What went wrong with autism research? Let’s start with lab mice.

In the world of neuroscience research, the mouse reigns supreme: in the US alone, tens of millions of mice are studied as a proxy for the human brain in labs. They’re small, they breed quickly, and they’re relatively easy to genetically manipulate, making mice ubiquitous in biomedical science. When studying something fundamental to biology, like how individual cells work, the leap from mouse to human doesn’t feel egregious. But when mice are used to study distinctly human conditions like autism, the parallels start to break down.
Fifteen years ago, researchers introduced the first two mouse models of autism, each carrying a genetic mutation linked to autism in humans. They claimed that these mice behaved like autistic humans, unusually preferring solitude over meeting new mice, and squeaking only around half as often as their non-autistic littermates.
Their results made major waves, inspiring researchers to experiment with other autism-related genes. Since the late 2000s, neuroscientists have bred over 20 types of mouse models with motor problems, sensory sensitivities, and repetitive behaviors. These each capture some hallmark of human autism — provided you buy that a mouse burying marbles is the same as, for example, an autistic child insisting on eating the same food every day.
As a freshly minted PhD working at the National Institute of Mental Health in the late 2000s, Jill Silverman ran experiments on mice missing part of their SHANK3 gene, a mutation found in about 1 in 100 autistic people. These SHANK3 mice seemed to show “autistic-like behaviors” like social discomfort and compulsive grooming, similar to the repetitive body movements, or stimming, seen in some humans with autism.
Silverman, now a principal investigator at the UC Davis Medical Investigation of Neurodevelopmental Disorders (MIND) Institute, still gets compliments on these mice — even though many of her original findings couldn’t be reproduced in future experiments. While she still works with laboratory mice and stressed their importance in understanding co-morbidities and rare genetic disorders tied to the condition, she noted that how “disrupted brain development manifests in a mouse is never going to be autism, because autism is uniquely human.”
But when people praise her earlier research on autism-like behavior in mice, “I’m like, that is the biggest mistake this field has ever made,” she said.
Billions of dollars have been poured into autism research over the last decade, funding a staggering number of experiments — including over 1,500 studies in the US in 2020 alone. Many of these studies use animal models, especially mice.
Whether you believe that animal testing is ethical (many don’t), scientists in numerous research fields — especially neuroscience, genetics, and other areas of biology — run experiments on animals. To understand how cells in the brain communicate to form thoughts and guide behavior, you need a living brain connected to a living body. Millions of rodents are used — and nearly all killed — for science experiments every year, many of which are preclinical tests of new drugs and other treatments with potential public health benefits, including for autism.
While mouse models have helped us understand how neural connections develop over time, and how they may be affected by autism, they haven’t led to the pharmaceutical treatments many researchers are looking for. So far, all attempts to make drugs that help people manage some of the more challenging effects of autism, like sensory sensitivity or self-harm, have failed.
When I asked senior scientist Brigitta Gundersen, who manages Simons Foundation Autism Research Initiative (SFARI) funding for autism studies involving rodents, for an example of a tangible quality of life improvement that this line of research has given us, she paused. “I struggle to think of examples across all of psychiatry, frankly.”
In theory, figuring out how autism manifests in the brain and body should help scientists develop better treatments for some of its more debilitating symptoms, like seizures, mobility challenges, and self-harm. Given how much we still have to learn about how the brain works, autistic or otherwise, this kind of research is “a really long game,” Gundersen said.
“There’s this overall idea that understanding biology and understanding mechanisms will lead to better interventions,” she said. Studying biological systems for the sake of understanding them led to advances in medicine we often take for granted today, like chemotherapy. But the brain the most complicated tissue in the body — so far, at least, the search for medications that effectively treat psychiatric and neurodevelopmental disorders “hasn’t totally panned out.”
Mouse models of autism-related gene mutations may help uncover the underlying biology of autism in the long run. But autistic people understandably want tangible support now, and research serving that need is hugely underfunded. “It barely matters to us what a mouse model says,” said Sam Crane, an advocate for people with disabilities and a public member of the federal Interagency Autism Coordinating Committee (IACC), a group that helps policymakers decide what types of autism research to pay for.
Others, including the parents of autistic children with very high support needs, fear that deprioritizing biological research will leave their loved ones behind, turning attention away from developing potentially lifesaving treatments. Massive funding agencies like the US National Institutes of Health (NIH) are also wary of those trying to shift autism research away from genetics and neuroscience, arguing that scientific breakthroughs often come from long-term studies of fundamental biology — even when those studies don’t seem to offer real-world benefits in the short term.
Looking at the numbers, though, research exploring how to help autistic people navigate everyday life — the research many autistic people say they’d like to see — is still only getting about a quarter of the money allocated for autism research in the US. At a moment when autism diagnoses are on the rise — for reasons scientists still don’t fully understand — why are we spending so much on mice that might help humans eventually, and so little on services that could help humans now?
Autism spectrum disorder (ASD), as defined by the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), is usually characterized by communication challenges, trouble navigating social interactions, and a high sensitivity to change.
It’s also defined by how different the roughly 5.4 million people diagnosed with autism in the US are from each other. The wide umbrella of ASD includes people who live independently, have fulfilling careers and relationships, and can advocate for their own needs. It also includes people who don’t speak much (or at all), use a wheelchair, and may require full-time support from a caregiver for the entirety of their lives.
Biologically speaking, autism — like the brain in general — is still poorly understood. In the 1950s and 1960s, medical professionals embraced the now-discredited “refrigerator mother” theory linking autism to cold, distant parenting, blaming mothers for their children’s condition. Later, psychologist Bernie Rimland presented evidence that autism is rooted in biology. Then, former physician Andrew Wakefield published a paper in 1998 incorrectly linking the measles, mumps, and rubella (MMR) vaccine to autism, fueling the modern anti-vaccination movement.
Today, most researchers believe that autism is strongly influenced by genetics. However, when symptoms can include everything from difficulty reading social cues to seizures to constipation, it’s hard to figure out what genes might be causing what — after all, over several decades of work, scientists have compiled a list of 100 or so genes that might be linked to autism.
To leaders at private funding agencies like SFARI, Autism Speaks, and the Autism Science Foundation (ASF), that complexity is precisely why we need basic research to explore the underlying biology and genetics of autism. The ultimate goal of these funders, several of whom have autistic children, is to find treatments for autism. Historically, some of these institutions even wanted to find “cures.”
But digging into the genetics of autism in the early aughts raised more questions than answers, forcing researchers to reconsider what autism even is. Meanwhile, in the absence of meaningful medical progress, some desperate parents turned to extreme DIY “cures” like making their autistic kids drink bleach.
“Despite the fact that they’re pointing in diametrically opposed directions, there’s a common theme with the refrigerator mother approach and the anti-vaccine approach,” said Ari Ne’eman, assistant professor at the Harvard School of Public Health and co-founder of the Autistic Self Advocacy Network (ASAN). “Namely, they both really emphasize the idea of causation as central to the business of autism advocacy.”
Framing autism as a disease that “happens” to otherwise healthy children as a consequence of their parenting, genetics, or environment makes it feel like something that science can fix, or even prevent in the first place. For many diseases — think deadly cancers — this wouldn’t be controversial.
But many autistic adults believe the “causation” framing is hugely misguided. Efforts to pinpoint genetic markers of autism have raised serious concerns about eugenics — namely, that if parents could get a prenatal test for autism, many of them would choose not to have those children.
Prenatal tests for many diseases, like cystic fibrosis and sickle cell disease, already exist, and the fears of autism advocates are not unfounded. In Iceland, for example, nearly 100 percent of parents who get prenatal tests for Down syndrome — a chromosomal condition affecting as many as 6 million people worldwide, many of whom live long, healthy, fulfilling lives — choose to abort their pregnancy if the results are positive, causing the population of Down syndrome children to almost completely disappear there. Even in the US, where abortion is politically fraught, over two-thirds of parents choose not to give birth after finding out their child will have Down syndrome. Should it also be acceptable for parents to abort a pregnancy if they learn that their child will be autistic?
“Autism research was really built with the assumption that the goal is a world without autism,” Ne’eman said. But a growing number of people embrace the neurodiversity movement, proposing that autism is simply another way to move through the world. To them, the condition is not something to cure with medication or prevent with prenatal testing. This shift has led to significant controversy in the world of autism research. Autism Speaks came under fire in the mid-2010s for portraying autism as a devastating disease that ought to be stamped out, before denouncing that rhetoric in 2016.
For now, an effective prenatal test is not widely available — while autism does seem to be strongly influenced by genetics, there isn’t a single gene that flags autism. Prenatal tests and emerging gene-editing tools like CRISPR seem to work best for conditions caused by a single genetic mutation, like sickle cell disease.
However, scientists have listed about 100 genes that all seem related to someone’s likelihood of being diagnosed with autism, making a target for potential screenings, drugs, or other therapies much harder to pin down. Mutations in any one gene don’t necessarily mean that a person will be autistic, or shape what autism will look like for them. While some single-gene mutations cause specific neurodevelopmental disorders that fall under the umbrella of autism spectrum disorder, like fragile X syndrome, they are relatively rare.
All considered, autism isn’t currently something that can be addressed by traditional drug development pipelines. Yet, funding for projects studying the biology of autism more than quadrupled since 2008, while funding for projects finding better ways to help autistic people in day-to-day life fell or remained stagnant.
Under the Combating Autism Act, which George W. Bush signed into law in 2006, Congress established the Interagency Autism Coordinating Committee (IACC). As the name suggests, the Combating Autism Act was focused on finding treatments to prevent or “cure” autism.
At the time, the vast majority of IACC members were not autistic — and their funding priorities were oriented accordingly. Their first set of recommendations, published in 2009, heavily skewed toward funding the search for causes and cures of autism. For example, they proposed spending $75 million on developing animal models of autism — nearly 50 times more than they suggested spending on studying everyday support services for autistic people.
In the world of biomedical research, where there are genetic risk factors, there are genetically altered mouse models. But by continuing to fall back on the rodents that they are so accustomed to studying, researchers are holding themselves back from fully understanding how autism manifests in humans.
Mice are small, reproduce quickly, and share about 85 percent of their functional genes with humans, making them desirable to geneticists hoping to study diseases outside of the human body. While non-animal models are slowly replacing animal testing in many areas of science, “you need a live animal to study a disorder that’s solely behavioral,” Silverman said. “Cells don’t behave.”
Mice behave, but their behavior is very different from ours. So, neuroscientists have had to stretch to draw parallels between the behavior of mice and autistic humans. If a mouse buries marbles with unusual fervor or over-grooms themselves, a study may qualify it as “repetitive behavior.” If a mouse prefers being alone to hanging out with a stranger mouse in its cage, it’s displaying “social deficits.” Studies have even measured changes in ultrasonic vocalizations in mice to try to understand speech problems in autistic humans, and recorded electrical activity from the brains of dogs with autism-related gene mutations to see whether LSD could improve their social interactions.
Animal behavior is finicky, though, especially when those animals are living in tiny laboratory cages, far from their natural habitat. The same mouse in the same marble-burying setup, for example, may bury fewer marbles than usual one day because it got distracted by the smell of whatever shampoo the experimenter used that morning.
Human error can play a role, too. An exhausted grad student may miscount the number of times two mice bump noses. Researchers in different labs may not even agree what that nose-bumping behavior means, or how to classify it in their papers. It “just lends itself to a lack of reproducibility,” Gundersen said.
It also makes preclinical trials for new treatments, which are often conducted in animals, challenging to translate to humans. Many symptoms, especially those related to social interactions and communication, are distinctly human — so much so that they’re nearly impossible to reproduce in mice. “You know,” Gundersen said, “no mice talk.”
Today, more scientists are rejecting the idea that mice can actually exhibit autistic-like behaviors. “Nobody thinks that mice are people,” Gundersen said. “Nobody thinks that mice are modeling autism.”
But the number of publications featuring “mouse model(s) of autism” in the title has steadily increased since they were first introduced in the mid-2000s. A cynic might wonder why scientists are continuing to pursue this line of research, when both autistic self-advocates and a growing number of leaders in biomedicine are saying that it doesn’t make any sense.
Ne’eman said that some people in the autistic community jokingly refer to autism research as a “geneticist’s Full Employment Act” — a parallel to the proposed Autism Full Employment Act, which would create incentives for workplaces to hire autistic people.
The grant application system is really competitive. To boost their chances of getting research funding, applicants increasingly have to twist their research proposals to align with whoever will give them money. A lab interested in studying how gene expression guides brain cells to form connections with each other, for example, could pitch it as an autism study to open up additional funding opportunities.
So, Ne’eman suspects that some scientists are “looking at the autism research agenda as exclusively or primarily a vehicle for a relatively small number of abstract questions of basic science,” which aims to expand knowledge without necessarily translating to new drugs or other practical applications. Just look at the mice: it’s been clear for years that they’re a bad proxy for autistic people, but many biomedical researchers have built their careers around using them. Moving away from dysfunctional models requires time, money, and critically evaluating old, imperfect findings — something scientists aren’t really incentivized to do.
People like Alycia Halladay, chief science officer at the Autism Science Foundation, worry that self-advocates like Ne’eman are too dismissive of basic science.
“At ASF, we strongly believe that the use of model systems is essential to understanding the biology of autism and to developing targeted, personalized interventions that will improve lives,” Halladay wrote in an email statement to Vox. “Use of animal models has been critical to our understanding of dozens of psychiatric disorders, and has helped to improve the lives of those who suffer from those disorders. Autism research must continue to include these important model systems. ASF remains absolutely committed to funding basic biology, genetic discovery, animal models, and pharmaceutical innovation.”
But it isn’t that autistic people don’t value science. Rather, many of them think the somewhat futile search for a “cure” to autism shouldn’t receive as much funding as it does, relative to other areas of research.
A more promising path for biomedical researchers could be studying rare neurodevelopmental disorders, like Angelman syndrome and Rett syndrome, caused by mutations in a single gene that exists in both mice and humans. People with disorders like these often have symptoms experienced by others with autism, like seizures, gastrointestinal issues, and insomnia — which are more easily quantifiable in mice than, say, language.
Silverman moved her lab in this direction entirely, after losing faith in models of other “autism-like behaviors.” She hopes that a clearer understanding of these specific genetic mutations will lay the foundation for things like better epilepsy medications down the line — not only for those with Angelman syndrome, but for anyone who experiences seizures alongside autism.
I asked Halladay what research she wanted to see, as a mother of an autistic daughter. She agreed that more investigations of conditions related to autism, like sensory sensitivity, would be incredibly helpful to families like her own. Halladay, like many other parents, doesn’t want her daughter’s autism to go away; she just wants more support — and possibly medicine — to help her child live the best life possible.
In general, Ne’eman thinks that “the average autistic person, as well as the average family member, doesn’t wake up in the morning thinking, ‘Have they found a better mouse model?’” They do think about whether they’ll be able to find a full-time caretaker who is covered by insurance, or what the newest adaptive communication devices will be capable of.
When autistic self-advocates were largely excluded from the decision-making process, funding for things that would help them immediately, like communication assistance or housing support, fell by the wayside.
That’s since changed — today, the IACC includes 23 non-autistic government employees and 22 public members, seven of whom are autistic themselves. Their budget priorities have shifted accordingly, centering research questions like “What services and supports are needed to maximize health and well-being?” in addition to basic biology studies.
At the same time, the gap between the committee’s proposed budget and how much funders actually spend has also grown. And while funding for services and support doubled between 2019 and 2020, it still only accounted for 8.4 percent of the money spent that year.
One big thing standing in the way of the IACC’s recommendations and reality: The biggest sources of science funding, public and private, weren’t really built to fund things other than biology research. Of the 28 organizations listed as funding autism-related projects between 2019 and 2020, the National Institutes of Health and SFARI — which only award grants for basic science and clinical research — together paid for over 80 percent of research.
Agencies like the Department of Education and the Administration for Community Living pay for projects studying interventions like how to help autistic adults avoid institutionalization and live as independently as possible — major priorities for autistic self-advocates. However, they only fund a tiny portion of autism research.
Solving this problem will likely require a major redistribution of funding, or a big overall increase in the pool of money available to everyone. “I’m not sure that you can totally fix it by just yelling at the NIH,” Crane said. In fact, she suspects that the Office of National Autism Coordination, which oversees management of the IACC, is aware that IACC members want to fund more studies about how to support autistic people. The problem is likely that the NIH — the federal body responsible for funding — may not be receiving grant applications for such studies.
One solution the IACC recommended involves growing the overall pool of money set aside for autism research to $685 million by next year. They specifically highlighted three research areas that need the most additional resources: lifespan issues, evidence-based interventions and services, and the development of culturally responsive services.
By “lifespan issues,” the IACC means anything related to big life transitions: access to higher education and employment, opportunities to live as independently as possible alongside non-autistic community members, and health care. Figuring out how to help autistic adults — including those with the most severe disabilities — find fulfilling jobs that they’re good at, stay out of harmful psychiatric institutions, and form healthy relationships doesn’t require mouse models. It requires piloting initiatives like new housing programs, building better assistive communication devices, and other community-oriented research.
Studying existing interventions to make sure they’re helping autistic people — not just making them appear non-autistic in public — is also crucial, Crane said. For example, applied behavioral analysis (ABA) therapy, which rewards “goal behaviors” like making eye contact or saying hello to people, is controversial in the autistic community because it can be experienced as abusive and coercive. Most existing studies on the effectiveness of ABA measured things like whether recipients behaved better in the classroom, rather than long-term outcomes like overall academic achievement or quality of life. With more money, Crane hopes this can change. “We need to be funding research that actually tracks the outcomes that matter to people.”
The bottom line is that we don’t need more mouse models of autism or of autism-like behaviors. Biomedical science has a role to play, especially in helping people manage symptoms of other autism-related health issues like epilepsy and sleep disorders — but it has claimed a disproportionately large chunk of autism research funding for too long.
Some people, especially the parents of children with intellectual and physical disabilities related to autism, argue that autistic self-advocates who push back against the biomedical research agenda are acting out of self-interest, leaving those with the most severe disabilities behind. People with different experiences of autism, Autism Science Foundation president Alison Singer argues, need different things. Specifically, she believes that people with the most severe disabilities need the kinds of pharmaceutical interventions that biomedical research aims to find — and that many autistic self-advocates want to deprioritize.
Ne’eman believes the opposite is true. “Those with the most severe impairments are especially poorly served by research that does not relate back to their needs,” he said. In its statement on genetic research, the Autistic Self-Advocacy Network emphasizes, “Autistic people with the highest support needs are some of the most vulnerable members of our community. They deserve good lives with the right to make their own decisions, not yet another round of ‘cures’ that will not work.”
Neuroscience still has a lot to offer the autism community, but neuroscientists need to listen to the people they’re claiming to serve. Ditching outdated behavioral tests on mouse models of “autism-like behavior” might be a great place to start.
Clarification, October 21, 12:50 pm ET: This story was updated to clarify the positions of Jill Silverman, Alycia Halladay, Brigitta Gundersen, and the Autism Science Foundation on animal models of autism.
Clarification, October 22, 5:30 pm ET: This story, originally published October 14, said the Office of Autism Research Coordination funds research within the NIH; it is part of the NIH but does not distribute funding.