Measured Wrong on Purpose: Gestalt Language, Core-Deficit Thinking, and the Limits of Autism Research

A critique of autism language research, core-deficit thinking, and why gestalt processing is still mistaken for impairment rather than recognised as a lifelong way of making meaning

Another year, another autism study measuring language through deficit frames. From lived experience as a gestalt processor and core-deficit critique, this piece asks why difference is still read as delay—and who that serves.

Opening — Another Year, the Same Frame

Another year, another paper that reads as if time has politely stood still.

Kritsotakis, G., & Morfidi, E. (2025). Reading Comprehension Challenges in Autism Spectrum Disorder: Linguistic Factors and Figurative Language Proficiency. Autism & developmental language impairments, 10, 23969415251371544. https://doi.org/10.1177/23969415251371544

The citations are newer. The statistics are tighter. The language has softened around the edges—heterogeneity acknowledged, variability noted, limitations carefully listed. And yet the conclusions land in exactly the same place they have for decades. Autism is still framed as a problem of missing pieces. Language is still treated as something that develops correctly only if it assembles in the right order. Difference is still read as delay.

There is a particular kind of fatigue that comes from reading autism research closely over time. Not the fatigue of ignorance, but the fatigue of recognition. You begin to know the ending before you reach the methods section. You can predict which findings will be called “surprising,” which skills will be labelled “intact,” and which divergences will be explained away as compensation rather than taken seriously on their own terms. The paper feels new only in its publication date.

Kritsotakis and Morfidi’s 2025 study on reading comprehension, morphosyntax, and figurative language in autistic children is, by any reasonable standard, careful work. The authors are thorough. They attend to variability. They avoid the cruder claims that once dominated the field. This is not bad research. And that is precisely the problem.

Because even here—perhaps especially here—the frame does not move.

This is not a hit piece. I am not interested in dismissing the authors, their methods, or their intentions. What follows is a critique of the structure within which this work takes place—a structure that cannot recognise gestalt language processing as language at all, except as something that has failed to become something else.

I write this from two positions at once. First, from my own lived experience as an autistic gestalt language processor, not in childhood but across an adult life—through study, teaching, writing, relationship, and time. And second, from within a growing body of developmental critique articulated clearly by Astle and Fletcher-Watson, who argued five years ago that the field must move beyond core-deficit thinking if it is to understand neurodevelopmental difference in any meaningful way.

Astle, D. E., & Fletcher-Watson, S. (2020). Beyond the Core-Deficit Hypothesis in Developmental Disorders. Current directions in psychological science, 29(5), 431–437. https://doi.org/10.1177/0963721420925518

The paper I am examining did not miss that argument because it was careless. It missed it because the frame it inhabits does not yet know how to see what it would require acknowledging.

And so, another year goes by.
Another careful study.
Another familiar conclusion.

The question is no longer whether autistic language development is heterogeneous. Everyone now agrees that it is.

The question is whether the field is prepared to accept that some of what it keeps measuring as impairment is, in fact, a perfectly ordinary human way of making meaning—just not the one it was trained to recognise.

What the Paper Claims to Study — and What It Assumes Instead

At face value, the aims of Kritsotakis and Morfidi’s study are modest and reasonable. The authors set out to examine reading comprehension in upper primary autistic children without intellectual disability, alongside three related domains: receptive vocabulary, morphosyntactic ability, and figurative language comprehension. They attend explicitly to heterogeneity, dividing the autistic sample into higher- and lower-achieving readers in an effort to move beyond a single, flattened profile.

This is, in many ways, exactly what contemporary autism research claims it wants to do.

Reading comprehension is treated as a complex outcome rather than a unitary skill. Morphosyntax is examined across a range of grammatical structures rather than through a single proxy. Figurative language is included as a pragmatic dimension often neglected in literacy research. Variability within the autistic group is acknowledged rather than dismissed as statistical noise. All of this reads as progress.

And yet the progress is circumscribed before it begins.

The study is published in Autism & Developmental Language Impairments, a venue whose title already performs an act of classification. Language differences are framed, by default, as impairments. Development is presumed to move toward a known endpoint. Deviation is something to be measured, explained, and—eventually—corrected.

Within that context, the authors’ choices begin to matter in quieter ways.

Reading comprehension is assessed through short narrative and expository texts followed by multiple-choice questions designed to probe literal and inferential meaning. Morphosyntactic knowledge is measured through forced-choice sentence completion. Figurative language is tested using decontextualised idioms and proverbs, presented in isolation and scored for conventional interpretation. These are familiar tools. They are also highly specific tools—ones that privilege analytic extraction, rapid selection, and alignment with standardised semantic targets.

What is not questioned is whether these tools are neutral.

The developmental framing is equally inherited rather than interrogated. Improvement is described in terms of age-related gains and proximity to ‘typically developing norms.’ Subgroups are defined by higher or lower achievement within the same measurement system. Reasoning ability is interpreted as compensatory when it predicts comprehension outcomes that language measures do not fully explain.

At no point does the study ask whether the order of operations it assumes—vocabulary, then syntax, then pragmatics, then comprehension—is itself culturally and cognitively specific. Nor does it consider whether meaning might be organised, accessed, or stabilised differently in some readers without constituting a deficit.

This is not a flaw unique to this paper. It is the inheritance of a field.

Before any child is assessed, language is already defined.

Gestalt Language Processing: Not a Deficit, Not a Phase

Gestalt language processing is often spoken about—when it is spoken about at all—as something children either grow out of or fail to resolve. A precursor. A workaround. A developmental stage that should eventually give way to something more properly linguistic.

That framing does not match lived reality.

For those of us who process language gestalt-first, this is not an early strategy abandoned with maturity. It is a lifelong architecture. Meaning arrives whole, patterned, and relational before it is decomposed—if it is decomposed at all. Language is organised around coherence rather than components. Synthesis does not happen sentence by sentence, but across time, context, and accumulated resonance.

This does not disappear in adolescence. It deepens.

As an adult autistic reader and writer, I do not approach text by assembling meaning from discrete grammatical cues and then checking my work against a list of expected inferences. Meaning arrives as a field. Tone, intent, trajectory, and implication register before individual parts are consciously identified. Details are filled in later, selectively, when the situation requires it.

This is particularly true for figurative language. Metaphor, for a gestalt processor, is not something to be decoded so much as something to be entered. Its meaning is felt in use—within a scene, a relationship, a cultural moment. Remove it from that context and demand a single, abstract interpretation, and you are no longer testing figurative understanding. You are testing willingness or ability to translate lived meaning into an analytic format.

There are many moments in my own education where comprehension arrived late—sometimes minutes, sometimes hours, sometimes days after the text was first encountered. The synthesis happened quietly, in the background, without announcing itself in the ways standard assessments are trained to recognise. When it arrived, it was complete. What it was not, always, was immediately demonstrable.

This is the core mismatch the paper never considers.

Gestalt processing is meaning-first, not part-first. It seeks coherence rather than extraction. It relies on resonance rather than compliance with task demands. When comprehension is measured through rapid selection among predefined options, the question being asked is not do you understand? but can you show your understanding in this specific way, on this timescale, using these units?

The distinction matters.

Difficulty demonstrating understanding under constrained conditions is not the same thing as difficulty understanding. When that distinction is collapsed, a perfectly functional language system begins to look like an impairment—simply because it refuses to perform on cue.

And from there, the conclusions write themselves.

Figurative Language, Decontextualised: When the Test Breaks the Meaning

One of the study’s central claims concerns figurative language. Autistic children, the authors argue, show particular difficulty with idioms and proverbs, and this difficulty is positioned as evidence of pragmatic weakness contributing to reduced reading comprehension.

The way figurative language is measured here matters.

Participants are presented with idioms and proverbs in isolation, stripped of narrative, social context, or speaker intent, and asked to select the correct meaning from a set of predefined options. This is a familiar paradigm in developmental research. It is also a profoundly specific one.

Figurative language does not function as a standalone object. It is not a riddle with a single hidden solution. It is a relational act—something speakers do with other people, inside situations that carry tone, history, and consequence. Its meaning emerges from use, not from abstraction.

For a gestalt processor, this distinction is not academic. Metaphor is understood by entering the scene it evokes. A phrase resonates because it aligns with a lived pattern—emotional, social, cultural—not because it can be translated into a paraphrase on demand. When an idiom is pulled out of context and presented as a multiple-choice puzzle, that resonance is disrupted. The meaning does not become clearer. It collapses.

This is not because the figurative meaning is inaccessible, but because the task has removed the very conditions under which that meaning coheres.

In my own experience, metaphor works most reliably when it is doing real work—when it sharpens a point, softens a blow, signals belonging, or reframes a situation that cannot be named directly. Asked to define that same metaphor in the abstract, I may hesitate—not from confusion, but from a sense that the question has severed the thing it is trying to measure from the context that gives it life.

Narrative scaffolding matters. Social intent matters. Timing matters. Without these, figurative language becomes something else entirely: a test of whether one’s internal sense-making can be flattened into an externally approved format.

When the paper reports lower figurative language scores, what it is really capturing is not a failure to understand metaphor, but a mismatch between how meaning is held and how meaning is demanded.

What is measured here is alignment with analytic norms, not figurative competence.

“Compensation” and the Politics of Sequencing

One of the study’s key interpretive moves is the claim that autistic readers rely on nonverbal reasoning to compensate for weaknesses in linguistic processing. Reasoning ability is positioned as something that steps in when morphosyntax or figurative competence falls short, helping to support reading comprehension despite underlying deficits.

The language is careful. The assumption is not.

Compensate for what, exactly?

The idea of compensation only makes sense if there is an agreed sequence of operations: language first, reasoning second. Syntax and pragmatic knowledge are treated as the proper foundation of comprehension, with reasoning invoked as a secondary support when that foundation is unstable. Meaning is presumed to flow upwards from parts to whole.

For gestalt processors, that is not how language works.

Reasoning is not a workaround. It is not an emergency measure brought in to patch a hole in an otherwise typical system. It is a primary route to meaning. Pattern recognition, analogy, and global coherence are how comprehension is established in the first place. Syntax is accessed when needed, not climbed step by step as a prerequisite for understanding.

When a gestalt processor reads, the overall structure of a text—its direction, its internal logic, its emotional and conceptual contours—often registers before individual grammatical details are consciously parsed. Reasoning does not rescue comprehension. It creates it.

To describe this as compensation is therefore to impose a hierarchy that does not belong to the reader. It treats analytic language processing as the default and all other routes as deviations. It converts architectural difference into strategic patchwork.

This matters because once reasoning is framed as compensatory, it cannot be recognised as legitimate. It becomes evidence of deficit rather than of design. Strengths are acknowledged only insofar as they soften the appearance of weakness.

The reframing here is simple, and unsettling for the field: what is being called compensation may in fact be the system working exactly as intended.

The problem is not that autistic readers are relying on the wrong tools. It is that the research insists on measuring language in an order that many humans never use.

Astle & Fletcher-Watson: Why This Keeps Happening

Five years before the study under discussion was published, Duncan Astle and Sue Fletcher-Watson laid out a clear diagnosis of the problem the field keeps circling. In Beyond the Core-Deficit Hypothesis, they argue that developmental psychology remains trapped in a model that seeks a single underlying impairment to explain complex, heterogeneous profiles—and that this model has failed repeatedly.

Their critique is not subtle. Core-deficit theories, they argue, are theoretically exhausted. They persist not because they explain developmental difference well, but because the methods used to study development continue to reproduce them. Highly selective samples, case–control designs, narrow task batteries, and univariate analyses all conspire to make difference look like deficit, even when the data themselves resist that interpretation.

Heterogeneity, in this framework, becomes a problem to be managed rather than a signal to be understood. Variability is smoothed, split, or averaged away. When it cannot be ignored, it is organised into hierarchies—higher and lower functioning, stronger and weaker profiles—without any accompanying theory of architecture.

Reading Kritsotakis and Morfidi through this lens is clarifying.

The study uses familiar, inherited measures—tasks already embedded in analytic assumptions about how language should work. It relies on case–control logic to define difference against a normative baseline. It addresses heterogeneity by dividing the autistic sample into higher- and lower-achieving readers, implicitly treating variability as a matter of degree rather than kind. And it interprets reasoning as compensatory because the theoretical framework has no other place to put it.

What is missing is not care or competence. What is missing is a theory capable of recognising multiple language architectures without ranking them.

Astle and Fletcher-Watson warn explicitly that until methods change, conclusions will not. Theoretical critique alone is insufficient if the same tasks, samples, and analytic habits continue to structure the work. The result is a literature that repeatedly rediscovers its own blind spots and then mistakes them for empirical findings.

Seen this way, the pattern is no longer surprising.

The paper does not misunderstand autism by accident. It misunderstands it by design.

And until the field is willing to abandon the idea that development must be explained through deficits in a presumed universal sequence, it will continue to generate careful studies that describe difference accurately—and interpret it incorrectly.

The Missing Lifespan: Autism as a Childhood That Never Grows Up

Like much of the autism–language literature, the study under discussion is confined to childhood. Its participants are upper primary school children. Its benchmarks are school-based. Its developmental claims are anchored to age norms derived from educational trajectories. Adulthood does not enter the frame, except implicitly—as an endpoint children may or may not approach.

This is so familiar it barely registers as a choice. And yet it is one of the field’s most consequential habits.

Autism is persistently treated as a ‘childhood condition’ with adult remnants, rather than as a lifelong way of being that continues to develop, reorganise, and deepen across decades. Language development is assumed to culminate in early adolescence, after which differences are either resolved or fossilised. There is little room for the possibility that autistic language continues to evolve—just not along the lines the field expects.

For gestalt language processors, this omission is not incidental. Gestalt processing does not stall, per se. It matures over time. Symbolic layering becomes denser. Connections across domains grow more precise. Meaning is held across longer spans of time, more complex contexts, and more nuanced social worlds. Coherence increases—not because language converges on an analytic norm, but because the internal system becomes more finely tuned.

The study’s interpretive frame, however, can only recognise development as convergence. Improvement is described as “approaching typical levels,” “bridging gaps,” or compensating more effectively for persistent weaknesses. Difference that does not narrow toward the norm remains, by definition, a problem.

This is where the “catching up” narrative does its quiet work. Autistic children are positioned as earlier versions of typically developing ones, progressing along the same path at a slower pace. What cannot be reconciled with that path is rendered delay, deficit, or failure to generalise.

That story does not survive contact with autistic adulthood.

I am not a delayed child. I am an autistic adult whose language has continued to develop—through study, through writing, through teaching, through relationship, and through time. My comprehension is not an approximation of something else. It is the outcome of a system that was always organised differently.

When research refuses to look beyond childhood, it mistakes absence of evidence for evidence of absence. It fails to see what emerges later, more slowly, and on different terms. And in doing so, it reinforces the idea that autistic language is something to be outgrown, corrected, or compensated for—rather than something that grows up in its own right.

Until the field is willing to follow autistic people across the lifespan, it will continue to misread development as delay, and difference as immaturity.

Heterogeneity Without Architecture Is Just Ranking

One of the paper’s stated strengths is its attention to heterogeneity. Rather than treating autistic readers as a single, undifferentiated group, the authors divide their sample into higher- and lower-achieving readers and analyse patterns within and across these subgroups. This move is presented as a way of respecting variability and avoiding overgeneralisation.

The intention is sound. The execution remains limited by the same underlying frame.

Splitting a group into high and low achievement categories still assumes a single developmental system, measured along a single axis, with individuals distributed at different points on a shared scale. It treats heterogeneity as a matter of degree rather than kind. Some children are closer to the normative endpoint; others are further away. The architecture itself is never in question.

This is why subgrouping so often collapses into ranking.

The higher-achieving group is described as approaching typical performance, with fewer and narrower weaknesses. The lower-achieving group is positioned as more impaired, more delayed, or more in need of support. Variability is acknowledged, but only to be ordered. What differs is how well the same system is assumed to be functioning.

From a gestalt processing perspective, this misses the point.

Different readers may be operating with different language architectures altogether—organising meaning through different sequences, stabilising comprehension at different scales, and prioritising coherence over extraction in ways that do not map cleanly onto achievement hierarchies. A child who performs poorly on decontextualised figurative language tasks may nevertheless demonstrate sophisticated narrative understanding when meaning is embedded in lived context. Another may show strong analytic skills without global coherence. These are not weaker and stronger versions of the same system. They are different systems doing different work.

Astle and Fletcher-Watson argue that until developmental science moves beyond core-deficit thinking, heterogeneity will continue to be mismanaged rather than understood. Variability must be theorised, not merely partitioned. This requires models that can accommodate multiple routes to comprehension, multiple developmental sequences, and multiple forms of coherence without forcing them into a single ranking scheme.

Without such models, acknowledging heterogeneity becomes a cosmetic gesture. The language softens. The categories multiply. But the underlying logic remains unchanged.

Difference is still measured against one presumed norm.
And diversity of architecture is reduced to degrees of success or failure within it.

What a Gestalt-Aware Study Would Have to Do Differently

If gestalt language processing were taken seriously—not as a complication, but as a legitimate language architecture—the changes required would not be cosmetic. They would be methodological, conceptual, and political.

First, tasks would have to be embedded in narrative and context. Meaning would be assessed where it actually lives: in stories, conversations, extended texts, and relational exchanges. Figurative language would be evaluated in use, not in isolation. Comprehension would be traced through coherence across time, not inferred from the ability to select a sanctioned paraphrase under artificial constraints.

Second, studies would have to follow development longitudinally. Not to track whether autistic children “catch up,” but to observe how meaning-making reorganises, stabilises, and deepens over years and decades. Gestalt processing does not announce its endpoints early. It unfolds. Any research design that cannot tolerate delayed synthesis is guaranteed to misread it.

Third, autistic adults would have to be included—not as outcomes of childhood intervention, but as language users in their own right. Adult GLP is not rare. It is simply under-theorised. The symbolic density, metaphorical sophistication, and narrative coherence visible in autistic adult writing, teaching, and theory-making do not emerge from nowhere. They are the result of developmental processes the field has largely chosen not to study.

Fourth, the measures themselves would have to change. Extraction would no longer be the gold standard. Coherence would be. Researchers would need tools capable of recognising meaning that is distributed, relational, and temporally layered—meaning that does not reduce cleanly to discrete responses but is nonetheless stable, precise, and real.

And finally, theory building would have to be autistic-led. Not “informed by lived experience” as a rhetorical flourish, but grounded in it. Gestalt processing has already been described in rich, developmentally coherent detail by practitioners and clinicians who actually work with gestalt processors across time. If the field is feeling adventurous, it could start by borrowing Marge Blanc’s data. 😳😉

This is not speculative. None of this requires a leap of faith. Work like this is already happening—just not, for the most part, in the journals that continue to define legitimacy in narrowly analytic terms. The knowledge exists. The methods exist. The developmental trajectories exist.

What is missing is not evidence.

It is willingness.

Closing — The Problem Is Not a Lack of Data

By the time you have read enough autism research, a pattern sets in. Not in the findings, but in the framing. Another year passes. Another paper is published. The language shifts slightly—more careful, more qualified, more attuned to variability—but the underlying story remains intact. Autism is still something to be explained by what is missing, delayed, or insufficiently developed.

This repetition is often mistaken for consensus. It is not.

The stagnation in the field is not a result of ignorance. The limitations of core-deficit models have been articulated clearly and repeatedly. Nor is it the result of malice. Most researchers working in this area are careful, well-intentioned, and genuinely concerned with improving outcomes for autistic people.

The problem is more structural than that.

It is a persistent refusal to relinquish deficit frames, even when they no longer explain what the data show. It is the continued use of methods that cannot see gestalt processing as language because they were never designed to. It is the quiet decision, made again and again, to interpret difference as approximation rather than as architecture.

The study by Kritsotakis and Morfidi does not stand out because it is unusually flawed. It stands out because it is so representative of where the field remains. Careful measurement. Thoughtful caveats. And a framework that still cannot recognise a perfectly ordinary human way of making meaning unless it resembles the one it already knows how to measure.

Gestalt language processing does not need to be defended as an exception.
It needs to be recognised as one of the ways humans make meaning—across a lifetime.

Until that recognition becomes possible within mainstream research, we will continue to see the same conclusions repackaged each year, and the same lives misunderstood—not because the evidence is lacking, but because the frame refuses to move.