Rebound Effect¶

Prime #: 1115
Origin domain: Pharmacology & Toxicology
Subdomain: pharmacodynamics → Pharmacology & Toxicology

Core Idea¶

The rebound effect is the structural pattern in which a system, responding to a sustained suppressing input, adapts by upregulating the very process being suppressed, and then, when the suppressor is removed, overshoots its original baseline in the opposite direction because the compensation decays more slowly than the input disappears. The structural ingredients are five: a target process at some baseline; a sustained external suppressor that holds it below baseline; an internal compensatory mechanism whose adjustment lags the input, building up to oppose the suppressor; removal of the input; and a transient overshoot before re-equilibration, as the now-unopposed compensation drives the process past its starting point. The decisive claim is that removing a "fix" can cause the original problem to reappear worse than before, purely because of the system's adaptive response to the fix itself. The withdrawal is not a return to the prior state; it is a distinct dynamic event whose trajectory is shaped by the compensator's state at the moment of removal, not by what the system would have been without the intervention.

This skeleton recurs across substrates as a process held below its natural level by an external force, an internal adjustment that prepares the system for the new equilibrium, removal of the force on a timescale faster than the readjustment, and a transient overshoot in the opposite direction. In pharmacology it is rebound hypertension on abrupt beta-blocker withdrawal and rebound insomnia after benzodiazepine cessation, because adapted receptor density does not normalize at the rate the drug clears. In psychology it is thought-suppression rebound and post-diet weight regain. In policy it is the post-austerity inflationary spike and the energy-efficiency rebound where cost drops induce usage increases that offset savings. In ecology it is predator-removal prey overshoot and fire-suppression megafires. In software it is the traffic surge beyond steady-state when a rate limit is lifted and queued retries fire together. Strip the substrate vocabulary and what remains is: a target process at baseline, a sustained suppressor, a lagging compensator, a removal event, and an opposite-direction overshoot whose amplitude scales with how much compensation had built up. The pattern is a pure feedback-transient, and its vocabulary travels intact.

How would you explain it like I'm…

Bounce-Back Ball

If you hold a beach ball under the water for a long time and then let go, it doesn't just float up calmly. It shoots up high into the air before settling. The Rebound Effect is when pushing something down for a while makes it bounce back even higher than where it started once you stop pushing.

When the Fix Backfires

Imagine your body or a machine has a normal level it likes to sit at. If something keeps pushing it down for a long time, the system fights back by pushing up harder and harder to balance it. The tricky part is that this fighting-back takes a while to fade. So when you suddenly remove the thing that was pushing down, the system is still pushing up hard, and it shoots past its normal level for a while. That is the Rebound Effect: the fix wears off faster than the body's reaction to it.

Overshoot After Withdrawal

The Rebound Effect happens when a system is held below its normal level by some outside force for a long time. While that force is applied, the system slowly builds up an internal push that fights back against it. The trick is that this push fades away more slowly than the force is removed. So when you suddenly take the force away, the leftover push is still active and shoves the system past its starting point in the opposite direction. This is why taking away a 'fix' can sometimes make the original problem come back worse — like quitting a sleep medicine and then sleeping worse than before you ever took it.

The Rebound Effect is a feedback transient with five ingredients: a target process sitting at some baseline, a sustained external suppressor holding it below baseline, an internal compensatory mechanism that lags the input and builds up to oppose the suppressor, removal of the suppressor, and a transient overshoot before the system re-equilibrates. The key asymmetry is timescale: the compensation decays more slowly than the suppressor disappears, so when the input is withdrawn the now-unopposed compensation drives the process past its original starting point. Crucially, withdrawal is not a passive return to the prior state — it is a distinct dynamic event whose trajectory depends on how much compensation had accumulated at the moment of removal. The overshoot's amplitude scales with that built-up compensation. You see this as rebound hypertension after abrupt beta-blocker withdrawal (adapted receptor density doesn't normalize as fast as the drug clears), thought-suppression rebound, post-diet weight regain, post-austerity inflation spikes, predator-removal prey overshoot, and retry storms when a rate limit is lifted. The decisive claim is that removing a fix can reproduce the original problem worse than before, purely because of the system's adaptation to the fix itself.

Structural Signature¶

the target process at baseline — the sustained suppressor holding it below baseline — the lagging internal compensator that builds to oppose the suppressor — the removal event — the timescale mismatch between fast removal and slow compensator decay — the opposite-direction overshoot scaled by accumulated compensation

A system exhibits this pattern when each of the following holds:

A target process at baseline. Some quantity has a natural resting level the system would occupy if left undisturbed.
A sustained suppressor. An external input holds the target below (or away from) baseline for long enough that the system can adapt to it.
A lagging compensator. An internal mechanism adjusts to oppose the suppressor, building up over time; its adjustment is slow relative to how quickly the suppressor can be removed.
A removal event. The suppressor is withdrawn — abruptly or faster than the compensator can stand down.
A timescale mismatch. The compensator decays more slowly than the suppressor disappears, leaving it transiently unopposed.
An opposite-direction overshoot. The now-unopposed compensation drives the target past baseline in the direction opposite the original suppression, with amplitude scaling with how much compensation had accumulated, before re-equilibration on the compensator's natural timescale.

These compose so that the rebound is forecastable: its size is set by the measurable compensator state at the moment of removal, not by the removal trigger.

What It Is Not¶

Not stressor-induced adaptation. stressor_induced_adaptation is the compensation building up during sustained stress; the rebound effect is the overshoot on withdrawal once the now-unopposed compensation outlives the stressor. Adaptation is the prerequisite; rebound is the transient when it is removed.
Not tolerance. tolerance is the reduced effect during sustained treatment; rebound is the opposite-direction overshoot after cessation. They share one mechanism but present at different times — tolerance while the suppressor is on, rebound when it comes off.
Not regression to the mean. regression_to_the_mean is a statistical return toward average from an extreme; rebound is a causal excursion past baseline driven by a real compensator. A mere return-to-baseline is not rebound.
Not reactance. reactance is a motivated psychological pushback against a perceived threat to freedom; rebound is a substrate-neutral dynamical transient that occurs in receptors, queues, and ecosystems with no agency involved.
Not hysteresis. hysteresis is path-dependence in which the state lags the input symmetrically on loading and unloading; rebound is a transient overshoot past the original baseline, not a lagged but bounded return.
Common misclassification. Diagnosing rebound where there was no compensator — treating any post-withdrawal worsening as adaptive overshoot. Catch it by asking whether removal caused an excursion past baseline in the opposite direction, scaling with how long suppression was held; if not, it is return-to-mean or coincident trigger.

Broad Use¶

Pharmacology (origin) — rebound hypertension on abrupt beta-blocker withdrawal, rebound insomnia after benzodiazepine cessation, rebound congestion after decongestant overuse, each because adapted receptors do not normalize at the rate the drug clears.
Psychology and habit — thought-suppression rebound, post-diet weight regain, post-quit cravings, where the suppressed process surges when active suppression ends.
Policy and economics — post-austerity inflationary spikes when released demand outpaces supply adjustment; the efficiency rebound where cost drops from efficiency induce offsetting usage increases.
Ecology — predator removal triggering prey overshoot beyond pre-predator levels; pesticide cessation followed by amplified pest resurgence; fire suppression followed by fuel-driven megafires.
Software and systems — rate-limit removal followed by a traffic surge as queued retries release together; circuit-breaker re-closing causing a thundering herd; cache expiry causing simultaneous backfill spikes.
Monetary policy — ending a sustained low-rate regime producing a rebound in inflation expectations and rapid term-structure re-pricing beyond pre-regime levels.
Organizational discipline — lifting a strict freeze (hiring, expense, code-review) producing a burst beyond the no-freeze counterfactual because demand built up under suppression.

Clarity¶

The rebound effect makes visible that cessation of an intervention is itself an intervention, and that the system you remove a treatment from is not the same system you applied it to — it has adapted. Without the prime, withdrawal looks like a simple return to baseline; with it, withdrawal is named as a separate dynamic event with predictable transient behaviour, so the analyst expects an overshoot rather than being surprised by it. A second clarifying move is the distinction between tolerance and rebound: tolerance is the reduced effect during sustained treatment, while rebound is the overshoot upon withdrawal, and although they share the same underlying compensation mechanism they present as different phenomena observed at different times. Naming the pair keeps an analyst from treating them as unrelated. The prime also sharpens the boundary of its own application through a decisive test: did the system adapt to the intervention while it was in place, and does removing the intervention cause an excursion in the direction opposite to the original intervention? If yes, the prime applies; if withdrawal merely returns to baseline, the system was not compensating and the pattern is something else. This test screens out mere returns-to-mean and isolates the genuine compensator-driven case.

Manages Complexity¶

The rebound effect decomposes a "the drug stopped working after we stopped it" or "the policy backfired when we ended it" puzzle into three concrete and tractable questions. What was the compensation that built up during the intervention? What is its decay timescale relative to the intervention's removal timescale? And what protective regime — taper, replacement, substitution — can keep the system within bounds during the transient? These three questions transfer across substrates unchanged, so the same diagnostic applies to a tapering medication, an ending stimulus program, and a lifted rate limit. The compression is valuable because it converts an apparently mysterious backfire into a quantitative reasoning problem: measure the compensation, compare two timescales, and design the withdrawal trajectory accordingly. An analyst who holds the schema knows that the size of the rebound is not arbitrary but scales with how much compensation accumulated, which means the rebound is forecastable from a measurable quantity — receptor density, backoff-queue depth, suppressed-demand backlog — observed before the suppressor is removed.

Abstract Reasoning¶

The rebound effect reveals the formal connection between adaptive systems and removal transients. Whenever a system has a homeostatic or feedback compensator that adjusts to a sustained perturbation, sudden removal of that perturbation creates a transient governed by the compensator's lag, and the general principle is that the withdrawal trajectory is shaped by the compensator's state at the moment of withdrawal rather than by the counterfactual no-intervention path. This connects to control theory as the open-loop response of an over-compensated plant, to economics as the impulse response after removing a policy stimulus, to ecology as the lag in predator-prey re-equilibration, and to operations as post-incident backlog draining. Recognizing the structure licenses a uniform set of interventions. Taper rather than stop abruptly, giving the compensator time to readjust as the suppressor is removed. Replace and substitute, transitioning to a different agent of the same effect to keep the compensator engaged. Anticipate and pre-mitigate, staffing the surge, queuing the requests, buffering the supply. Avoid sustained suppression in the first place where an intervention whose withdrawal produces no overshoot is available. And measure the compensation, not just the suppression, since knowing how much has built up tells you how big the rebound will be. The same five moves apply whether the compensator is receptor density, a backoff queue, or suppressed aggregate demand.

Knowledge Transfer¶

The inheritable structure is explicit: a target process at baseline; a sustained external suppressor shifting it below baseline; an internal compensator whose state moves to neutralize the suppressor; a lag in the compensator's adjustment, slower than the suppressor's removal; a removal event; an overshoot transient whose amplitude scales with the compensator's state at removal; and a re-equilibration on the compensator's natural timescale. With these fixed, the interventions transfer directly and recognizably. "Taper, do not stop abruptly" is the same move whether tapering a beta-blocker so receptor density can normalize before the antagonist clears, or gradually raising a rate-limit ceiling so queued retries do not fire at once. "Replace and substitute" maps from cross-tapering one medication to another with the same effect to phasing in a replacement control as an old one is removed. "Anticipate the rebound and pre-mitigate" maps from staffing for a post-freeze hiring burst to buffering supply ahead of released demand. And "measure the compensation" maps from quantifying receptor upregulation to monitoring backoff-queue depth as the leading indicator of rebound amplitude. A web service that lifts a month-long rate limit and sees traffic surge to four times baseline as queued retries fire — then learns to raise the ceiling gradually and watch the queue depth — is doing exactly what a clinician does in tapering a beta-blocker rather than stopping it: both recognize that the system adapted while the suppressor was in place, both measure how much compensation accumulated, and both design a removal trajectory slow enough for the compensator to catch up. A central banker exiting a low-rate regime, an ecologist removing a predator, and a manager lifting a spending freeze are all doing the same structural work: account for the compensation built up under suppression, compare its decay rate to the speed of removal, and shape the withdrawal so the overshoot stays within bounds.

Examples¶

Formal/abstract¶

Model the target process as a variable \(x\) with baseline \(x_0\), held below baseline by a sustained suppressor input \(u\), against which a lagging internal compensator \(c\) builds up. Let \(\dot{x} = -k(x - x_0) - u + c\) and let the compensator integrate the displacement, \(\dot{c} = \alpha\,(x_0 - x)/\tau_c\), so \(c\) rises slowly — on timescale \(\tau_c\) — to oppose the suppressor and restore \(x\) toward \(x_0\). At sustained-suppression steady state, \(c\) has grown to roughly cancel \(u\), so \(c^\* \approx u\). Now remove the suppressor abruptly: \(u\to 0\) while \(c\) is still near \(u\). Immediately \(\dot{x} = -k(x-x_0) + c^\*\), and the now-unopposed compensator drives \(x\) above \(x_0\) — the opposite-direction overshoot. The peak excursion scales with \(c^\*\), hence with how much suppression had accumulated; it then decays on the compensator's own slow timescale \(\tau_c\), not on the fast removal timescale. The structure makes the rebound forecastable from a measurable quantity: knowing \(c^\*\) at the moment of removal predicts the overshoot amplitude. And it prescribes the fix — ramp \(u\to 0\) over a time \(\gtrsim \tau_c\) (taper) so \(c\) stands down in step, keeping \(x\) near baseline throughout.

Mapped back: \(x_0\) is the baseline, \(u\) the sustained suppressor, \(c\) the lagging compensator, the \(u\to 0\) event the removal, \(\tau_c\) the timescale mismatch, and the post-removal excursion of \(x\) above \(x_0\) — scaled by accumulated \(c^\*\) — is the opposite-direction overshoot the prime names.

Applied/industry¶

A web service protects a downstream database with a rate limiter. The target process is request throughput; the sustained suppressor is the rate cap holding throughput below natural demand for a month. The lagging compensator is the accumulated backlog: clients hit the cap, back off, and queue retries, and well-behaved clients build up exponential-backoff state — a reservoir of deferred load whose size grows the longer the cap holds. When an operator lifts the cap abruptly, the suppressor vanishes in milliseconds but the queued retries do not decay at that rate; they fire together. The result is a thundering herd — traffic spikes to several times steady-state baseline, the opposite-direction overshoot, large enough to topple the very database the limiter protected. The prime's diagnosis is exact: the overshoot is not random, it scales with backlog depth, the measurable compensator state. The interventions transfer one-for-one from pharmacology. Taper: raise the ceiling gradually rather than removing it, so queued retries drain in step. Measure the compensation: monitor backoff-queue depth as the leading indicator of rebound amplitude before lifting anything. Pre-mitigate: jitter retry schedules and pre-warm capacity for the surge. The same shape governs an ecosystem after predator removal — prey overshoot beyond pre-predator levels because the prey's reproductive "compensation" outlives the predator — and a central bank exiting a long low-rate regime, where suppressed inflation expectations rebound past pre-regime levels once the suppressor lifts faster than expectations re-anchor.

Mapped back: the rate cap is the sustained suppressor, the backoff/retry backlog is the lagging compensator, lifting the cap is the removal event, and the thundering-herd traffic surge scaled by queue depth is the overshoot — with taper, measure-the-compensation, and pre-mitigation porting intact across software, ecology, and monetary policy.

Structural Tensions¶

T1 — Removal Timescale versus Compensator Decay (temporal). The whole phenomenon lives in a timescale ratio: rebound occurs only when the suppressor vanishes faster than the compensator stands down. The failure mode is treating an intervention's withdrawal as instantaneous-by-default and never asking how fast the system's adaptation decays, so the analyst is blindsided by an overshoot that a tapered removal would have erased. Diagnostic: estimate the compensator's natural relaxation time and compare it to the planned removal speed — if removal is faster, plan for rebound.

T2 — Compensated System versus Return-to-Mean (scopal). The prime applies only when the system adapted to the suppressor; a withdrawal that merely returns to baseline is something else entirely, and a competing measurement-noise view explains many apparent "rebounds" as regression. The failure mode is diagnosing rebound where there was no compensator, then prescribing a taper that does nothing because nothing built up. Diagnostic: did removal cause an excursion past baseline in the opposite direction? A simple return is not rebound.

T3 — Forecastable Amplitude versus Unmeasurable Compensation (measurement). The prime promises the overshoot is forecastable from the compensator's state at removal — but that state (receptor density, suppressed-demand backlog, anchored expectations) is often the hardest quantity to measure. The failure mode is a false sense of control: believing the rebound is predictable while having no instrument on the variable that sizes it. Diagnostic: name the concrete proxy for accumulated compensation and confirm it is observable before removal, not only inferable afterward.

T4 — Taper Cost versus Rebound Cost (sign/trade-off). Tapering keeps the system within bounds but prolongs exposure to the suppressor's own downside; abrupt removal ends that exposure but risks overshoot. The two costs point opposite directions, and the right choice depends on which is worse. The failure mode is reflexive tapering that needlessly extends a harmful suppression when the rebound would have been mild, or abrupt removal when the rebound is catastrophic. Diagnostic: weigh the integrated cost of continued suppression during taper against the peak cost of the overshoot.

T5 — Single Compensator versus Multiple Coupled Loops (coupling). The clean model assumes one lagging compensator, but real systems stack several adaptations on different timescales, and a feedback-style multi-loop view takes over. The failure mode is tapering against the slowest compensator while a fast one rebounds anyway, or seeing a multi-phase overshoot and mis-modeling it as one. Diagnostic: look for overshoot structure with more than one peak or timescale — multiple bumps signal multiple compensators that a single-taper schedule cannot all satisfy.

T6 — Removal-Driven versus Trigger-Coincident (sign/causal). The prime claims the rebound's size is set by accumulated compensation, not by the removal trigger — yet a precipitating event can coincide with removal and get blamed, inverting the causal story. The failure mode is attributing the overshoot to whatever small event accompanied withdrawal and intervening on the trigger rather than on the compensation. Diagnostic: ask whether the response magnitude scales with how long suppression was held (compensation) or with the trigger's size — disproportion to the trigger points at the prior gap.

Structural–Framed Character¶

Rebound Effect sits at the structural end of the structural–framed spectrum — structural, aggregate 0.0, every diagnostic reading zero. It is a pure feedback-transient: a process at baseline, a sustained suppressor, a lagging compensator, a removal event, and an opposite-direction overshoot whose amplitude scales with how much compensation accumulated. Every diagnostic points one way.

vocab_travels is zero because the pattern carries no home lexicon that must travel with it: the same shape is told as rebound hypertension in pharmacology, thought-suppression rebound in psychology, the efficiency rebound in economics, predator-removal prey overshoot in ecology, and a thundering-herd traffic surge in software, each in its own field's words, with no need to import "pharmacodynamics" to see it in a queue. evaluative_weight is zero — an overshoot is neither good nor bad in itself; the rebound that topples a database is a failure, the rebound exploited as a controlled discharge is merely a transient, and the prime supplies no approval or disapproval. institutional_origin is zero because the pattern is defined in purely dynamical terms — a timescale mismatch between fast removal and slow compensator decay — with no appeal to any human institution. human_practice_bound is zero: the effect runs indifferently in receptor populations, ecosystems, and retry queues, requiring no role, contract, or practice to exist; a flask of adapted receptors rebounds whether or not anyone is watching. And import_vs_recognize is zero because invoking the prime RECOGNIZES a compensator-driven excursion already wired into the system's feedback structure rather than IMPORTING any interpretive frame — naming the rebound adds nothing beyond noticing the dynamics already present. On every diagnostic it reads structural, and the formal compensator model (\(\dot x\) and \(\dot c\) equations) confirms the skeleton is medium-neutral.

Substrate Independence¶

Rebound Effect is maximally substrate-independent — composite 5 / 5 on the substrate-independence scale. Its domain breadth is maximal: the compensator-lag overshoot after suppressor removal recurs with identical structural force in pharmacology (rebound hypertension on beta-blocker withdrawal, rebound insomnia after benzodiazepine cessation), psychology (thought-suppression rebound, post-diet weight regain), policy and economics (post-austerity inflationary spike, efficiency rebound), ecology (predator-removal prey overshoot, fire-suppression megafires), software (thundering-herd surge when a rate limit lifts), monetary policy (inflation-expectation rebound on exiting a low-rate regime), and organizations (post-freeze hiring bursts). Its structural abstraction is maximal: the signature carries no medium-specific commitment whatever — a target at baseline, a sustained suppressor, a lagging compensator, a removal event, a timescale mismatch, and an opposite-direction overshoot — and the same words ("rebound," "overshoot," "compensation") land in each field without translation. The transfer evidence is maximal: the prime is carried by a shared formal compensator model (the \(\dot x\)/\(\dot c\) equations whose peak excursion scales with accumulated \(c^\*\)), and the documented interventions — taper, replace-and-substitute, anticipate-and-pre-mitigate, measure-the-compensation — port one-for-one from cross-tapering a medication to gradually raising a rate-limit ceiling to buffering released demand. Because the dynamics run indifferently in receptor populations, ecosystems, and retry queues with no agent in the loop, the prime is recognized rather than translated everywhere an adaptive system is un-stressed faster than it can stand down.

Composite substrate independence — 5 / 5
Domain breadth — 5 / 5
Structural abstraction — 5 / 5
Transfer evidence — 5 / 5

Relationships to Other Primes¶

Parents (2) — more general patterns this builds on

Rebound Effect is a kind of Constraint Release

Both files agree on a genus-species relation. constraint_release calls rebound_effect "the specific subspecies where the system had adaptively responded to the constraint... adaptation-driven rebound is only one of its cases," and frames itself as "the broader class." rebound_effect's file is consistent (the overshoot-on-withdrawal transient, of which the general revealed-baseline release may stabilize/oscillate/run-away). Direction verified: the general unmasking prime subsumes the adaptive-transient case. rebound_effect is a real candidate slug and the listed cross-ref. NOT a reparent to cascade (the 0.886 nearest — propagation vs unmasking, explicitly severed). Note: release_from_controlling_context (other cross-ref) is a lateral sibling (actor-moves-to-constraint-free-context, joint- attribution error), left untouched.
Rebound Effect presupposes Stressor Induced Adaptation

Rebound presupposes the compensation built up by stressor_induced_adaptation: 'adaptation is the prerequisite; rebound is the transient when the suppressor is removed faster than the compensation decays.' The build-up phase is its precondition (the 0.89 nearest).

Children (1) — more specific cases that build on this

Jevons Paradox is a kind of Rebound Effect

The file: Jevons is the super-rebound special case of the general rebound_effect (rebound above 100%, total use rises). Jevons is the sign-flipping child of the rebound family. rebound_effect is a candidate (CAND-R2-109-05).

Path to root: Rebound Effect → Constraint Release

Neighborhood in Abstraction Space¶

Rebound Effect sits among the more crowded primes in the catalog (11^th percentile for distinctiveness): several abstractions describe nearly the same structure, so a description that fits it will tend to fit its neighbors too — transporting it usually means disambiguating within this family rather than landing on it exactly.

Family — Memory, Records & Persistence (27 primes)

Nearest neighbors

Computed from structural-signature embeddings · 2026-06-14

Not to Be Confused With¶

The nearest neighbor, and the one most worth disentangling, is stressor_induced_adaptation. The two are halves of a single story and share the same compensatory machinery, which is exactly why they get merged. Stressor-induced adaptation captures the build-up phase: a system under sustained stress upregulates a counter-process, becoming better suited to the stressed environment — its invariant is a monotone accumulation of compensation while the stressor persists. The rebound effect captures the release phase: when the stressor is removed faster than that accumulated compensation can stand down, the now-unopposed compensation overshoots baseline. The decisive difference is the role of the removal event and the timescale mismatch — concepts that simply do not appear in adaptation, which is about the loaded state, not its discharge. A practitioner who knows only adaptation sees a system grow tolerant and stops there; the rebound prime adds the prediction that un-stressing such a system is itself a dynamic event with a forecastable overshoot. Adaptation answers "how is the system now suited to the stressor?"; rebound answers "what happens when the stressor leaves before the suiting decays?"

A second genuine confusion is with tolerance. Tolerance and rebound are often clinically paired and share the underlying receptor adaptation, but they are observations at different times and of opposite sign. Tolerance is the attenuation of effect while the agent is present — the same dose does less because compensation opposes it. Rebound is the opposite-direction excursion once the agent is gone — the compensation, suddenly unopposed, drives the variable past baseline. Tolerance is measured on the suppressor's own axis (the drug works less); rebound is measured on the rebounding variable's axis (the suppressed process surges). Treating them as one phenomenon causes an analyst to manage tolerance by escalating dose, never anticipating that the larger accumulated compensation now guarantees a larger rebound on withdrawal — the two are linked precisely through the compensator state the rebound prime tells you to measure.

A third confusion, easy at the level of surface phenomenology, is with regression_to_the_mean. Both look like "an extreme value that comes back," and a withdrawal followed by a swing can be misread as mere statistical return. But regression to the mean is a property of measurement and sampling — extreme observations are partly noise, and re-measurement tends toward the average with no causal mechanism. Rebound is a causal dynamical event: a real compensator, whose magnitude scales with how long suppression was held, actively drives the variable past baseline in a specific direction. The discriminating test is direction and dose-dependence: regression returns toward the mean and is independent of treatment duration, whereas rebound overshoots past baseline with amplitude that grows with accumulated compensation. Mistaking rebound for regression leads one to wait it out when a taper was needed; mistaking regression for rebound leads to spurious tapers against a compensator that never existed.

For a practitioner the cut points are operational. Adaptation tells you the system has changed; rebound tells you the withdrawal is the dangerous event and must be shaped (tapered, measured, pre-mitigated). Tolerance and rebound move together through the same compensator state, so escalating against tolerance silently loads a bigger rebound. And before prescribing any taper, confirm the swing is a causal overshoot scaling with suppression duration, not a statistical regression that no intervention can or should prevent.

Solution Archetypes¶

No catalogued solution archetypes reference this prime yet.