Fact Check: Is Autism a Byproduct of Human Cognitive Evolution?

Introduction

Autism spectrum disorder (ASD) is widely recognised as a neurodevelopmental disorder that results from the intricate interaction of several genes, environmental variables, and developmental processes. Although it is controversial, new research indicates that the theory that autism is brought on by changes in the human brain may merit careful consideration. According to a 2025 study by Starr et al. (published in Molecular Biology and Evolution), some neuronal types in the human cortex underwent unusually fast evolution, and these changes might have been caused by genes associated with autism. According to the authors, evolutionary forces that influenced human cognition might have unintentionally made people more susceptible to ASD. There are some limitations to the hypothesis. We need to carefully consider the data from comparative neuroscience, genomics, epidemiology, and developmental biology to see how plausible this hypothesis is, where its limits lie, and what predictions it generates.

Rising Autism Diagnoses: Real Increase or Better Detection?

Autism diagnoses have increased rapidly in recent decades in many parts of the world. According to the U.S. CDC, about 1 in 31 (≈ 3.2 %) children aged 8 have been identified with ASD. Globally, the World Health Organization estimates about 1 in 100 children are classified with ASD, though many countries lack robust data. However, an increase in reported prevalence does not necessarily mean a true biological increase. Many researchers argue that changes in diagnostic criteria, screening practices, awareness, and services play a major role. That said, there is ongoing debate about whether there is a genuine increase in incidence i.e., more children born with neurodevelopmental risk.

Hence, the evolutionary hypothesis must contend with the fact that much of the observed increase might be artifactual. Nonetheless, even if prevalence has not increased dramatically biologically, autism is still far more common in human populations than in other species, which raises the question: why do humans have such high neurodiversity risk?

The 2025 Study: What They Did and What They Found

Overview of Methods

Starr et al.’s work focuses on single-nucleus (or single-cell) RNA sequencing data from multiple mammalian species, comparing gene expression profiles of defined neuronal cell types across species (including humans, chimpanzees, and other primates). The logic is, if one can identify homologous neuronal types, one can ask how quickly gene expression patterns have shifted in each lineage, and whether those shifts implicate genes associated with neurodevelopmental disorders like autism.

The key cell type they focus on is layer 2/3 intratelencephalic excitatory neurons (L2/3 IT neurons) in the human neocortex. These neurons are abundant, connect across cortical areas, and are thought to play a major role in higher cognition. The surprising result was that these neurons appear to have evolved more rapidly in gene expression in humans than in other apes in contrast to what is expected under a more abundant cell types evolve more slowly rule. Further, the authors examined whether those gene expression shifts disproportionately affected autism‐associated genes. They found that many genes linked with high autism risk show down‐regulation in humans relative to chimpanzees in these neurons, and allele-specific expression analyses support the notion that these regulatory changes were driven by positive selection in the human lineage rather than genetic drift or neutral processes. In other words, the authors propose that evolving Homo sapiens lineages actively selected versions of regulatory elements that lowered baseline expression of autism‐linked genes in these neurons, perhaps because doing so conferred a fitness advantage for human‐specific cognitive traits.

The principle the authors propose is: “more abundant neuronal subtypes tend to change more slowly across species” a general evolutionary constraint. But L2/3 IT neurons behave as an outlier. They are abundant yet have accelerated evolution in humans. The shifts in expression in humans were strongly enriched for autism‐linked genes.

For example, they note DLG4 (encoding PSD-95) — an important synaptic scaffold protein — shows significantly lower expression in human L2/3 IT neurons compared to chimpanzees. About 2.5× lower in humans. Because of this lower baseline, further reduction might push expression below a functional threshold, making ASD more likely in humans than in non-humans.

In brief, the human lineage may have pushed gene expression in key neurons toward a state that fosters high cognitive capacity, but at the cost of making those neurons more vulnerable to developmental perturbations that manifest as ASD.

Critical Evaluation & Alternative Explanations

This evolutionary hypothesis is elegant and rich, but it also has uncertainties and competing models. Below I examine potential strengths, limitations, counterpoints, and what data would strengthen or weaken the claim.

Strengths of the Hypothesis

1. Biological Plausibility and Parsimony
   The idea of trade-offs in evolution is well established. In complex brains, small perturbations may have outsized effects; in a species with enhanced cognitive and linguistic demands, regulatory fine-tuning may be more precarious.
2. Cell-Type Specificity
   Many genetic/neurodevelopmental studies of autism complain about heterogeneity. Which neurons, which circuits? This study focuses on a specific neuronal class (L2/3 IT) that is plausibly linked to integrative cortical computation. That cell‐type specificity strengthens the mechanistic plausibility compared to more diffuse claims.
3. Comparative and Regulatory Angle
   By leveraging cross-species single-cell transcriptomics and allele-specific regulatory analyses, the authors avoid purely speculative narratives. The data provide a path to connect evolutionary shifts to disease‐relevant genes.
4. Consistency with Other Work on Human-Accelerated Regions
   Previous studies have identified human accelerated regions (HARs) in the genome. Segments with rapid sequence change in Homo sapiens that are enriched in regulatory, noncoding regions and sometimes linked to cognition and brain function. Some of those have been speculated to be involved in neuropsychiatric risk. The new neuronal-level results align with this broader theme of human-specific regulatory rewiring.

Limitations, Caveats, and Objections

1. Causation vs Correlation
   The fact that autism‐associated genes are differentially regulated in human vs chimp neurons does not prove that these differences cause autism vulnerability. It could be that many regulatory shifts occurred for other reasons and autism risk is some collateral effect—or even a statistical byproduct without adaptive significance.
2. Uncertain Selective Advantage
   What exactly would favor lower baseline expression of autism-linked genes? The authors propose slowed development or network stabilization, but these remain speculative. Without additional functional or phenotypic links, the selective story is weaker.
3. Relative Effect Sizes and Buffer Capacity
   The hypothesis implies that humans have less buffer capacity — less margin before deleterious effects in some neuronal systems. But that requires detailed quantitative modeling. How much expression noise or allelic loss can neurons tolerate? Are humans truly closer to failure thresholds than chimps? That remains unproven.
4. Confounding by Environment and Non-Genetic Factors
   Autism risk is influenced by environmental and epigenetic factors (maternal health, exposures, prenatal insults) in many models. The evolutionary hypothesis must be compatible with these. It neither requires them nor excludes them. But if non-genetic factors dominate risk variance, the evolutionary baseline vulnerability is a smaller piece of the puzzle.
5. Other Brain Regions and Cell Types
   The study focuses on one neuronal class in the neocortex. But autism likely involves complex circuits such as inhibitory neurons, glia, subcortical structures, connectivity, and synaptic plasticity across many regions. If the vulnerability derives from multiple cell types, this one pathway is incomplete.

Predictions and Falsifiable Implications

A good hypothesis should make predictions. Here are several:

1. Within-Human Variation: The regulatory alleles that reduce expression of ASD‐linked genes (predicted by the model) should be under detectable selective sweeps, in genomic signatures, or show clines in populations.
2. Resilience in Non-Humans: In non-human primates, introducing (via synthetic biology / organoids) human regulatory variants should show more susceptibility to perturbations in neuronal development compared to chimp or macaque baselines.
3. Correlation with Developmental Delay and Plasticity Traits: If one of the selective drivers was prolonged brain development, correlation analyses should show that ASD‐linked regulatory shifts track with genes known to affect developmental timing across species.
4. Comparative Prevalence: In rare cases of naturally occurring neurodevelopmental disturbances in non-human primates (if they occur), the rate should be dramatically lower or the phenotypes qualitatively different, consistent with greater buffer capacity.

If future data do not support these predictions, the evolutionary hypothesis will weaken.

Integrating with Other Models & Implications

The evolutionary view should not be seen as mutually exclusive with other models of autism risk. Rather, it may serve as a framework or substrate on which other risk factors act. Here is how we see the integration and implications.

Complementarity with Genetic and Environmental Models

1. Polygenic Risk Architecture
   Modern autism genetics shows that many genetic variants of small effect, plus rare variants of large effect, converge to modulate neurodevelopmental trajectories. The evolutionary hypothesis gives a contextual baseline: humans may have shifted regulatory baselines such that small perturbations carry more impact.
2. Gene × Environment Interactions
   Because humans may operate closer to thresholds, environmental insults (e.g. in utero exposures, maternal inflammation, toxic exposures) might push development over threshold more easily. Thus, the evolutionary background and environmental risk are not in competition but interact.
3. Epigenetics and Plasticity
   The evolutionary hypothesis centers on regulatory gene expression changes. Epigenetic modifications, chromatin state dynamics, and developmental plasticity will play key roles in mediating how much vulnerability manifests. These layers add complexity, but do not contradict the core idea.

Broader Implications

1. Reframing Autism
   If autism vulnerability is partially a byproduct of human brain evolution, it suggests that neurodiversity has deeper roots. The traits associated with ASD might have had adaptive or neutral roles in human history. That does not minimize the challenges, but encourages thinking of autism less strictly as a disease and more as a spectrum with trade-offs.
2. Therapeutic Approaches
   Understanding which regulatory pathways were shifted in human evolution might highlight which systems are especially fragile or dosage-sensitive. That could guide gene therapy, small molecules, or buffering strategies to restore stability.
3. Ethical and Social Perspectives
   If vulnerability to neurodevelopmental disorders is linked to what makes us uniquely human, then we should approach autism not just in clinical terms but as part of our species’ evolutionary heritage. That may encourage more acceptance, investment in support, and less stigma.

Potential Criticisms and Unanswered Questions

While promising, this hypothesis also raises challenging questions. Such as:

• Why Down-Regulation Rather Than Up-Regulation?
  It is interesting that many autism‐linked genes are down-regulated in humans versus non-humans. Why would evolution push lower expression? One might have expected up-regulation or novel paralogs. The authors propose network stabilization or plasticity reasons, but more empirical support is needed.
• Is Neurodiversity Vulnerability an Adaptation or Side Effect?
  Did evolution intend to ratchet toward vulnerability? Or is it simply collateral damage? That distinction matters for interpreting whether autism susceptibility has adaptive significance.
• Scope Beyond One Neuron Type
  Autism likely involves many cell types (inhibitory neurons, astrocytes, microglia) and circuits (long-distance connectivity). The hypothesis must eventually expand to or integrate additional neuronal classes and circuits.
• Quantitative Modeling
  The argument is qualitative (lower baseline, lower buffer). But to gain confidence, one needs mathematical models: how much expression change, margin of error, allelic variation, noise tolerance, perturbation distribution. Without that, the hypothesis remains plausible but untested in magnitude.
• Cross-Human Population Variation
  If baseline vulnerability is evolutionarily baked, one might expect limited variation across human populations. But we see variation in prevalence and penetrance. Genetic background modifiers might explain that, but it’s a complexity the hypothesis must contend with.
• Empirical Tests in Development
  The work is based on adult post-mortem or mature brain expression. But autism is a developmental disorder. How do these regulatory shifts manifest during critical prenatal, perinatal, and early postnatal windows? If differences arise only in mature states, their developmental relevance is uncertain.
• Alternative Evolutionary Models
  Other researchers might argue that autism vulnerability is not a byproduct but an unintended consequence of pleiotropy, gene drift, sexual selection, or epistasis. The authors’ model is one among many; it must be compared to alternatives.

A Tentative Narrative

Putting the pieces together, we can sketch a narrative:

1. In human lineage, evolving cognition, social complexity, language, and extended development imposed demands on cortical circuits, especially in integrative layers like L2/3 IT.
2. Natural selection fine-tuned regulatory networks in those neurons, favoring shifts that enhanced plasticity, connectivity, or stability under greater computational load.
3. Some of those shifts involved lowering baseline expression of certain synaptic or developmental genes — perhaps to maintain homeostasis in large, dense circuits.
4. Because human neurons now operate closer to the edge of perturbation tolerance, even minor allelic variation or developmental insults can push trajectories toward autism i.e. the ASD expression threshold is closer.
5. Over time, this background vulnerability sets the stage where genetic risk factors (common variants, rare variants) or environmental insults can more easily tip individuals into ASD phenotypes.
6. The rise in diagnosed autism over recent decades is mainly due to better detection, broader definitions, environmental changes, etc. but the baseline vulnerability is a deep, evolutionary legacy.

Conclusion & Next Steps

The hypothesis that human brain evolution helped create a baseline predisposition to autism is intellectually compelling. It elegantly links comparative genomics, neuronal cell type biology, and human cognitive evolution with disease vulnerability. But the gap between elegance and empirical proof is still significant.

To advance this idea, the following could help:

1. Developmental single-cell or spatial transcriptomic work in human and non-human primate fetal and early postnatal brain to examine when regulatory shifts emerge.
2. Quantitative modeling of expression buffering, noise, and threshold dynamics across species.
3. Functional tests (e.g. CRISPR perturbations, organoids, cross-species chimeric models) to test whether human regulatory variants increase sensitivity to perturbation.
4. Population genomics to search for signs of selection on regulatory alleles of autism-linked genes in humans.
5. Broader integration of other implicated neuronal types and circuits in ASD.

Until then, the evolutionary hypothesis should be viewed as a promising, but tentative, framework complementary to genetic, environmental, and developmental models of autism.