Brain networks

Functional connectivity is how closely activity in separate brain regions rises and falls together over time.¹ In 2004, a brain-scan study that measured activity during sentence reading (a task-based design, before later methods for handling head movement) found that language areas were less in step with each other in autistic participants without an intellectual disability than in a comparison group matched for verbal IQ. The authors proposed "a lower degree of information integration and synchronization" across large networks, and called this under-connectivity.¹

The same year, a review proposed that in the autistic brain, strong local connections may develop alongside weak long-range ones, and offered this as a framework that could tie together findings from genetics and brain tissue.²

The Autism Brain Imaging Data Exchange (ABIDE) pooled 1,112 resting-state scans from many sites: 539 autistic people and 573 non-autistic people aged 7 to 64. Its first analysis, of 360 autistic and 403 non-autistic males, found both weaker and stronger connections. Weaker ones were more common, especially between areas of the cortex and between the two halves of the brain.³

Later studies that combined several large samples also found both directions, but not in the same places. A 2019 study of four cohorts found weaker connections mainly in regions for sensing and movement, and more strongly connected hubs in the front and top of the brain. The pattern did not depend on age, sex or medication, though with few females a sex difference would be hard to detect, and autistic and non-autistic people overlapped a great deal.⁴ Studies that included more autistic females, also drawn from ABIDE, report differences by sex. In one, of 42 autistic males and 42 autistic females aged 7 to 30, autistic males showed weaker connectivity than non-autistic males and autistic females stronger connectivity than non-autistic females.⁴⁶ In another, of 362 autistic males and 82 autistic females aged 7 to 18, connectivity between the two halves of the brain in one region at the back was lower in autistic females, and this sex difference was repeated in one of two independent samples.⁴⁷ A 2023 analysis of 1,824 people aged 5 to 58, 796 of them autistic, found weaker links within sensory and attention networks, and stronger links between the default mode network and the rest of the brain and between the cortex and deeper structures.⁵

Results also depend on how the data are analyzed. In one multi-site study, no single type of connectivity measure gave results that held across sites, and results were consistent only in large samples.⁶ A careful study of autistic adults without an intellectual disability found largely typical connectivity, with a few differences in frontal and temporal cortex, nearly all weaker.⁷ Most of these samples were mostly male and left out autistic people with an intellectual disability or who could not lie still in the scanner, so the findings may not apply to all autistic people.^3,7,18 One review suggests that studies disagree partly because of participants' age and sex, scan design and analysis method, and because autistic people differ so much from one another.⁸

A separate line of evidence uses diffusion imaging to measure the white-matter tracts themselves. White matter is made of the bundles of fibers that link brain regions, and there water spreads more easily along the fibers than across them; fractional anisotropy (FA) measures how directional that spread is. A 2026 meta-analysis pooled 66 diffusion-imaging studies (2,231 autistic and 1,856 non-autistic people; study mean ages 2 to 50 years) across 12 tracts and found lower FA in nine of them, a moderate effect (Hedges' g = -0.30 to -0.50), and higher mean diffusivity, a larger difference (Hedges' g = 0.29 to 0.97), in eight. Both measures are taken to reflect properties such as fiber coherence, myelination and axonal density.⁵¹ The samples were mostly male and few reached later adulthood, and the authors note that group differences in head movement can inflate the effects. As with the connectivity findings, this is a group average, and autistic and non-autistic people overlap heavily.

The other half of the theory, stronger local connections, has less support. A systematic review of EEG and MEG studies, which record the brain's electrical activity, found relatively strong support for weaker long-range connectivity, but could not say whether local connectivity differs. The studies varied too much to be pooled in a meta-analysis.⁹

One fMRI study of adolescents found stronger local connectivity in regions at the back of the brain, and more of it went with more marked autistic traits. A subgroup with fewer traits showed weaker frontal connectivity instead.¹⁰ A 2024 meta-analysis of fMRI studies that measured local connectivity as regional homogeneity, in people aged 18 or under, found weaker local connectivity in parts of the default mode and sensorimotor networks, and no regions with stronger local connectivity. Its authors say the result needs confirming.¹¹ A study of 53 autistic people from one ABIDE site, split into children, adolescents and adults, found the pattern depended on region: weaker local connectivity in sensory regions and stronger local connectivity in regions for complex processing, at every age. In autistic children and adolescents, stronger local connectivity went with more marked autistic traits.⁴⁴

In 2013, one group proposed that the conflicting findings could be reconciled by age: young autistic children would show stronger connectivity and adolescents and adults weaker connectivity.¹² The same group found stronger connectivity in autistic children, across short- and long-range connections, in three samples it analyzed itself, and in these children stronger connectivity went with more marked social differences.¹³ A study from the same group that split ABIDE participants by age (26 children, 28 adolescents and 18 adults with autism) found stronger connectivity within networks only in children aged 11 and under, weaker links between networks in children and adolescents, and no differences in adults.¹⁴

The 2023 analysis of 1,824 people found no statistically significant change in the connectivity differences with age, or between the sexes; a study of this design, scanning each person once, has limited power to detect either.⁵ A 2024 study of 1,218 people aged 5 to 40 in ABIDE compared each person with what is typical for their age and found that connectivity far from the typical range, mainly in the default mode network, became more common with age in autistic people and less common in non-autistic people.¹⁵

The age-split study and the 2024 study compared different people at different ages, each scanned once.^14,15 One small study did follow 16 autistic and 22 non-autistic young people, almost all boys without an intellectual disability, scanning each twice about three years apart. As non-autistic youth grew older, the default mode and central executive networks became more opposed in their activity; in autistic youth they did not, so the groups differed only in late adolescence. The authors call the findings preliminary.⁴⁵

Small head movements during a scan distort connectivity measures. In two independent studies of non-autistic people, movement lowered many long-distance correlations and raised many short-distance ones, even after standard corrections.^16,17 That is the same shape as the under-connectivity theory. In one study of autistic adults, even small amounts of movement had large effects on connectivity measures.⁷

Removing people who moved too much creates a different problem. In one study of 545 children aged 8 to 13 without an intellectual disability, all scanned at one US center, autistic children were excluded more often: 28.5% against 16.1% under a lenient rule, and 81.0% against 60.1% under a strict one. The autistic children left were older, had milder social differences, better motor control and higher IQ than the group who were scanned.¹⁸ Results from such samples may not apply to autistic people who find it harder to stay still.

Correction methods differ in how well they work. A comparison of 14 methods in 393 young people found that removing the moving time points (censoring) reduced both the motion effect and its link to distance, at the cost of discarding data, while some other methods left a distance-dependent distortion.⁴⁸

The triple network model, proposed in 2011, holds that "deficits in access, engagement and disengagement" of three large networks appear across several conditions, including autism.¹⁹ The default mode network quietens during demanding outward tasks, which engage the central executive network. One group's fMRI studies suggest that a salience network, centered on the right anterior insula and the anterior cingulate cortex, switches between the two; this role was inferred statistically from the timing of activity.^20,19

A review describes altered organization and development of the default mode network in autism.²¹ A meta-analysis of 10 studies found weaker connectivity between two of its hubs, the posterior cingulate cortex and the ventromedial prefrontal cortex.²² A 2025 meta-analysis of 26 studies found differences in connections of the medial prefrontal cortex, also part of this network.²³ In a study of 22 autistic children aged 8 to 13 taking no medication, a weaker link between these two hubs went with more marked autistic traits, and the network was less separated from other networks.²⁴ Studies drawing largely on the shared ABIDE data report the default mode network more connected to other networks, so they are not independent confirmations. One found this at ages 7 to 18 but not 18 to 25; the larger 2023 analysis found no change with age.^5,25,26

Findings for the salience network point in both directions. In 20 autistic children aged 7 to 12, several networks including the salience network were more strongly connected, and salience network connectivity distinguished autistic from non-autistic children with 78% accuracy, and 83% in a second sample.²⁷ In 24 autistic adolescents with an average age of about 16, it was less connected, and weaker connectivity went with lower scores on social-cognition tasks.²⁸ An ABIDE study of 144 autistic people found more connectivity between the salience and default mode networks, and stronger coupling between the right insula and medial prefrontal cortex went with more social-responsiveness difficulties.²⁶ In 169 young people aged 8 to 17, autistic girls showed more connectivity between the default mode and central executive networks than autistic boys.²⁹

Methylation is a chemical mark on DNA that can turn a gene's activity down. The oxytocin receptor gene (OXTR) makes the receptor that the hormone oxytocin acts on. A 2009 study found more methylation at some sites of this gene in blood from one group of autistic people, and in temporal cortex from a separate group of 8 autistic and 8 matched non-autistic people after death, mostly male. One site, -934, was higher in both blood and brain. In a subset of 4 brain pairs, the male autistic brains had about 20% less of the gene's mRNA.³⁰

In 35 autistic men with IQs above 70, one site in the gene had more methylation in saliva than in 64 non-autistic adults of both sexes; because the groups differed in sex make-up, a sex effect cannot be ruled out. In 30 of the men, more methylation went with weaker resting connectivity between brain components involved in thinking about others' minds. Methylation at a second site, which did not differ between the groups, went with stronger connectivity between a reward area deep in the brain and the front of the brain.³¹ The authors call for replication in larger and more varied samples.³¹

Other studies disagree. One found less methylation in 27 autistic children, one found differences that depended on sex, one found no overall pattern in 80 children, and one that read the whole gene in 20 autistic adults found no site that differed.^32,33,34,35 A 2018 systematic review judged the evidence on this gene's methylation and social behavior too thin to draw conclusions.³⁶

Two findings often cited alongside these are about something else. Weaker reward-circuit connectivity in 41 autistic young people was linked to common variants in the gene's DNA sequence, not to methylation.³⁷ A link between methylation and more effort from the front of the brain during social attention came from 54 non-autistic young adults.³⁸

A popular way to picture the theory is a computer network: busy local nodes that process detail well, joined by a slow main line. The 2004 papers did tie weak long-range connections to weaker integration of information. The review also proposed that excess local connectivity produces strong but poorly selective local activity.^1,2 A review of EEG and MEG studies proposed something similar, less wide-ranging integrating activity offset by more focused local activity, while noting that its own evidence on local connectivity was unclear.⁹

Several experiences the picture is used to explain are documented. On a self-report questionnaire, about half of autistic people score in the range for alexithymia, difficulty identifying and describing one's own feelings, compared with about 5% of non-autistic people.⁴⁰ In one study, insula responses to another person's pain tracked alexithymia rather than autism.⁴¹ On a card-sorting task that requires switching rules, autistic people complete fewer sets and make more errors, with small to moderate differences across 31 studies.⁴² Monotropism, a theory developed with autistic authors, holds that autistic people allocate attention differently, with attention tied to a narrow range of interests.⁴³

No study found as of October 2026 has tested whether alexithymia, task switching or deep focus in autistic people track long-range connectivity. In fMRI, the weak-long, strong-short pattern is also what head movement produces, though EEG and MEG studies, which are affected differently, also lean toward weaker long-range connectivity.^16,9

In community writing, "wiring" is less a finding about connectivity than a way of saying that autistic brains are different, not broken. Steve Silberman records the journalist Harvey Blume, the first to use "neurodiversity" in the press, writing in 1998 that "NT is only one kind of brain wiring", meaning neurotypical, and asking "Who can say what form of wiring will prove best at any given moment?"^c1 Bianca Toeps writes that being wired differently from the average person is no reason to treat an autistic neurotype as something to be beaten out or eradicated.^c2 Orion Kelly puts the same point plainly: "Our brains as Autistic people are wired differently; we see, feel and experience the world differently."^c7

Nick Walker, an autistic scholar of the neurodiversity paradigm, questions the wiring picture from the other side. Walker argues that "neurotypical" does not name a normal kind of brain: "There's no such thing as a normal brain or a normal 'type of brain.'" In Walker's usage, a neurotypical person is one who lives and acts within the bounds of what the surrounding culture imagines a "normal mind" to be, and the idea of "types of brains" is "ultimately rather dubious, or, at the very least, overly reductionist."^c8 In an essay titled "What Is Autism?" in the same collection, Walker also writes that the evidence points to autistic brains having "particularly high levels of synaptic connectivity and responsiveness", which makes experience more intense.^c9 That is Walker's reading. The research sections above describe a less settled picture.

The same word is used on the other side of the debate. Eric Garcia quotes Jill Escher, a parent of two non-speaking autistic sons, who wants people to see that her son's behavior "comes out of a difference in his brain wiring", but adds: "But do I think his behavior and wiring is natural? Absolutely not."^c3 Garcia replies that accepting autistic people does not mean ignoring real difficulties, and that acknowledging needs and embracing autistic people are "complementary".^c3

Some writers take the connectivity research as a sign of individuality. Devon Price writes that "every Autistic person's brain is unique and exhibits its own distinct patterns of connectivity", and uses this to argue that autism has no single cause.^c4 That fits the wide variation described in the research sections above.

Others use wiring to explain daily experience. Sol Smith argues that extra neural connections and a busier resting brain cost energy: "on average, you can count on extra synapses being a liability day to day", and links this to rumination.^c5 Megan Anna Neff writes that task switching is harder for autistic people and that "this is related to hyper- and hypoconnectivity in our brains".^c6 These are the writers' explanations. As the last research section notes, no study found as of October 2026 has tested whether task switching or deep focus in autistic people tracks connectivity.