Looking to and Processing of Audiovisual Speech and Associations with Language in Infant Siblings of Autistic and Non-autistic Children

This study demonstrates that while differential audiovisual speech processing, indexed by N2 ERP amplitude suppression, is present in 12-18-month-old infants regardless of autism risk, its relationship with looking behavior and emerging language abilities is significantly moderated by group status, age, and biological sex.

The Gist

Imagine your baby's brain is like a busy construction site. Right now, the workers are trying to figure out how to build the "Language Tower." To build this tower, they need to understand how to combine two types of blueprints: the sound of a voice (hearing) and the movement of a mouth (seeing).

This study is like a sneak peek into that construction site, looking at babies between 12 and 18 months old. The researchers wanted to see if babies who have an older sibling with autism (let's call them "Group A") process these blueprints differently than babies with non-autistic siblings ("Group B").

Here is the story of what they found, broken down into simple parts:

1. The Experiment: The "Silent Movie" vs. The "Talk Show"

The researchers put babies in front of a screen and played two types of videos:

• The Talk Show: A woman saying "Ba" while her lips move naturally (Audio + Video).
• The Silent Movie: The same woman saying "Ba," but her face is frozen still, so you only hear the sound (Audio only).

While the babies watched, the researchers wore special hats with sensors (EEG) to listen to the electrical "buzz" of the babies' brains. They also used a special camera to track exactly where the babies were looking (at the eyes, the mouth, or the whole face).

2. The Big Discovery: The Brain's "Volume Knob"

When adults or older kids hear and see speech at the same time, their brains usually turn the volume down slightly because the two senses help each other out. It's like having two people tell you the same joke; you don't need to listen as hard because you can see their lips moving.

What they found in the babies:

• The "Volume Knob" works: Even at 12–18 months old, the babies' brains showed a specific signal (called the N2 wave) that changed when they saw the moving mouth. This proves that babies this young are already starting to combine sight and sound to understand speech.
• No "Group" Difference: Surprisingly, the average brain signal was the same for both Group A (autism siblings) and Group B. Just because a baby has an autistic sibling doesn't mean their brain is "broken" or working differently on average.

3. The Twist: It's All About the Individual

If the brains were the same on average, why did the researchers care? Because when they looked at individual babies, they found a fascinating pattern. It wasn't about the group; it was about the specific mix of who the baby was and what they were doing.

Think of it like a recipe. The ingredients (brain signals, looking behavior, language) are the same for everyone, but the amount of each ingredient changes the final dish.

• The "Looking" Factor:

  • In Group B (non-autistic siblings), babies who looked more at the speaker's mouth had a specific brain pattern.
  • In Group A (autism siblings), babies who looked more at the mouth had the opposite brain pattern.
  • Analogy: Imagine two cars driving down the same road. One car (Group B) speeds up when it sees a green light. The other car (Group A) slows down when it sees a green light. Both are reacting to the light, but in opposite ways.

• The "Age" and "Gender" Factors:

  • The link between the brain's "volume knob" and how well the baby speaks was only strong for boys and for babies older than 14 months.
  • For baby girls or younger babies, this link was weak or non-existent.
  • Analogy: It's like a radio signal. For some babies (older boys), the signal is crystal clear: "Better brain processing = Better talking." For others, the signal is fuzzy or static.

4. Why This Matters

This study teaches us three big lessons:

1. Babies are smart: By their first birthday, babies are already wiring their brains to combine sight and sound to learn language.
2. One size does not fit all: We can't just say "Autism siblings are different." Some are different, some are the same, and some are different in the opposite direction. The "average" hides the real story.
3. Context is King: To understand a baby's language development, we have to look at the whole picture: Are they a boy or a girl? How old are they? Where are they looking?

The Bottom Line

This research is like finding a new map for the construction site. It tells us that while the "Language Tower" is being built in all babies, the tools and the workers' habits vary wildly from child to child.

The researchers suggest that in the future, instead of treating all babies the same, we might need to tailor our help based on these specific details. For example, if we know a baby is a boy over 14 months old who struggles with language, we might focus on helping him look at the speaker's mouth, because that's the specific "key" that unlocks his brain's language potential.

Note: This is a preprint, meaning it's a fresh discovery that hasn't been fully checked by other scientists yet, but it offers a very exciting new way to look at how babies learn to talk.

Technical Summary

1. Problem Statement and Research Context

Autism Spectrum Disorder (ASD) is characterized by significant heterogeneity in language development. Early language ability is a strong predictor of long-term social and vocational outcomes. The cascading effects framework posits that early differences in sensory processing (specifically audiovisual speech) may cascade into higher-order language deficits.

While previous research has established links between audiovisual speech processing (measured via Event-Related Potentials or ERPs) and language in school-aged children, there is a critical gap in understanding these mechanisms in infancy (12–18 months). Specifically, it is unknown:

• Whether differential neural processing of audiovisual vs. auditory-only speech exists in infants.
• Whether these neural markers differ between infants at elevated likelihood for autism (Sibs-Autism) and those with non-autistic siblings (Sibs-NA).
• How neural processing, visual attention (looking to the mouth), and emerging language abilities interact, and whether these relationships are moderated by participant characteristics (group, age, sex).

2. Methodology

Participants:

• Sample: 54 infants aged 12–18 months (29 Sibs-Autism, 25 Sibs-NA).
• Matching: Groups were matched on biological sex and chronological age.
• Inclusion Criteria: Full-term birth, no genetic/neurological disorders, monolingual English household, and at least one older sibling.
• Group Definitions: Sibs-Autism required a confirmed ASD diagnosis in an older sibling (DSM-5/ADOS-2); Sibs-NA required no family history of ASD and low scores on the Social Communication Questionnaire.

Experimental Tasks:

1. EEG/ERP Task:

   • Stimuli: The syllable "ba" spoken by a female. Two conditions: Audiovisual (AV) (synchronized audio/video) and Auditory-Only (AO) (audio with a static face).
   • Procedure: 50 trials of each condition presented in random order.
   • Analysis: EEG recorded via 128-channel Geodesic sensor net. Data processed using EEGLAB (filtering, artifact removal, interpolation). Temporal Principal Component Analysis (PCA) identified two components of interest: P1 (100–200 ms) and N2 (350–500 ms).
   • Metric: Amplitude suppression (difference between AO and AV amplitudes) was calculated.

2. Eye Tracking Task:

   • Stimuli: 50-second video of an adult female using infant-directed speech.
   • Metric: Proportion of total looking time directed at the speaker's mouth (vs. eyes/face).

3. Language Assessment:

   • Measures: Mullen Scales of Early Learning (Mullen), MacArthur-Bates Communicative Development Inventory (MCDI), and Vineland Adaptive Behavior Scales (VABS-2).
   • Aggregation: Receptive and Expressive Language scores were aggregated into composite z-scores.

Statistical Approach:

• Design: Intact-group comparison and concurrent correlational design.
• Analyses: 2 (Group) × 2 (Condition) ANOVAs for ERP amplitudes.
• Moderation: Regression analyses using PROCESS in R to test if Group, Chronological Age, or Biological Sex moderated associations between ERP effects, looking behavior, and language.
• Handling Missing Data: Imputed using missForest (missing at random assumption).

3. Key Results

A. ERP Amplitude Effects (Research Questions 1 & 2)

• N2 Component: A significant main effect of Condition was found ($p = .042$). N2 amplitudes were significantly more negative (suppressed) in response to Audiovisual speech compared to Auditory-Only speech across both groups.
• Group Differences: No significant main effect of Group or Group × Condition interaction was found for N2 or P1. This indicates that, on average, Sibs-Autism and Sibs-NA show similar neural suppression patterns at this age.
• P1 Component: No significant differences were found for P1 amplitudes across conditions or groups.

B. Associations with Looking Behavior (Research Question 3)

• Moderation by Group: The relationship between looking to the mouth and ERP amplitude effects differed significantly by sibling group.
  • Sibs-Autism: Positive correlation between looking to the mouth and P1 amplitude effects ($r = .328$).
  • Sibs-NA: Negative correlation between looking to the mouth and P1 amplitude effects ($r = -.333$).
  • N2 Effects: Sibs-NA showed a negative correlation between looking to the mouth and N2 effects ($r = -.427$), while Sibs-Autism showed a negligible correlation ($r = .070$).

C. Associations with Language (Research Questions 4 & 5)

• Looking to Mouth & Language: The association between looking to the mouth and Expressive Language was moderated by Biological Sex.
  • Males: Moderate positive association ($r = .458$).
  • Females: Small negative association ($r = -.150$).
• ERP Effects & Language: P1 amplitude effects were significantly associated with both Receptive and Expressive language, but these associations were moderated by Age and Sex.
  • Age: Significant positive associations emerged only in infants older than ~14 months.
  • Sex: Significant positive associations were found in males ($r \approx .45–.53$), but negligible associations in females.
• N2 Effects & Language: No significant zero-order correlations or moderated effects were found between N2 amplitude effects and language.

4. Key Contributions

1. Developmental Timeline: This is the first study to demonstrate that differential ERP amplitude suppression (specifically at the N2 component) in response to audiovisual vs. auditory-only speech is present in infants as young as 12–18 months. This extends previous findings from school-aged children (who show P2 effects) to infancy.
2. Individual Variability vs. Group Means: The study highlights that individual differences in neural processing are more predictive of language outcomes than mean group differences between Sibs-Autism and Sibs-NA.
3. Complex Moderation: The research provides robust evidence that the links between sensory processing, visual attention, and language are not uniform. They are heavily dependent on participant characteristics:
   • Sibling Group: Determines the direction of the link between looking behavior and neural processing.
   • Age: Neural-behavioral links to language only become significant after 14 months.
   • Sex: Males show stronger positive associations between neural/visual metrics and language than females in this sample.
4. Methodological Integration: Successfully integrated EEG, eye-tracking, and standardized language assessments in a high-risk infant population to test the "cascading effects" framework.

5. Significance and Implications

• Mechanistic Insight: The findings support the hypothesis that early sensory processing differences (specifically how the brain integrates visual and auditory speech) contribute to the heterogeneity of language development in autism.
• Early Biomarkers: While group means did not differ, the specific patterns of neural suppression and their relationship to looking behavior could serve as early biomarkers for language risk, particularly when analyzed within specific subgroups (e.g., males >14 months).
• Intervention Targets: The study suggests that interventions targeting looking to the mouth and multisensory speech processing during the 12–18 month window could be crucial for optimizing language outcomes, especially for infants showing specific neural-behavioral profiles.
• Future Directions: The authors emphasize the need for larger, longitudinal studies to establish causality and to investigate why sex differences and age thresholds exist. They also note the need for more diverse samples (currently predominantly White/non-Hispanic) to ensure generalizability.

Limitations:

• Cross-sectional design prevents causal inference.
• Small sample size ($N=54$) limits power for some interaction effects.
• Lack of multiple comparison corrections in exploratory regression analyses.
• Demographic homogeneity limits generalizability.