A recent neuroimaging study (pdf) published in Molecular Psychiatry (2025) explored whether autism can be meaningfully divided into biological subtypes based on brain connectivity patterns. Using resting-state fMRI data from over 1000 participants (479 with autism, 567 without), the researchers identified two distinct “brain subtypes” of autism. Both subtypes had similar behavioral and diagnostic scores but showed opposite patterns of functional brain connectivity. These differences were also reflected in eye-tracking tasks focused on facial expressions and joint attention. The authors conclude that personalized care for autistic individuals should consider these neural subtypes rather than rely solely on traditional diagnostic categories
Some critical reflections
The study is impressive — methodologically sophisticated, data-rich, and careful in tone. But it also raises fundamental questions about how we conceptualize autism and who gets to define what it means.
1. Complex Technology, Simple Framing
Let’s start with the merits. The researchers use a cutting-edge statistical method known as normative modeling, which evaluates how much an individual’s brain function deviates from what is considered “typical.” They combine this with both static and dynamic connectivity analyses — and even eye-tracking data. The result: a nuanced classification that appears more precise than the traditional one-size-fits-all autism label.
But that very precision is deceptive. What we end up with is not a person-centered framework, but a system-centered calibration. Instead of asking, How do you experience the world?, the question becomes, How far do you deviate from the average brain?
2. From Heterogeneity to Reclassification
The paper rightly acknowledges that autism is heterogeneous. That’s good. But its response is to group people back into new categories — based on neurobiology this time. Two subtypes: one with higher-than-average connectivity in some regions (like the cerebellum and visual areas), and one with lower-than-average connectivity in those same networks — and vice versa in others.
So we’re replacing one overgeneralization with another — only now the defining feature is connectivity profiles, not lived experience. It’s as if our understanding of autism is being pushed deeper into the brain while moving further away from the individual.
3. What Are We Actually Measuring?
Here’s the crux: the people in both subtypes had similar diagnostic profiles. They scored similarly on the ADOS, SRS, and other behavioral tests. The only thing that separated them was how their brain networks connected — and how they moved their eyes during social tasks.
But is that enough to call them distinct “types”? Are we mapping meaningful differences or simply producing a new way to sort statistical noise? And are these classifications clinically helpful, or just scientifically elegant?
Eye-tracking, for example, is treated as a quasi-objective measure of “social attention.” Yet autistic people often learn to perform eye contact or gaze-following in social contexts — or avoid it for self-preservation. Are we measuring innate traits or learned adaptations? And do we risk medicalizing preference and coping?
4. The New “Precision” Mask
This research claims to move toward personalized care. But ironically, it risks replacing individualized understanding with a new kind of standardization: one brain-type fits this treatment, another that one. It’s medicine by algorithm, not by conversation.
Personalization here still starts from deviation from a normative model — not from a shared exploration of strengths, needs, and context. The individual is not the primary interpreter of their own experience, but a datapoint in a functional clustering map.
5. Subtypes or Subtexts?
What if the most important result of this study is not the existence of two brain types, but the confirmation that people with autism can be neurologically very different, even if they behave similarly?
That’s not a call to reclassify people. It’s a call to deepen our understanding of diversity — not just neural, but experiential, relational, contextual. This study’s findings should not lead to more segmentation, but to more humility: the idea that no single approach can fully grasp autism.
Conclusion: Don’t Let the Brain Replace the Person
This study is valuable. It shows that brain-based differences within autism are real and complex. But let’s not turn these insights into new boxes. Let’s use them to ask better questions — not just about what the brain does, but how the person feels, navigates, and gives meaning to their experience.
Because at the end of the day, an fMRI can’t tell you how it feels to stim under stress, to rehearse every sentence before a meeting, or to fight for your right to be different in a world that demands sameness.
Only the person can.
Liu, Q., Lai, H., Le, J. et al. Identifying brain functional subtypes and corresponding task performance profiles in autism spectrum disorder. Mol Psychiatry (2025). https://doi.org/10.1038/s41380-025-03086-x
The European Autist 