Common neural substrates of diverse neurodevelopmental disorders

Abstract

Neurodevelopmental disorders are categorized and studied according to their manifestations as distinct syndromes. For instance, congenital prosopagnosia and dyslexia have largely non-overlapping research literatures and clinical pathways for diagnosis and intervention. On the other hand, the high incidence of neurodevelopmental comorbidities or co-existing extreme strengths and weaknesses suggest that transdiagnostic commonalities may be greater than currently appreciated. The core-periphery model holds that brain regions within the stable core perceptual and motor regions are more densely connected to one another compared to regions in the flexible periphery comprising multimodal association regions. This model provides a framework for the interpretation of neural data in normal development and clinical disorders. Considering network-level commonalities reported in studies of neurodevelopmental disorders, variability in multimodal association cortex connectivity may reflect a shared origin of seemingly distinct neurodevelopmental disorders. This framework helps to explain both comorbidities in neurodevelopmental disorders and profiles of strengths and weaknesses attributable to competitive processing between cognitive systems within an individual.

Neurodevelopmental disorders are typically studied in isolation, obscuring potential commonalities. Sokolowski and Levine use the core-periphery framework to explore network-level commonalities across distinct disorders. Variability in multimodal association cortex connectivity may reflect a shared origin of seemingly distinct neurodevelopmental disorders.

Neurodevelopmental disorders—or congenital disorders of brain system development causing cognitive and behavioral impairments—have wide-ranging consequences for academic, social, and mental health outcomes.^1,2 These disorders—affecting ∼53 million children within the first 5 years of life (i.e. a prevalence of 12%)³—are the third-ranked form of child disability, after vision and hearing loss. The worldwide annual burden of childhood disability is estimated to be as high as $69 500 per child, annually. Although there is a heavier burden in developing countries,⁴ financial disparities between families with and without children with neurodevelopmental disorders are evident across all levels of socioeconomic status.⁵

Neurodevelopmental disorders can have effects confined to specific domains, such as reading (dyslexia) or facial recognition (congenital prosopagnosia), or domain-general functions such as attention (attention deficient hyperactivity disorder, ADHD) or social communication (autism spectrum disorder, ASD). Much of the accumulated knowledge concerning neurodevelopmental disorders has emerged from within-disorder research, rather than comparison across disorders. Indeed, as with case-control research in general, the presence of comorbidities can be grounds for exclusion as it contaminates the ‘purity’ of the phenotype of interest. This practice has promoted depth of knowledge concerning individual disorders at the expense of recognizing shared features across disorders.

In this update, we synthesize neuroimaging research in samples of putatively distinct neurodevelopmental disorders affecting specific perceptual, mnemonic, and academic functions. Many of these disorders mimic classical focal lesion syndromes (e.g. acquired dyslexia or prosopagnosia) such that they implicate discrete neurofunctional circuits (e.g. left-lateralized language circuits in dyslexia or ventral temporal circuits in prosopagnosia). Yet network models of brain organization hold that function is also supported by inter-regional connectivity, including stable trait-like differences in brain organization that are reliable⁶ and emerge during development,⁷ thus providing a rich source of information about brain-behaviour relationships.^8,9 Moreover, these models have the advantage of accommodating interactions within and between large-scale systems in addition to functions localized to discrete circuits.¹⁰

Examining commonalities in network-level function across diagnostic groups can provide insights into which connectome alterations are shared among brain disorders, as has been done with neurodegenerative diseases (Box 1)¹¹ and psychiatric disorders.¹⁷ The core-periphery model of brain organization distinguishes stable core perceptual and motor regions from multimodal association regions in the flexible periphery, with core regions more densely interconnected than those in the periphery.²⁴ The present review illustrates a common pattern of altered connectivity between core basic processing units and higher-level periphery association cortices across neurodevelopmental disorders affecting perceptual, academic, and mnemonic function, resonating with recent criticism of the core-deficit hypothesis in neurodevelopmental disorders.²⁵ This transdiagnostic approach helps to account for comorbidities and paradoxical extreme strengths and weaknesses within an individual.²⁶

Box 1. Pathology in neurodegenerative diseases.

In the early stages of neurodegenerative diseases, pathology aggregates in specific, localized brain regions with selectively vulnerable neuronal populations (e.g. Seeley et al.¹²), then progresses through anatomically linked regions,¹³ consistent with the network degeneration hypothesis (for review see Drzezga¹⁴). This hypothesis was systematically tested in a seminal paper in which network-sensitive neuroimaging methods were used to reveal that distinct neurodegenerative systems (Alzheimer’s disease; behavioural variant frontotemporal dementia; semantic dementia; progressive nonfluent aphasia; corticobasal syndrome) influenced human intrinsic functional connectivity systems present in healthy human adults¹¹ (see also Zhou et al.¹⁵). Even in healthy adults, structural atrophy mirrors characteristic patterns in neurodegenerative diseases, coupled with the individual’s predisposition to express that disease.¹⁶ The pressure of connectome-wide communication—with long-range connections maintained by centralized hub regions that allow for rapid transmission of information across functional domains– elevates the risk of local brain changes (such as neurodegenerative diseases) spreading easily across the network.¹⁷ Local network disruption (i.e. failure of a particular node in a network) initiates ‘cascading network failure’, in which nodes fail across the network over time,^{18, 19} such that the spread of neuropathological effects is mediated by structural and functional connectivity within the human connectome (for a review see Iturria-Medina and Evans²⁰). Neurodegenerative disorders provide information on connectivity through ‘fault lines’ that are vulnerable to spreading pathology. Intriguingly, network effects can be retrospectively traced, even to childhood,²¹ to cognitive strengths and weakness predating disease onset,²² suggesting a neurodevelopmental origin. For example, structural imaging and cognitive measures can be used to predict the onset of frontotemporal dementia 5–10 years before the onset of symptoms in adults at risk of developing the neurodegenerative disorder later in life.²³ Thus, insights into individual differences in normal function and neurodevelopmental syndromes are valuable in delineating the organizational principles of the human connectome.

Neurodevelopmental disorders across cognitive domains

Perceptual functions: congenital prosopagnosia and amusia

Individuals with a lifelong incapacity to perceive faces are classified as having congenital prosopagnosia. Across individuals, face-processing is supported by a network spanning the ventral occipital temporal cortex that includes the fusiform face area and occipital face area, structurally connected by the inferior longitudinal fasciculus and inferior occipito-frontal fasciculus (Fig. 1B). This core face-processing network is situated within an extended face-processing network that includes frontal regions, typically associated with executive functioning and attention (i.e. regions in the frontal lobes including the inferior frontal gyrus and orbital frontal cortex, and the anterior temporal lobe).³¹ Congenital prosopagnosia (i.e. the neurodevelopmental disorder in which people cannot process faces) is thought to arise from a failure to propagate neural signals between the functioning ventral occipito-temporal cortex and the extended nodes of the face-processing network,^31–37 in addition to dysfunction of activity and connectivity within the ventral occipito-temporal cortex.^38–41

Figure 1.

Long association fibres associated with perceptual, mnemonic, and academic functions. Individual differences, with neurodevelopmental disorders representing extreme ends of the distributions, have been studied in the context of specific processing domains (e.g. language, perception). Analysis of commonalities across these domains reveal general principles of human information processing and consequently, the architecture of the human brain connectome. (A) Visualization of anatomical connectivity within the human brain (left) and a schematic depiction of key white matter tracts (right). B–D illustrate white matter tracts forming distributed networks associated with perceptual (right hemisphere), mnemonic (left hemisphere) and academic abilities (left hemisphere). Each panel displays key white matter tracts associated with two exemplar cognitive domains (labelled in grey boxes) along with regions supporting relevant lower-level sensory processing. Individual differences that associate with variation in structural and functional network connectivity exhibit aberrant connectivity within these same networks in individuals with neurodevelopmental disorders (italicized text) (e.g. the inferior longitudinal fasciculus (ILF) and inferior occipito-frontal fasciculus (IoFF) relate to individual differences in face-processing and congenital prosopagnosia). (E) Schematic illustration of network topology applied to neurodevelopmental disorders. The small circles represent brain regions (i.e. nodes) and the lines represent connections between brain regions (i.e. edges). The connectome is comprised of modules (i.e. densely connected nodes depicted with green, orange, and yellow ovals). Networks with core-periphery organization also exhibit a set of tightly connected nodes (i.e. hub nodes) referred to as the core (four red inner circles) which are sparsely connected to a set of relatively isolated nodes referred to as the periphery (eight blue outer circles). Neurodevelopmental disorders share the common neural substrate of altered connectivity between the network core and periphery. These altered patterns may be characterized by weaker connections (dashed lines), but also instances of enhanced connections (bold lines) that may emerge through network reorganization or compensation. (Images in A–D were created using the anatomically curated white matter atlas generated by the O’Donnel Research Group (ORG), derived from 100 healthy human brain scans from the Human Connectome Project²⁷ and visualized using slicerDMR.^28,29 Panel E was inspired by Bassett et al.³⁰)

Individuals with a lifelong inability to perceive pitch are characterized as having congenital amusia. Pitch perception is supported by a frontal-temporal network that includes the primary auditory cortex, superior temporal gyrus, and inferior frontal gyrus,⁴² structurally connected by the superior longitudinal fasciculus (including the arcuate fasciculus) (Fig. 1B). Individuals with congenital amusia exhibit reduced functional and structural connectivity between auditory temporal regions and frontal regions, but intact bottom-up representations of pitch corresponding with typical patterns of connectivity between primary auditory cortex and the superior temporal gyrus.^43–50 This suggests that the pitch perception deficit arises from poor feedback control between the inferior frontal gyrus (i.e. a high-level processing region) and the superior temporal gyrus.⁴³ In direct contrast, individuals with absolute pitch show enhanced connectivity both within superior temporal lobe structures and across the frontal-temporal network.⁵¹

Mnemonic functions: autobiographical memory, imagery, and navigation

Encoding, storage, and retrieval of information on memory tasks are well-known to engage distributed cortical representations crucial for conscious apprehension of mnemonic content.⁵² Hippocampal-neocortical connectivity, through the fornix (the main efferent pathway from the hippocampus) and through long association fibres to frontal and parietal lobes, relates to performance on laboratory and naturalistic autobiographical memory tasks.^53–55 In Severely Deficient Autobiographical Memory (SDAM) (Fig. 1C), there is a lifelong inability to vividly recollect past autobiographical events, while other functions are preserved. Individuals with SDAM show evidence of intact basic perceptuo-mnemonic processing (e.g. intact performance on neuropsychological memory tests) and grossly normal medial temporal lobe anatomy, yet they exhibit reduced large-scale neural synchrony in relation to conscious recollection or re-experiencing of events encountered in both the laboratory and real life.^56,57 By contrast, individuals on the opposite extreme end of the spectrum of autobiographical memory ability [i.e. individuals with highly superior autobiographical memory (HSAM)⁵⁸] exhibit enhanced prefrontal-hippocampal functional connectivity.

The fundamental deficit in SDAM (subjective recollection of past events) is inaccessible to experimental verification. Visual imagery is a cognitive capacity that is closely related to autobiographical memory.^59,60 Individuals with aphantasia, a lifelong inability to voluntarily create mental images in the mind, report low visual (and other sensory) subjective imagery (including reduced autobiographical recollection) with a compelling objective behavioural correlate in binocular rivalry or perceptual competition between distinct, simultaneously- presented monocular stimuli. Priming a stimulus prior to the task normally induces a bias such that primed stimuli are perceived above chance. Aphantasics are immune to such perceptual priming effects.^61,62 That is, they fail to benefit from the primed image even though perceptual processing is intact. Similarly, they exhibit a physiological response in response to perceived, but not imagery-driven, fear-inducing stimuli.⁶³ The visual imagery deficit exhibited by aphantasics is not instantiated in perceptual processing units but rather in the feedback connections between the frontal and visual cortex.^64,65 As noted with other disorders, those on the other extreme end of the distribution of visual imagery ability (i.e. hyperphantasics) exhibit stronger functional connectivity between prefrontal regions and the visual network.⁶⁵

Spatial cognition in humans is supported by a network that includes the hippocampus, parahippocampus, retrosplenial cortex, and prefrontal cortex.⁶⁶ Developmental Topographical Disorientation (DTD) is characterized as a lifelong inability to navigate new and familiar environments.^67–69 Individuals with DTD exhibit intact processing in lower-level perceptual regions (e.g. the parahippocampal place area), but reduced structural and functional connectivity between the hippocampus, parahippocampal place area, retrosplenial cortex, and prefrontal cortex^67,69,70 and within the default mode network⁷¹ (Fig. 1C). Those with strong spatial orientation abilities exhibit increased levels of global efficiency within the spatial orientation network and increased node centrality in the hippocampus, supramarginal gyrus, and primary motor cortex.⁷² Similarly, graph theoretical techniques on low-density EEG data revealed that individuals with strong, compared to weak, spatial navigation abilities showed more functional connectivity.⁷³

Academic abilities: reading and mathematical competence

Reading is supported by a left-lateralized network, highly overlapping with the language network, which includes the superior temporal gyrus (Wernicke’s area), inferior frontal gyrus (Broca’s area), and fusiform gyrus, and is connected by the inferior longitudinal fasciculus, superior longitudinal fasciculus, inferior fronto-occipital fascicle, and corona radiata^74–76 (Fig. 1D). Individuals with developmental dyslexia, (i.e. a specific reading disability), exhibit intact phonetic representations in the auditory cortex, but a reduction in functional and structural connectivity between these temporal regions and the left inferior frontal gyrus.^74,77–79 Dyslexic individuals also present with reduced connectivity in the visual word-form area but increased connectivity within the right hemisphere.⁸⁰ Notably, individuals with dyslexia exhibit complex aberrant connectivity and even hyperconnectivity in brain systems beyond the reading network such as the limbic system and motor system, supporting the idea of a broad network-level origin of a behaviourally specific disorder.^80,81

Mathematical thinking, including basic number processing and higher-level symbolic manipulations and calculations, is another key skill within the academic domain predictive of later success.⁸² Developmental dyscalculia, a specific math learning disorder, is as equally prevalent as dyslexia but considerably less studied.⁸³ Mathematical thinking is subserved by a fronto-parietal network that includes the intraparietal sulcus, inferior parietal lobule, superior parietal lobule, and prefrontal cortex,⁸⁴ and is structurally connected by a range of white matter pathways including the inferior longitudinal fasciculus, superior longitudinal fasciculus, inferior fronto-occipital fascicle, corona radiata, corticospinal tract, and posterior segment of the corpus callosum^85,86 (Fig. 1D). Developmental dyscalculics present with reduced structural connectivity within these white matter tracts (for a review see Matejko and Ansari⁸⁶). Individuals with dyscalculia also show abnormalities in numerical representations and in functional connectivity within the mathematical network.^87,88 Functional connectivity findings in this domain are complex and differ in adults and children.⁸⁹ Adults with dyscalculia exhibit functional hyperconnectivity in temporo-occipital regions and abnormal functional activation in fronto-parietal regions during number processing, whereas children exhibit hyperconnectivity in the fronto-parietal network, but also the default mode network.^90,91 These network-level differences between dyscalculics and control participants may reflect the strategies that the individual uses to compensate for the underlying weaknesses.

Interim summary

Consideration of neuroimaging findings across perceptual, mnemonic, and academic domains, as outlined in Fig. 1, converge to support the claim that neurodevelopmental disorders across distinct cognitive domains share the common neural substrate of reductions in long-range projections between local perceptual regions and higher-level prefrontal regions that govern conscious access of information. Other disorders (e.g. alexithymia, developmental coordination disorder, dysgraphia) are not reviewed in detail due to the smaller number of studies, although they show similar patterns to the domains discussed above, namely aberrant activity and intrinsic connectivity within domain-specific localized regions and dysfunction in interconnectivity across large-scale brain networks.^92–94 Due to limitations in the inferences that can be drawn from case-control brain imaging data, the observed distributed effects cannot be conclusively separated from local (modular) effects. Our aim is not to rule out modular dysfunction, but to highlight the power of the network-level approach for explaining commonalities across neurodevelopmental disorders.

Applying network neuroscience to neurodevelopmental disorders

Diverse developmental disorders share network-level reductions in long-range cortical neuronal projections between local perceptual regions and higher-level prefrontal regions that govern conscious access of information. Deficits associated with neurodevelopmental disorders range from very specific limitations of learning to more widespread impairments of executive functions, social skills, or intelligence. The disorders reviewed above are defined by specific phenotypes in the context of relative preservation of other functions. Previous research interpreting domain-specific neurodevelopmental disorders from the perspective of encapsulated neurofunctional systems neglects the known dynamic interactivity across brain systems.^95,96 The non-random and unique organizational principles of brain networks as derived from neuroimaging data can be leveraged to formally characterize neural substrates of normal and disordered brain function.^97,98

Core-periphery organization (see Box 2) was recently proposed as a framework to characterize linkages between functional brain modules. The core-periphery model posits that the human connectome is composed of a stable core (i.e. sensorimotor and visual regions) with limited temporal connectivity variability and a flexible periphery (i.e. multimodal association regions) with frequently changing patterns of connectivity (Fig. 1E).²⁴ Like modularity, core-periphery organization consolidates and strengthens across developmental time, leading to an optimized modular yet integrated topology that simultaneously supports intra-modular functional specialization and inter-modular coordination.¹⁰³ Critically, core-periphery organization predicts individual differences in cognition across development more accurately than module organization alone.¹⁰⁴ Thus, core-periphery organization is an optimal model that can be used to understand neuropsychiatric disorders (for a review see Bassett et al.³⁰) and neurodevelopmental disorders.

Box 2. Network neuroscience and the core-periphery model.

Advances in neuroimaging methodologies have allowed for the modeling of structural and functional connections across brain regions that have been applied to the identification of brain networks supporting complex behaviors. Fig. 1E provides a schematic illustration of core-periphery network organization³⁰ applied to neurodevelopmental disorders. Network neuroscience models draw upon graph-theoretic frameworks whereby system elements or nodes supporting local processing are connected by edges. Nodes and edges combine to form network communities, linked by highly connected nodes in the brain network that occupy central positions in the overall organization of the network (i.e. hub nodes).⁹⁹ Networks tend to minimize the cost of wiring by forming locally dense clusters of nodes, referred to as ‘modules’ that are highly connected to each other, but sparsely connected to other clusters .¹⁰⁰ Modularity within the brain reflects the anatomical underpinnings of distinct functional systems that support information segregation. Across development, connectivity within modules strengthens while connectivity between modules weakens, with the exception of the strengthening of select hub edges linking modules.¹⁰¹ Network modules thus become more distinct across developmental time. This integrated modular topology present in adulthood supports functional specialization of brain networks across distinct areas of cognition. While network segregation is beneficial from a physical and metabolic cost perspective, networks must also contain attributes that support integration for global communication within the network. Long-range connections, typically passing through multiple hub regions, enable efficient communication between brain modules that are spatially and functionally distinct. Hub regions tend to be densely connected to each other, forming a central ‘core’ (sometimes referred to as a ‘rich club’) within a network.¹⁰² A network composed of a core made up of densely connected hubs, and a periphery of low-degree nodes that preferentially connect to the core is said to have core-periphery organization. Developmental disorders are characterized by weaker core-periphery connectivity, with instances of stronger core-periphery connections due to compensatory network reorganization, possibly accounting for the presence of extreme intra-individual profiles of strengths and weaknesses.

Core-periphery abnormalities are observed in domain-general disorders (e.g. schizophrenia, ASD, depression), emerging early, prior to the onset of clinical symptomology.^17,30 The findings reviewed above reveal that domain-specific neurodevelopmental disorders are linked to reduced connectivity between lower-level, domain-specific perceptual processing regions and higher-level processing networks (in addition to variation in the lower-level processing units themselves). This suggests that domain-specific disorders are a consequence of a developmental abnormality in the connectivity ‘between’ the core and periphery, rather than dysfunction of the core itself (as proposed for severe neuropsychiatric disorders³⁰). More specifically, network-level abnormalities associated with domain-specific neurodevelopmental disorders emerge from abnormal recurrent back-projections from the high-level processing networks (primarily within association cortex) that make up the periphery, into core perceptual regions (e.g. visual or auditory cortex). Connections in networks may consequentially reconfigure to optimize functioning within core-periphery organizational constraints, as is known to occur across distinct kinds of pathology¹⁷ (see Fig. 1E).

Harnessing connectivity in the transdiagnostic approach

The core-deficit hypothesis²⁵ (see Box 3) attempts to account for multifaceted neurobiological phenomena with a single and specific mechanistic impairment, such as the phonological deficit model of dyslexia.¹⁰⁷ (The term ‘core’ as used here is distinct from the above-described core-periphery model.) Yet such attempts to account for specific neurodevelopmental disorders via a single mechanistic core deficit have been unsuccessful. Accordingly, findings that specific neurodevelopmental disorders are underpinned by encapsulated neural regions or circuits are subtle and difficult to reproduce. Instead, researchers have proposed that developmental disorders should be reconceptualized as ‘few specific disorders and no specific brain regions’.²⁶

Box 3. The core-deficit hypothesis.

Children exhibit a complex combination of relative strengths and weaknesses across a wide range of cognitive domains. A developmental learning disability is diagnosed when a specific weakness is isolated (e.g. dyslexia in the case of low reading achievement). However, some children who reach the diagnostic criteria for multiple neurodevelopmental disorders are never formally diagnosed (e.g. Pearson,⁶²Bathelt et al.,¹⁰⁵ Siugzdaite et al.¹⁰⁶). This may be because the ‘core-deficit’ hypothesis is foundational to developmental psychology research, particularly research on neurodevelopmental disorders.²⁵ The core-deficit hypothesis posits that a single mechanistic impairment explains all observed cognitive and neural profiles within a particular diagnostic category. An example of the core-deficit model within the reading domain is phonological deficit theory, which argues that children with a reading impairment selectively struggle with phonemic awareness.¹⁰⁷ In the mathematical domain, the core-deficit model holds that developmental dyscalculia is a consequence of a deficiency within an evolutionarily ancient system specifically used to process quantities (i.e. the approximate number system).¹⁰⁸ This core-deficit account has been promoted across multiple cognitive domains in spite of serious limitations, such as its inability to explain many aspects of disorders (e.g. comorbidities¹⁰⁹). While a core-deficit model held promise to enhance understanding of complex behavioural phenomena by identifying fundamental cognitive or neural underpinnings, research supporting such models suffered from a range of methodological issues, including highly selective, small samples and measurements chosen using circular logic (e.g. using a phonological awareness task as the assessment tool for selecting a group of children who have dyslexia).²⁵ Neuroimaging methods used in developmental research drew upon studies of human adults that originated from research on focal brain damage. Empirical studies avoiding these pitfalls (e.g. Siugzdaite et al.¹⁰⁶) challenge the core-deficit model. Moreover, the core-deficit hypothesis cannot explain paradoxical advantages that are consistently reported in individuals with neurodevelopmental disorders.¹¹⁰ As a result of these limitations, researchers are embracing larger, diverse samples, a broader array of assessment methods, and network models of brain function (e.g. Bulthé et al.⁸⁷).

Our review suggests that while there is dysfunction within specific regions associated with neurodevelopmental disorders, a ‘common feature’ of domain-specific neurodevelopmental disorders is disruption in long-range projections between regions in the core to regions in the periphery, such that intact lower-level processing fails to ignite conscious awareness.^105,111 To the extent that this impaired conscious access is common across diverse neurodevelopmental disorders, domain-specific effects may be a product of the research approach (i.e. restricting samples to a single diagnosis) and therefore illusory, differing only in location and not in fundamental mechanisms.

Viewing neurodevelopmental disorders through a network lens helps to explain comorbidities that occur as a consequence of network disorganization.¹⁰⁶ Cognitive profiles of children with a wide range of neurodevelopmental disorders (i.e. both domain-general and domain-specific) including ADHD and ASD, but also dyslexia and selective language impairments, are associated with the connectedness of neural hubs as opposed to a one-to-one mapping between cognition and brain activation.¹⁰⁶ Children with comorbid dyslexia and dyscalculia can be distinguished from typically developing children and children with only dyslexia ‘or’ dyscalculia by structural aberrations within medial temporal lobe, and reductions in functional connectivity in circuits linking the medial temporal lobe to domain-specific regions critical for reading and math, respectively.¹¹² Relatedly, artificial neural networks can identify data-driven neurocognitive dimensions underlying learning difficulties that are unrelated to diagnoses but reflect distinct patterns of brain organization.¹¹³ Future research is needed to implement recently developed model-based approaches to uncover whether connectivity between core and particular periphery modules of brain networks account for differences between diagnostic groups of individuals with domain-specific neurodevelopmental disorders.

Reconceptualizing neurodevelopmental disorders through the lens of network neuroscience also provides insight into ‘paradoxical’ advantages exhibited by certain individuals with domain-specific developmental disorders (i.e. unexplained advantages in domains that are seemingly unrelated to the individual’s disorder),^110,114 broadening the perspective beyond disability to the proposed profile-based interpretation of strengths and weaknesses.

Network competition, such as a computational trade-off between encoding specific details of an individual experience and extracting regularities across experiences,^115,116 suggests that a deficit in one network may be accompanied by additional resource allocation to a distinct compensatory network. There are several examples of extreme strengths observed in individuals with domain-specific neurodevelopmental disorders. Within the mnemonic domain, individuals with SDAM and aphantasia display strengths in non-episodic processes (e.g. extracting meaning or regularities, enabling rapid and efficient acquisition of concepts with reduced interference from specific episodes) and are over-represented in high-level scientific professions.¹¹⁴ Within the academic domain, children with specific learning disorders sometimes exhibit paradoxical strengths in other areas of learning (e.g. individuals with dyslexia are gifted in other domains,¹¹⁰ (see also Eberi et al.¹¹⁷). The presence of extreme high and low abilities within the same individual is a consequence of competition and parallel processing between distributed systems that undergo network-level reorganization with the goal of optimization, given a deficit, within the brain’s core-periphery structure. Conceptualizing neurodevelopmental disorders under the guiding principles of network neuroscience and core-periphery organization enhances the potential for these ideas to support learners across the full spectrum of ability, rather than only those with neurodevelopmental disorders.

The identification of ‘learning styles’ (e.g. visual versus auditory learners¹¹⁸), an early profile-based approach in learning theory, has been dismissed as a ‘neuromyth’ due to a lack of scientific evidence.¹¹⁹ However, the notion that people have different profiles of cognitive strengths and weaknesses is—at least intuitively—self-evident. Indeed, it may be the intuitive appeal of learning styles—and the fact that they were originally identified in relation to focal lesion syndromes and not a neurodevelopmental framework—that led to an oversimplification of underlying neuro-anatomical mechanisms and overselling of related diagnostic and interventional products. A more nuanced anatomically-based approach to profiles would provide the basis for evidence-based diagnosis and intervention as well as contribution to theory across a wide range of abilities according to one’s unique cognitive profile.^113,120

Conclusion

Historically, neurodevelopmental disorders have been examined within, not across, cognitive domains. Synthesizing functional and structural connectivity findings across neurodevelopmental disorders suggests that distinct neurodevelopmental disorders are characterized by reductions in structural and functional connectivity between lower-level perceptual processing modules and higher-level control networks. The core-periphery network provides a useful framework for interpreting these findings. Specifically, seemingly distinct neurodevelopmental disorders share a dysfunction in long-range connections between the core and periphery of the human connectome. Implementing a network approach to identify a common origin of neurodevelopmental disorders across content domains may explain features such as comorbidity and paradoxical advantages within individuals in ways not evident from analyses focusing on surface-level differences in specific content domains. Such an approach would help to define a ‘connectome landscape of brain dysconnectivity’,¹⁷ valuable for the development and implementation of individualized prevention and intervention methodologies to support atypical learners across development.

Funding

This work was supported by operating grants from the Canadian Institutes of Health Research (CIHR) (Grant no. MOP-148940), the Social Sciences and Humanities Research Council of Canada (SSHRC) (Grant no. 430-2020-00215), as well as an SSHRC Banting Post-Doctoral Fellowship to H.M.S.

Competing interests

The authors report no competing interests.