"Pero existe algo que el tiempo no puede, a pesar de su innegable capacidad destructora, anular: y son los buenos recuerdos, los rostros del pasado, las horas en que uno ha sido feliz".
- Julio Cortázar
Autism has been treated as a single condition for decades. A major international study published in Nature Neuroscience in May 2026 found evidence that it may be at least two biologically distinct conditions that look the same from the outside.
Autism spectrum disorder is diagnosed in roughly 1 in 36 children in the United States. Despite decades of research, there is still no reliable biological marker no blood test, no brain scan that confirms the diagnosis. And despite the word spectrum acknowledging the enormous range of presentations, treatment has largely been one-size-fits-all.
This study suggests one reason that may be: what we call autism may include fundamentally different brain biologies grouped under a single label.
An international team led by the Istituto Italiano di Tecnologia and the Child Mind Institute in New York analysed brain imaging from 940 children and young adults with autism, compared with scans from more than 1,000 neurotypical individuals. They also studied 20 genetically different mouse models to investigate the underlying molecular biology.
They found two consistent subtypes.
In the first, brain regions communicate less than usual a pattern called hypoconnectivity, linked to disruptions in synaptic pathways, the physical connections between neurons.
In the second, brain regions communicate more than usual hyperconnectivity, linked to disruptions in immune-related systems.
These two subtypes accounted for roughly a quarter of the autistic individuals in the study. The researchers note that the full diversity of the autism spectrum likely includes additional subtypes that larger datasets may reveal these two represent the dominant patterns their analysis could isolate.
The two subtypes also showed modest but measurable differences on standardised autism assessments, with the hyperconnectivity group scoring somewhat higher on severity measures.
The patterns held across independent international datasets.
The implication is significant. Treatments and interventions have been designed and tested without knowing there may be at least two distinct biological groups. An approach that works for one biology may be ineffective for the other. The question of why autism interventions have such inconsistent outcomes now has a plausible structural explanation to investigate.
Autism has always been called a spectrum.
This study is the first large-scale evidence that the spectrum may contain biologically distinct lanes and that treating it as one condition may be part of why so many interventions have produced such different results in different people.