One powerful infant behaviour that can be leveraged in developmental research is infants’ looking patterns. Studies of infants’ ability to group visual stimuli have revealed that young infants are sensitive to things like outlines, contrasts or shapes that can be used to form distinct categories such as cat and dog (Quinn, Eimas & Tarr, 2001). What these studies can’t tell us is what underlying neural processes are driving the emergence of such behaviours. For this, neuroimaging can complement the progress made in developmental psychology. It enables us to answer questions about when and where in the brain a capacity to distinguish a category is present and to chart this emergence in different brain regions. In adults, there are dedicated areas on the bottom – or ventral – surface that demonstrate reliable patterns of responses to the different objects we see in the world (Grill-Spector & Weiner, 2014). This series of brain regions is organised as a hierarchy, processing increasingly complex visual features, from low-level perceptual features such as orientation (Hubel & Wiesel, 1962) to more abstract groupings by semantic concept (Huth et al., 2012) or animacy (Konkle & Caramazza, 2013). Although infants’ looking preferences have been shown to align with these organising principles (Spriet et al., 2022), it remained unclear whether activity in these brain areas is present from an early age.

Beyond knowing when something becomes functional in the infant brain, we also would like to know exactly how this comes to be. What learning mechanisms are at play? Is it the kind of structured information that can be learned from the statistics of the environment, or is a strong innate prior required? For this, a computational model of learning would be useful. Indeed, a further benefit of studying the development of the visual system, in addition to the wealth of research on infant looking behaviour to which we can compare, is the existence of advanced computational models of vision. For example, we can use deep neural network models that learn how to recognise objects through a training process during which they see many images. After training, the model’s hierarchical layers map onto the ventral visual hierarchy, processing increasingly complex visual features similarly to the brain (Yamins et al., 2014). We wondered whether such models would also capture visual responses in the infant brain. If so, we could test if the infant brain might map onto models at an earlier stage of their training, or perhaps earlier layers of the model that process simpler features. We believed that if such models could effectively capture infant brain responses, this would open new avenues for building advanced computational models of infant brain development. Models whose learning can be tightly controlled may provide insights into the unique processing of the infant brain that we cannot determine from comparison to adults alone.

To answer these questions we needed rich, visual responses from the infant brain that are suitable for comparison with computational models. We therefore required data from a large cohort of awake infants as they viewed a variety of visual categories. Collecting high-quality neuroimaging data in infants comes with many challenges. MRI requires participants to refrain from moving, but we cannot ask babies to lie still. Participants may also become fussy if a particular task is not interesting, or they may fall asleep mid experiment. Despite these challenges, pioneering work in awake infant fMRI laid the foundations for what was possible in this space (Deen et al., 2017, Ellis et al., 2020). These studies provided evidence that category selectivity to faces, scenes and body parts is present from early infancy (Kosakowski et al., 2022), as is the basic structure of early visual regions that reflect where in our visual field we perceive an object (Ellis et al., 2021). Given this, we sought to characterise infants’ responses to a broader set of objects, including animate and inanimate categories such as cats, rubber ducks, and trees. This diverse set of objects allowed us to measure the fine-grained patterns of response required for comparison with deep neural network models.
For our study, 134 families generously took the time to participate when their newest family member was just 2 months of age (see Figure 1). We successfully collected awake fMRI data from 130 of these infants as they watched our visual stimuli in the scanner. Encouragingly, many infants engaged with the images depicting various objects. After careful data preprocessing and quality control for motion we had a final dataset of 101 2 month olds who completed this experiment. The same infants were also scanned at 9 months, which gave us a final sample of 44 at this age group. This dataset allowed us to robustly characterise developing infant brain responses in ways that were not previously possible. Using patterns of activity across a given brain region, we found that category structure is present by 2-months of age. Although early visual regions were the most mature, infants’ high-level visual regions were organised very similarly to adults. There was also distinct processing for categories that was reliable across different examples of the same object. Fascinatingly, the infants’ brain responses already aligned with the deepest layers of a neural network, indicating complex processing beyond simple visual features. This suggests that the structure that’s present at this age is as complex as that learned by a model that has seen millions of images and can accurately label many objects.

Figure 1. 2-month-old participant with their mother participating in the study.

In addition to revealing that category structure is present in the youngest minds, much earlier than previously appreciated, this work proved that we can compare advanced computational models to developing brains. This is very exciting, as it unlocks a new toolkit that developmental scientists can use to reveal the mechanisms of information processing and learning during infancy, laying the groundwork for the emerging field of developmental cognitive computational neuroscience. Deep neural network models provide a computational language for asking questions such as: what rules does the infant brain use to learn new information? What information can be learned from experience alone? Which properties of the developing brain rely on innate structure or genetically predetermined wiring? Deep neural networks allow us to explore these questions by varying the model architecture, the data on which we train, and the rules that the models use to learn. By combining behaviour, neuroimaging, and computational modelling to provide a holistic account of development, this brings us one step closer to answering the question: what is going on inside an infant’s mind?

Deen, B., Richardson, H., Dilks, D. D., Takahashi, A., Keil, B., Wald, L. L., … & Saxe, R. (2017). Organization of high-level visual cortex in human infants. Nature communications, 8(1), 13995.
Ellis, C. T., Skalaban, L. J., Yates, T. S., Bejjanki, V. R., Córdova, N. I., & Turk-Browne, N. B. (2020). Re-imagining fMRI for awake behaving infants. Nature Communications, 11(1), 4523.
Ellis, C. T., Yates, T. S., Skalaban, L. J., Bejjanki, V. R., Arcaro, M. J., & Turk-Browne, N. B. (2021). Retinotopic organization of visual cortex in human infants. Neuron, 109(16), 2616-2626.
Grill-Spector, K., & Weiner, K. S. (2014). The functional architecture of the ventral temporal cortex and its role in categorization. Nature Reviews Neuroscience, 15(8), 536-548.
Hubel, D. H., & Wiesel, T. N. (1962). Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex. The Journal of Physiology, 160(1), 106–154.
Huth, A. G., Nishimoto, S., Vu, A. T., & Gallant, J. L. (2012). A continuous semantic space describes the representation of thousands of object and action categories across the human brain. Neuron, 76(6), 1210-1224.
Konkle, T., & Caramazza, A. (2013). Tripartite organization of the ventral stream by animacy and object size. Journal of Neuroscience, 33(25), 10235-10242.
Kosakowski, H. L., Cohen, M. A., Takahashi, A., Keil, B., Kanwisher, N., & Saxe, R. (2022). Selective responses to faces, scenes, and bodies in the ventral visual pathway of infants. Current Biology, 32(2), 265-274.
O’Doherty, C., Dineen, Á. T., Truzzi, A., King, G., Zaadnoordijk, L., Harrison, K., … & Cusack, R. (2026). Infants have rich visual categories in ventrotemporal cortex at 2 months of age. Nature Neuroscience, 1-10.
Quinn, P. C., Eimas, P. D., & Tarr, M. J. (2001). Perceptual categorization of cat and dog silhouettes by 3-to 4-month-old infants. Journal of experimental child psychology, 79(1), 78-94.
Spriet, C., Abassi, E., Hochmann, J. R., & Papeo, L. (2022). Visual object categorization in infancy. Proceedings of the National Academy of Sciences, 119(8), e2105866119.
Yamins, D. L., Hong, H., Cadieu, C. F., Solomon, E. A., Seibert, D., & DiCarlo, J. J. (2014). Performance-optimized hierarchical models predict neural responses in higher visual cortex. Proceedings of the national academy of sciences, 111(23), 8619-8624.

About the Author

In the 1990s, a wave of research changed how parents thought about early childhood. Developmental scientists demonstrated that early experience is critical for brain development and that enriched environments support children’s cognitive growth (Ramey & Ramey, 1998). The message was clear: Early experiences matter. What followed, however, could be considered a cultural overcorrection. Baby Einstein videos. Mozart for infants. Activity gyms, sensory toys, and structured play schedules designed to give babies the richest possible developmental diet. The underlying assumption was attractive: If some experience is good, more must be better. It didn’t work out that way. Studies showed that Baby Einstein babies learned no more language than babies who didn’t watch—and some learned less, likely because videos can’t provide the contingent, back-and-forth responding that a live caregiver does (Zimmerman et al., 2007). Ordinary home environments and responsive caregiver interactions were sufficient. Beyond a modest threshold, more stimulation didn’t produce proportional gains. Development is remarkably resilient.

New research from Tajikistan makes a strikingly similar point about motor development and early learning. Across Central Asia, it is common practice to swaddle infants tightly in a traditional cradle called a gahvora (Figure 1) for most of the day and night during the first two years of life (Karasik et al., 2018, 2023). At three months of age, many infants spend most of the day and night in the gahvora, their arms and legs immobilized with broad cloth bindings with brief windows outside the cradle for cuddling, feeding, and social interaction. From the perspective of Western developmental psychologists, this practice sounds alarming. (We know this because many expressed their alarm.) Modern theories of motor development emphasize that infants learn through movement—that they need to kick, reach, bat, and wiggle to discover how their bodies interact with the world (Adolph & Hoch, 2019; Campos et al., 2000). Without that physical freedom, shouldn’t infants’ development suffer?

Figure 1. Infant in a traditional gahvora cradle.

That question is exactly what we set out to answer. Karasik, Robinson, and Dodojonova (2026) brought a classic paradigm into Tajik homes: the mobile conjugate reinforcement task. A ribbon attaches one of the baby’s ankles to a colorful mobile overhead. Kick that leg, and the mobile moves. Within minutes, most babies figure it out and start kicking more deliberately—this reflects contingency learning, the basic cognitive capacity to connect actions with consequences. Typically, 80–90% of three-month-olds in North American samples learn the contingency in a single session (Rovee-Collier, 1997). What about gahvora-reared infants, who have spent almost their entire lives with their limbs bound? They learned it too! Despite their history of dramatically restricted movement, in a single session, Tajik infants increased their kicking across acquisition (with the ribbon attached), selectively kicking the tethered leg more than the untethered one. They also kept kicking at high rates even after we disconnected the ribbon—a hallmark of contingency learning. Finally, when the ribbon was detached, they protested—via negative vocalizations—just like babies growing up with full freedom of movement do (Fagen & Ohr, 1985). In sum, Tajik babies formed real expectations about what their kicks would produce, despite having a fraction of the unrestricted movement time of a typical Western baby.

The Timing Wrinkle and What It Reveals

One nuance in our findings illuminates just how the gahvora shapes learning. Infants naturally varied in how long they had been out of the cradle just before the task, from a few minutes to over an hour. Infants released right before the task barely moved during baseline—some hardly at all—while babies who had been free to move for a while were already kicking and vocalizing. But here’s the key finding: By the end of the session, both groups had learned equally well. Infants who started slowly needed time to get into the rhythm of free movement before they could exploit the contingency, but once they did—within minutes—their kicking spiked and stayed elevated through extinction. This very slight delay is not a failure of learning; it’s recalibration to an environment in which movement is newly unrestrained. As in Western infants faced with other motoric challenges (shoes, backpacks; Adolph & Avolio, 2000; Cole et al., 2014), the period of adjustment to new motor demands in Tajik infants happens fast, but it is not immediate.

What This Means for Parents (and for Developmental Science)

These findings shed light on a puzzle: Why do gahvora-reared infants show no long-term motor lags despite such extensive early restriction (Karasik et al., 2023, 2024), when, for example, orphanage-reared infants—who also experience restricted movement—show persistent delays (Gunnar, 2001)? The difference may not be the movement restriction itself, but the caregiving context that surrounds it. Whereas orphanage-reared infants often lack consistent, responsive caregivers altogether, gahvora-reared infants have dedicated caregivers who consistently attend to their signals and interact with them, both inside and during their windows outside the cradle.

This carries a message that runs counter to the assumptions underlying Baby Einstein and other super-enrichment schemes. Contingency learning—one of the foundational building blocks of cognitive and social development—doesn’t require months of unrestricted movement to kick in. It activates quickly whenever infants get the chance to act. What matters is not the sheer quantity of free movement, but whether infants have access to responsive caregiving and face real-world contingencies to act on the world when they are free to move. Quality over quantity.

A Word to Anxious Parents

Our findings have practical implications that deserve to be said plainly. If you’re not filling every waking moment with enrichment activities, your child is still going to learn to reach, to kick, to roll, to walk—and to figure out how the world responds to their actions. Babies don’t need a full-time event planner!

But the deeper point is not just that limited movement is harmless—it’s that infants are resourceful learners who adapt to what’s available. When one channel for exploration narrows, others stay open—watching, listening, tracking the social world. When free movement resumes, recalibration is fast. These aren’t moments lost; they’re moments spent differently.

The Baby Einstein hypothesis assumes that development is a bank account: the more you deposit, the richer the child. But it doesn’t work that way. Development doesn’t require optimization at every moment. What it requires is sufficient opportunity: some time to move, some time to interact with a responsive caregiver, and self-generated actions that produce meaningful consequences. Beyond that threshold, more is not necessarily better. Enough is enough.

Transporting a classic learning paradigm into homes halfway around the world, and discovering that infants defy our assumptions about what they need to thrive, requires sustained investment in basic developmental science. This work was funded by the National Science Foundation, and it is exactly the kind of question-driven, cross-cultural research that federal funding makes possible.

References

Adolph, K., & Avolio, A. (2000). Walking infants adapt locomotion to changing body dimensions. Journal of Experimental Psychology: Human Perception and Performance, 26(3), 1148–1166. https://doi.org/10.1037/0096-1523.26.3.1148

Adolph, K. E., & Hoch, J. E. (2019). Motor development: Embodied, embedded, enculturated, and enabling. Annual Review of Psychology, 70, 141–164.

Campos, J. J., Anderson, D. I., Barbu-Roth, M. A., Hubbard, E. M., Hertenstein, M. J., & Witherington, D. C. (2000). Travel broadens the mind. Infancy, 1, 149–219.

Cole, W. G., Gill, S. G., Vereijken, B., Adolph, K. E. (2014). Coping with asymmetry: How infants and adults walk with one elongated leg. Infant Behavior and Development, 37, 305-314.

Fagen, J. W., & Ohr, P. S. (1985). Temperament and crying in response to the violation of a learned expectancy in early infancy. Infant Behavior and Development, 8, 157–166.

Gunnar, M. R. (2001). Effects of early deprivation: Findings from orphanage-reared infants and children. In C. Nelson & M. Luciana (Eds.), Handbook of developmental cognitive neuroscience (pp. 617–629). MIT Press.

Karasik, L. B., Tamis-LeMonda, C. S., Ossmy, O., & Adolph, K. E. (2018). The ties that bind: Cradling in Tajikistan. PLoS ONE, 13, 1–18.

Karasik, L. B., Adolph, K. E., Fernandes, S. N., Robinson, S. R., & Tamis-LeMonda, C. S. (2023). Gahvora cradling in Tajikistan: Cultural practices and associations with motor development. Child Development, 94, 1049–1067.

Karasik, L. B., Schneider, J. L., Kuchirko, Y. A., & Dodojonova, R. M. (2024). Object play in Tajikistan: Infants explore objects despite bounds on play. Infancy, 30, e12627.

Ramey, C. T., & Ramey, S. L. (1998). Early intervention and early experience. American Psychologist, 53, 109–120.

Rovee-Collier, C. K. (1997). Dissociations in infant memory: Rethinking the development of implicit and explicit memory. Psychological Review, 104, 467–498.

Zimmerman, F. J., Christakis, D. A., & Meltzoff, A. N. (2007). Associations between media viewing and language development in children under age 2 years. Journal of Pediatrics, 151, 364–368.

About the Author

Previous research has shown that infant and adult preferences may in fact align. Infants are sensitive to low-level statistics— visual attributes of scenes or images consisting of properties such as edges, colours, and textures. Research has found that infants are sensitive to and prefer to look at certain image statistics over others (Newman et al., 2025). For example, in one study, infants tended to look longer at Van Gogh paintings that had a high variation in saturation and luminance (McAdams et al., 2023). Adults also tended to rate such paintings as being the most ‘pleasant’, displaying a shared preference for the same artworks that babies liked to look at. The authors tentatively suggested that these low-level image statistics that contributed to both infant visual preferences and adult pleasantness ratings are a form of ‘perceptual primitives’— the origins of aesthetic preference. Similar positive relationships between infant looking and adult aesthetics have been found for faces (Damon et al., 2019), colour (Skelton & Franklin, 2020), and building facades (McAdams et al., 2025).

In our research, we further investigated the relationship between infant looking and adult liking using baby book images. There is much evidence to suggest that early engagement with books is good for language acquisition (Franks et al., 2022), as well as social and emotional development (Briggs-Gowan, 2004). However, what kinds of books are most visually engaging for young babies, and are those books typically the ones that adults also like? We showed 54 infants (aged 2-12 months) 100 pages from baby board books across two experiments. A sample of 18 adults (aged 19-40 years) was also included in the second experiment. Infants were eye-tracked and their looking times, number of fixations, and saccade amplitudes were measured. Adults were also asked to rate the images on a sliding scale between 0 and 100, ranging from “I do not like at all” (0) to “I like a lot” (100). Scanned images of the book pages (see figure 1) were presented centrally on the screen, with the rating scale appearing below the stimuli in the adult experiment. The low-level image statistics of the book pages were also computed and were chosen based on their usage in previous research on both adult aesthetics (Berman et al., 2014) and infant looking behaviours (McAdams et al., 2023). These included saturation, luminance, straight and curved edge density, the proportion of different hues, and more (see figure 1 for examples).

Fig. 1. Examples of the image statistics used in the study (a) A book page used as a stimulus in Experiment 2 (Look Touch Learn Sky © 2022 Child’s Play (International) Ltd.), (b) The luminance matrix of the book page, (c) the edges of the book page highlighted by image analysis, (d) the pixel chromaticity plotted in the MacLeod & Boynton (1979) colour space, (e) the 8 hue segments (and the mean hue angles: 45°, red; 90°, cherry; 135°, yellow; 180°, chartreuse; 225°, green; 270°, teal; 315°, blue; 360°, violet) used to compute the mean saturation and proportion of pixels of different hues, with the dimension of saturation also shown. Figure from Newman et al. (2026).

Infant visual engagement in experiments one and two and adult liking ratings shared common predictors: variation in saturation, saturation of cherry and red hues, mean luminance, mean saturation, entropy, curved edge density, and fractal dimension. However, interestingly all predictors were in opposite directions for infants and adults. For example, infants visually engaged more with book pages that had a higher saturation, more colour contrast, more bluish hues, and a lower edge density, whereas adults liked these pages less.

In fact, infants were the most visually engaged with book pages that adults liked the least (see figure 2). Likewise, while pages with desaturated colours or black and white designs, low contrast, visual complexity, and detail were the most well-liked by adults, infants tended to visually engage with them the least.

Fig. 2. Scatterplot showing the relationship between the infant visual engagement principal component and adult liking ratings for each book page (each page is indicated by a black dot). The black line indicates the line of best fit. Figure from Newman et al. (2026).

While these results can act as a guide for baby book publishers, there are also implications for our understanding of what underlies infant visual preferences. Infant engagement and adult liking sharing common predictors echoes the notion of the ‘perceptual primitives’ of aesthetics. However, unlike previous research infants and adults responded in opposite ways for book images. These differences in infant visual preference and adult aesthetic preferences may be a result of infants’ immature visual system (Norcia & Tyler, 1985). While infants have functional colour vision by at least two-months-old (Teller, 1998), colours need to be highly-saturated to be visible to them. Infants’ poor visual acuity and colour sensitivities are not yet sufficient enough to properly engage in images with high detail and pastel colours, yet high contrast black and white designs may in fact be too ‘easy’ given that infants have colour vision, which may explain why infants visually engage the most with book pages that are highly colourful.

About the Author

Some researchers have begun to map structural brain development with growth charts using magnetic resonance imaging (MRI)³, but there has been much less work on comparable charts of early brain function. As we know, brain function also undergoes rapid change in early life, and this ongoing functional development supports emerging behavioral skills. But there has been very limited progress in identifying whether growth charts can be used to map brain function. Additionally, MRI is limited as a tool for scalable public health implementation due to its prohibitive cost and infrastructure. We sought to develop normative growth curves of infant brain function measured with electroencephalogram (EEG)⁴. EEG is a tool well-suited to this kind of public health developmental context, as it is a scalable, relatively low-cost modality that we can use to measure brain function in humans starting at birth.

We chose to characterize brain function in a key sensory domain: visual neurodevelopment. This domain is important for later visual functioning and may also scaffold later cognitive skills that matter across the lifespan. The visual-evoked potential (VEP) is a robust activity signal that has been measured across species and may reflect features of the neural architecture, such as myelination and underlying cortical circuit function. It can be collected with EEG while an infant watches a visual stimulus, like a flashing checkerboard. Even in infancy, the VEP has a canonical waveform, characterized by an initial negative peak (N1), a later positive peak (P2), and a final negative peak (N2), as shown in the bottom right panel of Figure 1.

Figure 1: Study Overview. 128-channel EEG data were collected in 802 infants (n=1374 EEG observations) between 57 and 579 days old across three continents. These data were a priori harmonized for paradigm, acquisition, and processing parameters across sites. Standard processing was used to extract the visual-evoked potential (VEP) from the EEG data and measure the amplitudes and latencies of the N1, P1, and N2 peaks. We created growth curves for each amplitude and latency.

We collected standardized EEG VEPs from 802 infants (ages 57 to 579 days old) from four longitudinal cohorts across three global sites (i.e., Cape Town, South Africa; São Paulo, Brazil; and Boston, USA), resulting in 1374 observations. These sites represent distinct cultural, linguistic (i.e., primarily speaking Xhosa, Brazilian Portuguese, and English, respectively), geographic, and socioeconomic contexts. For example, the socioeconomic diversity can be seen as more than 60% of the caregivers in the South Africa cohort reported an annual household income of less than $3,303 (60,000 ZAR); whereas 10.3% in the Brazil cohort fell under an HHI of $4,763 ($27,600 Brazilian Real) and only 5.4% in the US cohort reported making $50,000 or less. These differences allowed us to test whether developmental trajectories generalize across markedly different settings.

We developed growth curves for the different EEG signal components of this visual response. After extracting VEP waveforms from the EEG data, they can be characterized by the amplitudes (i.e., heights) and latencies (i.e., times from stimulus onset to peak) of the N1, P1, and N2 peaks. Growth curves for each of these parameters were created. Our modeling approach applied GAMLSS⁵ (generalized additive models of location, scale, and shape), a flexible centile-modeling method that can capture complex nonlinear trajectories and has been used in growth-chart work (see Figure 2). This approach creates flexible centile curves similar to those used in WHO growth standards. However, given the diversity of our three sites, we sought to test whether the developmental patterns we observed were similar across the distinct populations. When we modeled visual neurodevelopment features in two of the sites, we successfully applied the model to the held-out site, suggesting that these developmental trajectories were similar across populations.

Figure 2: Modeling Overview. We independently trained GAMLSS curves for each site and VEP feature (i.e., N1, P1, and N2 amplitudes and latencies). The GAMLSS curves were generalized across sites, indicating that a curve trained on the Boston data was valid for the Cape Town data. Individual centiles were extracted and linked to cognitive scores from a global public health measure (Global Scales of Early Development, Long Form).

Individual centile scores on these growth curves may also associate with broader cognitive development. We found that deviations from these trajectories were associated with Global Scales for Early Development scores, a WHO-developed measure of general development, in the South Africa cohort. Higher centile scores were associated with higher developmental scores at two years of age.

Our research suggests that key VEP developmental patterns are similar across these populations, despite substantial environmental, genetic, and cultural differences. Beyond this cross-site consistency, the VEP is a promising metric of neurodevelopment because of its accessibility. It can be collected in newborns using a simple, portable, and relatively low-cost EEG paradigm. This allows for greater diversity in both the data used to train the models and the settings where this technology could be deployed. Open-source preprocessing software, such as Harvard Automated Preprocessing Pipeline for EEG⁶ (HAPPE; https://github.com/PINE-Lab/HAPPE), can also facilitate standardized processing of EEG data. We have integrated our growth curves from the current dataset into HAPPE, making it easier to extract centile scores from new participants. We hope to continue updating these models as more data are collected and to improve our understanding of how the VEP relates to broader cognitive development. With further replication and prospective validation, VEP growth curves may provide a scalable framework for characterizing atypical developmental trajectories. In the future, they may help identify critical periods and support the assessment of intervention-related change.

Want to learn more about this study? Read our preprint here: https://doi.org/10.1101/2025.03.25.645314

1          Group, W. M. G. R. S. WHO Child Growth Standards based on length/height, weight and age. Acta Paediatr Suppl 450, 76-85 (2006). https://doi.org/10.1111/j.1651-2227.2006.tb02378.x
2          Kuczmarski, R. J. et al. 2000 CDC Growth Charts for the United States: methods and development. Vital Health Stat 11, 1-190 (2002).
3          Bethlehem, R. A. I. et al. Brain charts for the human lifespan. Nature 604, 525-533 (2022). https://doi.org/10.1038/s41586-022-04554-y
4          Margolis, E. T. et al. Growth charts of infant visual neurodevelopment generalize across global contexts. bioRxiv, 2025.2003.2025.645314 (2025). https://doi.org/10.1101/2025.03.25.645314

About the Author

When it comes to babies’ social worlds, one size does not fit all.

Even within the first year of life, infants differ in how their brains respond to social information, such as facial expressions, eye contact and speech. Rather than all babies attending to social cues in the same way, each infant appears to have their own unique pattern of neural engagement. Novel AI-based methods can now be used to personalise brain science and track these individual patterns in real time. This allows us to ask a simple but powerful question: what does each baby’s brain find most engaging in a social exchange?

Studying baby brains, one baby at a time

Traditionally, infant research has focused on group averages. While this approach has taught us a great deal, it can sometimes miss something important: individual differences. To address this, researchers at the Centre for Brain and Cognitive Development at Birkbeck, University of London used a cutting-edge method called Neuroadaptive Bayesian Optimisation (NBO), an artificial-intelligence technique that adapts in real time to each baby’s brain activity^1,2.

In a pre-registered study involving 61 infants aged 5 to 12 months³, babies looked at images of their caregiver’s face, a stranger’s face, or faces that were a mix of both. While babies viewed the faces, we recorded brain activity using electroencephalography (EEG), a safe and non-invasive technique (see Figure 1). We then used a machine learning algorithm that learned from each baby’s brain responses as the experiment unfolded. After each image was shown to the baby, the algorithm estimated which face had sparked the strongest response and used that information to choose what image to show next, allowing us to adjust the experiment moment by moment to each infant’s brain signals. This approach helped us identify which type of face triggered the strongest response for every individual baby.

Figure 1. Baby participants wearing EEG caps that measure their brain activity while they view images of different faces on a screen.

No “average” baby brain

What did we find? There was no single face type that worked best for all babies. Instead, around 85% of infants who completed the study showed strong individual preferences, with different babies responding most strongly to different faces.

An added bonus? This personalised approach also led to lower drop-out rates (about 15%, compared to the 22% typically seen in infant EEG studies). This demonstrates how tailoring experiments to babies by presenting images that elicit the strongest brain response can help keep them engaged.

The BONDS project: personalised approaches to social development

These findings are part of the wider BONDS project (Behaviour and Online Neuroimaging to study the Development of Socialisation), which involved over 120 infants and their families. BONDS combines wearable neuroimaging with artificial intelligence to test what social cues individual babies’ brains attend to.

A recent pre-registered study⁴ extended this work by showing that infants’ brains are sensitive to different combinations of social cues, such as eye gaze, head orientation, and emotional expression, and that the most engaging combinations vary from baby to baby.

Further, BONDS research has shown that personalised methods can be used not only with images on a screen, but also during live social interactions with an experimenter⁵. These studies demonstrate that:

• Infants differ in which social signals (e.g., gaze, infant-directed speech) elicit the strongest neural responses
• Brain engagement during social interaction reflects how motivated and attentive infants are in that moment
• Individual neural preferences can be detected reliably, even in dynamic, real-world social situations

Together, these findings suggest that early social brain development is shaped by individual experience and neural tuning, rather than following a single “typical” pattern.

Follow the baby’s lead

In summary, we showed, with a set of rigorous neuroimaging studies, that babies naturally differ in what captures their attention. Some may be especially drawn to expressive faces or direct eye contact, while others respond more strongly to subtler social cues. Because infants differ in what captures their attention, using personalised neuroimaging approaches to study these early differences in interests could open up new ways to support responsive parenting — that is, noticing what a baby naturally attends to and responding to it. Tuning into these individual preferences may help nurture babies’ attention and emerging social skills.

Looking ahead

By using personalised neuroimaging approaches, researchers are gaining a richer understanding of how babies engage with others from the very start of life. This work highlights the importance of recognising and respecting individual differences in early development — both in science and in everyday parenting.

The BONDS project is now expanding beyond infancy to study neurotypical and autistic toddlers, focusing on how children interact with their parents during real social exchanges. By studying parent–child interaction at this stage, the project aims to capture how diverse developmental pathways unfold, and how personalised approaches can help us understand and support children with different strengths and needs.

To learn more about this research and future findings, visit the BONDS project website: https://sites.google.com/view/bonds-project/

References:

1. Lorenz, R., Hampshire, A., & Leech, R. (2017). Neuroadaptive Bayesian optimization and hypothesis testing. Trends in cognitive sciences, 21(3), 155-167.
2. Gui, A., Throm, E. V., da Costa, P. F., Haartsen, R., Leech, R., & Jones, E. J. (2022). Proving and improving the reliability of infant research with neuroadaptive Bayesian optimization. Infant and Child Development, 31(5), e2323.
3. Throm, E., Gui, A., Haartsen, R., da Costa, P. F., Leech, R., Mason, L., & Jones, E. J. (2025). Combining Real‐Time Neuroimaging With Machine Learning to Study Attention to Familiar Faces During Infancy: A Proof of Principle Study. Developmental Science, 28(1), e13592.
4. Gui, A., Throm, E., da Costa, P. F., Penza, F., Mayans, M. A., Jordan-Barros, A., Haartsen, R., Leech, R., & Jones, E. J. H. (2024). Neuroadaptive Bayesian optimisation to study individual differences in infants’ engagement with social cues. Developmental Cognitive Neuroscience, 68, 101401.
5. Throm, E., Gui, A., Haartsen, R., da Costa, P. F., Leech, R., & Jones, E. J. (2023). Real-time monitoring of infant theta power during naturalistic social experiences. Developmental Cognitive Neuroscience, 63, 101300.

About the Author

On visits to my mother’s house, it is inevitable that we look through baby pictures, a tradition that has grown increasingly common since I began my research into memory development. I know the stories behind many of these pictures, mostly through my mother. This past winter, fixing her VHS player offered a new opportunity for reminiscing. The first unlabeled tape we played was of the day my twin sister and I arrived home from the hospital—the beginning of our life stories, and yet a day we could not remember.

Why are we so fascinated by our earliest memories? Memories for specific events in our lives (episodic memories) form the basis of our personal identities, and shape how we see ourselves in the world¹. Despite the immense amount of learning in the first few years of life, most of us report our first specific memories from around 3 to 4 years of age^2,*. This lack of specific memories from infancy and toddlerhood (termed infantile amnesia by Freud³) has been the subject of both scientific debate^4,5 and public interest⁶. More than just a philosophical question, understanding whether and how infants form memories for specific events has important implications for understanding why some early experiences have profound impacts on later development⁷. Excitingly, recent advances in neuroimaging methods in infants and young children⁸ are allowing us to probe the emergence and contents of early life memories.

Pictures can transport our family members back in time to specific moments from our infancy, even though our memories for those experiences are lost. Neuroimaging, and awake infant fMRI in particular, may help reveal the mechanisms by which we learn, remember, and forget our earliest experiences.

Infantile amnesia and brain development

Many theories of infantile amnesia^9–11 highlight developing brain systems, particularly the hippocampus¹², to explain why specific early life memories do not persist into adulthood. The hippocampus is critical for the encoding and retrieval of episodic memories in adults^13,14 and undergoes dramatic structural development in early life¹⁵. However, structural immaturity does not preclude function; many brain regions develop well into adolescence¹⁶ and are nonetheless associated with cognitive functions in childhood¹⁷. This leads to the question of whether specific memories from infancy are ‘lost’ during storage or retrieval, or were never formed in the hippocampus in the first place¹⁸. Elegant research with rodents^19–22 has begun to address this question, showing that memories for specific experiences are formed in the hippocampus in infancy and can be recalled via reactivating neurons with optogenetics in adulthood^21,22. Recent innovations in awake infant functional magnetic resonance imaging (fMRI)^23,24 has allowed translation of these findings to human infants.

Hippocampal contributions to infant memory

In our recent study²⁵, we scanned 26 infants aged 4 months to 2 years while they were awake and performing a memory task. We designed a task to examine ‘episodic-like’ memory^† for specific experiences tested at a delay outside of the window of working memory. In brief, infants viewed pictures of previously-unseen faces, objects, and scenes on top of a green kaleidoscopic background, and then saw the same picture an average of a minute later alongside a new picture from the same category in a visual paired comparison test. Using the logic of the ‘subsequent memory’ paradigm^26,27, we sorted encoding trials (i.e., the first time that infants saw a picture) based on infants’ looking time between the old and new picture at test. Hippocampal activity, particularly in the posterior subregion, was higher when infants first saw a picture that they would later show a familiarity preference for at test; moreover, infants with higher average familiarity preferences showed stronger memory-related hippocampal activity. Interestingly, there was age-related change, such that memory-related hippocampal activity was reliable only in infants older than one year of age. Because visual activation was reliable and not different between younger and older infants, this age-related change was not solely due to inadequate power to detect effects in younger infants. These results suggest that the human infant hippocampus is capable of encoding memories for specific experiences by at least the end of the first year of life, constraining theories of infantile amnesia based on deficits of memory encoding.

An early capacity for episodic-like memory encoding

The idea that infants are capable of forming specific memories early in life fits with a vast literature showing sophisticated episodic-like memory behaviors in infancy²⁸. Infant memory has been observed in tasks such as the visual paired comparison test²⁹, the mobile conjugate reinforcement paradigm³⁰, the deferred imitation paradigm³¹, and the relational binding task³², each of which exhibit one or more elements of episodic-like memory (e.g., one-shot learning, long-term retention, flexible associations, spatial/temporal context binding). One recent review³³ proposed that episodic-like memory may be even more ubiquitous in infancy than previously thought, given that paradigms that are not designed to study memory (e.g., surprise-induced learning³⁴ and social cognition³⁵) nonetheless require rapid and flexible encoding of events over extended delays.

If episodic-like memories can be formed in infancy, even in the hippocampus, why then, do they not persist into adulthood? We hypothesize that mechanisms of infantile amnesia may be more related to post-encoding processes, such as how infant memories are stored and/or later retrieved. Here, the rodent literature is quite suggestive. Neurogenesis in the hippocampus, which is prominent in early life^10,19,36, can induce forgetting by destabilizing memory engrams or altering their connections, suggesting that long term storage of infant memories may be disrupted. At the same time, infant memories can be recovered in adulthood through reminders²⁰ and optogenetic techniques^21,22, suggesting that some infant memories still exist in the brain, even if normally inaccessible for retrieval. Of course, memory encoding may still be impoverished in infancy. Indeed, we only tested memory for individual items and not relations (which engage the hippocampus more in adults³⁷), and hippocampal activity in our study tracked with familiarity preferences, which have been traditionally associated with partial or incomplete encoding³⁸. Assessing infant memories over time and across different paradigms will help reveal how different stage(s) of memory contribute to infantile amnesia.

Balancing between learning systems in early life

Hippocampal involvement in memory encoding in our study emerged around 9 to 12 months of age, which has been highlighted as an important time for memory development^39,40. Interestingly, this is notably later than hippocampal involvement in statistical learning (the ability to extract regularities from the environment), which we previously showed in infants aged 3 to 24 months⁴¹. Different developmental trajectories for statistical learning and episodic-like memory function in the infant hippocampus are consistent with anatomical development of the monosynaptic (entorhinal cortex to CA1 subfield) and trisynaptic (entorhinal cortex to dentate gyrus, CA3, and CA1) hippocampal pathways¹⁵, which have been computationally linked to statistical learning and episodic memory, respectively⁴². Subregion analyses also mapped onto known anatomical differences in the monosynaptic and trisynaptic pathways⁴³, with statistical learning effects strongest in the anterior (vs. posterior) hippocampus and memory encoding effects strongest in the posterior (vs. anterior) hippocampus.

Why would the hippocampus exhibit different developmental trajectories for statistical learning versus memory encoding? An ecological account proposes that developing organisms are adapted to meet their current environmental niche⁴⁴. When one has little prior knowledge (as is the case for an infant), it may be advantageous to extract general patterns in the environment before then encoding specific information^45,46. In other words, statistical learning may supply a ‘baseline’ from which novel information can then be compared⁴⁷. Just like perception starts broad and then narrows over development⁴⁸, cognitive processes like attention and learning may operate over broad (or ‘general’) information^†† before narrow (or ‘specific’) information⁴⁹. This may be adaptive for making predictions and guiding future behavior in a dynamic environment. At present, this is mostly theoretical, and longitudinal research is necessary for linking the emergence of statistical learning and episodic memory in the infant hippocampus. Nonetheless, these initial findings from awake infant fMRI are contributing to our understanding of how the infant brain may balance between learning systems in early life.

Looking forward

As much as my mother wishes I could remember our extravagant first birthday party, my specific memories from early life are sparse. But I still extracted meaning from the patterns in my early experiences⁵⁰, such as the fun we had and the love that was ever present. Awake infant fMRI is helping to reveal how we represent early experiences, with recent findings suggesting that the infant hippocampus is capable of forming episodic-like memories, even if such memories are later lost. This may constrain theories of infantile amnesia and inform our understanding of learning systems in early life. More broadly, this work highlights the promise of awake infant fMRI for characterizing mechanisms underlying cognitive development⁵¹ and contributing to our understanding of what it is like to be a baby.

Footnotes

*If you would like to contribute to our study of people’s earliest memories, please consider taking this anonymous, 5 minute survey: https://yalesurvey.ca1.qualtrics.com/jfe/form/SV_0IJ7NSrA1oYYNeK

^†Traditional theories define episodic memory as a type of declarative memory requiring explicit, conscious access — something that cannot be assessed in preverbal infants

^††General representations can arise from insufficient encoding and forgetting of details (‘fuzzy’ representations), or the abstraction of commonalities over multiple distinct episodes, which may exhibit distinct developmental trajectories⁵²

References

1. Nelson, K. & Fivush, R. The emergence of autobiographical memory: A social cultural developmental theory. Psychological Review 111, 486–511 (2004).
2. Akhtar, S., Justice, L. V., Morrison, C. M. & Conway, M. A. Fictional first memories. Psychological Science 29, 1612–1619 (2018).
3. Freud, S. Psychopathology of Everyday Life. (W. W. Norton & Company, 1966).
4. Nadel, L. & Zola-Morgan, S. Infantile amnesia: A neurobiological perspective. in Infant Memory (ed. Moscovitch, M.) 145–172 (Springer US, Boston, MA, 1984).
5. Alberini, C. M. & Travaglia, A. Infantile amnesia: A critical period of learning to learn and remember. Journal of Neuroscience 37, 5783–5795 (2017).
6. Daniel, A. Why don’t we remember being babies? Brain scans reveal new clues. NPR (2025).
7. Li, S., Callaghan, B. L. & Richardson, R. Infantile amnesia: Forgotten but not gone. Learning & Memory 21, 135–139 (2014).
8. Turk-Browne, N. B. & Aslin, R. N. Infant neuroscience: How to measure brain activity in the youngest minds. Trends in Neurosciences 47, 338–354 (2024).
9. Bauer, P. J. Constructing a past in infancy: A neuro-developmental account. Trends in Cognitive Sciences 10, 175–181 (2006).
10. Josselyn, S. A. & Frankland, P. W. Infantile amnesia: A neurogenic hypothesis. Learning & Memory 19, 423–433 (2012).
11. Gómez, R. L. & Edgin, J. O. The extended trajectory of hippocampal development: Implications for early memory development and disorder. Developmental Cognitive Neuroscience 18, 57–69 (2016).
12. Bevandić, J. et al. Episodic memory development: Bridging animal and human research. Neuron 112, 1060–1080 (2024).
13. Scoville, W. B. & Milner, B. Loss of recent memory after bilateral hippocampal lesions. Journal of Neurology, Neurosurgery & Psychiatry 20, 11–21 (1957).
14. Squire, L. R. Memory and the hippocampus: A synthesis from findings with rats, monkeys, and humans. Psychological Review 99, 195–231 (1992).
15. Lavenex, P. & Banta Lavenex, P. Building hippocampal circuits to learn and remember: Insights into the development of human memory. Behavioural Brain Research 254, 8–21 (2013).
16. Gogtay, N. et al. Dynamic mapping of human cortical development during childhood through early adulthood. Proceedings of the National Academy of Sciences 101, 8174–8179 (2004).
17. Casey, B., Tottenham, N., Liston, C. & Durston, S. Imaging the developing brain: What have we learned about cognitive development? Trends in Cognitive Sciences 9, 104–110 (2005).
18. Ellis, C. T. & Turk-Browne, N. B. Infant fMRI: a model system for cognitive neuroscience. Trends in Cognitive Sciences 22, 375–387 (2018).
19. Akers, K. G. et al. Hippocampal neurogenesis regulates forgetting during adulthood and infancy. Science 344, 598–602 (2014).
20. Travaglia, A., Bisaz, R., Sweet, E. S., Blitzer, R. D. & Alberini, C. M. Infantile amnesia reflects a developmental critical period for hippocampal learning. Nature Neuroscience 19, 1225–1233 (2016).
21. Guskjolen, A. et al. Recovery of “lost” infant memories in mice. Current Biology 28, 2283-2290.e3 (2018).
22. Power, S. D. et al. Immune activation state modulates infant engram expression across development. Science Advances 9, eadg9921 (2023).
23. Ellis, C. T. et al. Re-imagining fMRI for awake behaving infants. Nature Communications 11, 4523 (2020).
24. What infant fMRI is revealing about the developing mind. The Transmitter: Neuroscience News and Perspectives https://www.thetransmitter.org/cognitive-neuroscience/what-infant-fmri-is-revealing-about-the-developing-mind/ (2025).
25. Yates, T. S. et al. Hippocampal encoding of memories in human infants. Science 387, 1316–1320 (2025).
26. Wagner, A. D. et al. Building memories: Remembering and forgetting of verbal experiences as predicted by brain activity. Science 281, 1188–1191 (1998).
27. Brewer, J. B., Zhao, Z., Desmond, J. E., Glover, G. H. & Gabrieli, J. D. E. Making memories: Brain activity that predicts how well visual experience will be remembered. Science 281, 1185–1187 (1998).
28. Bauer, P. J., Souci, P. S. & Pathman, T. Infant memory. WIREs Cognitive Science 1, 267–277 (2010).
29. Pascalis, O., de Haan, M., Nelson, C. A. & de Schonen, S. Long-term recognition memory for faces assessed by visual paired comparison in 3-  and 6-month-old infants. Journal of Experimental Psychology: Learning, Memory, and Cognition 24, 249–260 (1998).
30. Rovee-Collier, C. K., Sullivan, M. W., Enright, M., Lucas, D. & Fagen, J. W. Reactivation of infant memory. Science 208, 1159–1161 (1980).
31. Meltzoff A.N. What infant memory tells us about infantile amnesia: Long-term recall and deferred imitation. Journal of Experimental Child Psychology 59, 497–515 (1995).
32. Richmond, J. & Nelson, C. A. Relational memory during infancy: Evidence from eye tracking. Developmental Science 12, 549–556 (2009).
33. Behm, L., Turk-Browne, N. B. & Kibbe, M. M. The ubiquity of episodic-like memory during infancy. Trends in Cognitive Sciences 0, (2025).
34. Stahl, A. E. & Feigenson, L. Observing the unexpected enhances infants’ learning and exploration. Science 348, 91–94 (2015).
35. Southgate, V., Chevallier, C. & Csibra, G. Seventeen-month-olds appeal to false beliefs to interpret others’ referential communication. Developmental Science 13, 907–912 (2010).
36. Ryan, T. J. & Frankland, P. W. Forgetting as a form of adaptive engram cell plasticity. Nature Reviews Neuroscience 23, 173–186 (2022).
37. Davachi, L. Item, context and relational episodic encoding in humans. Current Opinion in Neurobiology 16, 693–700 (2006).
38. Hunter, M. A. & Ames, E. W. A multifactor model of infant preferences for novel and familiar stimuli. in Advances in infancy research, Vol. 5. 69–95 (Ablex Publishing, Westport, CT, US, 1988).
39. Morgan, K. & Hayne, H. Age-related changes in visual recognition memory during infancy and early childhood. Developmental Psychobiology 53, 157–165 (2011).
40. Nelson, C. A. The ontogeny of human memory: A cognitive neuroscience perspective. Developmental Psychology 31, 723–738 (1995).
41. Ellis, C. T. et al. Evidence of hippocampal learning in human infants. Current Biology 31, 3358-3364.e4 (2021).
42. Schapiro, A. C., Turk-Browne, N. B., Botvinick, M. M. & Norman, K. A. Complementary learning systems within the hippocampus: A neural network modelling approach to reconciling episodic memory with statistical learning. Philosophical Transactions of the Royal Society B: Biological Sciences 372, 20160049 (2017).
43. Malykhin, N. V., Lebel, R. M., Coupland, N. J., Wilman, A. H. & Carter, R. In vivo quantification of hippocampal subfields using 4.7 T fast spin echo imaging. NeuroImage 49, 1224–1230 (2010).
44. Rovee-Collier, C. & Cuevas, K. Multiple memory systems are unnecessary to account for infant memory development: An ecological model. Developmental Psychology 45, 160–174 (2009).
45. Keresztes, A., Ngo, C. T., Lindenberger, U., Werkle-Bergner, M. & Newcombe, N. S. Hippocampal maturation drives memory from generalization to specificity. Trends in Cognitive Sciences 22, 676–686 (2018).
46. Ramsaran, A. I., Schlichting, M. L. & Frankland, P. W. The ontogeny of memory persistence and specificity. Developmental Cognitive Neuroscience 36, 100591 (2018).
47. Frost, R. et al. Statistical learning subserves a higher purpose: Novelty detection in an information foraging system. Psychological Review (2025).
48. Maurer, D. & Werker, J. F. Perceptual narrowing during infancy: A comparison of language and faces. Developmental Psychobiology 56, 154–178 (2014).
49. Forest, T. A., Schlichting, M. L., Duncan, K. D. & Finn, A. S. Changes in statistical learning across development. Nature Reviews Psychology 2, 205–219 (2023).
50. Tottenham, N. & Vannucci, A. Attachment as prediction: insights from cognitive and developmental neuroscience. Current Directions in Psychological Science 34, 195–206 (2025).
51. Yates, T. S., Ellis, C. T. & Turk-Browne, N. B. The promise of awake behaving infant fMRI as a deep measure of cognition. Current Opinion in Behavioral Sciences 40, 5–11 (2021).
52. Reyna, V. F. & Brainerd, C. J. Fuzzy-trace theory: An interim synthesis. Learning and Individual Differences 7, 1–75 (1995).

About the Author

How did these changes affect children and families?

Given school-closures and the transition to remote learning, much of the work on this question has understandably focused on school-aged children. This work has identified negative effects on family functioning^1,2, children’s mental health and emotional well-being^3,4, and broad learning delays^5,6. For instance, a meta-analysis by Betthäuser et al.⁵ found that during lockdowns children lost over a third of a school years’ worth of learning and this loss remained present even after returns to in-person schooling⁷.

But school-aged children were not the only ones impacted by the pandemic. Several studies suggest that preschoolers exhibited overall developmental delays and learning losses relative to pre-pandemic cohorts⁸, especially children from lower socioeconomic status (SES) backgrounds⁹. Changes in family dynamics and social interaction patterns also impacted younger infants and toddlers, who experienced increases in screen time^9,10 and slower vocabulary development¹⁰.

Yet little work has examined potential impacts of the pandemic on the development of social cognition skills in early childhood. This is a key oversight given substantial evidence that the development of such skills is tied to children’s social interactions^12,13,14, which shifted considerably during the pandemic.

In a recent study¹⁵, we sought to address this gap in the literature by focusing on preschooler’s false-belief understanding. We happened to be in the midst of a large-scale study on children’s social cognition when the lab was forced to close for pandemic lockdowns in March 2020. The study resumed in September 2021 when pandemic restrictions were lifted on our campus. We took advantage of this accidental cohort study to examine whether false-belief understanding differed in children tested before versus after pandemic lockdowns.

Children between 3.5 and 5.5 years of age completed a range of tasks that included two traditional elicited-response false-belief tasks^16,17, as well as a low-demand elicited-response task that children typically pass at younger ages^18,19. We found that children in the post-pandemic cohort performed worse on both types of tasks than children tested before the pandemic. In addition, the difference in performance between the two cohorts was larger for children from lower SES backgrounds.

Although our findings are cross-sectional, they suggest that the pandemic negatively impacted the development of children’s social cognition skills, especially children from lower SES backgrounds. Given that performance on traditional false-belief tasks is positively correlated with a range of other developmental outcomes such as cooperation and prosocial behavior^20,21, it is possible that these negative impacts extend beyond false-belief understanding.

Within the post-pandemic cohort, we did not find any differences based on the date that children were tested, which ranged from September 2021 to May 2024. In California, the shelter-in-place order was lifted in June 2021. This suggests that even children who were tested three years post-lockdown exhibited poorer performance relative to pre-pandemic cohorts. This finding is consistent with the finding that learning losses for school-aged children have persisted, even several years after children returned to in-person schooling^5,7. This in turn raises the question of whether children who experienced pandemic lockdown will continue to exhibit differences in social cognition at later ages, or whether these differences will diminish across middle childhood.

Similarly, it is an open question whether differences in social cognition skills will be evident for children born post-lockdown. Although many aspects of daily life have returned to pre-pandemic ‘normal’, some differences remain: remote work and online education have become more commonplace, and some studies suggest that levels of social interaction remain reduced relative to before the pandemic^22,23. These continuing differences in children’s social environment raise the possibility that children born post-lockdown might exhibit different developmental trajectories relative to pre-pandemic cohorts.

Our findings also have methodological implications. We used traditional false-belief tasks that have been widely used for decades with well-established patterns of performance. Yet we found that children tested post-pandemic performed below chance on these tasks at an age when they would typically succeed. It seems plausible that other paradigms, especially those that measure social cognition skills or rely heavily on social interactions, might also yield different patterns of performance post-pandemic. If so, then it is relevant to ongoing conversations regarding replication and reproducibility within the field. Our findings suggest that differences between studies conducted pre- and post-pandemic should be interpreted with caution, as such differences could reflect meaningful cohort differences in children’s behavior.

References:

1. Chung, G., Lanier, P., & Wong, P. Y. J. (2022). Mediating effects of parental stress on harsh parenting and parent-child relationship during coronavirus (COVID-19) pandemic in Singapore. Journal of Family Violence, 37(5), 801-812. https://doi.org/10.1007/s10896-020-00200-1.
2. Schmidt, A., Kramer, A. C., Brose, A., Schmiedek, F., & Neubauer, A. B. (2021). Distance learning, parent–child interactions, and affective well-being of parents and children during the COVID-19 pandemic: A daily diary study. Developmental Psychology, 57(10), 1719-1734. https://doi.org/10.1037/dev0001232.
3. Samji, H., Wu, J., Ladak, A., Vossen, C., Stewart, E., Dove, N., Long, D., & Snell, G. (2022). Mental health impacts of the COVID‐19 pandemic on children and youth–a systematic review. Child and Adolescent Mental Health, 27(2), 173-189. doi:10.1111/camh.12501.
4. Viner, R., Russell, S., Saulle, R., Croker, H., Stansfield, C., Packer, J., Nicholls, D., Goddings, A., Bonnell, C., Hudson, L., Hope, S., Ward, J., Schwalbe, N., Morgan, A., & Minozzi, S. (2022). School closures during social lockdown and mental health, health behaviors, and well-being among children and adolescents during the first COVID-19 wave: a systematic review. JAMA Pediatrics, 176(4), 400-409.
5. Betthäuser, B. A., Bach-Mortensen, A. M., & Engzell, P. (2023). A systematic review and meta-analysis of the evidence on learning during the COVID-19 pandemic. Nature Human Behaviour, 7(3), 375-385. https://doi.org/10.1038/s41562-022-01506-4.
6. Engzell, P., Frey, A., & Verhagen, M. D. (2021). Learning loss due to school closures during the COVID-19 pandemic. Proceedings of the National Academy of Sciences, 118(17), e2022376118. https://doi.org/10.1073/pnas.2022376118.
7. Kuhfeld, M. & Lewis, K. (2025, March 18). 5 years after COVID-19 hit: Test data converge on math gains, stalled reading recovery. Brookings. https://www.brookings.edu/articles/5-years-after-covid-19-hit-test-data-converge-on-math-gains-stalled-reading-recovery/
8. Sato, K., Fukai, T., Fujisawa, K. K., & Nakamuro, M. (2023). Association between the COVID-19 pandemic and early childhood development. JAMA Pediatrics, 177(9), 930-938. doi:10.1001/jamapediatrics.2023.2096.
9. González, M., Loose, T., Liz, M., Pérez, M., Rodríguez‐Vinçon, J. I., Tomás‐Llerena, C., & Vásquez‐Echeverría, A. (2022). School readiness losses during the COVID‐19 outbreak. A comparison of two cohorts of young children. Child Development, 93(4), 910-924. DOI: 10.1111/cdev.13738
10. Bergmann, C., Dimitrova, N., Alaslani, K., Almohammadi, A., Alroqi, H., Aussems, S., Barokova, M., Davies, C., Gonzalez-Gomez, N., Gibson, S.P., Havron, N., Horowitz Kraus, T., Kanero, J., Kartushina, N., Keller, C., Mayor, J., Mundry, R., Shinskey, J., & Mani, N. (2022). Young children’s screen time during the first COVID-19 lockdown in 12 countries. Scientific Reports, 12(1), 2015. https://doi.org/10.1038/s41598-022-05840-5.
11. Hartshorne, J. K., Huang, Y. T., Paredes, P. M. L., Oppenheimer, K., Robbins, P. T., & Velasco, M. D. (2021). Screen time as an index of family distress. Current Research in Behavioral Sciences, 2, 100023. https://doi.org/10.1016/j.crbeha.2021.100023.
12. Devine, R. T., & Hughes, C. (2018). Family correlates of false belief understanding in early childhood: A meta‐analysis. Child Development, 89(3), 971-987. https://doi.org/10.1111/cdev.12682.
13. Lane, J. D., & Bowman, L. C. (2021). How children’s social tendencies can shape their theory of mind development: Access and attention to social information. Developmental Review, 61, 100977. https://doi.org/10.1016/j.dr.2021.100977.
14. Suway, J. G., Degnan, K. A., Sussman, A. L., & Fox, N. A. (2012). The relations among theory of mind, behavioral inhibition, and peer interactions in early childhood. Social Development, 21(2), 331-342. https://doi.org/10.1111/j.1467-9507.2011.00634.x.
15. Scott, R. M., Nguyentran, G., & Sullivan, J. Z. (2024). The COVID-19 pandemic and social cognitive outcomes in early childhood. Scientific Reports, 14(1), 28939.
16. Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a “theory of mind”? Cognition, 21(1), 37-46. https://doi.org/10.1016/0010-0277(85)90022-8.
17. Gopnik, A., & Astington, J. W. (1988). Children’s understanding of representational change and its relation to the understanding of false belief and the appearance–reality distinction. Child Development, 59, 26–37. https://doi.org/10.2307/1130386.
18. Scott, R. M., Roby, E., & Setoh, P. (2020). 2.5-year-olds succeed in identity and location elicited-response false-belief tasks with adequate response practice. Journal of Experimental Child Psychology, 198, 104890. https://doi.org/10.1016/j.jecp.2020.104890.
19. Setoh, P., Scott, R. M., & Baillargeon, R. (2016). Two-and-a-half-year-olds succeed at a traditional false-belief task with reduced processing demands. Proceedings of the National Academy of Sciences, 113(47), 13360-13365. https://doi.org/10.1073/pnas.1609203113.
20. Caputi, M., Lecce, S., Pagnin, A., & Banerjee, R. (2012). Longitudinal effects of theory of mind on later peer relations: the role of prosocial behavior. Developmental Psychology, 48(1), 257-270. https://doi.org/10.1037/a0025402.
21. Imuta, K., Henry, J. D., Slaughter, V., Selcuk, B., & Ruffman, T. (2016). Theory of mind and prosocial behavior in childhood: A meta-analytic review. Developmental Psychology, 52(8), 1192-1205. http://dx.doi.org/10.1037/dev0000140.
22. Jarvis, C. I., Coletti, P., Backer, J. A., Munday, J. D., Faes, C., Beutels, P., … & Edmunds, W. J. (2024). Social contact patterns following the COVID-19 pandemic: a snapshot of post-pandemic behaviour from the CoMix study. Epidemics, 48, 100778.
23. Jeong, H., Park, S., Chun, J. Y., Ohmagari, N., Kim, Y., & Tsuzuki, S. (2025). Chronological trend of social contact patterns in Japan after the emergence of COVID-19. Journal of Infection and Public Health, 18(2), 102629.

About the Author

Previous studies have shown that human newborns and infants prefer simple, schematic face-like patterns (three dots arranged in a triangular configuration resembling two eyes and a mouth) over equally complex stimuli displaced in a neutral configuration (the same three dots arranged in an inverted pattern)^1-4. This ability is not unique to humans, but appears to be common across other species as well, including non-human primates⁵, chicks⁶, and even turtles⁷, suggesting it may have deep evolutionary roots.

But when, precisely, does this preference for face-like stimuli begin?

Recent research suggests that fetuses, while still in the womb, may already have the ability to differentially respond to and prioritize face-like patterns over other visual stimuli.^8,9 A pioneering study by Reid and colleagues (2017), using 4D ultrasound scans, demonstrated that third-trimester fetuses preferentially orient their heads toward light patterns resembling faces⁸.

A new study by Ronga and colleagues (2025), has recently replicated this research using a different methodological approach.¹⁰ The researchers questioned whether the preference for face-like stimuli observed in fetuses’ head movements could be observed using an alternative measure, namely, the fetal eye-lens^[1] movements.

The monitoring of eye movements and gaze behavior is already widely used in studies of infants after birth.^11-13 In fact, several studies have exploited looking preference paradigms to measure infants’ attention orienting to external events, showing a preference for face-like stimuli at birth.^1-4,13 In our recent prenatal study, we utilised eye movements – rather than head movements – to measure looking preference, allowing for maximum comparability between pre- and postnatal studies. This experimental approach consists of examining, via 2D ultrasound imaging, fetal gaze responses to visual stimuli projected through the mother’s abdomen to further investigate fetal preferential engagement with face-like stimuli.

The study also addresses two other key questions: When does the ability to discriminate such salient visual stimuli emerge? And in relation to which neural structure?

To answer these questions, we tested fetuses at various time-points during the third trimester: on average, at 26, 31, and 37 weeks of gestation. Moreover, we examined, at each time-point,  correlational effects with neurodevelopmental parameters, such as the growth of the thalamic nuclei and the insula, as well as the thickness of the cortical layers.

The results confirm previous findings, suggesting that the ability to visually differentiate between face-like and non-face-like stimuli is already present before birth. Further, even though a greater number of lens movements are performed overall at both 31 and 37 weeks of gestation versus  26 weeks of gestation, fetuses already showed a preference for face-like configurations at 26 weeks, the very beginning of the third trimester.

Although only exploratory, of particular interest are the results from the anatomical data. Specifically, we observed a correlation between the growth of the thalamic nuclei and the strength of the face-like preference at 26 weeks. This finding might be considered as a preliminary indication that, at the beginning of the third trimester – when the thalamocortical pathway has just established and the visual cortical layers are not fully developed^14,15 – the thalamic nuclei, once sufficiently developed, may represent a key neural structure for face‐like preference.

While our study did not directly compare face-like and top-heavy non-face-like stimuli, previous research in newborns (e.g., Buiatti et al., 2019) suggests that the preference is specific to face-like configurations. Investigating whether this distinction is already present before birth is an important direction for future research.

In sum, this research provides compelling evidence that certain aspects of visual recognition and brain development are already underway in the womb. Accordingly, the study indicates that fundamental aspects of visual processing begin taking shape prenatally, laying the groundwork for further postnatal development.

The idea that fetuses already pay more attention to face-like configurations is both intriguing and profound. However, further studies are necessary to fully understand the underlying neural mechanisms and the complex processes that shape prenatal vision: the very same processes that ultimately enable us, from the very beginning, to connect with others by seeking their gaze.

References

1. Morton, J. , and Johnson M. H.. 1991. “CONSPEC and CONLERN: A Two‐Process Theory of Infant Face Recognition.” Psychological Review 98, no. 2: 164–181. 10.1037/0033-295x.98.2.164.
2. Valenza, E. , Simion F., Cassia V. M., and Umiltà C.. 1996. “Face Preference at Birth.” Journal of Experimental Psychology. Human Perception and Performance 22, no. 4: 892–903. 10.1037//0096-1523.22.4.892.
3. Cassia, V. M. , Turati C., and Simion F.. 2004. “Can a Nonspecific Bias Toward Top‐Heavy Patterns Explain Newborns’ Face Preference?” Psychological Science 15, no. 6: 379–383. 10.1111/j.0956-7976.2004.00688.x.
4. Farroni, T. , Johnson M. H., Menon E., Zulian L., Faraguna D., and Csibra G.. 2005. “Newborns’ Preference for Face‐Relevant Stimuli: Effects of Contrast Polarity.” Proceedings of the National Academy of Sciences 102, no. 47: 17245–17250. 10.1073/pnas.0502205102.
5. Sugita, Y. 2008. “Face Perception in Monkeys Reared With no Exposure to Faces.” Proceedings of the National Academy of Sciences 105, no. 1: 394–398. 10.1073/pnas.0706079105.
6. Di Giorgio, E. , Loveland J. L., Mayer U., Rosa‐Salva O., Versace E., and Vallortigara G.. 2017. “Filial Responses as Predisposed and Learned Preferences: Early Attachment in Chicks and Babies.” Behavioural Brain Research 325, no. Pt B: 90–104. 10.1016/j.bbr.2016.09.018.
7. Versace, E. , Damini S., and Stancher G.. 2020. “Early Preference for Face‐Like Stimuli in Solitary Species as Revealed by Tortoise Hatchlings.” Proceedings of the National Academy of Sciences 117, no. 39: 24047–24049. 10.1073/pnas.2011453117.
8. Reid, V. M. , Dunn K., Young R. J., Amu J., Donovan T., and Reissland N.. 2017. “The Human Fetus Preferentially Engages With Face‐Like Visual Stimuli.” Current Biology: CB 27, no. 13: 2052. 10.1016/j.cub.2017.06.036.
9. Reissland, N. , Wood R., Einbeck J., and Lane A.. 2020. “Effects of Maternal Mental Health on Fetal Visual Preference for Face‐Like Compared to Non‐Face Like Light Stimulation.” Early Human Development 151: 105227. 10.1016/j.earlhumdev.2020.105227.
10. Ronga, I., Poles, K., Pace, C., Fantoni, M., Luppino, J., Gaglioti, P., Todros, T., & Garbarini, F. (2025). At First Sight: Fetal Eye Movements Reveal a Preference for Face-Like Configurations From 26 Weeks of Gestation. Developmental science, 28(2), e13597. https://doi.org/10.1111/desc.13597
11. Smith, N. A., Gibilisco, C. R., Meisinger, R. E., & Hankey, M. (2013). Asymmetry in infants’ selective attention to facial features during visual processing of infant-directed speech. Frontiers in psychology, 4, 601. https://doi.org/10.3389/fpsyg.2013.00601
12. Filippetti, M. L., Johnson, M. H., Lloyd-Fox, S., Dragovic, D., & Farroni, T. (2013). Body perception in newborns. Current biology : CB, 23(23), 2413–2416. https://doi.org/10.1016/j.cub.2013.10.017
13. Johnson, M. H. , Senju A., and Tomalski P.. 2015. “The Two‐Process Theory of Face Processing: Modifications Based on Two Decades of Data From Infants and Adults.” Neuroscience & Biobehavioral Reviews 50: 169–179. 10.1016/j.neubiorev.2014.10.009.
14. Kostović, I. , and Judas M.. 2010. “The Development of the Subplate and Thalamocortical Connections in the Human Foetal Brain.” Acta Paediatrica (Oslo, Norway: 1992) 99, no. 8: 1119–1127. 10.1111/j.1651-2227.2010.01811.x.
15. Frohlich, J. , Bayne T., Crone J. S., et al. 2023. “Not With a “Zap” But With a “Beep”: Measuring the Origins of Perinatal Experience.” Neuroimage 273: 120057. 10.1016/j.neuroimage.2023.120057.
16. Buiatti M, Di Giorgio E, Piazza M, Polloni C, Menna G, Taddei F, Baldo E, Vallortigara G. Cortical route for facelike pattern processing in human newborns. Proc Natl Acad Sci U S A. 2019 Mar 5;116(10):4625-4630. doi: 10.1073/pnas.1812419116. Epub 2019 Feb 12. PMID: 30755519; PMCID: PMC6410830.
17. ^[1] Eye movements refer to the movement of the entire eyeball within the orbit. Due to its highly echogenic appearance on 2D ultrasound, the crystalline lens of the fetal eye can serve as a visible marker for detecting eyeball motion. Thus, the movements of the eye lenses, which occur as a consequence of eyeball motion, were examined as an indicator of eye displacement.

About the Author

Attention-deficit/hyperactivity disorder (ADHD) is the most prevalent neurodevelopmental condition, affecting ~8% of children in the U.S.¹ With an average age of diagnosis of approximately 7 years, by the time it’s typically detected, intervention can already be a challenge. As shown in Figure 1, ADHD portends a host of other, later‐emerging forms of psychopathology and impairment,² and frequently co‐occurs with learning disabilities, depression, anxiety, oppositional defiant and conduct disorders, and substance use disorders. Long-term impairment into adulthood is common, including relationship problems, lower educational and occupational attainment, driving accidents, self-injury and high rates of suicide attempts, and early mortality.^3–5 Because ADHD is usually the first of these to develop⁶ and may exert a causal influence on later-occurring conditions,^7,8 it plays an outsized role in public health. Overall, because of the high prevalence and significant impairment associated with ADHD across the lifespan,^3,5,9,10 there’s an urgent need to identify, as early as possible, young children in need of support in order to improve longer-term outcomes.

Although it is commonly believed that ADHD originates earlier in development than it is typically diagnosed, there remains some controversy surrounding the idea of diagnosing ADHD in very young children. There are reasonable concerns about medicating young children, over-diagnosis, and differentiating typical from atypical preschooler behavior. Indeed, many of the symptoms of ADHD are expected in young children. The initial research on the early ADHD phenotype focused on the preschool period (3-5 years), and multiple studies have shown that careful, thorough diagnoses in preschoolers persist in the majority of cases.^11,12 But in our view, the preschool period is already late! This is because it is increasingly clear that etiological mechanisms associated with ADHD begin to exert effects prenatally, continuing through the infant-toddler period, a time characterized by heightened neural plasticity that could potentially amplify intervention effects. Thus, the mechanisms associated with ADHD act well before the age at which diagnoses can be made and symptoms can be reliably measured with current tools. As such, there is growing scientific interest in methods for identifying infants and toddlers at elevated likelihood for ADHD, with implications for early detection and intervention science. In the cascade model of ADHD shown in Figure 2, we illustrate bidirectional genetic-environmental influences which impact early development of brain networks related to self-regulation and attentional control across gestation and early development, impacting behavioral outcomes. The dashed arrows indicate effects of environment on genetic expression and brain/behavior outcomes, representing potential preemptive intervention opportunities, while solid arrows indicate genetic and epigenetic expression effects which may be harder to moderate.

Prior studies of early markers of ADHD in the infant-toddler periods generally highlight associations between non-specific factors and later ADHD symptoms like early motor and language delays, as well as temperament differences such as parent-reported overactivity. In 2022, Dr. Elizabeth Shephard published a meta-analysis in the Journal of the American Academy of Child & Adolescent Psychiatry¹³ and synthesized the existing literature focused on early life (<5 years) neurocognitive and behavioral precursors of ADHD prior to the preschool period. This meta-analysis revealed significant associations between ADHD and poorer motor and language development, social and emotional difficulties, early regulatory and sleep problems, sensory atypicalities, elevated activity levels, and executive function difficulties. These findings were critically important. They also revealed a number of gaps in the science and a need to remedy some of the methodological issues plaguing the existing literature. For example, studies have mostly been conducted in samples in which diagnostic outcomes had not been characterized, or samples not selected for elevated likelihood for ADHD, with some notable exceptions. The vast majority also relied on retrospective methods, using data from studies that were not originally designed to evaluate these questions, substantially limiting the availability of relevant measures. To remedy these limitations, we have begun using prospective methods to study these questions.

One of the primary designs utilized in early detection research is a familial likelihood design, which focus on recruiting infants at elevated and average likelihood for ADHD based on family history (i.e., presence or absence of first-degree relative with ADHD) so that the sample is enriched for the outcome of interest. Our groups typically enroll participants during gestation or through the first year of life and then follow them prospectively over the course of several years, collecting a variety of different types of data at each age and making initial clinical best estimate outcome determinations at the final study visit. This way, we can look backward at the previously collected data based on outcome group to determine when in development and in what domains the infants who went on to develop high levels of ADHD symptoms differ from those with typical development. Key findings across our groups thus far include:

• Infants at elevated familial likelihood for ADHD exhibit higher levels of ADHD-relevant behaviors (i.e., inattention/distractibility, hyperactivity, impulsivity) by 12 months of age compared to infants at average likelihood when rated by examiners or measured via second-by-second behavioral coding. Parents of infants in the ADHD-likelihood group are also more likely to endorse behavior concerns than parents of infants at average likelihood at 12 months of age, with increasing concerns from 12 to 18 months of age, versus stable concerns in the low-likelihood group.¹⁴
• Infants developing high levels of preschool ADHD symptoms are less likely to orient when their name is called at 12 and 18 months of age relative to infants without elevated ADHD symptoms in the preschool period, suggesting the possibility of early differences in attentional orienting mechanisms.¹⁵
• The interaction between negative emotionality and cognition measured at 9 months of age predicts toddler ADHD-related behaviors. Specifically, high or low levels of both negative emotionality and cognition predicted more ADHD-related behaviors in toddlerhood, suggesting there may be two affective-cognitive pathways to early inattention and hyperactivity/impulsivity.¹⁶
• Parent and observer ratings of activity level (but not attention) in 10-month-old infants with a family history of ADHD are positively associated with later preschool ADHD traits at 3 years.¹⁷
• 10-month-old family history infants show a lower EEG theta-beta power ratio than infants with no family history and this is positively associated with temperament dimensions conceptually related to ADHD at 2 years.¹⁸
• Infants with a family history of ADHD exhibit increased negative affect at 6 months of age. Thus affective response at 6 months of age may provide an early indicator of ADHD liability.¹⁹
• Several novel biomarkers of offspring ADHD risk can be assessed prior to the child’s birth, including maternal concentrations of inflammatory cytokines,²⁰ metabolic hormones (adiponectin and leptin),²¹ and serotonin system metabolites during pregnancy. For example, maternal inflammation in the 3rd trimester predicts child ADHD symptoms at age 4-6 years and mediates the effect of prenatal distress on child ADHD. Thus, maternal prenatal inflammation may be one common pathway by which prenatal risk factors (maternal distress, increased adiposity and poor nutrition) influence offspring mental health outcomes.²⁰
• Prenatal and postnatal factors interact in the prediction of child ADHD risk. For example, increased inflammation during pregnancy is prospectively associated with increased toddler ADHD symptoms, but only when children experience lower levels of maternal sensitivity during infancy.

Overall, we are learning that there is greater potential for earlier detection of ADHD-relevant behaviors than previously possible. But to move the field forward, we recognize that there is a need for collaborative, multisite studies focused on the early ADHD phenotype. To that end, we, along with other colleagues around the world, joined together to form the Early ADHD Consortium. This international network of investigators is engaged in prospective, longitudinal studies of elevated likelihood for ADHD beginning early in life. We are linked by our focus on developmental frameworks and centering developmental trajectories in our work, and through our incorporation of multimethod approaches. Our various teams have different areas of expertise, such as EEG, eye-tracking, behavioral and clinical methods, and preemptive intervention design. We published our first paper reviewing the state of the science and introducing the consortium in 2023 in JCPP Advances,²² and include a list of the measures used by our groups in case other investigators are interested in doing similar work. We are excited about opportunities to come together to pool data, resources, and ideas to increase the impact of this work.

References

1. Danielson ML, Bitsko RH, Ghandour RM, Holbrook JR, Kogan MD, Blumberg SJ. Prevalence of parent-reported ADHD diagnosis and associated treatment among U.S. children and adolescents, 2016. Journal of Clinical Child & Adolescent Psychology. 2018;47(2):199-212. doi:10.1080/15374416.2017.1417860
2. Barbaresi WJ, Colligan RC, Weaver AL, Voigt RG, Killian JM, Katusic SK. Mortality, ADHD, and psychosocial adversity in adults With childhood ADHD: A prospective study. Pediatrics. 2013;131(4):637-644. doi:10.1542/peds.2012-2354
3. Dalsgaard S, Østergaard SD, Leckman JF, Mortensen PB, Pedersen MG. Mortality in children, adolescents, and adults with attention deficit hyperactivity disorder: a nationwide cohort study. The Lancet. 2015;385(9983):2190-2196. doi:10.1016/S0140-6736(14)61684-6
4. Moffitt TE, Arseneault L, Belsky D, et al. A gradient of childhood self-control predicts health, wealth, and public safety. Proceedings of the National Academy of Sciences of the United States of America. 2011;108(7):2693-2698. doi:10.1073/pnas.1010076108
5. Kuriyan AB, Pelham WE, Molina BSG, et al. Young adult educational and vocational outcomes of children diagnosed with ADHD. Journal of Abnormal Child Psychology. 2013;41(1):27-41. doi:10.1007/s10802-012-9658-z
6. Kessler RC, Adler LA, Berglund P, et al. The effects of temporally secondary co-morbid mental disorders on the associations of DSM-IV ADHD with adverse outcomes in the US National Comorbidity Survey Replication Adolescent Supplement (NCS-A). Psychol Med. 2014;44(8):1779-1792. doi:10.1017/S0033291713002419
7. Riglin L, Leppert B, Dardani C, et al. ADHD and depression: investigating a causal explanation. Psychol Med. 2021;51(11):1890-1897. doi:10.1017/S0033291720000665
8. Treur JL, Demontis D, Smith GD, et al. Investigating causality between liability to ADHD and substance use, and liability to substance use and ADHD risk, using Mendelian randomization. Addiction Biology. 2021;26(1). doi:10.1111/adb.12849
9. Visser SN, Danielson ML, Bitsko RH, et al. Trends in the parent-report of health care provider-diagnosed and medicated attention-deficit/hyperactivity disorder: United States, 2003-2011. Journal of the American Academy of Child and Adolescent Psychiatry. 2014;53(1). doi:10.1016/j.jaac.2013.09.001
10. Wilens TE, Martelon M, Joshi G, et al. Does ADHD predict substance-use disorders? A 10-year follow-up study of young adults with ADHD. Journal of the American Academy of Child and Adolescent Psychiatry. 2011;50(6):543-553. doi:10.1016/j.jaac.2011.01.021
11. Lahey BB. Three-year predictive validity of DSM-IV attention deficit hyperactivity disorder in children diagnosed at 4-6 years of age. American Journal of Psychiatry. 2004;161(11):2014-2020. doi:10.1176/appi.ajp.161.11.2014
12. Riddle MA, Yershova K, Lazzaretto D, et al. The Preschool Attention-Deficit/Hyperactivity Disorder Treatment Study (PATS) 6-year follow-up. Journal of the American Academy of Child and Adolescent Psychiatry. 2013;52(3):264-278.e2. doi:10.1016/j.jaac.2012.12.007;
13. Shephard E, Zuccolo PF, Idrees I, et al. Systematic review and meta-analysis: The science of early-life precursors and interventions for attention-deficit/hyperactivity disorder. Journal of the American Academy of Child & Adolescent Psychiatry. 2022;61(2):187-226. doi:10.1016/j.jaac.2021.03.016
14. Miller M, Iosif, A., Bell, L.J., et al. Can familial risk for ADHD be detected in the first two years of life? Journal of Clinical Child & Adolescent Psychology. 2021;50(5):619-631.
15. Hatch B, Iosif AM, Chuang A, de la Paz L, Ozonoff S, Miller M. Longitudinal differences in response to name among infants developing ASD and risk for ADHD. Journal of Autism and Developmental Disorders. 2021;51(3):827-836. doi:10.1007/s10803-020-04369-8
16. Joseph HM, Lorenzo NE, Wang FL, Wilson MA, Molina BSG. The interaction between infant negative emotionality and cognition predicts ADHD-related behaviors in toddlerhood. Infant Behavior and Development. 2022;68:101742. doi:10.1016/j.infbeh.2022.101742
17. Goodwin A, Hendry A, Mason L, et al. Behavioural measures of infant activity but not attention associate with later preschool ADHD traits. Brain Sciences. 2021;11(5):524. doi:10.3390/brainsci11050524
18. Begum‐Ali J, Goodwin A, Mason L, et al. Altered theta–beta ratio in infancy associates with family history of ADHD and later ADHD‐relevant temperamental traits. Child Psychology Psychiatry. 2022;63(9):1057-1067. doi:10.1111/jcpp.13563
19. Sullivan EL, Holton KF, Nousen EK, et al. Early identification of ADHD risk via infant temperament and emotion regulation: A pilot study. Journal of Child Psychology and Psychiatry. 2015;56(9):949-957. doi:10.1111/jcpp.12426
20. Gustafsson HC, Sullivan EL, Battison EAJ, et al. Evaluation of maternal inflammation as a marker of future offspring ADHD symptoms: A prospective investigation. Brain, Behavior, and Immunity. 2020;89:350-356. doi:10.1016/j.bbi.2020.07.019
21. Sullivan EL, Molloy KR, Dunn GA, et al. Adipokines measured during pregnancy and at birth are associated with infant negative affect. Brain, Behavior, and Immunity. 2024;120:34-43. doi:10.1016/j.bbi.2024.05.018
22. Miller M, Arnett AB, Shephard E, et al. Delineating early developmental pathways to ADHD: Setting an international research agenda. JCPP Advances. 2023;3(2):e12144. doi:10.1002/jcv2.12144

About the Author

The formation of neural circuitry is an astonishing feat. Within a matter of months, two individual cells transform into a fully functioning nervous system. By adulthood, a single cubic millimeter of human brain tissue, roughly the size of a sharpened pencil tip, contains over 150 million connections.¹ But how does this incredibly complex organ develop? What are the levers that kickstart and propel this developmental cascade forward? How sensitive are these neurobiological processes to environmental influence? How much neural development occurs before versus after birth? Are individual differences in the initial development of neural phenotypes predictive of future health and disease?

My work, and the work of others, seeks to address these pressing scientific questions using fetal and infant magnetic resonance imaging (MRI). MRI is a noninvasive technology that leverages a strong magnetic field to estimate brain activity via changes in blood oxygen level, reconstruct detailed images of human anatomy, and quantify white matter properties such as myelination. Although commonly used in adults, adolescents, and children, MRI has only recently been used to study infant and fetal brain development. Many of the same analytic methods that are used in postnatal MRI can be applied to fetal MRI, including quantitative analysis of white matter tracts, resting-state functional connectivity analyses, and volumetric analysis of brain structure. The major difference between fetal and postnatal MRI lies in image processing. That is, fetal MRI data tends to contain higher levels of noise and thus requires specialized processing techniques. In addition to motion-related noise inherent to scanning very young ages, in vivo fetal MRI can be impacted by noise from maternal anatomy and lower signal-to-noise ratio since you are imaging the brain while it is inside the maternal compartment. These challenges have historically prevented widespread use of fetal and infant MRI. However, technological advances have made huge strides in overcoming these challenges, making it easier to collect and analyze infant and fetal MRI data than ever before. In addition to automated and semi-automated processing pipelines, fetal and infant MRI datasets are publicly available for researcher use. For example, the Developing Human Connectome Project (dHCP) issued their fourth data release earlier this year, which includes minimally processed functional, diffusion, and structural MRI data from 255 human fetuses and 809 human newborns at the time of this writing.²

In honor of this exciting data release, I wanted to highlight a few compelling scientific discoveries that have been made possible by fetal MRI in particular.

The beginnings of adult-like functional networks are in place prior to birth

The first paper on fetal resting-state functional MRI was conducted in only 16 fetuses that spanned 20 to 36 gestational weeks. Yet even at this young age, and prior to birth, two analogs of major resting-state networks were detected – a bilateral frontal network and a bilateral visual network.³ Identification of proto visual networks in the fetal brain is particularly remarkable given extremely limited, if any, visual input prior to birth. Since this first paper, analogs of other major canonical resting-state networks that are present postnatally have been detected in a larger fetal cohort⁴ and in preterm-born infants,⁵ including auditory, sensorimotor, default mode, and subcortical proto networks. In addition to identifying the presence of early networks, graph theory has been used to identify hubs, or highly connected regions, in the fetal brain. Interestingly, two major hubs that emerged from this analysis were bits of cortex that become specialized for perceiving faces and understanding language by adulthood.⁶ This finding raises the intriguing possibility that humans may come into the world prepared to process important social cues.

These studies do not suggest neural circuitry is fully formed at birth, despite significant overlap in fetal and adult connectomes.⁷ Rather, there is dramatic, nonlinear change in brain structure and function across gestation and into the postnatal period. For example, at least one study identified an inflation point in whole-brain functional connectivity metrics at 25-28 gestational weeks. Prior to this inflation point, mean functional connectivity values were low for resting-state networks. After 25-28 gestational weeks, rate of change in connectivity values increased by more than 90%, suggesting the third trimester is a key ‘expansion period’ for functional network development.⁸ Functional specialization, network restructuring, and anatomical changes continue into the postnatal period, with widespread changes in brain structure evident across infancy, childhood, adolescence, and even into adulthood.^9,10 Nonetheless, the fetal studies collectively suggest that the way we respond to the environment may be influenced by neural circuitry that is in place prior to birth. Such a possibility, and fetal MRI as a tool, open the door for critically important studies that will determine whether proto resting-state networks require experience to develop associated functions, and whether there are critical periods when structure/function mapping may be most primed to occur.

Fetal connectivity patterns correlate with infant and toddler behavior

Whether neural phenotypes early in life are informative for predicting future behavior is of immense clinical significance. To date, few human studies have prospectively linked individual differences in fetal neural connectivity with postnatal behavior. Yet studies are beginning to address this gap.

Using functional MRI data from 120 fetuses, Ji and colleagues showed that spontaneous coactivation of the supplementary motor area with visual cortex and other posterior brain regions in utero correlated with future motor behavior.¹¹ Specifically, more frequent supplementary motor area to posterior cortex co-activation associated with more mature motor behavior at 7 months postpartum. In addition, some of my own work has shown that neural correlates of emotion dysregulation are detectable in utero, well before the onset of clinical symptomatology. In this study (n=79), functional coupling between frontolimbic regions in the fetal brain correlated with future emotion dysregulation and aggressive behaviors at 3 years of age, even after accounting for co-occurring pre- and postnatal exposure to low income, maternal stress, maternal depression, and maternal anxiety.¹² These preliminary findings, along with others, underscore the behavioral, and potentially clinical, significance of individual differences in prenatal brain development. Excitingly, the new dHCP data release includes measures of infant behavior, so it will be possible for independent research groups to replicate and extend findings that tie in utero neural connectomics to future behavior.

Conclusions and future directions

Because fetal MRI allows examination of the initial wiring of neurocircuitry, this tool is well-poised to provide direct insight into the very foundations of human brain development and etiology of neurodevelopmental disorders. Examination of fetal brain development inherently occurs prior to postnatal exposures, such as parenting, that we know influence child growth and behavior. Thus, fetal MRI also enables researchers to disentangle impact of the gestational environment from postnatal exposures, with potential to isolate sensitive windows for exposure impact. Combined with genetically informed research designs, fetal MRI may eventually be used to answer fundamental questions about the extent to which neural phenotypes result from nature versus nurture.

References

1. Shapson-Coe A, Januszewski M, Berger DR, et al. (2024) A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution. Science, 384(6696). DOI: https://doi.org/10.1126/science.adk4858.
2. Karolis, V., Cordero-Grande, L., Price, A. N., et al. (2024). The developing Human Connectome Project fetal functional MRI release: Methods and data structures. bioRxiv Preprint, DOI: https://doi.org/10.1101/2024.06.13.598863
3. Schöpf, V., Kasprian, G., Brugger, P. C., & Prayer, D. (2012). Watching the fetal brain at ‘rest’. International Journal of Developmental Neuroscience, 30(1), 11-17. DOI: https://doi.org/10.1016/j.ijdevneu.2011.10.006.
4. Ji, L., Hendrix, C. L., & Thomason, M. E. (2022). Empirical evaluation of human fetal fMRI preprocessing steps. Network Neuroscience, 6(3), 702-721. DOI: https://doi.org/10.1162/netn_a_00254.
5. Doria, V., Beckmann, C. F., Arichi, T., et al. (2010). Emergence of resting state networks in the preterm human brain. Proceedings of the National Academy of Sciences, 107(46), 20015-20020. DOI: https://doi.org/10.1073/pnas.100792110
6. van den Heuvel, M. I., Turk, E., Manning, J. H., et al. (2018). Hubs in the human fetal brain network. Developmental Cognitive Neuroscience, 30, 108-115. DOI: https://doi.org/10.1016/j.dcn.2018.02.001
7. Turk, E., Van Den Heuvel, M. I., Benders, M. J., et al. (2019). Functional connectome of the fetal brain. Journal of Neuroscience, 39(49), 9716-9724. DOI: https://doi.org/10.1523/JNEUROSCI.2891-18.2019.
8. Jakab, A., Schwartz, E., Kasprian, G., et al. (2014). Fetal functional imaging portrays heterogeneous development of emerging human brain networks. Frontiers in Human Neuroscience, 8. DOI: https://doi.org/10.3389/fnhum.2014.00852.
9. Rutherford, S., Fraza, C., Dinga, R., Kia, S. M., et al. (2022) Charting brain growth and aging at high spatial precision. eLife, 11(e72904). DOI: https://doi.org/10.7554/eLife.72904.
10. Bethlehem, R.A.I., Seidlitz, J., White, S.R. et al. Brain charts for the human lifespan (2022). Nature, 604. https://doi.org/10.1038/s41586-022-04554-y.
11. Ji, L., Majbri, A., Hendrix, C. L., & Thomason, M. E. (2023). Fetal behavior during MRI changes with age and relates to network dynamics. Human Brain Mapping, 44(4), 1683-1694. DOI: https://doi.org/10.1002/hbm.26167
12. Hendrix, C. L., Ji, L., Werchan, D. M., et al. (2023). Fetal frontolimbic connectivity prospectively associates with aggression in toddlers. Biological Psychiatry Global Open Science, 3(4), 969-978. DOI: https://doi.org/10.1016/j.bpsgos.2022.09.003

About the Author