She had rehearsed the opening statement forty times. I had watched her do it — measured, precise, each pause deliberate. In the practice room, she could hold a hostile question, let the silence work, then answer with the kind of clarity that makes an audience lean forward.
Then the cameras switched on.
Within ninety seconds, she was speaking too quickly, answering questions that had not been asked, and reaching for phrases she had never used in rehearsal. The vocabulary was still there. The judgement was still there. She simply could not reach either one.
This is the problem that most conversations about talent, training and expertise fail to address. We spend enormous energy asking what a person can do. We spend very little asking what they can still do when it costs them something to do it.
Ability is not the same as access
The assumption behind most performance evaluation is straightforward: identify what a person knows and can do, and you have identified what they will deliver. Hiring processes test knowledge. Training programmes build skill. Performance reviews measure output. The entire architecture assumes a stable relationship between what someone possesses and what they produce.
The research tells a different story.
In a series of experiments at the University of Chicago, cognitive scientist Sian Beilock discovered something that should unsettle anyone who selects, trains or evaluates people for high-pressure roles. She gave participants mathematical problems under two conditions: low pressure and high pressure. The high-pressure condition involved financial incentives, social evaluation and time constraints — the ordinary furniture of any consequential professional moment.
The result was not simply that people performed worse under pressure. That would have been unsurprising. What Beilock found was that the people with the highest working memory capacity — the strongest cognitive resources — experienced the greatest performance decline. The individuals with the most to offer lost the most access to what they had.
This is not a minor finding. It inverts the comfortable assumption that stronger ability provides a buffer against pressure. In Beilock's experiments, stronger ability provided a larger target.
The mechanism is specific. Under pressure, anxious thoughts begin to consume working memory — the same cognitive resource that makes complex reasoning possible. The people with the most working memory have the most to lose, and lose more of it to the interference created by the pressure itself. As Beilock described it in her book Choke, the information logjams in the brain are not random failures. They are predictable results of a system turning against itself.
The architecture of interference
To understand why capable people lose access to their abilities, it helps to understand what "access" requires at the level of the brain.
Michael Eysenck, the British psychologist, proposed what he called attentional control theory to explain the relationship between anxiety and cognitive performance. The theory distinguishes between two attentional systems. The first is goal-directed: it keeps attention on the task, filters irrelevant information, and coordinates the cognitive resources needed to perform. The second is stimulus-driven: it responds to whatever is most salient in the environment — a sudden noise, a hostile face, a sense of being watched.
Under normal conditions, the goal-directed system governs performance. A surgeon focuses on the incision. A speaker tracks the argument. A negotiator reads the room.
Under anxiety, the balance shifts. The stimulus-driven system begins to dominate. Attention drifts toward threat — toward the evaluator's expression, the imagined consequences of failure, the awareness of being observed. The goal-directed system does not shut down entirely. But it is now competing for resources it used to have to itself.
Eysenck made a further distinction that matters for practice. Anxiety does not always reduce the quality of performance. It reduces efficiency — the ratio between the quality of the output and the cognitive resources required to produce it. Under moderate pressure, people compensate. They invest more effort, recruit more resources, work harder to maintain the same standard. The performance looks the same from the outside. But the internal cost has increased.
This is why pressure-related collapse often appears sudden. It is not. The degradation has been happening for some time, hidden behind compensatory effort. The collapse arrives when the effort required exceeds the resources available. The person was losing access to their abilities well before anyone — including themselves — noticed.
The paradox of expertise
If losing access to ability under pressure were simply a problem of insufficient skill, the solution would be obvious: train more. Build stronger capability. Become more expert.
But the research suggests something more uncomfortable. Expertise itself can become a vulnerability.
Beilock's work with athletes illustrates this with particular clarity. She studied skilled golfers attempting a simple three-foot putt — a shot they had made thousands of times. Under pressure, these golfers began to think about their putting stroke. They analysed the angle of the club face, the rotation of the wrist, the weight transfer through the hips. They attended consciously to processes that had been automated through years of practice.
The result was predictable to anyone who has tried to explain how they ride a bicycle. Conscious attention to an automated skill disrupts it. The golfers who were given strategies to stop analysing their stroke — to focus on a single word, or to putt more quickly — performed better under pressure than those who brought their full analytical capacity to bear.
This is not limited to sport. In operating theatres, research has documented a parallel phenomenon. A systematic review published in Surgical Endoscopy found that acute stress degrades both technical and non-technical surgical skills — higher error rates, longer completion times, diminished instrument handling. One study found that forty per cent of surgeons surveyed had personally witnessed an intraoperative error caused by stress. The cognitive mechanism is the same: overwhelming stress produces what researchers call attentional narrowing — a form of tunnel vision in which the surgeon's focus collapses onto an immediate detail at the expense of the broader operative field.
The surgeon's knowledge does not disappear. Their training does not evaporate. But their access to the integrated, flexible thinking that expert surgery requires is compromised by the very conditions under which surgery most demands it.
Jana Novotná understood this, though she might not have described it in these terms. In the 1993 Wimbledon final, she led Steffi Graf 4-1 in the third set. She was the better player on the day. Then something shifted. She lost five consecutive games and the match. The ability to hit a tennis ball had not left her body. The tactical understanding had not left her mind. But the access — the fluid, automatic, confident connection between intention and execution — had closed.
When identity becomes the interference
The research on choking under pressure focuses largely on situational anxiety — the pressure created by a specific high-stakes moment. But there is a deeper form of interference, one that Claude Steele's work at Stanford exposed with uncomfortable precision.
In 1995, Steele and Joshua Aronson gave Black and white students the same challenging verbal test, drawn from the Graduate Record Examination. The students had been matched for ability. In one condition, the test was described as a diagnostic measure of intellectual capacity. In the other, it was described as a problem-solving exercise.
When the test carried no identity-relevant weight, the performance gap between the two groups disappeared. When the test was framed as a measure of intelligence — activating the stereotype that Black students might confirm — Black students performed significantly worse than equally qualified white students. Not because they lacked ability. Because awareness of the stereotype consumed the cognitive resources needed to demonstrate it.
Steele called this stereotype threat. Over three hundred published experiments have since replicated and extended the finding across gender, age, socioeconomic status and professional domain. The mechanism is the same one Beilock identified: working memory, the engine of complex thought, is hijacked by a concern that has nothing to do with the task itself. The person is performing two operations simultaneously — solving the problem and managing the threat of confirming a negative judgement about who they are.
This is capability access at its most structurally constrained. The interference is not created by the difficulty of the task or the pressure of the moment alone. It is created by the social context in which the person operates. A surgeon who is also the only woman in the theatre. A political leader who is also the first from an underrepresented background. A student who is also carrying the weight of a cultural expectation they did not choose.
The ability is present. The access is conditional.
What access looks like when it holds
If the research on choking described only collapse, it would be interesting but incomplete. What makes the field valuable is that it also describes the conditions under which access holds.
Beilock's work points to several. Practising under the conditions that create pressure — rather than in their absence — builds what she calls "stress inoculation." The goal is not to eliminate anxiety but to train performance in its presence, so that the cognitive system learns to function with interference rather than being derailed by it. The golfers who practised putting in front of an audience choked less than those who practised alone.
Eysenck's attentional control theory suggests another route. If anxiety shifts control from the goal-directed system to the stimulus-driven system, then techniques that strengthen goal-directed attention — focusing on process rather than outcome, externalising attention rather than internalising it — should protect access under pressure. The evidence broadly supports this, though the effect sizes vary and the research is far from complete.
There is also a structural dimension that the individual-focused research tends to understate. If stereotype threat consumes working memory through social context, then changing the context is not merely a matter of fairness. It is a performance intervention. Steele's experiments demonstrated that simply removing the identity-relevant framing of a test eliminated the performance gap entirely. The students did not need coaching, training or therapy. They needed a context that did not tax their cognitive resources before the task had even begun.
This is where the concept of capability access becomes most practically useful. It shifts the question from "how do we build more capable people?" to "how do we build conditions in which capable people can access what they already have?"
Both questions matter. But most organisations, institutions and selection processes focus almost entirely on the first.
The question that training does not answer
In my work with leaders preparing for high-pressure situations — parliamentary questions, crisis communications, board presentations where careers are at stake — the central problem is rarely a lack of knowledge. Most know what they need to say. They have rehearsed the position, understood the data, anticipated the objections.
The problem is that between the preparation room and the pressurised moment, something closes. The voice that was measured becomes tight. The argument that was structured becomes scattered. The judgement that was clear becomes clouded — not by ignorance, but by the weight of consequence.
What I have come to call capability access is the distinction between possessing an ability and being able to use it when conditions demand it. It is not the same as competence. A person can be deeply competent and still lose access to that competence under pressure. It is not the same as confidence. A person can feel confident and still find that the confidence does not translate into performance when the environment shifts.
Capability access is a specific, observable, researchable dimension of human performance. It can be measured — Beilock measured it. It can be degraded — Eysenck described how. It can be structurally constrained — Steele demonstrated that. And it can, within limits, be trained.
But it cannot be assumed. The presence of ability does not guarantee the presence of access. And until we treat access as a separate question from capability, we will continue to be surprised when capable people fail at the moments that matter most.
Every assessment of human performance asks one question:
What can this person do?
Performance under pressure answers the harder one:
What can this person still do when it costs them something to do it?
The gap between those two questions is where most performance failures live. Not in the absence of ability. In the absence of access.
ALTIUS Note
This article belongs to the ALTIUS editorial library: evidence-based essays on Human Performance, Leadership and Society.
Tags: Human Performance, Psychology, Leadership
Selected Research Foundations
Beilock, S. L. — Choke: What the Secrets of the Brain Reveal About Getting It Right When You Have To (Free Press, 2010)
Beilock, S. L., & Carr, T. H. — "When High-Powered People Fail: Working Memory and 'Choking Under Pressure' in Math" https://doi.org/10.1111/j.0956-7976.2005.00789.x
Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. — "Anxiety and Cognitive Performance: Attentional Control Theory" https://doi.org/10.1037/1528-3542.7.2.336
Steele, C. M., & Aronson, J. — "Stereotype Threat and the Intellectual Test Performance of African Americans" https://doi.org/10.1037/0022-3514.69.5.797
Lebares, C. C., et al. — "The Effects of Stress on Surgical Performance: A Systematic Review" https://doi.org/10.1007/s00464-024-11389-3
Mobbs, D., et al. — "Choking Under Pressure: The Neuropsychological Mechanisms of Incentive-Induced Performance Decrements" https://doi.org/10.3389/fnbeh.2015.00019
