Nina Park: Hey, glad you're — okay wait, I always say that, ignore me. How was the rest of your week?
Walt Garner: Quiet. Which I suspect is about to change.
Nina Park: Yeah, the topic today has a certain weight to it — I want to ease into it right, because I think it's easy to either oversell or kind of flatten what's actually going on here.
Walt Garner: Quite so. And the science itself resists easy framing.
Nina Park: So here's the plain version first. Imagine your brain has something like a face detector — a system that fires whenever it sees a human face. In most kids, that detector gets sharper as they grow up, year by year, recognizing faces faster and more precisely. This study — John McPartland at Yale School of Medicine, published in Nature Mental Health, reported by ScienceDaily on October 27th and 28th of 2026 — found that autistic children's brains produce a less neurally distinct signal when they see faces. And unlike neurotypical kids, that signal doesn't get sharper with age. It's on a totally different developmental arc.
Walt Garner: That word — arc — is, I think, the key. Griffin, who was quoted in the ScienceDaily report, put it precisely: 'their face-specific processing is not following the same trajectory.' And that framing matters enormously. It displaces the older deficit language.
Nina Park: Not less able — differently timed, differently shaped. Yeah. And this was 400 kids from the Autism Biomarkers Consortium for Clinical Trials, the ABC-CT dataset, which — I mean, for EEG autism research that's a substantial number.
Walt Garner: Now, what the study adds beyond the trajectory claim is that the neural differences appear to emerge before the N170 signal — which has been, since the early 2000s, the primary EEG marker everyone uses to study face processing. So the divergence begins upstream of where researchers have been looking.
Nina Park: Before the N170. Okay that actually — huh, that recalibrates everything, doesn't it.
Walt Garner: It potentially does, yes. The N170 fires at roughly 170 milliseconds post-stimulus. If the distinguishing signal in autistic children is even earlier than that, then the decades of comparative work anchored to N170 response rates was, at minimum, missing the opening act.
Nina Park: And the framing the researchers used — they were very deliberate about this being a difference in signal quality, not evidence of avoidance, not about the kids looking less. That distinction feels important to me in a way I want to sit with a little longer.
Walt Garner: Indeed. Because if we conflate looking behavior with neural distinctiveness, we've already slipped back into a behavioral explanation dressed in neuroscience clothing.
Nina Park: But that's actually the part I want to pull on — because saying 'different signal quality' sounds clean, but the methodology is doing a lot of work there. The old Yale fMRI studies from the early 2000s were looking at fusiform cortex activation, a specific region. This 2026 study is scalp-wide EEG — like, the whole head, not one spot.
Walt Garner: And that distinction is not cosmetic. fMRI resolves *where* in the brain — spatially precise — but it's temporally sluggish. You can't catch a signal that fires in the first hundred milliseconds of face perception with fMRI, structurally you just can't. EEG gives you that temporal resolution, and when McPartland's team spread the measurement across the whole scalp, they caught something distributed and early that the region-specific approach was simply blind to.
Nina Park: So the old work wasn't wrong, it was — incomplete? Or actually wrong?
Walt Garner: Both, maybe. And I say that carefully. The early 2000s Yale finding — lower fusiform cortex activation — was real. But it named a location and called it the mechanism. What the 2026 scalp-wide EEG is showing is that the divergence begins even before the N170 fires, before that 170-millisecond window the field has been anchoring to for twenty years. So the fusiform finding may have been a downstream echo of something that started earlier and more broadly.
Nina Park: A downstream echo. That's — yeah, that reframes it completely.
Walt Garner: Well, and that's precisely why the ABC-CT dataset matters here — roughly 400 children is large enough that you can actually trust the earlier signal isn't noise. You'd expect noise to wash out at that sample size.
Nina Park: But here's what I'm stuck on — and I don't know if I'm right — we're saying the signal is less neurally *distinct*, not absent. But does the scalp-wide EEG actually let you say something about what the experience of that is? Like, the mechanism is different, but is the child perceiving a face any differently?
Walt Garner: Now that is the gap the imaging cannot close. You can measure neural distinctiveness — how sharply the brain marks 'face' versus 'not-face' — but you cannot read from that whether the child's subjective experience of seeing a face is impoverished, or simply organized differently.
Nina Park: Which matters enormously if we're about to hand parents a brain scan and say 'see, less distinct signal' — because that slides very easily into deficit language with a neuroscience badge on it.
Walt Garner: Indeed. And the neurodiversity frame is genuinely trying to resist that — to say different developmental trajectory, not broken developmental trajectory. But the imaging itself is neutral on that question. The 2026 study shows the arc diverges. It doesn't say the divergent arc produces worse outcomes for the child.
Nina Park: Right — but the part that doesn't fit is that the age-related sharpening, the thing neurotypical kids get automatically, is absent. So it's not just a different curve, it's a curve that isn't accelerating the same way. Is that neutral?
Walt Garner: That's the honest tension in the finding, and I don't think McPartland's team resolves it — they can't, not from EEG alone. What they've done is move the question upstream. The mechanism starts earlier than N170, it's more distributed than the fusiform story suggested, and now we actually have to ask what that earlier divergence *does* — which is a different and harder question than the one the field has been asking.
Nina Park: And that question — what the earlier divergence actually *does* — that's where I want to go, because I keep thinking about this concrete scenario. Picture an eight-year-old autistic kid at a birthday party. She's scanning the room, she's looking at faces. Her brain is not ignoring faces. But every social encounter that a neurotypical kid's brain is quietly using to sharpen its face template — her brain isn't doing that same sharpening. The circuitry is on a different curve.
Walt Garner: Yes, and that image captures something the data alone cannot. The neurotypical child's face-recognition system is being *calibrated* by every birthday party, every classroom, every dinner table — passively, automatically.
Nina Park: She's not falling behind. She's just — not on that road.
Walt Garner: Which is why the Floris et al. study in Nature Mental Health becomes relevant here — and the finding is, I think, genuinely clarifying. Dorothea Floris, Simon Baron-Cohen, the EU-AIMS LEAP group — they examined 204 autistic and nonautistic individuals, aged seven to thirty, across four imaging modalities. And the fusiform gyrus changes they found were associated specifically with social functioning. Not nonsocial functioning. That distinction is — well, it's doing a great deal of work.
Nina Park: Wait — specifically social? Not just face recognition generally?
Walt Garner: Specifically social. The region everyone pointed to as the 'face recognition area' turns out to be tied to *relationships*, not simply to visual identification of a face.
Nina Park: Okay that actually — I mean, that reframes the whole thing. Because the old story was: fusiform fires for faces, autistic kids have weaker fusiform response, therefore face recognition problem. But the Floris data is saying the fusiform is doing something much more social than that. It's not a visual decoder, it's more like — a relationship builder?
Walt Garner: That may be overstating it slightly, but the directional implication holds. If the multimodal fusiform component predicts social outcomes rather than perceptual ones, then the birthday-party child isn't struggling to *see* the faces. Something about the social weight those faces carry is being processed differently.
Nina Park: Which — and this is where I think the heterogeneity question bites back hard — because my friend Marcus would say he processes faces fine and I actually believe him. The 204-person Floris sample, the 400-child ABC-CT dataset — are we averaging over a range of different things and landing on one number?
Walt Garner: That is, I think, the honest limitation neither study fully resolves. The population of autistic individuals is profoundly heterogeneous, and what reads as a coherent divergent trajectory in aggregate may, at the individual level, be several different trajectories.
Nina Park: And that's actually the part I think we haven't gotten to yet — because the difference-versus-deficit reframe is warranted by this data, genuinely, but the distributed-network picture also means this isn't one broken region, and I don't think either framework should flatten what the individual variation actually looks like.
Walt Garner: Now, that is precisely where I think this conversation needs to go next.
Nina Park: But flattening is exactly what happens — and the distributed-network data is what makes it so easy to do by accident. Because if it were one region, one broken thing, you could at least say 'this specific system, this specific person.' But the meta-analysis of 25 fMRI studies — 632 autistic participants, 568 nonautistic — found decreased activation spread across inferior frontal gyrus, fusiform gyrus, middle temporal gyrus, cerebellum, insula. That's not a region. That's a whole conversation happening across the brain.
Walt Garner: And cerebellum, notably. Which no one's face-processing story ever leads with.
Nina Park: Right — the cerebellum is in there and I just — why is nobody talking about that?
Walt Garner: Well, because the fusiform story is cleaner to tell. But the inverted-face fMRI study — PMC3047502 — found that autistic and typically developing children showed *similar* fusiform face area responses. Similar. The difference was in prefrontal cortex and amygdala activity. So we've been centering the wrong region for, arguably, two decades.
Nina Park: Hold on. Similar fusiform responses?
Walt Garner: Similar. Which means the visual identification machinery may be intact — it's the frontal-limbic circuitry, the part that assigns weight and meaning to what's been identified, that diverges.
Nina Park: Okay that's — I mean, that's actually the 'oh, that's why' moment for me. Because if the fusiform is doing its job and the divergence is in prefrontal and amygdala, then we're not talking about recognition at all. We're talking about what the brain does with a face once it's recognized.
Walt Garner: And then Taeyeon Kim's Communications Biology study from October 26th, 2026 adds another layer — hierarchical distortions in structure-function coupling, with atypical connectivity specifically in the thalamus and putamen. Which are subcortical. We're now below cortex entirely.
Nina Park: So the network keeps getting wider every time someone looks carefully.
Walt Garner: Which is precisely why replacing 'avoidance' with 'weaker neural distinctness' sounds like progress — and may not be. We've traded a behavioral explanation for a neurobiological one, and the neurobiological one keeps expanding its footprint. At some point you have to ask whether 'face-processing difference' is still a coherent unit of analysis, or whether we've named a cluster and called it a mechanism.
Nina Park: And the Frontiers in Psychiatry editorial from October 2026 — the one on sensory processing as a fundamental dimension — makes that cut even deeper, because it points out that Leo Kanner's original clinical descriptions already included sensory atypicalities. We sidelined them for decades in favor of the social-communication frame.
Walt Garner: Yes. And now we're rediscovering them with thalamus connectivity data and calling it new. The mechanism is genuinely new — but the observation is seventy years old.
Nina Park: Which loops back to the heterogeneity problem, because the researchers themselves say — explicitly — findings don't yet establish diagnostic tests or treatment pathways, and differences are heterogeneous across autistic individuals. So 'different not deficient' can't just become the new universal claim either. It's not a cleaner story. It's a more honest acknowledgment that there might not be one story.
Walt Garner: Now that, I think, is where the science actually lands in 2026. Not a reframe that resolves the deficit debate — a distributed, temporally early, heterogeneous network that resists any single frame. That's genuinely uncomfortable. And it may be the most useful thing the data has produced.
Nina Park: And that's the thing I can't settle — because the ABC-CT sample, 400 kids, that's real, that's substantial, but it's still a particular slice. Kids who were identified, enrolled, assessed. Which means the heterogeneity we keep invoking is probably even wider than what McPartland's team could see.
Walt Garner: Indeed. And the prevalence numbers compound that. The Molecular Biomedicine review is quite explicit — rising ASD prevalence reflects broadened diagnostic criteria and improved screening, not a true incidence increase. So the population we're generalizing about keeps expanding at the edges, even as the datasets we're studying from remain bounded.
Nina Park: So we might be on the edge of something — mechanism-informed support, not just behavior correction — but only if the people building interventions resist the pull to take this distributed, early, heterogeneous picture and just... bolt it onto the same deficit scaffold with sharper imaging.
Walt Garner: That is the risk. New images, same assumptions.
Nina Park: Yeah. And I don't know how you guard against that — I mean, honestly. That's the part I'm still holding.
Walt Garner: Quite so. As is the field, I think — whether it knows it or not.
