Four EEG studies (Seery et al., 2013, 2014; Finch et al., 2017; Riva et al., 2018) examined differences in ERP components elicited through auditory oddball paradigms, where a series of repetitive, standard auditory stimuli are occasionally interrupted by infrequent, deviant sounds (odd balls; Näätänen et al., 2007). Another study implemented a classic McGurk effect paradigm to examine neural correlates of audiovisual integration among EL vs. LL infants (Riva et al., 2022). Finally, Arslan et al. (2020) examined differences in neural responses to infants’ hearing their own name vs. an unfamiliar name.

Three studies used the same oddball paradigm (from Rivera-Gaxiola et al., 2005) which consisted of a stream of three consonant-vowel syllables: a standard stimulus (voiced, unaspirated, retroflex stop;/?a/) 80% of the time, a native deviant stimulus (voiceless, aspirated retroflex palatal stop;/ta/) 10% of the time, and a second non-native deviant (voiced, unaspirated dental stop;/da/) 10% of the time (Seery et al., 2013, 2014; Finch et al., 2017). In Seery et al. (2013), ERP responses to speech stimuli were recorded at 6 and 12 months to examine whether perceptual narrowing (i.e., increased sensitivity for native language-specific speech processing and decreased sensitivity for non-native language-specific processing) may be predictive of later autism symptomology (language processing, in particular). The results of this study found that at 6-months the P150 amplitude (occurring 150–300 ms after the onset of the stimulus) was similar for both native and non-native stop-consonant contrasts compared to the standard/da/for both groups of infants. At 12 months, both groups of children (including those diagnosed with ASD at 36 months of age) ceased to show a difference in their P150 amplitude of the non-native contrast, relative to the standard, in support of the perceptual narrowing hypothesis. In contrast, significant group differences were found for lateralization of ERP responses to speech between 6 and 12 months. While the EL group showed no difference in their responses to the speech stimuli across hemispheres, the LL group exhibited a more negative response over the right hemisphere relative to the left (Seery et al., 2013).

Building on Seery et al. (2013)’s findings, Finch et al. (2017) examined whether group differences (non-ASD, EL-non-ASD, vs. EL-later ASD) in lateralization of ERPs to speech sounds exist at 12 months of age and whether these patterns related to later developmental profiles. At 12 months, the EL-non-ASD group showed a more positive later-going slow wave (LSW) laterality index in the left hemisphere, compared to the EL-later ASD group. When infants with lower language scores at 12 months were excluded from analyses, non-ASD participants also exhibited a more positive LSW mean amplitude over the left hemisphere, compared to ASD infants (Finch et al., 2017).

In contrast to Seery et al. (2013, 2014) and Finch et al. (2017) aimed to determine whether the P150 of EL infants differed from LL infants in response to the first, second, and third consecutive repetitions of a standard stimulus (as described above). Additionally, this study examined the relation between these 9-month-old responses and subsequent 18-month-old behavior, as well as examining infants’ responses to a deviant stimulus to understand whether any potential atypicality was specifically related to stimulus repetition or more generally linked to speech processing (Seery et al., 2014). The findings revealed that EL infants exhibited a larger P150 amplitude to repeated standard speech stimuli compared to LL infants in the frontal region; and in EL infants only, higher P150 amplitudes to the standard stimuli over the frontal and central regions were associated with higher expressive language scores (Seery et al., 2014).

Riva et al. (2018) examined the P3 to “deviant” stimuli and the mismatch response (MMR) between the deviant vs. standard stimuli. This study aimed to determine whether the P3/MMR components (1) differentiated the LL and EL groups and (2) related to later language development and autism symptomology at 20 months. EL- ASD infants exhibited an overall higher P3 amplitude compared to LL infants across stimulus types and hemispheres. EL-ASD infants and EL infants with later language impairment (EL-LI) had a significantly longer peak latency relative to the TD group. Twelve-month-olds’ peak latency of the mismatch response duration deviant (MMRD) was associated with expressive vocabulary scores at 20 months of age. That is, faster MMRD at 12 months was associated with more words produced at 20 months (Riva et al., 2018).

In a follow-up study, Riva et al. (2022) investigated differences in audiovisual (AV) integration between EL and LL infants using ERP responses within a McGurk paradigm, which demonstrates the interaction between auditory and visual perception in speech processing by presenting mismatched auditory and visual speech stimuli, leading to the perception of a fused syllable (McGurk and MacDonald, 1976). This paradigm included four AV stimuli: congruent PA, congruent KA, incongruent fusion, and incongruent mismatch. Mean ERP amplitudes were analyzed in the following time windows: 100–250 ms for the visual effect in the occipital area and 350–650 ms for AV stimuli in the frontal (left and right) and temporal areas (left and right). They found that, in the left temporal hemisphere, LL infants exhibited a higher mean amplitude for the mismatch condition compared to both congruent conditions, while EL infants showed no differences between ERP responses to congruent and incongruent stimuli. This suggests that the LL group demonstrated AV integration effects over the left temporal area with larger mean amplitude responses to the mismatch condition, whereas the EL group showed no difference in their ERP responses to congruent and incongruent AV stimuli. However, both groups showed no differences in the earlier stages of visual processing (occipital early negative response), indicating that the impairment in the EL group was specific to AV speech integration.

Finally, Arslan et al. (2020) investigated whether EL and LL infants exhibit differences in ERPs in response to their own names at 10 and 14 months of age. LL infants displayed an overall increased positive neural activity in the frontal regions, compared to EL infants. As compared to LL infants, EL infants exhibited significantly reduced positive-going activity in response to their own name at 14 months. In addition, only 14-month-old EL infants displayed reduced neural responses to hearing their own name vs. an unfamiliar name. Frontal positive responses and the difference scores in the frontal areas between hearing one’s own name and an unfamiliar name correlated with the receptive language scores at 10 months (Arslan et al., 2020).

Five EEG investigations (Righi et al., 2014; Levin et al., 2017; Kolesnik et al., 2019; Wilkinson et al., 2020; Tran et al., 2021) examined EEG measures of spectral power and/or phase coherence within frequency bands or trials (i.e., inter-trial coherence; Kolesnik et al., 2019).

Two studies examined EEG power in 3-month-old EL and LL infants: one in a resting state EEG paradigm (Levin et al., 2017), and another in a passive listening paradigm (Tran et al., 2021). In the study using a resting state EEG paradigm, EL infants at 3 months showed significantly lower power over the frontal region in the high-alpha and beta bands (Levin et al., 2017). Another study that scanned infants during passive exposure to a stream of trisyllabic pseudo-words, however, found that 3-month EEG power was not significantly different between infants with (ASD-concern) and without (no-ASD-concern) autism concerns at 18 months. In contrast, 3-month phase coherence differentiated these groups. Indeed, at 3 months, infants later categorized in the ASD-concern group [groups defined using the ADOS Toddler Module (ADOS-T; Luyster et al., 2009) calibrated severity score (CSS; Esler et al., 2015) ranging from 1 to 10 where CSS > 4 = ASD-concern and CSS < 4 = No-ASD-concern] showed reduced theta and alpha coherence at left frontal-central electrodes (F9-C3) as compared to infants in the no-ASD-concern group (Tran et al., 2021). Using a double oddball paradigm, a third study examined linear coherence in gamma frequency bands between frontal and temporoparietal regions at 6- and 12 months. While no neural differences were significant between likelihood groups at 6 months, at 12 months, LL infants displayed higher linear coherence than EL infants. LL infants without later ASD diagnosis displayed higher linear coherence compared to all EL infants. In addition, EL-noASD infants showed higher coherence compared to EL-ASD infants (Righi et al., 2014).

In another study (Kolesnik et al., 2019), eight-month-old EL and LL infants were tested in a non-linguistic auditory oddball paradigm. The paradigm consisted of a pure tone at 500 Hz, presented with a 77% probability (standard tone), and two infrequent deviant tones—a white noise deviant and a pure tone of 650 Hz (pitch deviant)—each presented with an 11.5% probability. Compared to EL infants with typical development, infants who later developed ASD showed decreased neural habituation, as evidenced by an increase in gamma activation and significantly greater 10–20 Hz inter-trial coherence for repeated tones.

Of the five reported EEG studies in this section, four reported brain-language associations. In one resting-state study, reduced high-alpha power in frontal regions at 3 months was associated with lower expressive language scores at 12 months (Levin et al., 2017). Consistent with these findings, across all outcome groups, Tran et al. (2021) found that left frontal-central alpha coherence at 3 months was correlated with word production scores at 18 months (as measured by the MB-CDI; Tran et al., 2021). In contrast to these findings, a resting-state EEG study found that EEG-language associations differed by ASD outcome in EL infants, such that higher estimated 6-month gamma power was correlated with increased 24-month language functioning in HR-NoASD infants, but decreased 24-month language functioning in HR-ASD infants (Wilkinson et al., 2020). Somewhat consistent with findings from Wilkinson et al. (2020), using a non-linguistic auditory oddball paradigm, Kolesnik et al. (2019) found that 8 month cortical reactivity scores (derived by averaging z-scores for the evoked gamma and inter-trial coherence responses) were associated with reduced growth in receptive language skills between 8 and 36 months of age in high risk infants. Taken together, the differences in patterns of brain-language associations across the studies summarized here may potentially reflect early dynamic neurodevelopmental changes. Indeed, the former two studies, although using different EEG paradigms, focused on 3 month infants, while the latter two EEG studies focused on relatively older infants aged 6 and 8 months, respectively. Future work is needed to determine whether the reported brain-language associations are stable during early development or change dynamically over time.

Three investigations used novel EEG paradigms and/or analytic approaches beyond looking explicitly at more traditional EEG time- or frequency-domain metrics.

Using non-linear language-related EEG measures collected during an auditory oddball paradigm, Peck et al. (2021) aimed to predict ASD diagnosis in 6- and 12-month-old infants. For this study, the paradigm from Seery et al. (2013) was implemented but only EEG data recorded to the standard English phoneme (voiced, unaspirated dental/da/) were used. Employing a combination of Pearson correlation feature selection and a support vector machine (SVM) classifier, 100% predictive diagnostic accuracy was achieved at both ages. However, the predictive features differed between the models trained on 6-month versus 12-month data. At 6 months, features were biased toward measures from central electrodes, power measures, and frequencies in the alpha range. In contrast, at 12 months, features were more evenly distributed between power and non-linear measures and biased toward frequencies in the beta range. Notably, the diagnostic prediction accuracy substantially decreased in a larger, more behaviorally heterogeneous 12-month sample, from 100% with the matched dataset to 7% with the full dataset, suggesting an inability to effectively separate the two diagnostic outcome classes after expanding the EL-ASD group.

Another study investigated the association between neural processing of continuous speech (i.e., speech tracking), language development, and later autism symptoms. Infants at 10- and 14-months-old listened to five, sung nursery rhymes while continuous EEG data were recorded. Speech-brain coherence to these nursery rhymes served as a measure of infants’ ability to align neural activity with ecologically relevant speech input, with an emphasis on stressed syllables, syllables, and phonemes. The association between speech-brain coherence at 10 and 14 months and receptive and expressive vocabulary scores at 24 months [as measured by the parent-report, MacArthur-Bates Communicative Development Inventories (MCDI)] and autism symptoms at 36 months (as measured by the ADOS) was also examined. Results revealed no significant difference in speech-brain coherence between the EL and LL groups or any association between speech-brain coherence and later autism symptoms. However, a significant association was found between speech-brain coherence and later vocabulary development, with higher coherence in the stressed syllable rate correlating with higher parent-reported receptive and expressive vocabulary. This association was more pronounced in the 10-month-old EL group and is suggested to reflect individual differences in infants’ word segmentation skills, which have been found to be critical for language acquisition.

Finally, Hedenius et al. (2022) moved beyond the auditory domain and investigated the associations between global motion processing (GMP) and global form processing (GFP) with language and motor skills in 5-month-old EL and LL infants, assessed again at 18 months. GMP involves the integration of local visual motion signals into a global perception of motion, while GFP focuses on form processing and object recognition (Hedenius et al., 2022). The study found a significant negative relation between GMP laterality at 5 months and receptive language subtest on the MSEL at 18 months. The association between GMP laterality and language skills was consistent regardless of the infant’s likelihood for autism status, suggesting that atypical lateralization of GMP may be associated with lower motor and language scores at 18 months. These findings highlight the role of GMP in the development of motor and language skills, with potential implications for autistic children who may show atypical lateralization and processing challenges.