There is a pattern that most people misread at the moment it matters most. You have been making progress. Things have been shifting. And then the anxiety spikes — harder than it has felt in a long time, possibly harder than it has ever felt. The natural interpretation is that something is going wrong. The research says the opposite: that spike is the signal that the change is arriving.

The extinction burst: why behaviour escalates before it stops

Skinner’s foundational operant conditioning research documented a consistent and counterintuitive pattern. When a previously reinforced behaviour stops being reinforced, it does not fade quietly. It intensifies — dramatically and temporarily — before declining. The pigeon that received food pellets for pecking pecked more rapidly and more forcefully immediately after the pellets were withdrawn than it ever had when they were arriving. This temporary spike — the extinction burst — precedes and predicts extinction. It is not a sign that extinction is failing. It is the mechanism through which the organism tests whether the change in reinforcement contingency is real.

The logic is a rational prediction-testing process. The organism has learned that a specific behaviour produces a specific outcome. When the outcome stops arriving, it does not immediately conclude that the contingency has permanently changed. It increases the intensity of the behaviour to test whether the problem is insufficient effort. Only when sustained, intensified effort produces no reinforcement does the behavioural system update its model and begin to disengage.

Applied to anxiety: the person who has learned that avoidance produces relief, and who stops avoiding in exposure or any other change process, experiences their anxiety peak as their behavioural system tests with escalating intensity whether avoidance will restore the relief. When it does not — when they stay in the feared situation and the catastrophe does not occur — the extinction process completes. The peak is the test. Without it, the learning cannot happen.

The predictive brain: peak anxiety is maximum schema revision pressure

Friston’s predictive processing framework establishes that the brain’s primary function is to continuously generate predictions and update them in response to prediction error — the discrepancy between what was expected and what actually occurred. Anxiety, in this framework, is the signal that prediction error is high and unresolved — the brain’s alert that its current model of a situation is inaccurate and that significant revision is required.

The peak of anxiety before a breakthrough is the moment of maximum prediction error — the moment when the existing schema (avoidance works, the feared outcome is certain, the situation cannot be tolerated) is being most strongly violated by incoming evidence. The brain cannot update a model without first generating the prediction error signal that indicates revision is necessary. The more entrenched the existing model — the more deeply a fear belief has been encoded through years of avoidance — the larger the prediction error required to revise it, and the larger the anxiety signal that prediction error generates. Peak anxiety is the computational signature of the old model under maximum revision pressure, not the signal that the situation is maximally dangerous.

Why leaving at the peak is the worst possible moment

The SUDS rating protocol used across exposure therapy research documents the peak-then-decline anxiety pattern with measurable precision. Participants rate their anxiety throughout exposure sessions on a 0–100 scale. The consistent finding is that anxiety rises to a peak in the early phase of each trial, then declines over continued exposure — with the decline occurring faster across successive trials as the extinction memory builds.

Participants who leave during the peak fail to complete the extinction trial. They also negatively reinforce their anxiety by escaping at its highest point, which teaches the behavioural system that escalation produces relief. The peak must be sustained through to the natural decline for the extinction learning to occur. The single most replicated finding in the extinction burst literature is that reinforcing the escalated behaviour during the burst produces a more entrenched and resistant behaviour than existed before the change attempt began. The person who applies avoidance strategies at the exact moment of peak anxiety has specifically confirmed to the behavioural system that maximum intensity is what it takes to generate the avoidance that produces relief. The next encounter will begin with a higher anxiety baseline.

The frustration state that precedes reorganisation

Amsel’s frustration theory established that the removal of an expected reward produces a distinctive motivational state — frustration — that energises behaviour before disengagement. The organism in a frustration state shows increased arousal and behavioural persistence before withdrawing. This is the motivational mechanism underlying the extinction burst: the frustration produced by unrewarded effort generates the escalated responding that characterises the peak.

The person who is making genuine progress but feels worse than ever is experiencing Amsel’s frustration state. Their existing coping strategies are failing to produce the relief they previously provided, and the resulting frustration generates the arousal and negative affect that feels like deterioration. The most difficult moment in any change process — the moment most commonly interpreted as evidence of getting worse — is the peak frustration state that immediately precedes behavioural reorganisation.

The clinical evidence: temporary deterioration before breakthrough

Research on sudden gains in cognitive-behavioural therapy found that a substantial proportion of patients show sudden deterioration — a rapid worsening of symptoms — in the sessions immediately preceding their most significant and durable improvements. Tang and DeRubeis found that sudden gains were often preceded by sessions in which patients engaged in particularly intensive processing of core maladaptive beliefs — the computational equivalent of maximum prediction error preceding model revision. The peak precedes the change, consistently enough to constitute a meaningful pattern rather than individual variation.

The patient who deteriorates sharply in one session and improves dramatically in the next is not experiencing random fluctuation. They are experiencing the frustration-arousal state and prediction error peak that immediately precede genuine psychological reorganisation.

What this means in practice

For any entrepreneur navigating a psychological change process — addressing anxiety that has been constraining their professional reach, working through a pattern of avoidance that has been limiting their decisions, or engaging with a difficult conversation rather than deferring it — the practical implication is specific: the moment the anxiety spikes hardest is the moment to stay, not to leave. That spike is the behavioural system testing whether the change is real. The spike is the prediction error that precedes schema revision. The spike is the frustration state that precedes reorganisation. Leaving at the peak teaches the system that escalation works. Staying through it teaches the system that it does not.

Book worth reading on this

Redirect by Timothy Wilson examines the psychological conditions under which narrative change — reframing the story one tells about a difficult experience — produces genuine behavioural and emotional change. Wilson’s account of why reappraisal is most effective when it is accurate rather than merely positive is the applied complement to the predictive processing framework: the brain updates its model in response to genuine prediction error, not in response to reassurance that does not match the evidence. Understanding that the peak anxiety is the mechanism of change rather than the evidence of failure is precisely the kind of accurate reframe Wilson documents as producing durable change.

If the dynamics described here are significantly affecting your wellbeing, speaking with a psychologist is the right next step. UK: Samaritans (116 123, free, 24/7). Mind (0300 123 3393). BACP: bacp.co.uk/search/Therapists. Crisis Text Line — text HOME to 741741 (US, UK, Canada, Ireland). International: internationaltherapistdirectory.com.

This article is for educational and informational purposes only. Sources: Skinner, B.F. (1938), The Behavior of Organisms, Appleton-Century-Crofts. Ferster, C.B. & Skinner, B.F. (1957), Schedules of Reinforcement, Appleton-Century-Crofts. Friston, K. (2010), The Free-Energy Principle: A Unified Brain Theory?, Nature Reviews Neuroscience, 11(2), 127–138. Amsel, A. (1958), The Role of Frustrative Nonreward in Noncontinuous Reward Situations, Psychological Bulletin, 55(2), 102–119. Foa, E.B., Huppert, J.D. & Cahill, S.P. (2006), Emotional Processing Theory: An Update, in Pathological Anxiety, Guilford Press. Lerman, D.C. & Iwata, B.A. (1996), Developing a Technology for the Use of Operant Extinction in Clinical Settings, Journal of Applied Behavior Analysis, 29(3), 345–382. Tang, T.Z. & DeRubeis, R.J. (1999), Sudden Gains and Critical Sessions in Cognitive-Behavioral Therapy for Depression, Journal of Consulting and Clinical Psychology, 67(6), 894–904. Craske, M.G. et al. (2014), Maximizing Exposure Therapy, Behaviour Research and Therapy, 58, 10–23. Harris, R. (2008), The Happiness Trap, Exisle Publishing. Sapolsky, R. (2017), Behave, Penguin. Wilson, T.D. (2011), Redirect, Little, Brown.