A year and a half ago neuroscientist Anila D’Mello scanned the brains of a dozen autistic women who had just entered her study at the Massachusetts Institute of Technology. The all-female cluster immediately threw off the results. “When we analyzed their data, we realized that it looked really different than the data we had collected up until that point” from their all-male pool, recalls D’Mello, then a postdoctoral fellow in neuroscientist John Gabrieli’s lab at the McGovern Institute for Brain Research at M.I.T. The study, which has not yet been published, involved looking at how the brain of an autistic person responds to seeing the same face or object—or hearing a word—over and over again. The focus was not on sex differences, but the researchers now wondered whether there were some.

Yet when they tried to recruit more women and girls from M.I.T.’s Autism Research Participant Database, which contains several hundred people with a clinical autism diagnosis, there was a problem: To qualify for the study, prospective participants had to take a standard activity-based assessment for autism to confirm their diagnosis. After testing, half of the 50 girls and women who would otherwise be eligible for the scientists’ study did not meet the test’s criteria for autism. “We realized, ‘Wow, we only are retaining 25 of them, 25 of the 50.’ These are all women who came in with a diagnosis,” says D’Mello, now an assistant professor at the University of Texas Southwestern Medical Center in Dallas. She and her colleagues wondered: Could this diagnostic test, which is widely used in autism research, help explain why so few women are included in studies of autism?

The standard lore holds that about four boys are diagnosed with autism for every girl, an estimate that is largely based on community diagnoses: those from, say, office visits to a physician who looks for autism traits from the widely used fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5). In recent years, researchers have recognized a diagnostic sex bias and recalibrated the ratio of boys to girls with autism as closer to 3:1.

In studies of autism biology, traits or treatments, however, the ratio of male to female participants is often far more skewed. Large-scale review articles of autism neuroimaging studies have reported ratios as high as 15:1; others find male participants predominate 6:1 or 9:1. What’s more, a 2021 review reported a huge preponderance of male-only studies in two decades of research on the brain structure of autistic people.

Historically, many autism researchers have overlooked sex differences in some cases deliberately excluding women and girls to keep samples uniform, says Kevin Pelphrey, an autism researcher at the University of Virginia. Seeing autism through a male lens, combined with an actual dearth of autistic girls and women, has been blamed for the lopsided sex ratios in autism research, he says. But in a new study in Autism Research, the M.I.T. team found support for another explanation: the Autism Diagnostic Observation Schedule (ADOS). This gold-standard test, which is routinely used to confirm autism diagnoses in research participants, filters out proportionally more female participants than male ones. “It’s a really beautiful paper,” Pelphrey says. “I always believed that there was a hidden cluster of people that we don’t study because they don’t quite fit criteria.”

Scientists use the ADOS because it standardizes what is considered autism across studies and because “we have to,” Pelphrey says. There is a kind of “groupthink” in the field that insists, he says, “if you didn’t use these measures to confirm the diagnosis, your paper is not worthy.” It affects whether a researcher can get funded and where they can publish their research.

The M.I.T. study suggests that this way of thinking could be stymieing discovery of sex differences in autism. “There is a biological reality to the sex differences that’s very important to understand,” Pelphrey says. “We don’t fully understand it yet, and we are not going to unless we study a lot of girls.”

“If you exclude [female participants] or have a low sample size in that group, you wouldn’t be able to capture those unique patterns that are dependent on sex or gender,” says Meng-Chuan Lai, associate professor of psychiatry at the University of Toronto. Capturing those patterns is important for developing therapies and designing services for autistic women and girls, Lai says. Not everyone, however, is convinced that the ADOS is filtering out female participants. “I absolutely think we are missing women, but I don’t think it’s because of the ADOS,” says Catherine Lord, a professor of psychiatry at the David Geffen School of Medicine at the University of California, Los Angeles. Lord, who helped develop the ADOS, points to a large multisite study, among other data showing that girls with community diagnoses of autism score similarly to boys on the ADOS. “If you go back and look at the papers of large ADOS samples,” she says, the conclusion that the ADOS is a major factor in the lack of women in studies “is not justified.”

In their study, D’Mello and her colleagues first looked in the M.I.T. database at the ADOS scores of 50 female volunteers and 95 male ones who were verbally fluent (because many people with autism speak few or no words) and aged 16 years or older. They found that just 19 percent of the men and boys did not meet ADOS criteria for autism, compared with half of the women and girls. An updated ADOS fared only slightly better, filtering out 41 percent of the women and girls and 18 percent of the men and boys. “We were quite stunned to see the very large percentage of women who volunteered and then were excluded on that basis [of an ADOS score],” Gabrieli says.

To see how pervasive this problem was, the researchers looked at the proportions of men and boys and women and girls older than age 16 in eight national or international databases that differed by whether they admitted participants based on the ADOS or on a community diagnosis of autism. As the team predicted, the community samples included proportionally more women and girls. In the largest of those samples, called Simons Foundation Powering Autism Research for Knowledge (SPARK), for example, the researchers found 7,708 autistic men and boys and 4,504 autistic women and girls who met their age and other criteria—a sex ratio of approximately 2:1. By contrast, in the Autism Brain Imaging Data Exchange (ABIDE) databases, which rely largely on the ADOS, 189 men and boys and 25 women and girls fit the researchers’ inclusion criteria, a ratio of more than 7:1. “It wasn’t just our modest sample [at M.I.T.],” Gabrieli says. “In large national samples, it was exactly the same story.”

Why might fewer diagnosed women and girls meet ADOS criteria for autism? The gold-standard test was developed largely using data from boys, which could mean it is geared to detect autistic traits as they appear in males.  Autism may look somewhat different in girls, experts say. Girls may have repetitive behaviors such as hair twirling that are more socially acceptable, for example, D’Mello says. And because girls may be better at masking, or camouflaging, autism traits, they may be less likely to show these traits during ADOS testing, Lai says. This does not mean that they do not have distress or functional impairment stemming from autism—which could show up in a clinical history. “Based on DSM-5 criteria for autism, you don’t necessarily have to show everything behaviorally for autism at this moment,” Lai says.

Another interpretation of the findings, however, is that women and girls are overrepresented in the databases that rely on community diagnoses. Women are more likely to sign up for large survey studies such as SPARK, Lord says. Women are also more likely to seek help from a mental health professional, she adds. “I don’t think [women] are excluded,” Lord says. “I think you could reverse it and say, ‘Why do these other samples have so many women?’” Lord asks. "Where are all the men?"

The debate forces people to confront “the elephant in the room,” Lai says. “Are we talking about the same autism?” The different inclusion criteria—a community diagnosis or an ADOS cutoff score—may, in fact, include different populations of people on the spectrum, he says. Some researchers believe it is important to learn who these individuals are. “Who are those people who clinically appear to have it but don’t pass research criteria?” Gabrieli asks. “If they are excluded from research, we can’t answer that.”

The existing testing framework also means leaving out a lot of the people who identify as autistic. “I personally am interested in studying the experience of autistic people not filtered through some additional confirmatory criteria,” D’Mello says. “If you think about it, the whole idea of having to confirm someone’s diagnosis is kind of sad.”