You go under in seconds, and come back by a different road

Every year millions of people are handed a drug that removes them from the world for a few hours. General anaesthesia “affects your whole body,” the US National Library of Medicine explains, and “feels like a deep sleep, but you do not feel anything” [1]. The going-under part is famously quick. What almost nobody is told is that the coming-back part is not the same road run backwards.

Going in and coming out cross different lines

Anaesthetists have a name for this. A 2024 review of what general anaesthesia still cannot explain lists, among the open questions, “the hysteresis of induction versus emergence” [2]. Plain version: the drug level at which you lose consciousness is not the level at which you get it back. You go under at one point on the dial and return at a lower one.

That is not a rounding error. It means the brain’s state depends on where it has been, not only on how much drug is in it. At the same concentration, on the way down you are awake and on the way up you are not.

The mismatch shows up in brain recordings too. A large open imaging dataset was released for exactly this kind of work. Its authors note that earlier analyses found “asymmetric neural dynamics” between losing consciousness and recovering it [4]. A case report published in January 2026 puts it flatly: recovery of consciousness and loss of consciousness “are different processes with specific mechanisms in the brain” [3].

Physicists have started to ask why. A 2025 paper argues that this is general. Systems sitting close to an abrupt, all-at-once kind of transition lose their balance quickly and recover slowly. The authors show that pattern in the loss and recovery of consciousness under anaesthesia, and in the collapse and recovery of stock markets in 2008 [5].

Coming back is a job, not a wearing-off

The tidy assumption is that waking is just the drug leaving. The evidence says the brain has to climb back.

In rats, an injected stimulant drives animals out of anaesthesia produced by two very different drugs [9]. The stimulant was methylphenidate, which raises the brain’s own alerting chemicals. The anaesthetic had not left. The arousal machinery was simply switched on, and that was enough.

Which arousal machinery matters is still being sorted out. One 2025 study followed the locus coeruleus, a small alerting cluster in the brainstem. Its activity tracked anaesthetic depth closely. But turning it up or down “does not significantly affect anaesthetic emergence” [10]. An honest negative result, and a useful one.

The return also runs on different wiring from the one sleep uses. In mice, researchers watched the two small cell clusters that promote sleep and wakefulness. They behaved one way across natural sleep transitions, and another way across losing and recovering responsiveness under anaesthetic gas. The two processes, the authors conclude, “partly do not share similar functional connectivity patterns” [6].

Recordings inside the brains of 34 people being assessed for epilepsy surgery found the same split from the other side. Some anaesthetics produce more sleep-like brain activity than others, and the commonest ones are not the most sleep-like [7]. A separate scan study compared two anaesthetics, natural sleep, and brain-injured patients. Five parts of the thalamus - the brain’s central relay - changed in state-specific ways, rather than all dimming together [8].

What “under” is actually like inside

The unconscious brain is not blank. Researchers placed fine recording probes into the hippocampus, a deep memory structure, in seven patients anaesthetised with propofol. Propofol is the milky drug most often used to send people under. Neurons there still sorted expected sounds from odd ones, and got better at it over ten minutes of playback. Played a podcast, individual cells fired differently for nouns than for verbs. They grouped “cat” with “dog” and away from “pen”, and predicted the next word in real time [11]. None of the patients remembered a thing.

So the anaesthetic did not stop the brain working. It stopped something else, and finding a reliable marker for that something is still live work. In January 2026 Nature reported a candidate: a distinctive brain-wave pattern - the brain’s electrical activity falling out of step - that marks the slide into unconsciousness. If it is confirmed, it could help avoid sedating people too deeply or too lightly [12].

Both edges of that are real. Waking up during surgery is rare, with estimates spanning 1 in 1,000 to 1 in 20,000 depending on how hard you look for it [13]. The UK’s fifth National Audit Project found it in about 1 in 8,200 cases where muscle relaxants were used, and 1 in 135,900 without them [13]. Relaxants remove the patient’s ability to move - the clearest sign that an anaesthetic is running too light.

Dreaming, by contrast, is common and mostly pleasant. A scoping review pooled 157 studies covering 87,866 people. Large clinical trials found 3% to 8% dream recall; experiments that asked immediately and carefully found 40% to 80% [14].

The bill for the return trip

The asymmetry is not free, and older brains pay most of it. Postoperative delirium is sudden confusion, with disordered attention and thinking, in the days after surgery. It affects 10% to 60% of major-surgery patients, and a longer-lasting decline in thinking affects 10% to 25% [15]. An earlier review put the elderly figure at 20% to 50%, and concluded that the type of surgery matters more than the type of anaesthetic [22].

It is also routinely missed. A prospective study followed 304 patients aged 65 and over in Enugu, Nigeria. Delirium appeared in 24%, over 90% of it within 48 hours. Of those cases, 90.4% went undetected by the surgical and nursing teams looking after them [16]. Hospitals in the US describe the same thing to patients in plainer words: temporary confusion or memory trouble, lasting hours to days [23].

In 111 older adults in Bangkok, thinking scores fell by a meaningful margin in 40.5% at one week. At three months, 40% were still below their own baseline - though nearly half of those affected had recovered [17].

People whose brains were already unsteady fare worst. A 2025 review reports a three- to five-fold higher risk of a delayed wake-up in patients with psychiatric or chronic brain illness, and delirium rates reaching 60% [21]. And the return can simply stall. The January 2026 case report describes a 28-year-old woman with undiagnosed narcolepsy, who took roughly ten hours to come round after two routine internal camera examinations [3].

What helps, and how sure we are

Two lines of evidence now point the same way, and both are about not going deeper than needed.

A meta-analysis pooled 12 trials and 7,441 older patients. Steering the anaesthetic by brain-wave monitoring, rather than by habit, cut delirium after surgery by roughly a quarter [18]. A second pooled 8 trials that split patients at random, 1,957 in all, comparing deliberately lighter anaesthesia with deeper. Delirium was about half as likely, the longer thinking decline likewise, and recovery-room time about nine minutes shorter [19].

Both findings are modest, not miraculous. And the instrument is imperfect in exactly the wrong place. The brain-wave indices were built mostly on data from younger adults. One brain wave the monitors lean on, the alpha rhythm, fades with age. That pushes the number up even when the brain is deeply suppressed, so the readings can understate depth in the patients most at risk [20]. Sleep-like brain activity during the return, meanwhile, looks protective - less agitation on waking, less time in recovery [6].

The honest bottom line: the going-under half is well controlled and steadily better measured. The coming-back half is a separate process, less understood, and still listed in the literature as unexplained [2]. Anyone facing surgery should take their own case to their anaesthetist. That conversation, not an article, is where the specifics belong.