Is the Autism Spectrum Broken? One Pioneer Says

There's a particular kind of intellectual courage in a scientist who spends decades building a framework and then turns around and says: we may need to burn this down and start over.

That's roughly where Uta Frith finds herself. In a recent conversation with New Scientist editor Thomas Lewton, the developmental psychologist—who helped establish two of the field's landmark cognitive theories and was among the first researchers to point brain scanners at the autism question—argued that the autism spectrum, as currently constructed, may have become scientifically incoherent. The diagnostic criteria, she says, have been "gradually widened... diluted. The bar was becoming very low."

The claim is not that autism isn't real. It's that the word "autism" may now be doing too much work—covering experiences so different from one another that grouping them together tells us less than it should.

How We Got Here

Frith's career is essentially a guided tour of how profoundly the field has shifted. When she first encountered autistic children at London's Maudsley Hospital in the 1960s, the dominant explanation was psychogenic: the so-called "refrigerator mother" hypothesis, which blamed cold, unloving parenting for a child's withdrawal. Frith found this unconvincing immediately—she'd spoken with those mothers and found the theory more about narrative than evidence.

What followed was decades of empirical work trying to identify what was actually going on at the cognitive level. Her framing here is worth pausing on. Frith describes a layered model of explanation: observable behavior at the surface, cognitive mechanisms beneath it, brain structure below that, and genetics and biology underlying everything. The psychoanalytic tradition, she says, was interesting "in the sense that reading a novel is interesting"—but it didn't give you something you could actually test.

The shift came in the 1970s, when twin studies provided the first hard evidence of a strong genetic component. That moved the conversation out of the mother's kitchen and into the genome. From there, Frith and her colleagues spent years constructing cognitive models that could explain autism's particular signatures.

Two Theories, Two Signatures

The first theory Frith helped develop—mentalizing, sometimes called Theory of Mind—emerged in the 1980s and is probably best explained by the Sally-Ann test, a paradigm she describes in the conversation. The setup is simple: two dolls, a marble, a basket, a box, and a key question about what Sally believes when she returns to a scene she didn't witness.

Most children by age five handle this without much difficulty. They understand that Sally's belief is based on what she saw, not on what's actually true. Autistic children, at the age they were originally tested, pointed to where the marble really was—as if Sally's knowledge matched their own. They solved the factual question rather than the mental-state question.

The finding was striking. But Frith is careful not to overstate it: "We don't see autism as rooted in that. We see the particular difficulties in social communication, which are only one aspect of autism, are possibly rooted in a difficulty in mentalizing." Even she frames her own landmark theory as a partial explanation, one that "may only be true for a small subgroup."

The second signature is something completely different: a strong pull toward detail over big picture. Frith noticed early on that autistic children could complete complex jigsaw puzzles upside down—without the image to guide them—by processing individual piece shapes rather than seeking a gestalt. They could spot hidden figures embedded in busy patterns. What looked like an unusual perceptual strength turned out to be connected to a processing style—one she believes is normally distributed across the population, but which, at the extreme end, shows up heavily in autism.

This is also the territory of savant abilities: the narrow, deep mastery of drawing, music, or recall that a small proportion of autistic individuals develop. Frith suspects the genetic underpinnings of this processing style are probably entirely distinct from those driving the social communication differences. In other words: autism may not be one thing with a spectrum of severity. It may be several different things that historically arrived in the same diagnostic room.

The Spectrum Problem

This is the friction point—and the one with real stakes beyond academic debate.

Since the 1990s and accelerating through the 2000s, autism diagnoses have risen sharply, especially among women and girls. Much of this reflects genuinely valuable progress: recognition that autism presents differently in women, that masking and camouflage had been hiding diagnoses for decades, and that the neurodiversity framework—treating cognitive difference as variation rather than deficit—opened space for people who'd spent years being told something was wrong with them.

Frith's challenge to this trajectory is uncomfortable, and she knows it. Her argument is that broadening the diagnostic criteria without an objective biological marker has made the category so elastic that it now contains experiences with, in her words, "absolutely no overlap" at opposite ends. Someone who cannot communicate verbally and requires full-time support, and someone who struggles socially but holds a job and navigates independent life—are these the same condition? Or have we drawn a circle so large it no longer carves nature at its joints?

The absence of an objective diagnostic measure is central to this problem. "We don't have an objective way of diagnosing autism," Frith says plainly. "That is the reason why there is so much confusion and debate and vagueness." Behavioral criteria, she notes, are malleable. Without a biological ground truth, the criteria shift with cultural pressure, clinical consensus, and advocacy—all of which have legitimate roles, but none of which are the same as a blood test.

Who Gets to Define It

Here's where the conversation gets genuinely complicated, and where Frith's perspective runs into serious friction from another direction.

The neurodiversity movement—and many autistic people themselves—has argued forcefully that the definition of autism shouldn't rest exclusively with clinicians and researchers. Lived experience has diagnostic weight. The expansion of the spectrum wasn't just a scientific accident; it reflected advocacy by people who recognized themselves in the criteria and found the label validating and useful.

Frith's skepticism about masking—she reportedly doubts its existence as a distinct mechanism—is perhaps the sharpest edge of her position. For many late-diagnosed autistic women, masking is not a theoretical construct. It's a description of an exhausting, decades-long performance of neurotypicality that cost them enormously. Dismissing it isn't a neutral scientific move; it lands as an erasure.

The tension is real and it doesn't resolve neatly. Frith is not arguing that recently-diagnosed people aren't struggling or don't deserve support. She's arguing that lumping their experiences under the same diagnostic umbrella as profoundly autistic individuals may serve no one well—scientifically or practically. But the people most likely to lose something if the spectrum contracts are often the ones with the fewest other diagnostic options.

Frith calls for scientists to be "brave enough" to rethink the whole framework. What she doesn't fully address is what a dismantled spectrum gets replaced with—and who makes that call.

That might be the question the field needs to sit with, rather than answer too quickly.

— Nadia Marchetti, Unexplained Phenomena Correspondent, Buzzrag