Stress identified as neurological mechanism behind the recall-to-application gap in learning

The recall-to-application gap

You studied. You revised. You could recall the facts. But you didn’t ace the exam. Unless you’re lucky enough to have an eidetic memory, the chances are that you’ve had this experience at least once in your life, or your child has. What’s going on there, and why?

My eldest is currently navigating her last year of school, the dreaded make-or-break matric year. Right now, she is in the midst of writing exams, with all the stress and worry that involves for both parents and young adult. This makes today’s topic, the effect of stress on integration and application of learning, quite a biggie for me, and I suspect for other parents of children in their final year of school. It might also be of interest if you’re a teacher, as I once was, and have struggled with the gap between learners’ recall and application of knowledge, and wondered often how to help them close that gap.  Let’s take a look at the science.

We have known for some time that stress impairs learning. What we did not have, until this month, is a precise neural account of exactly where in the learning process the damage occurs, with arguably significant implications for how we structure education in South Africa. Research by Schüren et al. published in Science Advances  (May 2026) reports how a combined team from the University of Hamburg and the University of Texas at Austin used fMRI to observe what stress actually does to the learning brain in real time, and the implications of what they found for how we understand learning, effort, and academic performance are difficult to overstate.¹

How does stress affect the brain during learning?

In the research literature, this capacity is called associative inference - the ability to recognise that separately learned pieces of knowledge are related, and to draw correct conclusions from that relationship. Associative inference has a specific mechanism, reliant on the functioning of the hippocampus.

What the hippocampus actually does during learning

To understand what the research found, it helps to understand what the hippocampus is doing when we learn something genuinely new — not just facts, but knowledge.

The hippocampus is the brain's central structure for memory integration.² When you encounter new information, the hippocampus does not simply file it away as an isolated entry. It actively searches for existing memories that overlap with what is being encoded, reactivates them, and then weaves the new experience into the same representational network. The result is that what you learn today becomes genuinely connected to what you already knew — which means you can later make inferences, draw comparisons, and apply what you have learned in contexts you have never encountered before.

This is the mechanism behind conceptual understanding, and is what separates higher-order learning from recall. The mechanism relies on the successful encoding of new material in the moment of learning itself, not just afterward during consolidation.

The Schüren et al. study tested this directly.¹ Participants first learned associations between a set of paired images (A-B). Twenty-four hours later, before learning a second overlapping set (B-C), one group was exposed to a well-established psychosocial stressor, a mock job interview designed to induce genuine acute stress, complete with elevated cortisol, increased heart rate, and subjective anxiety. Then both groups learned the new B-C pairs. Afterward, they were tested on associative inference - specifically, A-C inference: could they connect the dots between information they had learned on separate occasions but that shared an overlapping element?

Both groups learned the B-C pairs equally well. On basic recognition memory, stress made no measurable difference. However, on the inference test (the measure of whether the new learning had been genuinely integrated with prior knowledge) the stressed group performed significantly worse. Their brains had absorbed the new content without connecting it to the knowledge structure it belonged to. They had acquired information without building understanding.

This is the neural basis of a failure most teachers will recognise immediately: the learner who can do multiplication accurately but cannot identify that a word problem requires multiplication, or who can recite the themes of Macbeth but cannot recognise those same themes in an unseen poem. The capacity being measured (associative inference) is the ability to recognise that what you already know is relevant to a new problem, and to apply it accordingly

Acute stress impairs learning by disrupting memory integration, not memory acquisition. Under stress, learners can still absorb and recall new information, but the hippocampus fails to connect that new information to existing knowledge. The result is a student who can recite facts but cannot apply their new knowledge, i.e. has not achieved transferrable learning. This is the neurological difference between recall and understanding, and it is why stressed learners so often underperform relative to the effort they have put in.

What stress does to the hippocampus during encoding

Using multivariate decoding of the fMRI data, the researchers were able to see precisely why this happened. There were two primary findings in the study. The first finding was that during the B-C learning phase, while the control group's hippocampi were actively reactivating their prior A-B memories — essentially retrieving the relevant existing knowledge to link it with what was being learned — the stressed group's hippocampi showed significantly reduced reactivation.¹ The mechanism was suppressed, and the lower the reactivation, the worse inference performance was. Performance could be accurately predicted by just the degree of reactivation.

The second finding was equally important. The researchers used representational similarity analysis to examine how the hippocampus was representing linked memories after learning. In the control group, overlapping memories — information connected through shared elements — became more neurally similar to each other after learning, indicating integration into a shared representational network. In the stressed group, the opposite happened. Memories that should have been connected became more neurally dissimilar, stored as separate, unrelated events rather than as part of a coherent knowledge structure.¹

The hippocampus was actively differentiating, coding related experiences as if they had nothing to do with each other.

Stress does not only prevent the hippocampus from integrating related memories — it actively drives them apart. fMRI research by Schüren et al. (2026) found that after learning under stress, memories that shared overlapping content became more neurally dissimilar. The brain was coding related experiences as separate, unconnected events. This stress-induced pattern differentiation is the neural mechanism behind the gap between a student's effort and their ability to apply what they have studied.

The mechanism behind this involves the hippocampus's high density of receptors for stress hormones — glucocorticoids and norepinephrine — which are released in significant quantities during acute stress. These hormones directly disrupt hippocampal function, and the hippocampus appears to be particularly vulnerable to their effects on integrative encoding specifically.^3,4 Under stress, the hippocampus shifts from its integrative mode — linking new experiences to related prior memories — into a more conservative, differentiated mode that treats each event as distinct.

There is likely an adaptive logic to this. In genuinely threatening situations, precise, discrete memory encoding may be more immediately useful than building conceptual structures. But in an educational context, and in the chronically stressed lives of many young adults at this point in history, this adaptive mechanism works directly against the kind of learning that education is trying to produce.

Why this matters in a classroom — and beyond it

The implications here extend further than revision stress before a single exam. Chronic psychological stress (such as stress from ongoing financial pressure, family instability, anxiety disorders, unresolved trauma, or simply the ambient pressure of a performance-oriented educational culture) creates a sustained state in which this hippocampal integration mechanism is persistently compromised.

Consider what that means in practice. A learner in a state of chronic stress may be sitting in every lesson or attending every lecture, completing every reading or piece of homework, and genuinely trying to learn. They are acquiring information. Their basic recognition memory may be largely intact — they could pass a multiple-choice test on the material. But the building of connected knowledge structures, the process through which information becomes something they can reason from, apply, and transfer to unfamiliar contexts, is systematically undermined at a neural level. The gap between their effort and their performance is indicative of a brain operating in a state in which a core learning mechanism has been degraded.

The research also found that within the control group, inference accuracy was higher for material that participants subjectively rated as emotionally positive — consistent with existing evidence that positive emotions during and about the learning process facilitate associative learning.¹ Stressed participants showed no such modulation: their inference performance was uniformly impaired regardless of how they had experienced the emotional content of the material. The implication is that stress may remove a naturally occurring facilitator of integrative encoding, on top of the direct disruption it causes to hippocampal reactivation.

These findings connect directly to a distinction that sits at the heart of the Human Capacity Architecture (HCA) framework developed at the Illumin-Ed Institute: the difference between the infrastructure that makes learning possible, and the content that learning operates on. HCA organises development into two tracks - Adaptive Capacity (the neurological and psychological foundations support and facilitate complex thinking possible) and Cognitive Architecture (the higher-order reasoning and application capacities that tertiary education demands, and that enable us to create new knowledge rather than just consume it). The framework's central argument is that although development of each track and their component elements is not sequential, Track 1 must run slightly ahead, because it creates the conditions Track 2 requires. The Schüren et al. findings give that argument a precise neurological basis. Operating under severe or chronic stress, the hippocampus fails to perform the integrative function on which higher-order learning depends. Within HCA, the specific capacity that addresses this is called Psychoadaptive Resilience.

The problem with content-focussed interventions such as extra lessons and subject tutoring

When a learner is underperforming despite apparent effort, the usual educational response is to add more content support such as tutoring, additional resources,  and practice exercises. The Schüren et al. findings suggest that in many cases this is insufficient, supporting what I discovered first hand and how that shaped much of what I developed over the decade I spent leading the academic development programme for my department as a university academic.

When stress is preventing the hippocampus from integrating new learning with existing knowledge, adding more content simply gives the student more isolated facts that are also failing to integrate. The architecture of the problem sits at the regulatory level, and the solution has to sit there too.

This is one of the clearest neuroscientific illustrations of a distinction central to the Human Capacity Architecture (HCA) framework: the difference between content-level intervention and infrastructure-level intervention. Tutoring addresses content, coaching addresses motivation, and clinical support addresses symptoms. All of these have genuine value in the right context, but none of them acknowledge or address the underlying neural infrastructure on which learning itself depends — the regulatory foundations that determine whether a brain can do what learning requires of it in the first place.

The HCA framework's Adaptive Capacity track exists precisely because of this. Psychoadaptive Resilience — the capacity to remain regulated and functionally present under pressure — is a precondition of the integrative encoding that turns acquired information into genuine, transferable knowledge. Without it, a student can work as hard as they like, revise as much as they like, and still battle with integration and application of learning.

When a student underperforms despite genuine effort, the most common educational response is more content support — tutoring, extra resources, additional practice. But research on stress and hippocampal memory integration suggests this is often the wrong intervention. When stress is preventing the brain from connecting new learning to existing knowledge, more content simply produces more factual information, without developing the capacity to see the relevance of that content or apply it to novel problems and situations. The problem lies at the level of neural infrastructure (the regulatory foundations that determine whether the brain can do what learning requires) and it calls for an infrastructurally-informed response.

What this means for how we think about learning environments

The research has implications not just for individual learners but for the environments in which learning happens. An educational culture that generates chronic stress (be it through unrelenting performance pressure, high-stakes assessment formats, competitive social dynamics, or the elimination of recovery time from the school or university day) is systematically degrading the neural mechanism through which students build the connected knowledge structures that education is supposed to produce. In the South African context specifically, where learner populations carry disproportionately high stress loads from poverty, trauma exposure, overcrowded classrooms, and language-of-teaching-and-learning barriers, this is not a marginal concern.

This matters especially during adolescence and early adulthood, the developmental window that Human Capacity Architecture focuses on most directly. This is the period during which the prefrontal cortex is still maturing, hippocampal plasticity is at its highest, and the foundational knowledge architectures that support a lifetime of learning are being built.⁵ Chronic stress during this window does not just affect today's exam result, but shapes the neural infrastructure that determines how effectively a person will be able to learn and reason throughout their adult life.

It also raises a pointed question for assessment design. When acute stress immediately before or during learning prevents the hippocampus from integrating information, the gap between what learners actually know and what they demonstrate under stressful exam conditions may be far larger — and far more systematic — than we currently assume. High-stakes, time-pressured assessment formats may be measuring stress tolerance and regulatory capacity as much as they are measuring knowledge, and conflating the two. If a learner's regulatory capacity is actually the binding constraint on their performance, then more preparation and more content review will not address the problem. In such cases, the intervention point is the nervous system, and assessment design that ignores this may be producing a systematic underestimate of learning.

So, what we can do about it?

The Schüren et al. findings point toward practical responses at multiple levels.

For individuals, the most direct implication is that stress regulation is a learning strategy — one of the highest-leverage interventions available, because it operates at the level of the mechanism that makes learning ‘stick’ and facilitates higher order learning and application. This is not about eliminating pressure or difficulty, which these serve genuinely important developmental functions. It is about building the capacity to remain neurologically functionally regulated under pressure, to keep the hippocampus in an integrative mode even when the environment is demanding. That capacity can be developed. It responds to deliberate, effortful practice of the kind that builds genuine understanding rather than surface familiarity, and it pays compound returns across every domain in which learning and reasoning matter.

For young adults themselves, particularly those in the 16-25 window where the stakes and the opportunity converge most sharply: this research is further evidence that when stress seems to be blocking your thinking, what you are experiencing is a neurobiological process that is well understood and, with the right kind of deliberate work, addressable. The brain you have is not fixed but changeable, and you can take charge of changing it with the right information and tools. The regulatory capacity that makes integrative learning possible is infrastructure, and infrastructure can be deliberately built.

For educators and institutional designers, the implication is that learning environment design is neurological design. The conditions under which students are asked to learn — the stress load of their environment, the availability of recovery time, the presence or absence of supportive social relationships — are part of the architecture of learning outcomes, not peripheral to them. Creating environments where the hippocampus can do its integrative work is as much a pedagogical responsibility as curriculum design.