Altered Neural Responses to Familiar and Unfamiliar Speech in Six-Week Old Infants at High Risk for ASD

Background: Based on prenatal exposure to speech, newborns already show a preference for their native language (e.g., Mehler et al, 1988). Infants’ native language preference (or lack thereof) can thus provide an index of their attention to, and uptake of, the input that is available to them. A growing number of studies have demonstrated that fMRI during natural sleep can successfully be used to map neural responses to speech in infancy (e.g., Dehaene-Lambertz et al., 2010; Redcay & Courchesne, 2008b), with reliable activity in fronto-temporal language areas reported as early as 2 days after birth (Perani et al., 2011). Notably, altered activity during speech processing has also been observed in toddlers with autism spectrum disorders (ASD) who underwent fMRI during natural asleep (Redcay & Courchesne, 2008; Eyler et al., 2012).

Objectives: Here we examine the neural circuitry subserving language processing in 6-week-old infants at high (HR) and low risk (LR) for ASD in order to identify patterns of brain activity that may predict altered trajectories of language development, as well as an ASD diagnosis, well before overt delays in language acquisition and ASD symtomatology can be observed.

Methods: fMRI was conducted during natural sleep. Stimuli were speech samples produced by different female native speakers of English and Japanese (8 segments for each language). Both English and Japanese stimuli were previously used in behavioral studies in LR infants (Sundara et al, 2008) and were matched for duration, intensity, peak amplitude, pitch, and pitch range. According to a traditional block-design, the alternating English and Japanese speech segments (18s each) were interspersed with periods of silence (12s each). Preprocessing (including motion scrubbing) and statistical analyses were conducted in FSL except for registration to a neonate template implemented in AFNI. Regression analyses were conducted using language measures collected at 9 (MacArthur-Bates Communicative Development Inventories; CDI) and 12 (Mullen Scales of Early Learning; MSEL) months of age.

Results:   Robust activity in bilateral language areas was detected in both groups for both English and Japanese segments (vs. baseline), with LR infants showing overall stronger activity than HR infants. Both groups showed differential activation patterns for English vs. Japanese. However, these patterns differed between the two groups. LR infants showed significantly greater responses to Japanese than English in temporal language regions, as well as in parietal attentional areas, suggestive of a novelty bias. In contrast, HR infants showed no significantly different activity for Japanese and English in canonical language areas, showing instead greater activity in subcortical regions for English vs. Japanese. Neural responses to speech at 6-weeks of age predicted CDI and MSEL language scores at 9 and 12 months of age, respectively. 

Conclusions: These data suggest that natural sleep fMRI is well suited to probe the extent to which infants have already learned the prosodic contour of their native language. Furthermore, these findings suggest that early patterns of brain activity in response to familiar and unfamiliar speech are predictive of subsequent language development and may thus provide an early biomarker of future risk for ASD.