Altered Activation and Functional Connectivity during Implicit Language Learning in 9-Month-Old Infants at Risk for ASD

Background: Word segmentation is a fundamental aspect of language learning, since the identification of word boundaries in continuous speech must occur before the acquisition of word meanings can take place. We previously used fMRI to show that children with autism spectrum disorder (ASD) are less sensitive to statistical and speech cues that guide implicit word segmentation. Prior studies have shown that neural activity related to language processing can be detected in 2-day-old neonates during natural sleep; to date, however, very little is known about the neural mechanisms underlying language learning during infancy and how this may be associated with risk for developing ASD.

Objectives: We examined the early neural signatures of language-related learning to characterize the neural networks subserving language acquisition in 9-month-old infants at high (HR) and low risk (LR) for ASD. We also related early measures of brain activity to longitudinal behavioral measures of language development and ASD symptom severity.

Methods: During natural sleep, infants were exposed to speech streams consisting of concatenated syllables where the statistical regularities and prosodic speech cues to word boundaries were manipulated across three conditions: 1) Stressed Language containing prosodic cues (i.e., stress; higher pitch, increased amplitude, and longer duration for the first syllable of the trisyllabic words used to created the speech stream) and strong statistical regularities, 2) Unstressed Language containing solely strong statistical regularities, and 3) Random Syllables containing weak statistical regularities and no prosodic cues. Risk status was determined by virtue of having one or more older siblings with a confirmed ASD diagnosis. A linear function was used to model changes that occurred over the course of exposure to each condition to estimate learning-related signal increases and compared these across groups. Next, a psychophysiological interaction (PPI) analysis examined differences in functional connectivity during exposure to the speech stimuli. Lastly, parameter estimates of brain activity and connectivity were correlated with behavioral measures collected at later time points.

Results: Compared to HR infants, LR infants showed more learning-related signal increases in left temporal regions during the Stressed Language condition, which contained both strong statistical regularities and speech cues. During this condition, the PPI analysis revealed greater functional connectivity between bilateral primary auditory cortex and right anterior insula in LR infants compared with HR infants. Interestingly, learning-related signal increases at 9 months correlated with later language outcome (CDI Receptive Advantage; 12 months) as well as ASD symptom severity (ADOS-2; 36 months); in addition, functional connectivity was associated with later language outcome (MSEL Verbal Score; 36 months).

Conclusions: In line with prior evidence in older children with ASD, HR infants already show diminished neural responses associated with implicit language learning by 9 months of age. Compared to the HR group, LR infants show greater brain activity as well as stronger functional connectivity in regions implicated in salience detection. Our findings indicate that early differences in the neural networks underlying language learning may predict altered trajectories in language development and ASD symptomatology before delays in language acquisition can be observed overtly at the behavioral level.