Travelling together: Uncovering how social interaction shapes social brain development

Identifying and responding to social cues is one of the critical characteristics of being human, and infants' behaviour will trigger responses from their caregivers and family from the first minutes following birth. Previously, studies used functional near-infrared spectroscopy (fNIRS) to investigate responses in the temporal lobe of neonates between 1 and 5 days old. In a post-hoc analysis of a small cross-sectional sample, preliminary evidence indicated that distinct cortical responses to social video stimuli emerged within the initial days following birth. 

Dianna Ilyka at the University of Cambridge has delved into understanding the associations between cortical specialisation to social stimuli, caregiver-infant interactions, and the emergence of social capacities in infants in the first five months of life. Dianna analysed data from three longitudinal projects to investigate the interplay between cortical specialisation for social stimuli and its relationships with behaviours sampled from infant social interactions. 

In her first study, Dianna devised and validated a coding scheme to capture the behaviours of both infants and caregivers during interactions when the infants were approximately one month old and then again at five months old. Through this scheme, she unveiled a reciprocal dynamic in caregiver-infant interactions at one and five months of age. Furthermore, she observed that dyadic behaviours at one month, such as mutual gaze and dyadic engagement, predicted infant later behaviour beyond individual caregiver or infant behaviours, suggesting an early developmental impact of mutual interaction patterns, even in neonates. In the following two studies, Dianna studied typically developing infants and those at elevated likelihood (EL) for developing autism spectrum disorder (ASD) and/or attention deficit hyperactivity disorder (ADHD) at around five months. Using fNIRS and cluster permutation analyses (CPA) to analyse the brain data, she assessed how their temporal and frontal cortices responded to social stimuli (i.e., faces) compared to non-social stimuli (i.e., toys), extending previous research findings with the paradigm. She revealed that EL infants showed reduced social brain responses (visual and auditory) to social stimuli compared to TL infants. Although measures of caregiver-infant interactions did not significantly differ between the two groups, Dianna revealed, across two studies, that in typically developing infants, caregiver-infant interaction behaviours were linked to selective cortical responses to social visual and vocal stimuli. However, crucially, these associations were not observed in the elevated likelihood group, highlighting differences in susceptibility to social experiences in this group. 

Dianna also played a role in establishing the PIPKIN study, which gathered neuroimaging data from newborns using the high-density-diffuse optical tomography (HD-DOT) device, focusing on the critical early phase of life when sensory and interactive changes are most prominent. The study aimed to leverage the unique resources available at the Cambridge maternity hospital to attempt to replicate previous findings on selective cortical responses in neonates with a larger longitudinal sample (N=60). Additionally, this study aimed to extensively monitor newborns' social experiences and interactions in their home environments over this period to determine the relationship between very early experiences and the emergence of specific social responses in the temporal lobe. Unfortunately, due to the COVID-19 pandemic, data collection has been delayed, leading to preliminary analyses of the results. Nevertheless, Dianna's initial findings demonstrated the feasibility of conducting HD-DOT assessments with newborns in home settings. Her preliminary findings also revealed early changes in how infants processed vocal and non-vocal stimuli, which may be potentially linked to concurrent measures of neonatal alertness (i.e., readiness, motivation and ability to engage with others). Furthermore, she provided evidence of the development of social visual responses in neonates. Importantly, as part of HD-DOT data analysis, Dianna developed an algorithm tailored to automate the identification of optimal channels in HD-DOT. Her primary objective was to enhance the preservation of valuable data and ensure replicability. This innovation has marked a notable improvement compared to the more manual processes employed previously. In addition to working on these studies, Dianna published a Review Article titled "Infant Social Interactions and Brain Development: A Systematic Review" (Ilyka et al., 2021) and established the PIPKIN journal club, connecting researchers and students passionate about developmental research. 

Moreover, Dianna has also been involved in various academic roles, including the co-supervision of Mphil students within the PIPKIN project and the supervision of undergraduates in the Cambridge Access to Mental Health & Psychology (CAMP) 2022 program, showing a strong commitment to both academic research and a supportive engagement towards students' academic progress from a diverse socioeconomic backgrounds and ethnic minorities. During her secondment at CBCD in Birkbeck, University of London, Dianna helped collect data for a different ESR project. She gained experience in real-time social paradigm design and contributed to developing the fNIRS real-time processing pipeline. Within the framework of the PIPKIN project, she obtained certification in utilising the Neonatal Behavioural Assessment Scale (NBAS) following extensive training completed in the Rosie Hospital's maternity ward in Cambridge. 

Dianna's work has unveiled the significant association between early social interactions and social brain functions during infancy, encompassing the early prenatal and postnatal phases. It highlights the significance of conducting research throughout the neonatal period, shedding light on how the behaviours and interactions of newborns in their earliest stages of life contribute to the development of social behaviour and brain functions. Her project has been valuable in addressing the limitations of prior research on social interactions and brain development as measured with the fNIRS technique. Within the context of the PIPKIN project, her valuable contributions have played a pivotal role in pushing the boundaries of HD-DOT technology application in home settings for longitudinal studies involving newborns. Her fNIRS analysis approach, featuring cluster permutation analyses, extends our understanding of the selectivity of the social brain in infants. Her consistent and methodical application of a behavioural coding scheme across these studies has yielded a systematic and comprehensive investigation of the variables that impact early social brain functions, taking neurobiological factors into account.

Tanaya’s project has investigated how parental social responsiveness and infant sensory reactivity interact to shape social brain development. Daily social interactions play a pivotal role in shaping the emerging abilities of infants to regulate perceptual input and deploy attentional resources efficiently (Niedzwiecka et al., in press). Early difficulties in top-down regulation of perceptual input have been proposed as a mechanism driving atypical development in autism and other Neurodevelopmental Disorders (NDDs) (Pellicano, 2013; Jones et al., 2017). High levels of sensory reactivity may be associated with improved social attention and development (Jones et al., 2017), but this relationship may only hold true in optimal social environments where parents tailor their behaviors to the child's optimal level of stimulation. 

To address these questions, Tanaya combined behavioral observation, questionnaires, and EEG data to understand the interplay between reactivity, parental social responsiveness, and social brain development. Notably, Tanaya’s project also delved into investigating the role of individual differences in sensory processing, such as sensation avoidance and sensory thresholds, in shaping the emerging functional organization of the brain through interactions. Tanaya’s project utilized an existing longitudinal dataset comprising typical infants (n=100, Warsaw study) and those at a higher risk of autism (BASIS, n=250). In this cohort, infant-parent interactions and resting brain EEG activity were longitudinally measured across the first two years of life at multiple points. The primary objective of this project was to identify the dyadic measures of early infant-parent interactions and sensory reactivity that most robustly predict later measures of social brain development (EEG, eye-tracking, behavior) using advanced machine learning algorithms. 

The secondary aim was to employ advanced statistical modeling to examine reciprocal associations between sensory reactivity, social interaction, and social brain development throughout the first and second years of life. The first study from Warsaw Cohort was a longitudinal study using quantitative coding of dyadic behavior (mutual gaze, parallel attention) from free play episodes in lab between parents and infant at six months,  non-invasive electroencephalography (EEG) data during resting state from 12 month, eye-tracking measures during Gap Overlap Task from 12 month, parent-report measures of IBQ and ECBQ temperamental measures and its association to various markers of cognitive development in typically developing infants. One hundred twenty-two infants were recruited for the study, followed by 51 infants who participated in both the 6-month visit and 12-month visit, and 38 infants who participated in the 6-month, 12-month and 24-month visit. After completion of the study, linear mixed modeling (using SPSS), EEG power analysis (using Matlab/EEGlab), and longitudinal analysis (case inclusive) using data from different time points to analyze the longitudinal data. The first major finding is that measures of parallel attention and mutual gaze showed association in the central and frontal areas of the brain in beta and gamma ranges. This indicates that top-down regulation and bottom-up processing are affected by attentional training episodes experienced while sharing visual attentional cues during dyadic interaction. It also indicated that neural correlates of visual attention could predict attentional measures of engagement and disengagement while performing a Gap overlap task at 24 months. Moreover, results suggest that temperamental profiles assessed in the 24th month associate with attentional correlates in the 12th month. Thus, the study elucidates that early dyadic experiences have a waterfall effect on later stages of brain development and skills that an infant acquires to process and react to sensory stimuli in later stages of life. 

The second longitudinal study from London Cohort aimed to assess how the risk status of siblings born to elder siblings with Autism and without Autism is associated. Specifically, between their quality of dyadic behaviour at six months, neural correlates of social attention at the 12th month, and behavioural outcomes at 36 months during clinical assessment. Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by social behavior disruption and uncontrolled motor behavior in the early stages of development. In the current longitudinal study, we investigated whether younger infant siblings born to elder siblings with a history of typical likelihood (TL-ASD) (n=41) and elevated likelihood (EL-ASD) (n=91) to ASD have a differential association between its dyadic experiences at six months, oscillatory features in theta and alpha band at 12 months (while resting state EEG is acquired during attentive viewing to social and nonsocial dynamic videos for frontal and occipital channels) and prognosis in behavioral outcomes measured by Autism Diagnostic Observation Schedule (ADOS) Total and component scores of Social Affect (SA) and Restrictive and Repetitive Behavior (RRB) at 36 months. First, we checked for the association between qualitative assessment of dyadic mutuality from Manchester assessment of caregiver–infant interaction (MACI) and oscillatory features (power spectral density differences during social versus nonsocial condition of theta and alpha in frontal and occipital channels) in both TL-ASD and EL-ASD. Secondly, we checked for the association between dyadic mutuality and behavioral outcomes measured by ADOS total and component scores (wherever applicable, details under research question) in both TL-ASD and EL-ASD. 

Thirdly, we checked for association between power spectral density differences during social versus nonsocial condition of theta, alpha (frontal and occipital channels), and behavioral outcomes measured by ADOS total and component scores (wherever applicable, details under research question) in both TL-ASD and EL-ASD. Fourthly, we checked whether there is a mediatory effect played by the power spectral density differences during social versus nonsocial condition of theta and alpha in frontal and occipital channels between dyadic mutuality and behavioral outcomes measures by ADOS total and component scores in both TL-ASD and EL-ASD (whichever group shows association). The alternate fifth hypothesis, in case the fourth hypothesis gets rejected, was to check if there is any additive effect played by the separate associations between dyadic mutuality and ADOS (total and component scores) and power spectral density (differences between social versus nonsocial condition in theta and alpha band in frontal and occipital channels) and ADOS (total and component scores). The results indicate that only EL-ASD showed a predictable linear association between dyadic mutuality, the theta power differences (frontal and occipital), and ADOS Total and Social Affect (both in theta occipital but only SA during theta frontal power differences between social versus nonsocial). 

These findings replicate previous studies demonstrating a higher genetic predisposition to ASD because having an elder sibling diagnosed with Autism predisposes an infant sibling to ASD diagnostic outcomes at later stages of life. It also replicates previous studies in EL-ASD that early dyadic behavior at six months can predict ASD diagnostic outcomes at 36 months. Moreover, the current study shows that oscillatory features in theta band power differences in occipital predicted ASD total and social affect score outcomes at 36 months. Additionally, theta band power differences in frontal partially predicted ASD scores by associating with only SA score outcomes. The final results associated with checking with additive effect showed improved predictability of ADOS total and SA scores by using dyadic mutuality as an additional predictor along with theta power differences between social versus nonsocial in frontal and occipital electrodes. Thus, the results indicate that early quality of dyadic interaction in elevated likelihood siblings of autism within half of the first year and neural oscillatory theta power (frontal, occipital) difference between social versus nonsocial conditions within 1st year of life can be considered as early indicators of ASD total and social communication outcomes at 36 months.

Pedro explored how infants detect social cues like smiles, humming, speech, singing as initial building blocks for establishing social relationships with their caregivers and the broader social circle. How do infants come to choose these social cues from multiple options available within their environment? Pedro’s project uses a technology that already exists in marketing - designed to exploit our individual preferences to sell us more products - and turn it on its head to help identify and select the social cues that infants find the most rewarding. 

To achieve this, Pedro has built on recently developed real-time fMRI method, Neuroadaptive Bayesian Optimization (NBO; Lorenz et al., 2016), which uses AI to optimize a cognitive task based on patterns of brain activity. By combining machine learning with real-time data analysis NBO maps out the best task or task parameters that activate a specific brain circuit from a large task parameter space in a highly efficient way that would not be possible using standard neuroimaging approaches. Specifically, Pedro has used this approach with electroencephalography (EEG), techniques that can be used to measure functional brain activity in awake infants. By employing new advances in artificial intelligence, real-time feedback and neuroimaging during live social interaction, this project has aimed to identify the specific cues that produce socially motivated behaviour in development that will provide fundamental insights into the mechanisms that drive the creation of human society. This line of research is important in the context of neurodevelopmental conditions because by focusing on understanding variations on what infants enjoy the most or what they like to attend to, it will enrich our knowledge on individual differences in early development and to identify individuals’ strengths of children who are on typical and neurodivergent developmental trajectories. 

Pedro’s revolutionary approach has combined sampling algorithms, data modelling, and real-time analysis of neural responses to stimuli, aiming to enhance our understanding of social brain development. Pedro has presented his innovative research “Bayesian Optimization for real-time, automatic design of face stimuli in human-centred research”. ICML2020, AutoMLat The International Conference on Machine Learning (ICML) in 2020, an international and fast-growing research conference. Pedro successfully defended his PhD thesis in 2022, and has now secured a highly competitive position within an AI company based on the interdisciplinary skills acquired during his time within SAPIENS.

Hanna's project aimed to test the associations between familial factors and infant development in families with a history of autism and/or ADHD. Structural equation modelling (SEM) was used to investigate how proband (older sibling) traits of autism and ADHD associate with infant sibling’s development of temperament, adaptive behaviour and later traits of autism and ADHD in early childhood. Determining how familial factors uniquely contribute to the complex developmental pathways of these conditions will clarify the underlying mechanisms that produce developmental heterogeneity in autism and ADHD and offer opportunities for better individualistic support for families with history of these conditions. 

To do this, Hanna worked with the British Autism Study of Infant Siblings (BASIS) and from the Studying Autism and ADHD Risks (STAARS) cohorts that included infants with family history of autism (n = 250), ADHD (n = 31), both (n = 21) or neither (n = 106). Infant siblings were seen at 5, 10, 24 and 36 months who had been comprehensively ‘deep phenotyped’ at each age point (measures of cognitive and brain function, developmental and clinical measures, parent child interaction and demographic, and older sibling and parental traits). Hanna has contributed to ground-breaking work on the familial associations between familial factors and autism and ADHD phenotypes and clinical outcomes in infant siblings at 3 years of age. Hanna's project has demonstrated the following novel and clinically relevant outcomes:

Proband traits of ASD and ADHD associate with separable aspects of infant sibling temperament development, with infant fear being specifically associated with proband ASD traits and infant activity level and attention are specifically associated with proband ADHD traits. Understanding the contribution of familial liabilities to the development of temperamental traits that are associated with later neurodevelopment conditions may help us better understand the shared versus distinct heritable developmental substructures of ASD and ADHD and inform approaches to early intervention and support to families (Halkola et al., in preparation).

Profiles of outcome in infants with a family history of autism and ADHD are not restricted to autism outcome only but a broader range of behavioural and developmental outcomes affecting autism, ADHD and anxiety and general developmental ability and adaptive behaviour. This will inform clinical services to be aware of the vulnerability of these infants to later clinical outcomes that could be targeted and ameliorated by prodromal interventions. Charman et al. (2023, JCPP Advances).

Whilst recurrence rates for autism in infants with an autism family history have long been accepted may be as high as 20%, we have shown for the first time that when followed through to mid-childhood recurrence rates of autism might be as high as 40%. Some of these children had later detected autism – particularly girls – and this has clinical implications in terms of the need for ongoing monitoring of these children within clinical child development and CAMHS services. Bazelmans et al. (in preparation).

In addition, Hanna did a 4-week offline and 4-week online secondment at Uppsala University (Terje Falck-Ytter) for a collaborative project using data from Babytwins Study Sweden (BATSS). The study looked at genetic and environmental effects on parent-rated adaptive behaviour in infancy, finding that shared environment influence explained most variability for both motor and social-communication skills with 45% shared variance, and while both had low but significant genetic influence, they did not share genetic variance. Halkola et al. (in preparation).

Twin studies examine phenotypic similarity in Monozygotic (identical) and Dizygotic (fraternal) twins to distinguish the extent to which genetic factors contribute to a specific trait compared to environmental factors. Most existing twin studies have studied older children and adults, meaning there is very little information about the heritability of different traits in early infancy. This project aimed to disentangle child contributions and parent contributions to parent-child interaction (PCI). Although it is a challenge, genetically informed studies could help us unfold the role of heritability.  

Irzam's project has used a unique dataset to answer fundamental questions about the etiology of early PCI, and about the nature of the putative relations between early interaction measures and later development. We have collected data on parent(mother)-child interaction (PCI) from 620 5-month-old twins in an ongoing infant twin study recruited from the general population [Baby Twins Study Sweden, BATSS; 50% monozygotic (MZ), 50% same sex dizygotic (DZ)] were collected. The infants were followed until 3 years of age and assessments on a wide range of skills such social, communication and language were collected at multiple time points.  The PCI videos (5 min free interaction between infant and parent) were coded by using a scheme that produces two broad categories of measures: 1) child characteristics (e.g., attentiveness), 2) parent characteristics (e.g., sensitive responsiveness). Further, a classical twin design (comparing similarity in MZ and DZ twins) to address the following: Aims 1) Establish the extent to which genetic factors influence the three types of PCI measures at 5 months of age, including “evocative genetic effects” whereby genes affect characteristics of the parent. 2) evaluate how variability in child, parent and interaction characteristics during PCI in infancy relate to long term outcomes at a phenotypic level (social skills, communication, language). 3) Examine the degree to which genetic factors explain the putative correlations between early PCI and later development. 

Irzam’s secondment was with professor Petra Warreyn (Ghent University, Belgium), who together developed the PCI analysis and the manuscript. The secondment was very helpful in this respect. This project is now conducted as planned and the report is about to be submitted to a scientific journal (estimated October 2023). In addition, Irzam Hardiansyah has used the twin study mentioned above to study the contributions of genes and environments to individual differences in visual perception, more specifically brain (EEG) activity during global motion perception. He has also analysed data from another study to check if atypicalities in visual perception and motion processing is associated with later autism. These results, both from the twin sample and the autism sample, are published in journal Communications Biology (Hardiansyah et al., 2023), a Springer Nature journal. Irzam has also co-authored two other papers about visual perception and twin study methodology, published in scientific journals (Journal of autism and other developmental disorders; Twin research and human genetics). In addition to these main projects and the scientific publications, Irzam has participated in several national and international conferences, including the International meeting for the Society for Autism Research (INSAR, 2023), where he presented his latest results on early markers of autism.