The walls Jeff Hawkins encountered were not accidents. They were the product of an academic and institutional system that had spent decades narrowing the kinds of questions behavioral science was structured to ask.
Hawkins approached the brain with the instincts of an engineer. Observable behavior was evidence, but it was not an explanation. If working memory, attention, perception, or executive function appeared in human behavior, the next question was mechanical: What is the physical system doing that produces what we observe?
That progression—from observation to mechanism to causation—is routine in the harder sciences. Researchers can name an unexplained phenomenon without pretending that naming it explains it. Dark matter and dark energy are obvious examples. Their effects can be measured and modeled while their underlying physical nature remains unresolved. The classification gives scientists a common language for discussing what they observe. It does not close the causal question.
American psychiatry made a consequentially different choice.
In 1980, the American Psychiatric Association published the third edition of the Diagnostic and Statistical Manual of Mental Disorders. DSM-III was transformative. It introduced explicit diagnostic criteria, a multiaxial system, and a deliberate attempt to remain neutral about the causes of mental disorders. Robert Spitzer and his colleagues emphasized improved diagnostic reliability as one of the edition's major achievements. Contemporary psychiatric literature described the new system as a descriptive nosology intended to remain largely atheoretical about etiology.
There was a legitimate problem they were trying to solve. Psychiatry had competing theoretical schools using different explanations and sometimes different language for the same patients. If psychiatrists could not agree about causation, they could at least attempt to agree about what they could observe and what patients reported.
That was an enormous improvement in reliability. But reliability and causation are two very different scientific achievements.
Naming the Output
The harder sciences classify unexplained phenomena all the time. The difference is what happens next.
Astronomers can measure gravitational effects attributed to dark matter without knowing precisely what dark matter physically is. Cosmologists can observe the accelerating expansion of the universe and use the term dark energy without claiming that the name itself explains the underlying cause. Those labels identify unresolved phenomena and allow scientists to build models, compare observations, generate predictions, and continue searching for physical mechanisms.
The observation can be objectively real before its cause is known. The mistake comes when the name given to the observation begins functioning as though it explains the cause.
DSM psychiatry created a diagnostic architecture capable of operating without resolving that second question. A disorder could be identified through sufficiently reliable clusters of symptoms, behaviors, reported experiences, developmental history, duration, context, and impairment without identifying the biological mechanism producing them.
That distinction remains particularly visible in neurodevelopmental disorders. ADHD and autism are diagnosed primarily through characteristic patterns of behavior, development, experience, and impairment. There is no blood test, brain scan, genetic panel, or other biomarker that identifies the individual biological mechanism responsible for either diagnosis.
Consider the child who repeatedly loses schoolwork, abandons sequential tasks, appears not to listen, or becomes captured by something outside the window. Ask why the child repeatedly loses the worksheet and the answer may be ADHD. Ask how we know the child has ADHD, and among the evidence will be the fact that the child repeatedly loses things necessary for tasks.
The diagnosis may be clinically useful. The behavior is real, and the impairment can be profound. But the diagnostic label has not answered the causal question. It has classified a recurring output of a biological system without explaining what happened inside that system to produce it.
It has named what came out of the black box.
When the Comparison Group Becomes the Model
There is another consequence of building a taxonomy this way. The neurotypical population becomes more than a statistical comparison group. It can quietly become the reference model for how cognition is supposed to operate.
If most children can sit through a particular instructional environment, maintain the expected sequence of tasks, shift attention when instructed, organize materials in the expected fashion, and respond appropriately to delayed consequences, then failure to reproduce those behaviors becomes evidence of impairment. Sometimes it unquestionably is impairment, and that should not be minimized.
But there is a scientific assumption hiding underneath the comparison: that the neurotypical behavioral outcome represents not merely the statistically common outcome, but the standard against which the underlying cognitive system should be understood.
That does not necessarily follow.
Two nervous systems can arrive at the same observable result through different processing strategies. They may require different environmental conditions to reach that result, and the cognitive cost of producing the same result may be radically different. Conversely, the same nervous system can perform poorly under one set of conditions, conventionally under another, and exceptionally under a third.
If we measure only the output, those differences can disappear.
Autism research provides an important example. Laurent Mottron, Michelle Dawson, and colleagues developed the Enhanced Perceptual Functioning model from findings that included enhanced low-level perceptual discrimination and different recruitment of neural systems during complex perceptual tasks alongside well-documented areas of difficulty. Their argument was not that autism produces universal superiority. It was that neurological difference could produce both diminished and enhanced performance depending upon the operation being measured.
That changes the scientific question. Instead of asking only whether an autistic or ADHD person can perform like the neurotypical comparison group, we also need to ask how that nervous system produced the performance, what conditions were required to produce it, what it cost to produce it, and what happens when those conditions change.
Behavioral equivalence is not necessarily mechanistic equivalence. And behavioral difference is not automatically evidence that the neurotypical mechanism is present but defective.
The Missing Half of ADHD
ADHD makes the consequences particularly visible because a diagnostic system organized around impairment is necessarily very good at finding impairment. The unfinished worksheet matters. The forgotten assignment matters. The lost backpack, interruption, missed deadline, abandoned sequence, and inability to sustain engagement under particular circumstances all matter.
But the same nervous system may behave radically differently when the circumstances change.
A person who cannot remain engaged with an administrative task for fifteen minutes may spend hours absorbed in a difficult problem. A student whose performance collapses during conventional instruction may demonstrate extraordinary persistence when an activity provides immediate feedback, movement, novelty, competition, creativity, challenge, or genuine interest. Someone who appears incapable of regulating attention in one environment may exhibit remarkable concentration in another.
Those observations do not erase ADHD impairment. They create a scientific problem that a complete theory of ADHD should have to explain.
Yet research into the positive or high-performing side of ADHD remains comparatively limited. Sedgwick, Merwood, and Asherson's study of successful adults with ADHD identified reported themes including cognitive dynamism, energy, divergent thinking, and hyperfocus. Schippers and colleagues later examined self-reported positive characteristics in adults with ADHD and identified perceived strengths involving creativity, dynamism, flexibility, socio-affective abilities, and higher-order cognitive skills.
Neither study establishes that ADHD inherently confers these advantages. That would simply replace one overgeneralization with another. What the studies demonstrate is that the other side of the performance distribution exists and has received far less systematic attention than impairment.
That asymmetry should not surprise us. If the doorway through which a population enters the scientific taxonomy is impairment, researchers inherit a population defined by impairment. Research questions naturally accumulate around the characteristics used to define that population in the first place: What goes wrong? How badly does it go wrong? How does performance differ from neurotypical controls? Which deficits can be measured? How can those deficits be reduced?
Every one of those questions is legitimate. They are simply not all of the questions.
The Pathology Loop
This is how a descriptive taxonomy can gradually become a pathology loop without anyone deliberately designing one.
The diagnostic system identifies people according to recurring impairments. Researchers recruit participants carrying those diagnoses and compare their performance with neurotypical controls. Differences emerge. Those differences become candidate deficits associated with the disorder, which generates additional research designed to measure those deficits more precisely. The resulting literature then informs clinical practice, educational policy, professional training, and the next generation of research questions.
Each individual step can be scientifically legitimate while the cumulative effect still creates a profound blind spot.
The system becomes extraordinarily sophisticated at describing circumstances under which ADHD and autistic nervous systems perform worse than neurotypical ones while devoting considerably less attention to circumstances under which their performance converges with—or sometimes exceeds—the comparison group. When exceptional performance appears, it can be treated as an interesting exception, a special interest, a talent, or an anecdote rather than as evidence demanding a mechanistic explanation.
But the failure state and the exceptional state are outputs of the same nervous system. If an ADHD nervous system performs poorly under one condition, ordinarily under another, and exceptionally under a third, a biological theory of ADHD should eventually explain all three states. The transition itself may contain more information about the underlying mechanism than another experiment demonstrating that the ADHD group performs worse on a task constructed around neurotypical expectations.
Instead of asking only What is wrong with this person?, the mechanistic question becomes What changed in the system?
That opens an entirely different research program.
The Cost of Ignoring Causation
This is where the DSM's deliberate neutrality toward etiology produces consequences far beyond the diagnostic manual itself.
Once the behavioral category becomes the organizing unit, researchers can continue refining the category without resolving the biology beneath it. Educational researchers can measure executive-function deficits. Psychologists can construct increasingly sophisticated behavioral tasks. Clinicians can quantify symptom severity. Schools can document accommodations and compliance. Researchers can compare diagnosed populations with neurotypical controls and identify increasingly subtle statistical differences.
Useful knowledge can emerge from every one of those approaches, but the causal question can remain unanswered.
This becomes particularly consequential when a psychological construct begins to sound like a physical explanation. Attention, working memory, inhibition, and executive function are useful descriptions of cognitive performance. They do not automatically identify the neural computation producing that performance.
That distinction will matter later when we return specifically to working memory and the cortical mechanisms that may underlie it. For the moment, the important point is institutional: a field organized around behavioral taxonomy can become extraordinarily productive without ever being forced to open the black box.
And if the black box remains closed, maximizing human development becomes much harder. Knowing that a child performs worse than the neurotypical mean tells us something important. It does not necessarily tell us why the child performs worse, why the same child performs normally somewhere else, why performance may become exceptional under another set of conditions, or what environmental changes might move that nervous system from one state to another.
Those are questions of mechanism, but they are also questions of development.
When the Phenomenon Does Not Fit the Box
Dr. William Dodson's work provides another window into this institutional problem because his clinical career and his relationship with the Cleveland Clinic placed him partly within a medical structure organized very differently from the conventional university tenure system.
The Cleveland Clinic was established as a group-practice model influenced by the Mayo Clinic. Physicians work within a salaried institutional structure rather than building careers around the conventional academic progression toward tenure. Cleveland Clinic physicians operate through renewable employment arrangements rather than lifetime academic tenure. That difference in structure matters because institutions inevitably influence which questions people can afford to pursue.
In the traditional academic system, professional survival and advancement can depend heavily upon publication, grant funding, established research programs, departmental priorities, and ultimately promotion and tenure. None of this makes academic research invalid. Peer review, replication, methodological discipline, and competition for funding remain essential scientific safeguards. But the structure creates incentives. A young researcher attempting to build a career has powerful reasons to pursue questions that funding agencies recognize, study sections understand, established instruments can measure, and journals already know how to classify.
Research built around an established DSM construct has an institutional home. A clinical phenomenon that does not fit the taxonomy—and is extraordinarily difficult to capture experimentally—has a much harder path.
Dodson's work on Rejection Sensitive Dysphoria provides an unusually clear example. Over decades of treating people with ADHD, he has described a recurring pattern of acute emotional pain associated with perceived rejection, criticism, failure, or loss of approval. RSD is not a DSM diagnosis, and the independent empirical literature surrounding the construct remains limited. Those qualifications matter. Clinical observation is not proof of a distinct biological entity.
But clinical observation is often where biological investigation begins.
There is an additional problem with studying RSD that goes beyond its absence from the DSM. As Dodson describes these episodes, they occur spontaneously. They are intensely personal, dependent upon circumstances and meaning, and cannot simply be summoned on command for a laboratory experiment.
This creates what I think of as a kind of Heisenberg Uncertainty Principle of RSD research. Not in the literal quantum-mechanical sense, but as an analogy for the measurement problem.
To observe an authentic episode, a researcher would need to be watching at precisely the moment something genuinely meaningful to that individual produces the response. But the researcher cannot simply schedule an RSD episode for 2:00 Tuesday afternoon, place someone inside an fMRI scanner, manufacture an authentic rejection from someone whose opinion actually matters, and assume that the resulting state is biologically equivalent to the spontaneous event.
By the time you know the real thing has happened, much of what you wanted to measure may already be gone.
That creates an enormous problem for conventional behavioral research. The phenomenon is difficult to summon, difficult to standardize, difficult to observe prospectively, and difficult to reproduce across subjects. The trigger is not merely a stimulus. Its meaning depends upon the individual, the relationship, the circumstances, and the perceived stakes.
The measurement problem can then become self-reinforcing. Something is difficult to measure, so relatively little research is conducted. Because relatively little research exists, the phenomenon is treated as insufficiently established to justify substantial research attention. Because researchers devote little attention to it, the validated instruments and experimental paradigms required to study it remain underdeveloped.
The measurement problem becomes an evidence problem, and the evidence problem becomes an institutional problem.
Dodson's pharmacological observations make the unanswered question even more interesting. He has reported that when guanfacine or clonidine is appropriately used, roughly half of the patients in whom he identifies this phenomenon can become essentially asymptomatic for the RSD response. That observation does not establish the prevalence of RSD, prove that it represents a distinct disorder, or establish its biological mechanism.
It does, however, generate an obvious scientific question:
What changed in the nervous system?
If a recurring emotional phenomenon changes dramatically when alpha-2 adrenergic signaling is altered, a mechanism-first science should want to know why. The appropriate response is neither to declare the phenomenon proven because a clinician observed it nor to ignore the observation because the phenomenon cannot yet be conveniently reproduced inside an established experimental paradigm.
The observation is where the investigation should begin.
This is precisely where the institutional structure surrounding Dodson becomes relevant. An integrated, salaried medical system such as Cleveland Clinic or Mayo Clinic does not eliminate bureaucracy, evidence standards, or professional pressure. But it changes the professional architecture surrounding the clinician. A career does not depend upon winning tenure by establishing a publication program inside an already recognized academic lane.
That creates more room for bedside observation to push against taxonomy rather than always flowing from it.
RSD may ultimately prove to be one construct, several overlapping phenomena, a particular expression of emotional dysregulation, or something requiring different vocabulary entirely. That is what research is supposed to determine. The absence of an established box is not evidence for the absence of a phenomenon.
The larger lesson extends far beyond RSD. Institutions are exceptionally good at investigating phenomena that fit the categories, instruments, funding mechanisms, and experimental designs they already possess.
Breakthrough science often begins precisely where those tools stop working.
The Institution Eventually Recognized the Problem
The irony is that the National Institute of Mental Health eventually created an entire research framework in response to many of these limitations.
The Research Domain Criteria project, or RDoC, was developed to move research beyond conventional diagnostic categories and toward dimensions of functioning grounded in behavioral science and neuroscience. Rather than requiring researchers to begin with DSM diagnoses, RDoC organizes investigation across multiple levels of analysis, including genes, molecules, cells, neural circuits, physiology, behavior, self-report, and experimental paradigms.
In 2013, then-NIMH director Thomas Insel made the problem unusually explicit. DSM diagnoses, he wrote, were based on consensus about clusters of clinical symptoms rather than objective laboratory measures. NIMH would begin reorienting its research toward underlying biological and behavioral dimensions rather than treating DSM categories as unquestioned natural boundaries.
RDoC also introduced something especially important for the argument here: dimensionality. Its constructs are intended to be investigated across the range of human functioning rather than only after behavior crosses a diagnostic threshold. Development and environment are explicitly part of the framework.
That is a fundamentally different scientific orientation. Instead of beginning with the category and asking how the diagnosed population differs, researchers can begin with the system itself and ask how it operates across different people, developmental stages, environments, and performance states.
The black box begins to open.
The Other Side of the Coin
That missing half matters enormously for ADHD and autism because these are not nervous systems that simply perform some fixed percentage below normal across every domain and every environment. Performance changes with context, task, motivation, environment, development, feedback, and the particular operation being demanded.
If the goal of science is merely to identify where an ADHD or autistic person diverges from the neurotypical mean, a deficit framework can accomplish a great deal. If the goal is to understand the nervous system, it is not enough.
We have to investigate the failure state, the ordinary state, and the exceptional state with equal seriousness. We have to ask why a person cannot engage here but cannot disengage there, why one environment exposes profound impairment while another reveals remarkable ability, why two people can produce identical behavioral outcomes through different cognitive routes, and why changing the conditions surrounding a nervous system can sometimes change performance more dramatically than the diagnostic label would lead us to expect.
This is not an argument for replacing pathology with a romanticized story about neurodivergent “superpowers.” Disability and impairment are real, and people deserve treatment, accommodations, and support for both. Nor does exceptional performance occur everywhere or in everyone. The entire point is that performance can be conditional, uneven, and highly dependent upon circumstances.
That variability should not be treated as noise surrounding the disorder.
It may be evidence about how the system works.
A science that studies a nervous system primarily when it fails will become very good at describing failure. A science that understands the machinery can begin asking the other question:
Under what conditions does this nervous system perform at its best—and why?
That is where pathology ends and development begins.
References
American Psychiatric Association. (1980). Diagnostic and Statistical Manual of Mental Disorders (3rd ed.). American Psychiatric Association.
American Psychiatric Association. DSM History. American Psychiatric Association.
Dodson, W. W. Draft chapter on the regulation of emotions and Rejection Sensitive Dysphoria. Unpublished manuscript/personal communication.
Insel, T. (2013). Transforming Diagnosis. National Institute of Mental Health.
Insel, T., Cuthbert, B., Garvey, M., Heinssen, R., Pine, D. S., Quinn, K., Sanislow, C., & Wang, P. (2010). Research Domain Criteria (RDoC): Toward a new classification framework for research on mental disorders. American Journal of Psychiatry, 167(7), 748–751.
Klerman, G. L., Vaillant, G. E., Spitzer, R. L., & Michels, R. (1984). A debate on DSM-III. American Journal of Psychiatry, 141(4), 539–553.
Mottron, L., Dawson, M., Soulières, I., Hubert, B., & Burack, J. (2006). Enhanced perceptual functioning in autism: An update, and eight principles of autistic perception. Journal of Autism and Developmental Disorders, 36(1), 27–43.
National Institute of Mental Health. Research Domain Criteria (RDoC). National Institute of Mental Health.
Schippers, L. M., Horstman, L. I., van de Velde, H., Pereira, R. R., Zinkstok, J., Mostert, J. C., Greven, C. U., & Hoogman, M. (2022). A qualitative and quantitative study of self-reported positive characteristics of individuals with ADHD. Frontiers in Psychiatry, 13, 922788.
Sedgwick, J. A., Merwood, A., & Asherson, P. (2019). The positive aspects of attention deficit hyperactivity disorder: A qualitative investigation of successful adults with ADHD. Attention Deficit and Hyperactivity Disorders, 11(3), 241–253.
Spitzer, R. L., Williams, J. B. W., & Skodol, A. E. (1980). DSM-III: The major achievements and an overview. American Journal of Psychiatry, 137(2), 151–164.