga”) and non-patterned (ABC; “ba-ga-lo”) speech, while wearing a fNIRS cap. HRA- infants had an older sibling with ASD, but no ASD diagnosis by 24-months. All infants provided data from 16+ trials. Change in oxyhemoglobin concentration (OxyHb) over the 16-second trial period was extracted and averaged across the first four trials and last four trials (trial blocks) after motion artifact correction. This process was repeated for two regions of interest (ROIs), left anterior and right anterior (Figure 1.b), and for both syllable sequence types. At 24-months, infants returned to complete the Mullen Scales of Early Learning to assess language.

Results: Results of a 2x2x2x2 mixed factorial ANOVA revealed a significant interaction effect between ROI, time, and group (F(1,36)=4.277, p=.046). Collapsing across both syllable sequence types, LRC infants had greater OxyHb response (M=.121x10^-3) than HRA- infants (M=-.013x10^-3) during the first trial block within the left anterior ROI (Figure 1.a). LRC infants had decreased OxyHb in response to speech during the last trial block (M=.010x10^-3) versus the first trial block (M=.121x10^-3) within the left anterior ROI. There were no significant differences in OxyHb response during the first trial block (M=-.013x10^-3) versus the last trial block (M=.084x10^-3) for HRA- infants within either ROI. Additionally, OxyHb response to patterned speech during the first trial block within the left anterior ROI was significantly correlated with Mullen receptive language scores at 24-months (r(30)=.438, p=.016; Figure 2).

Conclusions: 6- and 7-month-old TD infants exhibited left-lateralized suppression of brain activity as they listened to repeated stimuli, which may reflect successful stimulus encoding of both patterned and non-patterned speech (Issard & Gervain, 2018). In contrast, HRA- infants did not exhibit changes in brain activity overtime, which may indicate atypical statistical learning within the broader ASD phenotype. Future studies should continue to investigate whether early brain response to speech predicts later language abilities in high risk populations.