Are more people becoming neurodivergent?
Sometimes a population appears to change before we know what, exactly, has changed.
Over the last few decades, autism has become far more frequently identified. In CDC surveillance of participating U.S. communities, roughly 1 in 150 eight-year-olds were identified with autism in 2000. Just two decades later, that figure was about 1 in 31. (CDC ADDM Network, 2000–2022)
ADHD diagnoses have also risen. More adults are being diagnosed later in life, while words such as autism spectrum, dyslexia and neurodivergence have moved far beyond clinics and classrooms into everyday language.
The numbers invite a simple story:
Are more people becoming neurodivergent?
But the word tying these observations together does not describe one biological condition.
Neurodivergent belongs to a broader framework of neurodiversity that developed collectively within neurodivergent communities rather than as a new clinical diagnosis. (Botha et al., 2024)
So before asking why the numbers changed, there is another question underneath them:
What are we actually grouping together?
An umbrella can be useful without describing one thing underneath it.
The language of neurodiversity did not begin as a new clinical category. It developed through neurodivergent communities as a way of describing variation in human minds and nervous systems. Over time, neurodivergent became a broader term that could include people diagnosed with autism, ADHD, dyslexia, Tourette syndrome and other neurological or developmental differences. (Botha et al., 2024)
But sharing an umbrella does not mean sharing one biological explanation.
Even closely related diagnoses can contain both overlap and difference. Large genetic studies of autism and ADHD, for example, find some genetic influences shared between them and others that distinguish them. Research including dyslexia finds another pattern of partial overlap: ADHD and dyslexia cluster strongly around attention and learning, while autism and ADHD also share broader neurodevelopmental genetic influences. (Mattheisen et al., 2022; Ciulkinyte et al., 2025)
The boundaries begin to look less like separate boxes and more like partially overlapping paths.
So perhaps the commonality is not one gene, one brain region or one type of sensory difference.
A different question begins to open:
Could different biological starting points sometimes produce recurring differences in how developing nervous systems receive, coordinate and organize information?
If so, the place to look for similarity may not be where the paths begin.
At what level do different developmental paths begin to look alike?
What changed besides the numbers?
A number is easiest to compare across time when the thing being counted stays the same.
Here, it didn’t.
Autism entered the DSM as a distinct diagnosis in 1980. By 1994, Asperger’s disorder and several related categories had been added. In 2013, many of those categories were folded into a single autism spectrum disorder, while autism and ADHD could now be diagnosed together. (Rosen et al., 2021; American Psychiatric Association, 2013)
The boundary around what could be identified—and recorded—had moved.
At the same time, identified autism became far more common. Across CDC surveillance sites, the recorded prevalence among eight-year-olds rose from 6.7 per 1,000 in 2000 to 32.2 per 1,000 in 2022. Those estimates also vary substantially between communities, where access to evaluation and identification practices differ. (CDC, 2025)
Some of the historical rise can be measured directly in those changes. In a large Danish cohort, a change in diagnostic criteria and the later inclusion of outpatient diagnoses together accounted for an estimated 60% of the recorded increase in autism prevalence within that population. (Hansen et al., 2015)
But the categories were not the only thing changing.
The environment in which attention develops and operates has changed too. Longitudinal studies have found that heavier digital-media use, and later problematic social-media use, can precede small increases in ADHD-like symptoms in adolescents. The effects are not large enough to explain ADHD as a whole, but they add another moving variable to the picture. (Ra et al., 2018; Nagata et al., 2026)
So a rising number can contain several histories at once:
changes in who is recognized, changes in what gets classified, and changes in the environment in which nervous systems are developing and behaving.
Before treating prevalence as evidence that one biological thing has increased, we need to separate those histories.
Where could different paths meet?
Different beginnings do not always lead to completely different systems.
Genes help build receptors, channels, signaling molecules and other cellular machinery. But those components do not act alone. Their effects depend partly on what other signals are present, when they arrive, and the state of the cells receiving them.
Ion channels give us one place to see this shift in scale.
By controlling the movement of charged particles across cell membranes, ion channels help determine how easily neurons fire, how calcium signals move through cells, and how developing circuits communicate and organize. Alterations in ion-channel function have been implicated across several neurodevelopmental disorders. (Shi et al., 2026)
But an ion channel is still only one component.
Michael Levin’s work on developmental bioelectricity looks at what happens when electrical states are considered across groups of cells. Ion channels help create membrane voltage; gap junctions allow neighboring cells to become electrically coupled; together, populations of cells can form larger patterns of electrical activity that participate in regulating gene expression, growth and anatomical development. (Levin, 2021)
The distinction matters:
an ion channel is not a bioelectric pattern.
The pattern belongs to the relationships among many cells.
Recent work with Anthrobots pushes this idea further. Adult human airway cells, without genetic editing, can reorganize into motile multicellular structures with new morphology and behavior. In that new collective state, thousands of transcripts are expressed differently from the cells they came from. (Gumuskaya et al., 2025)
The genetic material did not have to become different for the system to begin behaving differently.
This gives us another way to think about neurodevelopment.
Different genetic or molecular changes may begin in different places while still entering systems that regulate cellular signaling, development and eventually neural organization.
So perhaps the useful question is not only which component changed?
It is also:
At what point does a change in one component become a change in the organization of the larger system?
When timing becomes part of the signal
Not every difference in a system is about what is present.
Sometimes it is about when.
Seeing a mouth move and hearing a voice gives the nervous system two different streams of information. They do not have to arrive at exactly the same moment for us to experience them as one event. Somewhere in that process, the system allows a window in which separate signals can still belong together.
That window appears to vary.
A 2020 review and meta-analysis found measurable differences in temporal processing in both autism and developmental dyslexia, including judgments about whether events happened at the same time or in a particular order. Multisensory temporal processing was affected in both groups. (Meilleur et al., 2020)
In a separate line of research, Julia Mossbridge and colleagues measured how accurately adults could detect whether auditory and visual events occurred together. Greater sensitivity to audiovisual synchrony was associated with better reading comprehension in both behavioral testing and EEG measurements. (Mossbridge et al., 2017)
The studies meet around a broader problem:
How does a nervous system decide which signals belong to the same event?
One way researchers formalize that problem is through Bayesian causal inference. The system has to work with signals that differ not only in timing, but in reliability and uncertainty, while estimating whether they were likely produced by one source or several. (Jones & Noppeney, 2024)
Timing therefore does more than place events in sequence.
It can help determine which information is allowed to interact in the first place.
Change that relationship enough, and the system may begin grouping the same incoming information differently.
How far can timing or reliability shift before a different pattern begins to form?
When the pattern belongs to the network
A pattern can belong to a system without belonging to any one of its parts.
Once signals begin interacting, their effect depends not only on which neurons are active, but on the relationships between them: which cells are connected, how strongly they influence one another, and what activity is already moving through the network.
Change those relationships, and the same incoming information can lead somewhere different.
A Hopfield network gives us a simplified mathematical way to see this.
In the original model, information is stored across the strengths of connections among many interacting units. Give the network part of a familiar pattern, and those relationships can guide its activity toward the larger stored pattern. No individual unit contains the memory by itself; the network’s collective organization makes retrieval possible. (Hopfield, 1982)
That changes where we might look for similarity.
A difference in sensory timing could change which neurons repeatedly become active together. A difference in cellular excitability could change how strongly some signals influence the rest of a circuit. Developmental changes could alter which connections form in the first place.
Each difference begins somewhere else.
But each can eventually change the relationships through which a larger pattern forms.
We explored this kind of distributed information more closely in Sensations of Smell, where odor identity is represented across populations of neurons and a simplified Hopfield model shows how a familiar pattern can be recovered from incomplete input.
Here, it leaves us with another possibility:
Could different developmental paths begin to look alike when they create similar relationships at the level of the network?
Where does the similarity begin?
We began by asking whether more people are becoming neurodivergent.
That question is harder to answer than the numbers first suggest.
The categories have changed. Recognition has changed. The environments in which attention and learning develop have changed. And the conditions grouped beneath neurodivergence do not point back to one shared biological origin.
But different origins do not necessarily remain different at every level.
Changes that begin in genes, signaling pathways, ion channels or development can enter larger systems. Timing can change which signals are grouped together. Network relationships can change what patterns those signals are able to form.
The commonality, if there is one, may therefore not sit at the beginning of the path.
It may appear later—in the way information is coordinated, constrained and organized across a developing system.
That leaves us with a different question than the one we started with:
At what level do divergent biological paths begin producing similar patterns?
Not one gene.
Not necessarily one brain region.
Perhaps the more useful place to look is where the parts begin acting together.
Show the source trail Follow the evidence Identification → development → timing → collective state
The evidence behind this piece does not all sit at the same level. Some sources track changing diagnostic language and identified prevalence. Others measure genetics, sensory timing, cellular signaling, or collective behavior in biological systems. Mathematical models then give us a way to ask what can happen when many uncertain or interacting parts are organized into a larger system.
This source trail keeps those levels separate: what has been measured in neurodevelopment, what biological mechanisms are established, what provides a useful parallel, and what remains an open connection.
Terminology & identification
Sources that ground what the umbrella means, how clinical categories changed, and what rising identification rates can and cannot tell us.
Context · History of the concept Botha et al., 2024 Neurodiversity emerged collectively from autistic communities
What it supports
The authors trace the neurodiversity concept to multiple contributors within autistic communities in the 1990s rather than to a single originator. Their historical review separates the broader neurodiversity framework from the histories of the clinical diagnoses discussed in this article.
What it doesn’t establish
This historical account does not define one fixed modern boundary for who should or should not be described as neurodivergent, nor does it make the umbrella equivalent to a single clinical diagnosis.
Why it matters here
The article begins by asking what we are actually grouping together. Understanding that the umbrella and the diagnoses beneath it have different histories prevents us from treating them as one biological category from the start.
Botha M, Chapman R, Giwa Onaiwu M, Kapp SK, Stannard Ashley A, Walker N. (2024). The neurodiversity concept was developed collectively: An overdue correction on the origins of neurodiversity theory. View source ↗
Human evidence · Longitudinal cohort Nagata et al., 2026 Problematic social-media use and later ADHD symptoms
What it supports
In a large U.S. adolescent cohort followed across five annual study waves, years in which young people reported more problematic social-media use than usual were followed by small increases in parent-reported ADHD symptoms during part of mid-adolescence. The association was most evident among males.
What it doesn’t establish
The study does not show that social media causes ADHD or that digital media explains rising ADHD diagnoses. The observed within-person effects were small, appeared during limited developmental periods, and varied by sex and reporter.
Why it matters here
Diagnostic categories are not the only variables changing over time. The environments in which attention and self-regulation operate are changing too. This study gives us longitudinal evidence that aspects of that environment can move alongside attentional symptoms without reducing ADHD to an environmental explanation.
Nagata JM, Otmar CD, Nayak S, et al. (2026). Problematic Social Media Use and Attention-Deficit/Hyperactivity Disorder Symptoms in Adolescents. JAMA Network Open, 9(7), e2623748. View source ↗
Context · Diagnostic history Rosen, Lord & Volkmar, 2021 Autism’s diagnostic category changed repeatedly across DSM editions
What it supports
This historical review follows autism from its first official inclusion in DSM-III in 1980 through later revisions, including the expansion of diagnostic categories in the 1990s and their consolidation into autism spectrum disorder in DSM-5.
What it doesn’t establish
A changing diagnostic definition does not by itself tell us how much of an observed prevalence increase reflects classification, recognition, biology, environmental influence, or some combination of these.
Why it matters here
Historical diagnosis counts are not measurements made with one unchanging ruler. The category being counted has itself changed over time.
Rosen NE, Lord C, Volkmar FR. (2021). The Diagnosis of Autism: From Kanner to DSM-III to DSM-5 and Beyond. View source ↗
Direct evidence · Population surveillance CDC ADDM Network, 2022 surveillance Identified autism prevalence rose from about 1 in 150 to about 1 in 31
What it supports
Across CDC Autism and Developmental Disabilities Monitoring Network sites, identified autism prevalence among 8-year-old children was about 1 in 150 in 2000 and 32.2 per 1,000, or about 1 in 31, in 2022.
What it doesn’t establish
ADDM measures identified prevalence in participating communities. The 2022 estimate is not a census of every child in the United States, and the rise in identification is not by itself a measurement of an equivalent rise in underlying biological prevalence.
Why it matters here
It establishes the observation that opens the article: the number of children being identified with autism has changed dramatically over the surveillance period.
Shaw KA, et al. (2025), reporting ADDM Network surveillance data from 2022. Prevalence and Early Identification of Autism Spectrum Disorder Among Children Aged 4 and 8 Years — Autism and Developmental Disabilities Monitoring Network, 16 Sites, United States, 2022. View source ↗
Direct evidence · Diagnostic epidemiology Hansen, Schendel & Parner, 2015 Reporting changes explained a substantial part of one historical autism increase
What it supports
In a Danish birth cohort, the authors estimated that a 1994 diagnostic change together with the 1995 inclusion of outpatient diagnoses could explain about 60% of the observed increase in reported autism spectrum disorder prevalence for children born from 1980 through 1991.
What it doesn’t establish
The 60% estimate belongs to this Danish cohort, these reporting changes, and this historical period. It should not be generalized as an explanation for 60% of autism prevalence increases everywhere.
Why it matters here
It demonstrates that changes in how a condition is defined and recorded can produce large changes in measured prevalence without requiring an equally large change in underlying biology.
Hansen SN, Schendel DE, Parner ET. (2015). Explaining the Increase in the Prevalence of Autism Spectrum Disorders: The Proportion Attributable to Changes in Reporting Practices. View source ↗
Human evidence · Longitudinal association Ra et al., 2018 Frequent digital-media use was associated with later ADHD symptoms in adolescents
What it supports
In adolescents without significant ADHD symptoms at baseline, more frequent use of modern digital-media activities was modestly associated with a greater likelihood of subsequent ADHD symptoms across a two-year follow-up.
What it doesn’t establish
The study does not show that digital media causes ADHD, that screen exposure explains rising ADHD diagnoses, or that the association runs in only one direction.
Why it matters here
Diagnostic systems are not the only thing that changed. The attentional environment also changed, giving us a reason to keep environmental influences in the question without treating them as a single explanation.
Ra CK, Cho J, Stone MD, et al. (2018). Association of Digital Media Use With Subsequent Symptoms of Attention-Deficit/Hyperactivity Disorder Among Adolescents. View source ↗
Different origins, overlapping biology
Evidence that neurodevelopmental conditions can have distinct biological starting points while still intersecting in shared genes, proteins, and cellular processes.
Direct evidence · Cross-disorder genetics Mattheisen et al., 2022 Shared and distinct genetic architecture in autism and ADHD
What it supports
Large cross-disorder genetic analyses identified both overlap and separation between autism spectrum disorder and ADHD, including seven loci shared by the disorders and five loci that differentiated them in this analysis.
What it doesn’t establish
Shared genetic architecture does not imply one autism/ADHD gene, one biological pathway, or one underlying condition. Genetic overlap can coexist with substantial biological heterogeneity.
Why it matters here
It gives us a concrete example of commonality without sameness: two diagnostic categories can share some biological influences while remaining distinguishable.
Mattheisen, et al. (2022). Identification of shared and differentiating genetic architecture for autism spectrum disorder, attention-deficit hyperactivity disorder and case subgroups. View source ↗
Human genomic evidence · Cross-disorder genetics Ciulkinyte et al., 2025 Dyslexia, ADHD and overlapping genetic structure
What it supports
Using genome-wide association data across dyslexia and several psychiatric and neurodevelopmental traits, the researchers found partially overlapping genetic structure rather than one shared biological category. ADHD showed genetic overlap with both an autism-linked neurodevelopmental factor and a separate attention-and-learning factor containing dyslexia.
What it doesn’t establish
Shared genetic influences do not mean that dyslexia, ADHD and autism arise from the same genes or developmental pathway. Genetic overlap also cannot tell us by itself which later biological processes produce similar cognitive or behavioral patterns.
Why it matters here
It gives us a concrete example of overlap without equivalence. A condition such as ADHD can share genetic influences with multiple developmental traits while remaining biologically heterogeneous—supporting the question of whether commonality might appear at more than one level.
Ciulkinyte A, Mountford HS, Fontanillas P, et al. (2025). Genetic neurodevelopmental clustering and dyslexia. Molecular Psychiatry, 30, 140–150. View source ↗
Mechanism review · Ion-channel biology Shi, Li & Jin, 2026 Ion channels link molecular variation to excitability, signaling, and circuit development
What it supports
This review describes the roles of voltage-gated and ligand-gated ion channels in neuronal excitability, synaptic transmission, and coordinated brain-circuit development, and surveys channel dysregulation across multiple neurodevelopmental disorders.
What it doesn’t establish
Ion-channel involvement does not identify one shared channel abnormality across neurodivergent people, nor does it show that ion-channel differences explain every feature of autism, ADHD, dyslexia, or other conditions.
Why it matters here
Ion channels give us one route by which a molecular difference can alter a larger signaling system: changing how cells respond, communicate, and participate in developing neural circuits.
Shi Z, Li S, Jin X. (2026). Ion Channel Dysregulation in Neurodevelopmental Disorders: Mechanisms, Models, and Therapeutic Advances. View source ↗
Development, state & collective organization
Experimental and mechanistic work showing that biological outcomes can depend on physiological state and collective organization, not only on the identity of individual components.
Mechanism review · Developmental bioelectricity Levin, 2021 Ion channels and gap junctions can form tissue-level bioelectric networks
What it supports
Endogenous membrane-potential patterns produced by ion channels and gap junctions can participate in coordinating gene expression, growth, morphogenesis, development, and regeneration across groups of cells.
What it doesn’t establish
This work does not show that a particular tissue-level bioelectric pattern causes autism, ADHD, dyslexia, or neurodivergence as a whole.
Why it matters here
It provides an experimentally grounded example of biological information existing at a collective physiological level: individual channels help create the state, but the larger pattern is a property of the interacting cell network.
Levin M. (2021). Bioelectric signaling: Reprogrammable circuits underlying embryogenesis, regeneration, and cancer. View source ↗
Parallel experimental evidence · Pattern state Durant et al., 2017 A temporary bioelectric perturbation changed later regenerative outcomes
What it supports
Brief modulation of endogenous bioelectric signaling in regenerating planaria produced a persistent alteration in regenerative patterning. Some worms later appeared anatomically normal yet regenerated differently after being cut again, revealing a hidden change in patterning state.
What it doesn’t establish
Planarian regeneration is not human neurodevelopment, and this experiment does not show that the same bioelectric mechanism explains neurodivergent traits.
Why it matters here
It demonstrates that present anatomy and genomic sequence do not necessarily exhaust the state of a biological system. A change in physiological organization can alter what the same tissue does later.
Durant F, Morokuma J, Fields C, Williams K, Adams DS, Levin M. (2017). Long-Term, Stochastic Editing of Regenerative Anatomy via Targeting Endogenous Bioelectric Gradients. View source ↗
Parallel experimental evidence · Collective state Gumuskaya et al., 2025 Wild-type adult human cells entered a radically different multicellular state
What it supports
Anthrobots self-assemble from genetically wild-type adult human airway epithelial cells without genomic editing. During that reorganization they acquire new morphology, motility, and a distinct transcriptomic state. In the study’s RNA-seq comparison, 8,992 of 22,518 measured transcripts showed significant up- or downregulation relative to the progenitor cells.
What it doesn’t establish
Anthrobots are engineered multicellular constructs, not developing human nervous systems. Their gene-expression changes do not provide evidence for an Anthrobot-like process in neurodivergence.
Why it matters here
The genetic starting material remained human and wild-type while the collective organization changed dramatically. It is a striking experimental example of the same components participating in a different system state and expressing a very different biological program.
Gumuskaya G, et al. (2025). The Morphological, Behavioral, and Transcriptomic Life Cycle of Anthrobots. View source ↗
Timing, integration & network organization
Human measurements and formal models that ask how nervous systems decide which signals belong together and how relationships across a network can shape the pattern that forms.
Direct evidence · Systematic review & meta-analysis Meilleur et al., 2020 Temporal-processing differences appear in autism and developmental dyslexia
What it supports
Across the studies included in this review and meta-analysis, both autism and developmental dyslexia showed measurable differences in temporal processing, including temporal-order and simultaneity judgments. Multisensory temporal processing differences were found in both groups.
What it doesn’t establish
The findings do not show that autism and dyslexia share one timing mechanism, that every person in either group has the same temporal-processing profile, or that timing differences are the cause of either diagnosis.
Why it matters here
It gives us direct evidence that timing is one measurable level at which two different neurodevelopmental conditions can show partially overlapping patterns.
Meilleur A, Foster NEV, Coll SM, Brambati SM, Hyde KL. (2020). Unisensory and multisensory temporal processing in autism and dyslexia: A systematic review and meta-analysis. View source ↗
Parallel human evidence · Audiovisual timing Mossbridge et al., 2017 Audiovisual synchrony sensitivity was associated with reading comprehension
What it supports
In separate behavioral and electrophysiological experiments, greater sensitivity to auditory-visual synchrony predicted better reading comprehension. The reported measures accounted for 16% and 25% of variance in reading comprehension in the two experiments.
What it doesn’t establish
The study was not a test of autism or ADHD, and the association does not show that audiovisual timing causes reading ability or that one timing process explains dyslexia.
Why it matters here
It independently measures the problem we are interested in: how precisely a nervous system detects whether signals from different senses belong to the same moment or event.
Mossbridge J, Zweig J, Grabowecky M, Suzuki S. (2017). An Association between Auditory-Visual Synchrony Processing and Reading Comprehension: Behavioral and Electrophysiological Evidence. View source ↗
Computational framework · Bayesian causal inference Jones & Noppeney, 2024 Multisensory integration can be modeled as inference about common causes
What it supports
This review presents multisensory integration within a Bayesian framework in which a system must estimate whether uncertain sensory signals share a common cause and weight them according to reliability. It also examines whether atypical populations differ in parameters such as sensory noise or perceptual priors, or in broader computational principles.
What it doesn’t establish
Bayesian causal inference is a computational framework, not proof that every neural system literally implements the equations in one fixed way. It also does not identify one Bayesian parameter that defines neurodivergence.
Why it matters here
It turns our timing question into a formal problem: given signals that differ in timing, uncertainty, and reliability, how strongly should the system infer that they came from the same event?
Jones SA, Noppeney U. (2024). Multisensory Integration and Causal Inference in Typical and Atypical Populations. View source ↗
Mathematical model · Collective network behavior Hopfield, 1982 A network can exhibit memory as a collective property of its interactions
What it supports
Hopfield showed mathematically how a network of many simple interacting units can exhibit content-addressable memory as a collective property. Under the model, a sufficiently complete part of a stored pattern can guide the network toward the larger stored state.
What it doesn’t establish
A Hopfield network is not a model of neurodivergence and is not a literal description of the neural circuits discussed in this article.
Why it matters here
It provides a formal example in which the important system-level capability belongs to relationships among components rather than to any one component considered alone.
Hopfield JJ. (1982). Neural networks and physical systems with emergent collective computational abilities. View source ↗
Where the trail ends
Human research gives us evidence for heterogeneous genetics, ion-channel involvement, differences in temporal processing, and differences in how multisensory information is integrated.
Developmental biology independently shows that the same genetic cellular material can occupy very different physiological and collective states, and that changing those states can alter what a biological system does later. Network models show mathematically how information-processing properties can arise from relationships among components rather than from an individual component alone.
What we do not yet have is evidence that these findings belong to one causal chain.
We do not know whether the developmental mechanisms demonstrated in planaria or Anthrobots contribute to the sensory-timing differences measured in autism or dyslexia. We also do not know whether different neurodevelopmental routes converge on a common network state.
What the evidence supports is the question:
When different biological starting points produce overlapping cognitive patterns, at what level does the commonality emerge?
Leave a Reply