Tolerance, dependence & down-regulation

When a compound over-stimulates a receptor day after day, the neuron protects itself. The first response is fast and local: the receptor is chemically modified so that it no longer couples efficiently to its signalling machinery, and is then pulled inside the cell where the drug cannot reach it. For opioids this sequence of desensitisation, phosphorylation and internalisation is worked out in unusual detail and is the first layer of tolerance ^[2]. Now the same dose lands on fewer, duller receptors, so the effect shrinks.

That is only the first of four layers, and the layers run on very different clocks. Below them all sits a fifth thing that is not tolerance at all but shapes how tolerance is experienced: learning. Cues that reliably precede a drug come to trigger the counter-adaptation in advance, which is why the same dose in an unfamiliar setting can hit far harder than in a familiar one.

Tolerance is not uniform across a drug's effects, and this is the detail most worth carrying away. Benzodiazepines are the clearest case: tolerance develops relatively quickly to the sedative and anticonvulsant actions, while tolerance to the anxiolytic and amnesic effects probably does not develop at all ^[7]. The practical implication is uncomfortable. Someone taking a sedative for sleep needs steadily more of it for the same sleep, and every increase brings the effects that have not tolerated along with it. The same asymmetry explains why stimulant euphoria fades long before the cardiovascular load does, which is what makes escalation dangerous rather than merely disappointing.

The mechanism at the receptor level also varies more than the word down-regulation suggests. Prolonged exposure to GABA-A modulators changes subunit expression, intracellular trafficking and phosphorylation state rather than simply removing receptors, and the resulting receptor is a different receptor rather than a scarcer one ^[8]. That is why tolerance to one GABAergic compound transfers imperfectly to another, and why compounds selective for particular receptor subtypes have shown no convincing tolerance in the same assays ^[7].

These three are related but genuinely distinct, and conflating them is the source of a great deal of bad advice in both directions.

Tolerance is the effect fading. Physical dependence is the same adaptation showing up as withdrawal or rebound when the drug is removed: rebound anxiety after a calming drug, crushing fatigue after a stimulant, pain and autonomic upheaval after an opioid. Withdrawal symptoms are very often the mirror image of the drug's effects, because they are the counter-adaptation running unopposed. Addiction is a behavioural disorder; compulsive use despite harm, driven by circuitry that spans binge and intoxication, withdrawal and negative affect, and preoccupation and anticipation ^[1].

The clean way to hold them apart: you can be tolerant without being dependent, and dependent without being addicted. Someone stable on a prescribed medication for years may have both tolerance and physical dependence and no addiction at all. Conversely, addiction can persist long after any physical dependence has resolved, which is why detoxification on its own is not a treatment.

The oldest theory here still explains most of what people notice. The opponent-process account holds that any affective state triggers an opposing process, that the opposing process strengthens with repetition while the original weakens, and that what you eventually feel is the sum ^[3]. That is tolerance and withdrawal in one sentence, and it predicts the specific shape of the problem: the high gets smaller and the trough gets deeper even though nothing about the dose has changed.

The modern version adds a piece that matters. Rather than simply returning to baseline between doses, the system settles at a new baseline; counter-adaptive processes that are supposed to bring reward function back into range stop doing so, and the set point drifts, maintained by stress systems as much as reward ones ^[4]. This is why the honest description of long-term heavy use is not chasing a high but paying to feel ordinary.

One more separation is worth making, because it explains behaviour that otherwise looks irrational. Wanting and liking are not the same system. The incentive-sensitisation account holds that repeated use sensitises the circuitry that makes a thing feel worth pursuing, while the circuitry that makes it feel good does not sensitise and may weaken ^[14]. Craving growing while enjoyment shrinks is exactly what that predicts, and it is what imaging in active cocaine users shows: stimulant-induced dopamine increases were markedly blunted compared with controls, and yet the same stimulant still triggered intense craving ^[17].

Risk is not evenly spread, and the differences are large enough to matter more than almost any other consideration when choosing what to take regularly.

The table below is the practical summary. The column to read first is the last one, because a compound that is merely unpleasant to stop and a compound that can kill you to stop are in different categories entirely and are frequently discussed as though they were not.

Two general rules run underneath it. Speed of onset predicts trouble. The faster a compound raises its target, the stronger the counter-adaptation and the more addictive the pattern tends to be, which is most of why route of administration matters so much. And cross-tolerance follows mechanism, not name. Tolerance to one GABA-A modulator transfers substantially to another, so alcohol, benzodiazepines and Z-drugs are not independent risks stacked in a list; they are the same risk counted several times. See anxiolytics and GABA.

One phenomenon deserves its own section because it is genuinely counterintuitive and it changes what safe use means.

For most unpleasant experiences, repetition makes them easier. Withdrawal from alcohol and from strong GABAergics goes the other way. Each episode of withdrawal makes the next one more severe, a process called kindling, borrowed from the observation that a weak repeated electrical stimulus which initially produces nothing eventually produces seizures ^[11]. Someone on their fourth unassisted detox is at higher risk than they were on their first, at the same level of use, and the risk includes seizure.

That has two consequences that follow directly. It is the reason clinical guidance is to treat even mild withdrawal properly rather than waiting for it to become severe ^[11], and it is the reason repeated cycles of quitting and restarting are worse than either sustained use or one properly supported stop. Kindling may also contribute to relapse risk and to the cognitive impairment seen after long heavy use, so the damage is not confined to the acute episode.

The general management principles for drug and alcohol withdrawal are well established: taper rather than stop, substitute a longer-acting drug of the same class where one exists so the fall is gradual, and treat the autonomic symptoms directly ^[10]. For benzodiazepines specifically, the evidence supports gradual dose tapering with psychological support where needed, and studies consistently find that mental and physical health and cognitive performance improve after withdrawal, particularly in older people taking them as hypnotics ^[9]. That last point is worth stating plainly because it cuts against the fear that makes people stay on them: the outcome of getting off, done properly, is usually better function rather than worse.

This is not a do-it-yourself procedure for the dangerous classes. Withdrawal from alcohol or high-dose benzodiazepines is a medical situation, and the correct move is a doctor rather than a schedule found online.

The practical takeaways are unglamorous and they work.

Use the lowest effective amount. Tolerance is driven by the size and persistence of the signal, so the dose that is merely enough produces less adaptation than the dose that is comfortably more than enough. Escalating to chase a fading effect is the mechanism of the problem rather than a solution to it.

Take real breaks. This is what cycling is actually for. It is worth being honest that the evidence base for specific schedules is thin to nonexistent for most compounds, and the few hard numbers that exist are humbling: cortical CB1 receptors took about four weeks of continuous abstinence to normalise ^[18], and dopamine transporter density in methamphetamine users recovered measurably only between 12 and 17 months of abstinence ^[16]. A two-day weekend off is not a receptor holiday for anything with a slow adaptation.

Watch the interval, not just the dose. Something taken every third day produces far less adaptation than the same weekly total taken daily, because the counter-adaptation needs persistence to build. Where an effect genuinely requires daily dosing, that is a cost to be priced in rather than a detail.

Be much more cautious with the classes where dependence forms fast and withdrawal is dangerous: strong GABAergics such as benzodiazepines and phenibut, opioids, and heavy stimulants. Combining two compounds that push the same system multiplies their risks rather than adding them, which is exactly what this site's interactions and stacks tool is for.

Treat a compound that has stopped working as information. The useful response to a fading effect is a break or a stop, not an increase; an escalating dose with a flat effect is the signature of a problem forming. And taper rather than stop anything taken daily for more than a few weeks, whatever the class. Not medical advice.

Confident recovery timelines circulate widely in this space and almost none of them rest on measurement. Four gaps are worth knowing.

Recovery timelines are mostly unmeasured. The two numbers quoted above, four weeks for CB1 and 12 to 17 months for dopamine transporters, are notable precisely because they are rare ^[16][18]. For almost every compound discussed on this site there is no imaging study, no post-mortem series and no timeline, so any specific claim about how long receptors take to recover from it is extrapolation at best.

Receptor recovery is not the same as functional recovery, and the gap is measurable. In the methamphetamine study, transporter density in the caudate and putamen rose significantly over protracted abstinence, while performance on neuropsychological testing did not improve to the same extent ^[16]. Whatever the persistent deficit is made of, it is not fully explained by the number of transporters.

The molecular basis of the long-lasting part of addiction is unresolved, and the field says so. The behavioural abnormalities are close to permanent, which means something stable must underlie them; but the specific molecular changes identified so far, in gene transcription, RNA and protein processing and synaptic structure, are not long-lasting enough to account for them ^[5]. That gap has been open for over two decades.

Tolerance is uncharacterised for most of what this site covers. There is no controlled human study of tolerance to any racetam, most peptides, or the great majority of research chemicals. Claims that a given compound does or does not build tolerance are reports rather than findings, and the absence of a tolerance warning carries no information. See research chemicals and harm reduction.

What is solid is the shape of the thing: adaptation is real, it is layered, it runs on different clocks at different levels, it is not uniform across a drug's effects, and for a small number of classes the exit is genuinely hazardous. Everything more precise than that should be read as an estimate.