Last month, researchers in Australia led by Andrew Whitehouse published a new study: a trial of a “pre-emptive” intervention for infants assessed to have an elevated likelihood of later developing autism.  Internet furor immediately ensued, with some headlines proclaiming that the new intervention could prevent as many as two-thirds of autism diagnoses. Many neurodiversity advocates were naturally horrified by these ominous headlines, but was the media portrayal accurate?

What were the researchers trying to do?

The researchers who conducted the study claimed that it “chershes neurodiversity”: that instead of being about suppressing autism, the intervention aimed to help the caregivers of these possibly-neurodivergent babies better understand their children and adapt the environment to be more supportive of the infants’ needs.  The researchers said that the “inaccurate” media headlines “misrepresented” what they were trying to do.

If the study really was about helping caregivers understand neurodivergent children, it could be very much in line with the neurodiversity movement’s goals. Damian Milton’s concept of the “double empathy problem”—which is supported by empirical research—highlights that it is not only autistic people who struggle to understand neurotypicals, but also neurotypicals who struggle to understand autistics.

So if the goal of this controversial study was really trying to address the double empathy problem for caregivers of young children, in a manner aligned with the neurodiversity movement and with autistic people’s priorities, did the researchers succeed?

How close did the researchers come?

Although the researchers did come quite a bit closer to succeeding in this than many of the headlines about their work would suggest, in many ways their study also illuminates just how far autism research still has to go, in order to reach the standards the community can trust. Nevertheless, the intervention itself was probably the area in which the researchers’ study came closest to meeting the community’s standards and expectations.

The intervention—which is summarized in one of the paper’s supplements—does indeed explicitly aim to help parents understand and have empathy for the child’s point of view. In particular, it aims to promote parental attunement to the child’s unique, individual way of socially connecting with the caregiver, making sure that the child does not become discouraged or disinterested after efforts to connect are overlooked or misunderstood. This is very well-aligned with the goals and objectives of the neurodiversity movement.

The researchers also tried to keep track of whether the intervention caused adverse events (they reported it didn’t), which is something that autism intervention research has historically tended to ignore. Thus, although it’s not quite clear how the researchers judged whether adverse events were caused by the intervention, their work made progress over prior research by simply considering the possibility the intervention could cause harm.

Furthermore, the intervention involved weekly sessions with parents that were only about two hours long each.  A lot of pressure gets placed on parents of young neurodivergent children.  While this intervention did recruit parents to drive its effects, it doesn’t seem like this time commitment would cut into other important parent-child activities.

One potentially concerning part of the intervention protocol says that it includes “reinforcement” of interpersonal interactions between babies and caregivers. That word often conjures up images of artificial manipulation, as in the case of children being rewarded with food and praise for doing things they would not otherwise be willing to do. This can often encourage the child to “mask” their autistic traits, hiding their true feelings and needs in order to receive approval or tangible rewards from others. Pressure to mask is a common concern among autistic advocates and masking is associated with a number of negative mental health outcomes.

However, reinforcement also occurs naturally, when the natural outcome of a behavior encourages the person to repeat their behavior in the future. For example, if a baby points at something and receives it, they are naturally reinforced for pointing, meaning that they are encouraged to continue doing so because it helped adults accurately understand and meet their needs.

We emailed the authors of the study to learn more about their intent, and they clarified that parents were encouraged to respond naturally to baby-initiated interactions: if the baby’s behaviour suggests that they are trying to communicate or request something, the caregiver should respond accordingly. This teaches the baby that reaching out to others can be a positive and worthwhile experience. Thus, the “reinforcement” in the protocol seems to be nothing more than good, empathetic parenting.

Where did they fall short?

Although the intervention itself may have been consistent with the community’s expectations and goals, the researchers probably did fall short by failing to make this clear in the main body of the paper, or even the supplement. Ideally, community members shouldn’t have to send follow-up emails to reassure themselves of key details about research studies.

Moreover, although the authors of the paper were careful to frame their work in a neurodiversity-aligned way when they responded to backlash from the autistic community, their original paper is written in a very traditional, medical model-aligned manner. The first sentence describes autism as a “disorder” characterized by “impairments,” and much of the rest of the paper is devoted towards explaining how the intervention may have reduced autism “symptom severity” and made children less likely to receive an autism diagnosis. This is, of course, totally inconsistent with the goals of the neurodiversity movement. It might even risk resurrecting the spectre of the “refrigerator mother” by implying that parents, as the deliverers of the intervention, can control whether their children become autistic or not.

Why are the authors saying one thing in their paper, while talking about how the intervention “cherishes neurodiversity” elsewhere?  They probably weren’t trying to be duplicitous, especially since the intervention itself seems inoffensive, but many of the institutions of academia do tend to encourage orthodoxy—one might even say rigidity and insistence on sameness!—in academic papers. For example, the authors may have decided to write their paper the traditional way in order to get it through peer review more quickly; one of the downsides of academic peer review is that any departure from convention tends to elicit knee-jerk criticism. Moreover, researchers unfamiliar with the neurodiversity movement would have needed a neurodiversity framing to be justified and explained, lengthening the paper, and medical journals (such as the one the authors published in) tend to require short, simple papers.

Unfortunately, the authors made a choice to focus on how their intervention affected “autism symptoms” and autism diagnosis. Admittedly, studies take time to run—they chose their outcome measures more than five years ago, and they may well have regretted the choice later—but it was clearly not a choice in line with the priorities of neurodiversity advocates.

Did the study actually reduce autism diagnoses?

One important question is whether the media hype was justified: Did the study actually reduce the odds of autism diagnosis?  Statistically, it’s quite a bit more complicated than some of the simplistic media headlines suggest.

Crucially, the claim that “two thirds” of autism diagnoses were prevented is based on a tiny number of children. Of the 89 children that were examined by the researchers when they reached the age of three years old, only 12 children could clearly be diagnosed with autism.  Three of the diagnosed children had received the intervention, and nine were controls. These are extremely small numbers!

Could it have been a fluke?  If you repeatedly toss twelve coins, chance alone will sometimes give you more heads than tails, or more tails than heads…

Indeed, fewer children who received the intervention were indubitably neurotypical—only five—than children in the control group, eight of whom were assessed as clearly neurotypical.  Does this mean the intervention, while reducing the odds of autism, also reduced the odds of being neurotypical?

The researchers used a statistical test (i.e. a “Fisher exact test”) to determine whether children’s likelihood of being clearly autistic, clearly neurotypical, or potentially neurodivergent differed between intervention and control groups more than would be expected by chance. The result did not quite achieve statistical significance, meaning that the intervention group wasn’t necessarily either less autistic or less neurotypical than the control group: both trends might well have been flukes. As such, the media headlines about two-thirds of autism diagnoses being “prevented” were very premature!

It’s also worth noting—as the researchers did—that not all children can be reliably diagnosed as autistic when they are three years old. It’s possible that some children in the very large potentially-neurodivergent group, who fell short of meeting the formal autism diagnostic criteria, will still receive an autism diagnosis later in life. The intervention might have reduced autistic traits enough to delay diagnosis, but not prevent it.

Unfortunately, news media that need to sell headlines don’t always closely scrutinize research papers for these sorts of caveats. Moreover, sometimes it’s hard to spot the caveats unless you have the specialized training and knowledge that comes from being a researcher. Since plenty of people in the autism world who aren’t researchers are very interested in—and very affected by—autism research, we as a field need to do a better job clearly and transparently communicating the limitations of our work.

Did the study actually reduce autism traits?

Only follow-up monitoring will tell whether more study participants will come to meet autism diagnostic criteria as they grow older, but the researchers did say that they already have evidence showing that so-called “autism symptoms” were lower in the intervention group than the control group.

But what does it actually mean to have fewer “autism symptoms”? The main measure used by the researchers—the Autism Diagnostic Observation Schedule-2 (ADOS-2)—assesses a wide variety of autism-related traits, ranging from functional skills (e.g., pointing to communicate) to harmless differences (e.g., unusual phrasing), to possible stress responses (e.g., repetitive movements). On a positive front, a child could look “less autistic” because they are more skilled or less stressed, while still maintaining a fundamentally autistic neurology. Alternatively, a child could look “less autistic” because they have learned to suppress unusual behaviors, (i.e., mask their autism), without improving—and perhaps even hurting—their quality of life. ADOS-2 scores alone do not give us enough information to assess whether the outcome is desirable.

What do researchers need to do better?

Too often, neurodiversity and community perspectives are treated as an afterthought or a distraction, rather than a scientific priority. The aftermath of this study poignantly illustrates the need for widespread change. No matter how well-designed an intervention is in the laboratory, stigmatizing messages of “disorder” and “prevention” will damage the public’s trust in its potential. How can the field of autism research begin healing the rift with the autistic community?

On an individual level:

The authors of this study did many things right that should be repeated by future work. In addition to directly addressing the double empathy problem, they reportedly partnered with autistic people, and they investigated whether the intervention caused harm. One area of improvement would be being more transparent about these efforts so that the community can know how thoroughly their concerns were addressed. Researchers should also take care to communicate clearly, so that their work cannot easily be distorted or misrepresented in the news media.

More broadly, researchers need to fundamentally shift how they think about autism. Researchers often continue to conceptualize the neurotype solely in terms of deficits, despite ample evidence of associated strengths as well as harmless/neutral differences. A neurodiversity-informed framework would not ignore the challenges associated with autism, but rather balance that perspective with acknowledgement of valued aspects. A neurodiversity-informed researcher would select or develop outcome measures that reflect the autistic community’s goals for themselves, such as quality of life and mental health, rather than assuming that fewer “autism symptoms” or lack of an autism diagnosis is inherently good. Indeed, a neurodiversity-informed researcher might worry about whether falling short of diagnostic criteria could leave people without necessary supports.

Partnering with the community

It is very difficult to achieve the competence to navigate these complex issues without direct community voices. Although the researchers said their study emerged out of ongoing community partnerships, it’s unclear what this entailed. For researchers to secure community trust and establish genuine collaborations, autistic partners must represent a range of perspectives (e.g., race, gender, level of support need), and have meaningful power to challenge traditional researchers’ biases. If community collaborators are meaningfully included in the research team, community collaborators could, to name just a few examples:

  • Ensure that outcome measures and intervention goals reflect community priorities;
  • Ensure that intervention methods are likely to be helpful, not harmful; and
  • Anticipate how community members will engage with and respond to research papers, and revise the framing to be respectful and sensitive.

Perspectives of autistic parents may be particularly valuable and insightful regarding how to create supportive environments for autistic children.

Systemic changes needed in our field

Researchers may be frustrated by the slower pace of community collaboration. This is understandable, because many researchers face intense career pressure to publish lots of papers quickly (“publish or perish,” some say). However, this systemic issue merely emphasizes how our field must change. Partnering with the community should not be seen as “slowing down progress” any more than a thorough literature review is “slowing down progress.” Research results cannot be considered meaningful if important context has been excluded in the name of saving time.

Unfortunately, there is no shortage of systemic barriers facing those idealistic researchers who are already attempting to partner with the community or re-evaluate traditional assumptions. It can be difficult to secure funding or publication for efforts that require readers to rethink their assumptions. Journal editors often require tight word counts, leaving little room for nuance.

It will be hard for the community to trust research until the autistic perspective is meaningfully represented at every level, from grant funding to ethics approval to journal peer review. Some of these changes are beginning to occur, with autistic researchers and other community members reviewing grant applications at the AIR-P Network, journal submissions at Autism in Adulthood, and to some extent NIMH grant applications, but much more needs to be done. This will require not only greater inclusion of community members in these processes, but also proactive support for developing autistic autism experts, who are often excluded from education and employment by lack of accommodation or even direct stigma. These sorts of large-scale systemic changes would help researchers, autistic laypeople, and parents gain confidence that research is truly working towards the best interests of the people it aspires to help.

In summary, it appears that many concerns about this infant intervention were misplaced. The core message of Whitehouse et al. is no different than what neurodiversity advocates have been saying all along: neurodivergent children, like their neurotypical peers, will be more socially engaged when they are treated as human beings with meaningful perspectives and intentions.

At the same time, we do not intend to say that the outcry from the autistic community was unreasonable. This incident has highlighted much deeper issues in the field of autism research. Why did the authors feel the need to dress up their message in the language of pathology and prevention? What if the headlines had read, “Autism doesn’t have to be a problem if children are understood and supported”? Exciting new visions are emerging for research that can be easily embraced by the neurodiversity movement. We must all do our part to further these efforts and help create a future where new generations of neurodivergent children are accepted, understood, and able to thrive.