Two Tuesdays in the same chair
Ravi is 41 and teaches physics at a school in Kochi. He has volunteered for a study at a university hospital, and on two Tuesdays a week apart he lies in the same reclining chair with the same EEG cap on his head. Ravi, the hospital and everyone in his story are invented for this paper. The drugs and the findings are real.
On the first Tuesday the anaesthetist, Dr Leela Thomas, starts a slow infusion of propofol. Every few seconds a recorded voice says his name and asks him to press a button in his right hand. For a while he presses. Then the presses come late, then not at all. His eyelids stop fluttering. On the EEG screen, the fast, small waves of waking give way to big, slow ones, and over the front of his head a strong rhythm near ten cycles a second appears.
On the second Tuesday she gives ketamine instead. The voice asks the same thing. Ravi stops pressing at about the same point in the session. His eyes stay partly open and drift a little. He does not track a finger. To someone at the foot of the chair, the two Tuesdays look almost the same: a man in a chair who has stopped answering.
The EEG says otherwise. Under ketamine the screen does not fill with slow waves. It stays busy, and in the higher frequencies it is busier than when he was awake.
When Ravi comes round from the first session, Leela asks what he remembers. "Nothing," he says. "You said my name, and then you were saying it again, and I was in the recovery bay." When he comes round from the second, he is quiet for a while, then says that he has been somewhere. There were corridors, a sense of rising, a long conversation with someone he could not see, and colours he struggles to describe. "It went on for ages," he says. The drug was stopped twenty minutes earlier.
Same chair, same silence, two different inner nights. That is the puzzle this paper takes on. Why does propofol switch experience off, and why does ketamine keep it running while the person stops responding?
The question is old, but the tools to answer it are new. For most of the history of anaesthesia, the only test was whether a patient moved or remembered. Over the last two decades, researchers have learned to record from many brain regions at once, to knock on the cortex with magnetic pulses, and to wake people at chosen moments and ask what was happening. Those tools are what make a paper like this possible.
My answer has three parts. Propofol strengthens inhibition until the cortex falls into slow, sweeping silences that stop distant regions from exchanging information. Ketamine weakens the inhibitory cells themselves, so the cortex remains active and varied, while its link to the senses and to the body is cut. And the best single measure that tells these states apart, the perturbational complexity index, does so because it tests the cortex's capacity to sustain widespread and varied activity, not its contact with the outside world.
Two ways to stop answering
The clinical word for what Ravi did on both Tuesdays is unresponsiveness. Anaesthetists measure it carefully because it is what they can see. A patient who no longer moves to command, or to a voice, is usually assumed to be unconscious.
That assumption has been questioned for years. In a 2012 paper in Anesthesiology, Robert Sanders, Giulio Tononi, Steven Laureys and Jamie Sleigh argued that unresponsiveness and unconsciousness come apart in both directions. A person can be unresponsive and still be having experiences, as in dreaming. A person can respond without being aware of responding, as in some reflexes and some forms of sleepwalking. They suggested a three-way split: being conscious and connected to the environment, being conscious but disconnected, and being unconscious.
The split matters because each drug can be placed in it. On the evidence that follows, propofol at doses that stop responses tends to produce the third state. Ketamine at similar doses tends to produce the second. The outward sign is the same in both: no answer.
Connected, disconnected, or gone
The cleanest test for connected consciousness under anaesthesia is old and simple. In the isolated forearm technique, a cuff on one arm is inflated before a muscle relaxant is given, so that the forearm can still move while the rest of the body is paralysed. The anaesthetist then asks the patient to squeeze a hand.
A prospective study of 260 adults across six centres, led by Sanders and published in 2017, found that 12 of them, 4.6 per cent, squeezed on command after the breathing tube had gone in. The authors called this a conservative estimate. Most of those patients did not remember it afterwards. The finding is a reminder that the absence of a later report is not proof that nothing was happening at the time, which will matter when we come to the strongest case against my argument.
The isolated forearm test can show connected consciousness. It cannot show disconnected consciousness, because a dreaming patient does not hear the request. That is the gap ketamine falls into. A ketamine patient can be having rich experiences while every behavioural test, including the squeeze, comes back negative.
What the monitors already know
The clinical world learned this lesson from its machines before it had a theory for it. The bispectral index, one of the most widely used depth-of-anaesthesia monitors, turns the frontal EEG into a number from 0 to 100, with values between 40 and 60 usually regarded as adequate for surgery. It was developed largely on patients receiving GABA-acting drugs such as propofol and the volatile agents.
Ketamine does not fit the scale. A clinical review in StatPearls notes that ketamine and nitrous oxide can increase BIS values while causing significant sedation, and calls the monitor unreliable with those drugs. In one study summarised in a 2013 pharmacology review, a ketamine dose of 0.5 mg per kg given to patients already anaesthetised with propofol and fentanyl raised the BIS from 40 to 63.
The monitor was not broken. It was calibrated to one way of becoming unresponsive and was then shown another. That is the point of this paper in miniature. Unresponsiveness is one outward sign with at least two very different brain states underneath it.
Why the lookalike matters
The difference between these two states is not only a puzzle for theorists. It matters in at least three practical places. The first is the operating theatre, where anaesthetists choose drugs partly on what patients will experience and remember. The second is intensive care, where sedated patients may spend days in states that look alike from the bedside but may differ inside. The third is the clinic for disorders of consciousness, where the central question is whether an unresponsive person is having any experience at all.
In each place, the lookalike problem is the same. Responsiveness is easy to test and tells you less than it seems. Propofol and ketamine are the cleanest natural experiment we have for separating the two, because the cause is known, the dose is controlled, and the person usually wakes to tell us what happened. Lessons learned here can be carried, with care, into settings where none of that is true.
Propofol: a cortex in rolling blackouts
Propofol is the most widely used intravenous anaesthetic in the world. It acts mainly by enhancing GABA-A receptors, the main fast inhibitory receptors in the brain. When GABA binds to these receptors, a channel opens and chloride ions flow into the neuron, which makes it harder to fire. Propofol holds the channel open for longer. The net effect is more inhibition, almost everywhere.
More inhibition sounds as though it should simply turn the brain down, like lowering the volume. What happens is stranger and more informative.
The slow oscillation is the key. Its large swings are not random. In the troughs of the slow wave, neurons fall silent together. In the peaks, they fire. Purdon's team found that, at loss of consciousness, alpha power was largest at the troughs of the slow oscillation, a pattern they called trough-max. At deeper levels the relationship flipped to peak-max. The researchers suggested that these phase-amplitude patterns mark states in which cortical activity is organised around the slow cycle and cannot be sustained freely.
Fragments of a cortex
A year earlier, Laura Lewis, Brown, Purdon and colleagues had looked at the same transition from inside the brain. They worked with patients who already had intracranial electrodes in place for epilepsy surgery, and they recorded both individual neurons and local field potentials while propofol was given. Their PNAS report found that a slow oscillation appeared abruptly at the moment of loss of consciousness, within seconds.
During that oscillation, neurons fired only in a limited window of each slow cycle and were silent otherwise. Neurons less than 4 mm apart kept the millisecond-scale patterns of connection they had shown while awake. Across the wider cortex, though, the slow oscillations were out of step with one another, and functional connectivity between areas broke down. The authors described a cortex in which local networks stayed active while their ability to communicate with distant regions broke down. They called this fragmentation.
The cortex under propofol does not go quiet; it goes out of sync with itself. Each patch still works, but in short bursts on its own timetable, so a message started in one region usually finds the next region in a silent phase.
Silences that travel
The slow waves under propofol are not just simultaneous swings in each patch. They move. Sayak Bhattacharya, Earl Miller, Emery Brown and colleagues at MIT ran one such study, published in the Journal of Cognitive Neuroscience. They recorded from electrode arrays in the frontal and auditory cortex of two macaque monkeys and tracked waves of activity before and during propofol. After loss of consciousness, slow waves around 1 Hz became more frequent, faster, more spatially coherent and more stereotyped, sweeping across the arrays in flatter, more regular fronts. Faster beta-band waves became less frequent.
A wave that sweeps the cortex in one direction, again and again, brings silence with it. Each area falls into its down-state when the wave arrives. The more regular the travelling wave, the less room there is for the irregular, back-and-forth exchanges between areas that waking cognition relies on.
The layer that comes apart
There is a further mechanism, at the level of single cells, that may explain why so many different anaesthetics end in the same place. Large pyramidal neurons in layer 5 of the cortex have long dendrites that reach up to the surface. Their cell body near the bottom receives input from below, and the top of the dendrite receives input from other areas and from the thalamus.
In a 2020 study in Cell, Mototaka Suzuki and Matthew Larkum reported that general anaesthesia decoupled these two ends of the neuron in mice. Signals arriving at the top of the dendrite no longer reached the cell body. They found the same effect with three different anaesthetics, and the same decoupling when they blocked certain glutamate and acetylcholine receptors or silenced the higher-order thalamus. They argued that, if feedback signals arrive mainly through these dendrites, the mechanism explains how anaesthesia can block feedback while leaving feedforward signals relatively intact. If the result generalises, part of what anaesthesia does is to cut a bridge inside individual neurons, not just between distant regions.
The propofol dreamers
The tidy version of this story says that propofol switches experience off. The data say something more careful, and the careful version is more interesting.
In a 2023 study in the British Journal of Anaesthesia, Katja Valli, Harry Scheinin, Antti Revonsuo and colleagues in Turku gave healthy men either propofol or dexmedetomidine, raising the dose until they stopped responding. Then they called each participant's name, nudged a shoulder, and, if he woke, asked whether he had been dreaming. They repeated the procedure several times, and compared the answers with awakenings from natural NREM sleep in the same people.
Recall was common. Across 76 awakenings from drug-induced unresponsiveness, 53, or about 70 per cent, produced reports of some content. Among participants given propofol, 73.7 per cent recalled an experience at least once. About 62 per cent of the content reports were dreamlike, and most wove in elements of the research setting. Reports of awareness of the actual surroundings were rare, three of 53. The rates were similar to those from NREM sleep. The authors concluded that both states are characterised by disconnected conscious experiences, and noted that their doses corresponded to moderate sedation rather than surgical anaesthesia.
Surgical patients tell a similar story. In a 2007 cohort of 300 patients, Kate Leslie and colleagues found that 22 per cent reported dreaming on emergence from general anaesthesia. Dreamers were more likely to have received propofol maintenance, and their dreams were mostly pleasant and unrelated to surgery. The authors argued that this dreaming probably happens during recovery, when patients are lightly sedated or in a natural sleep state.
These findings refine the claim rather than overturn it. Near the edge of unresponsiveness, propofol behaves like natural NREM sleep: sometimes there is a disconnected dream, sometimes nothing. At the deeper doses used in the Sarasso study, PCI fell below 0.31 and no volunteer reported anything. The right statement is graded. As the dose rises, propofol moves the cortex from connected experience, through a sleep-like zone of occasional disconnected dreams, to a state with low complexity and no reports. Ketamine at anaesthetic doses does not seem to pass through that last stage at all.
The xenon puzzle
Xenon is the third drug in the Sarasso study, and it is a puzzle for any explanation that runs only on receptors. Xenon is a noble gas that produces anaesthesia at high concentrations. According to the 1998 Nature paper by Nick Franks and colleagues, its effects on GABA-A receptors are negligible; instead it inhibits NMDA receptors, reducing NMDA-activated currents by about 60 per cent at a concentration that maintains surgical anaesthesia.
So xenon blocks the same receptor family as ketamine and largely spares the receptor propofol acts on. On a receptor-only story it should look like ketamine. It does not. In the Sarasso study, PCI under xenon fell from 0.47 to 0.17, lower than under propofol, and the xenon volunteers reported no experience.
A 2023 systematic review in Frontiers in Neuroscience offers part of the answer. Xenon acts at the glycine site of the NMDA receptor, competing with the co-agonist glycine, and it also inhibits other glutamate receptors and influences several other channels. The review concludes that the nature of xenon's NMDA antagonism distinguishes it from ketamine and nitrous oxide, and that this is reflected in its distinct behavioural and electrophysiological effects.
The receptor a drug blocks does not decide the state the cortex ends up in. What decides it, on the evidence here, is the dynamics that follow. Ketamine's action, apparently concentrated on interneurons, releases pyramidal cells and keeps the cortex busy. Xenon's broader dampening of excitatory signalling pushes the cortex towards the same low-complexity state propofol produces, by a different route. This is a counterexample to any account that sorts anaesthetics by receptor class alone, and it is the strongest reason I prefer an explanation in terms of cortical capacity.
Turning consciousness back on
If propofol switches experience off by disrupting communication through thalamus and cortex, then restoring that communication should switch it back on, even with the drug still present. That prediction has been tested.
In a study published in eLife in 2021, André Bastos, Earl Miller, Emery Brown and colleagues kept macaque monkeys under propofol and stimulated frontal thalamic nuclei, the intralaminar and mediodorsal nuclei, through implanted electrodes at 180 Hz. High-current stimulation produced what the authors describe as a strong behavioural wake-up while the propofol infusion continued. Slow-frequency power fell, higher-frequency activity and spike rates rose, and slow-wave synchronisation between thalamus and cortex dropped by 71 to 77 per cent of the way back towards the awake level.
This is the most direct evidence that propofol's effect on consciousness runs through a particular circuit, and that the circuit can be driven past the drug. It also bears on the question of what is lost. The monkeys showed the outward signs of waking. Whether they also had experiences, the study cannot say, but restored responsiveness and restored cortical dynamics arrived together.
Ketamine: a cinema with the doors shut
Ketamine came out of a search for a safer cousin of phencyclidine, a drug that produced anaesthesia but also long, disturbing emergence reactions. The new compound was first made in 1962 at the Parke-Davis company. It was shorter acting and its psychological effects were milder. That made it worth testing in people.
That last phrase carries most of the story. A ketamine patient may look awake. Their eyes may be open. Their muscle tone may be normal, or even raised. Yet they do not respond. The name chosen in a kitchen in Michigan has turned out to be close to the neuroscience.
Removing the brake on the brake
Propofol strengthens inhibition. Ketamine works the other way round, at least at first. Its main target is the NMDA receptor, one of the receptors for the brain's main excitatory transmitter, glutamate. Blocking an excitatory receptor might be expected to quieten the brain. Under ketamine, the cortex becomes more active.
The leading explanation is that ketamine acts more strongly on certain inhibitory interneurons than on the excitatory cells they control. In a 2017 review in Current Opinion in Neurobiology, Oluwaseun Akeju and Emery Brown describe ketamine as blocking NMDA receptors on fast-spiking cortical interneurons, reducing the GABA those interneurons release onto pyramidal neurons. The result is less inhibition of the pyramidal cells and a marked increase in their activity.
Ketamine releases the cortex by taking the brakes off its brakes. The interneurons that normally keep pyramidal cells in check are themselves held back.
A 2023 mouse study in Nature Neuroscience by Joseph Cichon, Alex Proekt and colleagues added a twist. Under ketamine, neurons that had been active during waking became suppressed, and neurons that had been silent became spontaneously active. The switch happened in all cortical layers and in different cortical regions. The authors traced it to the suppression of two classes of interneurons, parvalbumin and somatostatin cells, together with NMDA receptor and HCN channel block. They suggested that two largely separate populations of neurons exist, one engaged in wakefulness and one contributing to the ketamine state.
If that picture holds in humans, the cortex under ketamine is not running the waking programme with the input disconnected. It is running a different programme, on different cells.
Gamma bursts and the busy trace
The EEG under ketamine is distinctive. Akeju and Brown describe gamma oscillations at around 30 to 45 Hz as a characteristic feature of ketamine anaesthesia. At anaesthetic doses these gamma oscillations alternate with slow oscillations in a pattern they call gamma bursts. As the drug wears off, the slow oscillations are the first to go.
A busy EEG is one sign. A more general one is the diversity of the signal. In a 2017 study in Scientific Reports, Michael Schartner and colleagues measured the Lempel–Ziv complexity of spontaneous MEG signals, a count of how many distinct patterns a recording contains. Under psychoactive doses of ketamine, LSD and psilocybin, spontaneous signal diversity rose above the level seen in normal waking.
That study used sub-anaesthetic doses, in people who stayed awake. But it points in the same direction as the anaesthetic data: ketamine does not simplify the cortex. It appears to diversify it.
The decisive comparison
The study that ties these strands together came from the Milan group in 2015. Simone Sarasso, Marcello Massimini and colleagues gave 18 healthy volunteers one of three anaesthetics at doses that made them unresponsive: six received propofol, six xenon and six ketamine. They measured the perturbational complexity index before and during unresponsiveness, using TMS pulses and high-density EEG. Then they woke each volunteer and asked what they had experienced.
The results, reported in Current Biology, divide cleanly. Under propofol, mean PCI fell from 0.50 awake to 0.24, range 0.20 to 0.30. Under xenon it fell from 0.47 to 0.17. Under ketamine it barely moved: 0.48 awake and 0.44 during unresponsiveness, range 0.35 to 0.55, statistically indistinguishable from waking.
The reports followed the numbers. All twelve volunteers in the propofol and xenon groups reported no conscious experience or no recall. All six ketamine volunteers reported full dreams during unresponsiveness. In every case the reports were vivid and rich in visual and emotional content, had a clear narrative, were extended in time, and were unrelated to the anaesthesia setting.
AwakeNo report (squares)Dream reports (diamond)
Six of six dreamed, and twelve of twelve remembered nothing. The samples are small, and the propofol and ketamine groups were different people. But the pattern lines up with the benchmark data used to set the PCI threshold, in which ketamine and REM sleep, both followed by reports, scored above 0.31 and all no-report conditions scored at or below it.
Ketamine also cuts feedback
There is a complication that any tidy story must face. A widely discussed idea is that anaesthetics cause unconsciousness by disrupting feedback signals from frontal to posterior cortex. In a 2013 study in Anesthesiology, UnCheol Lee, George Mashour and colleagues measured directional connectivity between frontal and parietal EEG in surgical patients given ketamine, propofol or sevoflurane.
All three drugs selectively reduced frontal-to-parietal feedback connectivity after loss of consciousness, while feedforward connectivity was preserved. Under ketamine, the effect built up gradually and became significant after loss of consciousness. The authors concluded that diverse anaesthetics disrupt frontal-parietal communication despite their molecular differences.
This is a counterexample to the simplest version of the feedback theory. If ketamine patients are often dreaming, and ketamine cuts frontal-parietal feedback as propofol does, then frontal-parietal feedback cannot be necessary for experience in general. It may instead be necessary for connected experience: for experience that is about the world outside and can be reported to it. The feedback measure would then track disconnection. That reading fits ketamine better than the alternative, which is that ketamine dreams are not experiences at all.
High gamma entropy and the experience
The question this paper was set asks why ketamine preserves vivid, disconnected experience "amid high gamma entropy". The phrase deserves a careful look, because it hides two different claims.
The first claim is descriptive. Under ketamine, the EEG contains more high-frequency activity than under propofol, and measures of signal diversity are higher. That is well supported, from Akeju and Brown's gamma bursts to Schartner's diversity measures.
The second claim is explanatory: that the high gamma, or the high entropy, is what carries the experience. That claim is much weaker, and I want to explain why I do not rely on it.
Three reasons for caution
The first reason is that gamma on the scalp is a noisy signal. Muscles in the scalp and face produce electrical activity at the same frequencies. Ketamine is not a muscle relaxant, and it leaves airway reflexes and breathing largely intact, so muscle activity is a live concern. Careful studies filter or model this activity, but any argument that leans on scalp gamma alone has to show it is not partly muscle.
The second reason is that entropy rewards randomness. A recording of pure noise has very high entropy. A cortex in which every neuron fired independently, with no coordination at all, would score near the top of most diversity measures. Diversity without integration is noise, and noise is not experience on anyone's theory. Spontaneous entropy cannot by itself distinguish a cortex that is richly coordinated from one that is merely disorganised.
The third reason is that the gamma bursts under ketamine alternate with slow waves. A purely spectral story would have to explain why some slow-wave activity coexists with dreaming under ketamine but goes with oblivion under propofol. The likely answer is that what matters is not the presence of slow waves but whether they impose widespread, synchronised OFF-periods, and that is a question about dynamics, not about frequency bands.
Why the knock does better
This is where the perturbational approach earns its place. PCI does not ask how varied the cortex's spontaneous activity is. It sends in a single known cause, a TMS pulse, and asks whether the response spreads widely and stays varied. A disorganised cortex would produce a varied but local response, or no response. A cortex in synchronised OFF-periods produces a large but uniform response. Only a cortex that is both integrated and differentiated produces a high value.
Under ketamine, PCI stays near waking levels. Under propofol, it falls below the threshold. The best evidence that ketamine preserves experience is not the gamma; it is the echo. The gamma and the entropy are consistent with the echo, and they help explain the busy EEG, but they are not the deciding measurements.
The answer in one chain
The question asked why propofol abolishes consciousness through travelling slow waves while ketamine preserves vivid, disconnected experience. Pulling the evidence together, the answer runs as a chain.
Propofol increases inhibition at GABA-A receptors almost everywhere. Neurons in cortex and thalamus become harder to fire and drift into slow cycles of activity and silence. Those cycles are not aligned across regions, and the silent phases sweep across the cortex in regular waves. Local networks keep working, but a signal started in one region usually arrives at the next during a silence, so widespread, varied patterns cannot build. A TMS pulse produces a large, simple response, PCI falls, and at deep enough doses there is nothing to report.
Ketamine blocks NMDA receptors, with an apparently strong effect on the interneurons that hold pyramidal cells in check. Pyramidal cells are released, and the cortex switches to a different pattern of activity, with gamma bursts and, in mice, a partly different set of active neurons. Feedback from frontal to parietal cortex is reduced, and the senses no longer drive the cortex in the usual way. The capacity for widespread, varied activity remains. A TMS pulse produces a complex response, PCI stays near waking, and the person wakes describing a long, internally generated dream.
High gamma activity and high signal entropy belong to the second half of that chain as signs of a busy, disinhibited cortex. They are not, on the evidence, what makes the difference between experience and its absence. The capacity to sustain a varied, integrated response does.
What the toy shows, and what it does not
To make the logic concrete I built a small simulator, the instrument for this part of the paper. It has 24 units and 200 time steps, and two controls. One, labelled GABA, adds a shared slow wave whose silent phase sweeps across the units, a travelling OFF-period. The other, labelled NMDA block, makes units fire more often and more independently, and closes the gate through which an external input reaches them.
The simulator reports three numbers. Diversity is a normalised Lempel–Ziv count of the whole activity pattern. The wave index is the fraction of moments in which a block of at least 16 neighbouring units is silent together. Responsiveness is the share of units that react to an external input.
I ran each setting with 25 random seeds. With both controls at zero, an awake-like state, diversity had a median of 1.01, the wave index 0.30 and responsiveness 0.71. With the GABA control at 0.9, propofol-like, diversity fell to 0.79, the wave index rose to 0.62 and responsiveness fell to 0.17. With the NMDA control at 0.9, ketamine-like, diversity rose slightly to 1.03, the wave index fell to zero and responsiveness fell to 0.17, the same as the propofol-like setting.
That is the toy version of the two theatres. Responsiveness, the behavioural readout, is identical in the two drug settings. The internal measures point in opposite directions. A negative control matters here: with both controls at 0.9, diversity fell further to 0.69 and responsiveness to zero, so the toy does not simply reward any change with higher diversity.
The toy shows how readouts can separate. It does not show that real brains work this way, because I built the separation into its rules. It also shows the problem with entropy: the ketamine-like setting scores highest on diversity partly because its units become more independent, which is exactly the property that, pushed further, would produce noise.
Off, elsewhere, here: the RTM reading
The Recursive Transcendence Model, RTM, is the working hypothesis behind this series. It describes consciousness in five layers. L1, spanda, is basic responsiveness: a system's capacity to be moved and to move. L2, prakāśa, is manifestation: the showing-up of anything at all. L3, vimarśa, is reflexive awareness: experience that registers itself. L4, pratyabhijñā, is self-recognition. L5, cit, is the ground of awareness that the contemplative traditions describe. RTM is a hypothesis, and this paper tests one of its seams.
The propofol and ketamine data fit the layers in a specific way. Under propofol, L1 responsiveness survives locally. Neurons still fire; a TMS pulse still produces a response, often a large one. What fails is the capacity for widespread, varied activity, which RTM ties to L2. Under propofol, nothing shows up.
Under ketamine, the picture is different. The coupling of L1 to the world is cut: the senses no longer drive the cortex in the usual way, and the body no longer answers. But the cortex keeps its capacity for widespread, varied activity, and the volunteers report that things showed up: corridors, colours, voices, a sense of movement. On the RTM reading, L2 is intact and has been turned inward.
Propofol removes the stage; ketamine keeps the stage and locks out the street.
L3 is less clear. A dream is experienced, but is it registered as experience? Most dreams carry little reflective awareness: the dreamer rarely thinks "I am dreaming". The ketamine reports are rich, but the Sarasso data do not tell us whether the volunteers had any sense, at the time, that they were in an altered state. Some reports from sub-anaesthetic ketamine describe a loss of the sense of self, which would be a change at L4. The anaesthetic data cannot settle either question.
A prediction from the layers
If the RTM reading is right, L2 can survive when L1's coupling to the world is cut, but L3 reflexivity need not. That suggests a test. Ketamine dream reports, collected immediately on waking with structured questions, should rarely include the sense of knowing at the time that one was dreaming or drugged, which is the mark of lucidity. Reports from lucid dreamers in REM sleep include it by definition. If ketamine reports regularly included reflective awareness, RTM's separation of L2 from L3 would need rethinking for this state. I have not found a study that asks this question directly, and I list it among the unknowns.
| Māṇḍūkya verse 4 | Ketamine laboratory | Status |
|---|---|---|
| Conscious of the internal (antaḥprajña) | Dreams unrelated to the room | ✓ holds Both describe content made from within. |
| Senses withdrawn | Sensory input gated off | ✓ holds Both describe disconnection. |
| A natural nightly state | A drug state with its own physiology | ✗ breaks Ketamine is not REM sleep. |
| The same self across all states | Group measurements per session | ✗ breaks The lab cannot test identity of the experiencer. |
Where the Upaniṣad and the lab part ways
The parallel has limits. The Māṇḍūkya places dream between waking and dreamless sleep, as a natural state every person passes through each night. Ketamine produces something different: a state the cortex does not normally enter, with a distinctive pattern of activity and, according to the mouse work, partly different neurons. Natural dreaming happens mostly in REM sleep, which has its own chemistry. The ketamine state shares its disconnection and its rich content with dreaming. It does not share its physiology.
The text also makes a claim the lab cannot test. For the Upaniṣad, the experiencer in dream and in waking is one and the same self, and the deep-sleep state that follows is itself a kind of consciousness without objects. Neither propofol nor ketamine data can confirm or refute that. They show when content appears and when it does not. They are silent on whether anything persists when it does not.
Two cases and one checklist
The instrument for decisions in this paper is a checklist with three possible verdicts: off, elsewhere, or here. "Off" means the evidence favours no experience. "Elsewhere" means it favours experience disconnected from the surroundings. "Here" means it favours connected experience: the patient is aware and the environment can reach them. The checklist uses four signals, in order: responsiveness, including the isolated forearm squeeze; the EEG pattern; a perturbational measure if one is available; and the report on waking.
Worked case one: Ravi under propofol
On the first Tuesday, Ravi stops pressing the button. The isolated forearm test, had it been run, would most likely have been negative. His EEG shows large slow oscillations and a strong frontal alpha rhythm, the pattern Purdon's group linked to propofol unconsciousness. Suppose a TMS-EEG session had been run during unresponsiveness and gave a value of 0.25, in the range the Sarasso study reported for propofol. On waking he reports nothing.
All four signals point the same way. The checklist ends at "off". The strongest single signal is the perturbational value, because it is measured during the state and does not depend on memory. If the PCI value had been above 0.31 while the EEG showed slow waves, the checklist would have sent the team to "uncertain: repeat", because that combination is outside the pattern the benchmark data describe.
Worked case two: Ravi under ketamine
On the second Tuesday, Ravi again stops pressing. The forearm test would again most likely be negative. His EEG is busy, with gamma bursts alternating with slow waves. Suppose a perturbational value of 0.44 had been recorded, close to the Sarasso ketamine mean. On waking he reports a long dream unrelated to the room.
Here the signals disagree with each other, and that disagreement is the information. Responsiveness says no. The EEG says the state is not propofol-like. The perturbational value is above the line. The report describes rich content that had nothing to do with the surroundings. The checklist ends at "elsewhere": experience likely present, disconnected from the world.
Running the checklist on published groups
The checklist can be run on real numbers as well as on Ravi. The Sarasso study gives group values for all three drugs, and the reports collected on waking. Treating each group as a single patient is crude, but it shows whether the checklist gives sensible answers on data it was not written around.
For the ketamine group, the answers are: no response; a single, calibrated drug; a perturbational value of 0.44, with a range of 0.35 to 0.55. The checklist ends at "elsewhere". All six volunteers reported long, vivid dreams on waking. That is the positive case, and the verdict matches.
For the propofol group, the answers are: no response; a single drug; a value of 0.24, with a range of 0.20 to 0.30. The checklist ends at "off". None of the six reported any experience. For the xenon group, the value was 0.17 and the verdict is again "off", with no reports. Those are the negative cases, and both match.
This is not an independent test, because the threshold of 0.31 and the drug data come from the same research group. A fair test would apply the checklist, fixed in advance, to a new group of patients at a different centre and compare its verdicts with serial awakenings during the state. The point of running it here is narrower. It shows that the checklist can separate the groups in both directions, rather than saying "elsewhere" or "off" to everyone.
A counterexample from Turku
The Turku propofol study supplies a case the checklist handles badly. Picture one of its participants just past the point of unresponsiveness. He does not respond. A single, calibrated drug is the cause. No perturbational value was measured, so the checklist falls through to question four. His EEG, at this light dose, would most likely show slow waves and frontal alpha, the propofol signature. And when his name is called, he wakes and describes a dream about the laboratory.
The checklist cannot give a clean verdict here. Its "slow waves, no report" branch does not apply, because there is a report. Its "gamma bursts, dream report" branch does not apply, because the EEG is propofol-like. The honest output is the one I added after reading that study: uncertain, with a note that light propofol sedation behaves like NREM sleep, where dreams come and go under a broadly similar EEG.
This is useful rather than embarrassing. It shows where the EEG branch is weakest. A spectral signature tells you which drug is acting and roughly how deeply. It does not tell you, moment to moment, whether a sleep-like cortex is dreaming. In natural sleep, as the Siclari work shows, that seems to depend on local activity in posterior regions, which a standard frontal EEG montage does not resolve well. The perturbational value, measured during the state, remains the stronger signal, and the checklist ranks it above the EEG for that reason.
A negative control and a failure boundary
A good checklist must be able to say "off" to a ketamine-like EEG when the evidence calls for it. Suppose a patient has received ketamine together with a high dose of propofol, a common clinical combination. The EEG may still show some fast activity. But if the perturbational value is low and the slow-wave signature is strong, the checklist should end at "off" or "uncertain", not at "elsewhere". The toy simulator mirrors this: with both controls high, diversity fell to 0.69, below the propofol-like setting.
The failure boundary is clear. The checklist was built from healthy volunteers and surgical patients given single drugs. It does not transfer automatically to drug combinations, to children, to older patients with brain disease, to sub-anaesthetic doses, or to other drugs such as dexmedetomidine, xenon or nitrous oxide. Each of these would need its own calibration of what the EEG and the echo look like. A checklist is only as broad as the population that calibrated it.
The strongest case against
The strongest objection is that the whole argument rests on memory. Propofol volunteers report nothing; ketamine volunteers report dreams. But a report is a memory, formed at or after waking. If propofol blocks memory formation more effectively than ketamine, the reports would differ even if the experiences did not.
The objection has two versions. The first says propofol volunteers may have had experiences they cannot recall. The isolated forearm data make that possible for at least a few patients at the start of anaesthesia. The second, Dennett's cassette version, says ketamine volunteers may not have had the experiences they report; the dream might be composed and loaded on waking.
I think the objection is right that report alone cannot settle the matter, and that this paper's claims would be much weaker if they rested on reports. It is also right that propofol is a potent amnesic, so "no report" is weaker evidence of "no experience" than it looks.
Where I disagree is on what follows. The perturbational measure does not depend on memory. It is taken during the state, from the cortex directly. In the Sarasso study, the propofol and ketamine groups differed sharply on PCI during unresponsiveness, and the direction of the difference matched the reports. If reports were the only evidence, the cassette theory and the amnesia story would be strong rivals. With a measure taken during the state, both have to explain why the cortex under ketamine kept the capacity for widespread, varied activity and the cortex under propofol lost it.
The cassette theory has a reply: the busy cortex under ketamine is composing the cassette. That reply makes a prediction. If dreams are composed without being experienced and loaded at waking, then waking people at random points should yield dream reports that are only loosely related to the moment of waking. If dreams are experienced as they happen, reports should reflect what was going on just before waking, and their content should track the EEG in the seconds before. Something close to this has been tested in natural sleep. In a 2017 serial-awakening study in Nature Neuroscience, Francesca Siclari, Giulio Tononi and colleagues found that dream reports followed periods of reduced low-frequency activity in a posterior cortical "hot zone".
High-frequency activity there was linked with specific contents such as faces, speech and movement. They could predict, in real time, whether a sleeper would report a dream. Critics have argued, in a response hosted on HAL, that this shows the correlates of dream recall rather than of dreaming. That critique is fair as far as it goes. But content-specific activity in the seconds before waking is harder to square with a cassette composed at the moment of waking. I take the experienced-dream reading as the default until a ketamine study tests it directly.
The objection from the access camp
A second rival line comes from theories that tie consciousness to global access, such as the global workspace view and some higher-order views. On these views, a state is conscious when its content is broadcast widely enough to be available for report, reasoning and memory. Ketamine raises a problem for them that is the mirror image of the problem it raises for memory sceptics.
Under ketamine, frontal-to-parietal feedback is reduced, according to the 2013 study by UnCheol Lee and colleagues, and the person cannot report anything at the time. A strict access theorist might therefore say that whatever the ketamine cortex is doing, it is not conscious, and that the dream reports describe content assembled when access returns at waking. That is a version of the cassette theory, motivated by theory rather than by doubt about memory.
The perturbational data cut against this line too, but less decisively. A high value shows that the cortex can sustain a varied, widespread response to a pulse. It does not show that the response is globally broadcast in the access theorist's sense. An access theorist can say that complexity is necessary but not sufficient, and that broadcast to frontal systems is what matters. On that view, the low frontal feedback under ketamine predicts limited or absent experience, whatever PCI says.
I do not think the present evidence settles this, and I say so in the claim ledger. What would help is the serial-awakening test from the ledger: if ketamine dream reports can be predicted from posterior activity during the state, before any access is restored, the access reading is harder to hold. If they cannot, it gains ground.
Where the objection still bites
Even granting all this, two parts of the objection remain. The first is that all the human data come from small groups: six per drug in the Sarasso study. The second is that the ketamine and propofol groups were different people. A crossover study, in which the same people received both drugs on different days, has not, as far as I know, been published. That is one reason Ravi's two Tuesdays are invented: the cleanest version of the experiment is still hypothetical.
What would change my mind
Several results would make me revise the argument.
-
High PCI under propofol with no report. If a large study found PCI values above 0.31 during propofol unresponsiveness, together with reports of no experience, the link between perturbational complexity and experience would weaken.
-
Low PCI under ketamine with dream reports. If, in a larger sample, ketamine volunteers reported dreams after sessions where PCI fell well below the line, the measure would be failing in the other direction.
-
Experience-sampling against the cassette theory, reversed. If interrupted awakenings under ketamine showed that report content was unrelated to activity in the seconds before waking, the experienced-dream reading would lose its main support.
-
A crossover study with a different pattern. Suppose the same volunteers were given propofol and ketamine on different days. If they showed the PCI difference without the report difference, or the reverse, then I would treat the current pattern as an artefact of group differences.
-
A receptor-level explanation that predicts xenon. If someone showed that a simple receptor rule predicted the PCI and report pattern for propofol, xenon, ketamine and nitrous oxide without reference to cortical dynamics, my preference for a capacity explanation would weaken.
-
Deep propofol with frequent reports. The Turku study found frequent dream reports near the edge of unresponsiveness. If a similar protocol at surgical depths, with PCI below the line, produced report rates like those near the edge, the link between low complexity and absent experience would be in trouble.
- A crossover PCI study, same people, both drugs
- PCI at sub-anaesthetic ketamine doses
- Serial awakenings under ketamine with content mapping
- Does the mouse neuron switch occur in humans?
- How much reflexive awareness survives in ketamine dreams
- Brief forgotten experience under light propofol
- Whether a cassette and a lived dream differ for the dreamer
- Whether anything persists when content is absent
What I still don't know
I do not know how much reflective awareness survives in ketamine dreams. I do not know whether the gamma bursts carry content or are a by-product of disinhibition. I do not know whether the mouse finding, a switch to a different population of neurons, holds in human cortex. And I do not know whether low-dose propofol, at which people are sedated but responsive, produces brief disconnected experiences that are then forgotten.
Three further gaps matter for the argument. I do not know where on the propofol dose curve dream reports stop. The Turku study found frequent reports near the edge of unresponsiveness, and the Sarasso study found none at deeper doses. But no study I have found maps reports and perturbational values together across a range of doses in the same people. I do not know whether xenon's low complexity reflects its action on NMDA receptors, its other targets, or both. And I do not know how far any of this applies to older patients, whose brains respond to anaesthetics differently and who are most at risk of confusion after surgery.
Each gap points to a study that could be done with existing tools. That is the encouraging part. The question of whether anyone is at home under anaesthesia is no longer only philosophical. It has become a set of measurements, some made and some waiting.
Claim Ledger
| ID | Claim | Status | Evidence |
|---|---|---|---|
| MF41-C1 | Propofol at doses causing unresponsiveness produces slow oscillations that fragment cortical communication within seconds. | established | Lewis 2012; Purdon 2013 |
| MF41-C2 | Under propofol, slow waves travel more regularly and are more stereotyped. | likely | Bhattacharya 2022 (two monkeys) |
| MF41-C3 | Ketamine reduces inhibition of pyramidal cells by acting on interneurons, raising gamma activity. | likely | Akeju and Brown 2017; Cichon 2023 (mice) |
| MF41-C4 | PCI remains near waking levels under ketamine and falls below 0.31 under propofol. | established | Sarasso 2015; Casarotto 2016 |
| MF41-C5 | Ketamine unresponsiveness is usually accompanied by disconnected experience. | likely | Sarasso 2015 (6 of 6 dream reports) |
| MF41-C6 | Frontal-parietal feedback disruption tracks disconnection, not unconsciousness in general. | speculative | Lee 2013 (inference) |
| MF41-C7 | High spontaneous gamma or entropy is not sufficient evidence of experience. | likely | argument in this paper |
| MF41-C8 | Under RTM, propofol removes L2 while ketamine preserves L2 and cuts L1 coupling to the world. | speculative | interpretation |
| MF41-C9 | The off, elsewhere, here checklist transfers only to single-drug states it was calibrated on. | likely | argument in this paper |
Dated predictions:
- MF41-P1 (by end of 2030): a crossover TMS-EEG study in which the same volunteers receive propofol and ketamine will find PCI above 0.31 under ketamine and at or below it under propofol in at least 80 per cent of participants.
- MF41-P2 (by end of 2032): a serial-awakening study under ketamine will find that dream-report content relates to cortical activity in the 20 seconds before waking, as it does in natural sleep.
Established (solid)Likely (striped)Speculative (hatched)
Glossary
- GABA-A receptor: the main fast inhibitory receptor in the brain; propofol makes it stay open longer.
- NMDA receptor: a glutamate receptor; ketamine blocks it.
- Interneuron: a local inhibitory neuron that controls the firing of excitatory pyramidal cells.
- Slow oscillation: a large brain rhythm below 1 Hz in which neurons alternate between firing and silence.
- Travelling wave: a pattern of activity that moves across the cortex.
- Gamma oscillation: fast brain activity above about 30 Hz.
- Perturbational complexity index (PCI): a measure of how widely and variedly the cortex responds to a TMS pulse.
- Isolated forearm technique: a method that leaves one arm unparalysed so a patient can squeeze on command during surgery.
- Disconnected consciousness: experience without awareness of the surroundings, as in dreaming.
- Spanda (स्पन्द): vibration, pulsation; RTM's L1, basic responsiveness.
- Prakāśa (प्रकाश): light, manifestation; RTM's L2.
- Vimarśa (विमर्श): reflexive awareness; RTM's L3.
- Pratyabhijñā (प्रत्यभिज्ञा): recognition, self-recognition; RTM's L4.
- Cit (चित्): consciousness as ground; RTM's L5.
- Taijasa (तैजस): "the luminous", the dreamer in the Māṇḍūkya Upaniṣad.
- Antaḥprajña (अन्तःप्रज्ञ): conscious of the internal; the dreamer's mode of knowing.
Elsewhere is not nowhere
Propofol and ketamine bring patients to the same outward point, and in doing so they have given consciousness science one of its best natural experiments. Propofol strengthens inhibition until the cortex falls into slow, sweeping silences in which local networks keep working and distant regions stop exchanging information. Ketamine weakens the inhibitory cells, releasing pyramidal activity and closing off the senses. The cortex under propofol loses its capacity for widespread, varied responses. The cortex under ketamine keeps it.
The reports follow: nothing after propofol, long vivid dreams after ketamine. And the perturbational complexity index, measured during the state, separates the two in the same direction. Gamma and entropy are consistent with the difference, but they are not the best evidence for it. The echo of a single knock is.
For clinicians, the lesson is that unresponsiveness is not a single state. For theories of consciousness, it is that contact with the senses is not necessary for experience, and that a measure of capacity tracks experience better than a measure of contact. Through RTM's layers, the finding is that L2 can survive the loss of L1's link to the world, which is what the Māṇḍūkya says of the dreamer, conscious of the internal.
I take no position on whether any artificial system could have L5, the ground the contemplative texts describe. The narrower lesson for anyone building or testing such systems is about calibration. The tests that tell propofol from ketamine were learned from human brains and human reports, and they would need new benchmarks before they could say anything about a system built differently.
There is a practical point under the philosophy. Patients who receive ketamine are sometimes told that they will "just go to sleep". The evidence here suggests that is not accurate. Many will dream, and some of those dreams will be strange or frightening on the way out. A truthful sentence before induction, such as "you may have vivid dreams, and that is expected", costs nothing and may reduce distress. Patients who receive propofol can be told, with fair confidence, that at surgical depth there is most likely nothing to experience, while light sedation may bring ordinary, mostly pleasant dreams.
The deeper lesson concerns what we measure. Responsiveness was the first test of consciousness in medicine, and it remains the most practical. These two drugs show plainly that it is not enough. A person can be unresponsive and absent, or unresponsive and somewhere else entirely. Telling those states apart needed a measure taken during the state and independent of report. That measure exists, it is imperfect, and it is the best tool we have for asking an unresponsive brain whether anyone is at home.
Sources
Anaesthesia and EEG
- Purdon PL, Pierce ET, Mukamel EA, et al. (2013). Electroencephalogram signatures of loss and recovery of consciousness from propofol. PNAS. https://www.pnas.org/doi/10.1073/pnas.1221180110
- Lewis LD, Weiner VS, Mukamel EA, et al. (2012). Rapid fragmentation of neuronal networks at the onset of propofol-induced unconsciousness. PNAS. https://pubmed.ncbi.nlm.nih.gov/23129622/
- Bhattacharya S, Donoghue JA, Mahnke M, et al. (2022). Propofol anesthesia alters cortical traveling waves. Journal of Cognitive Neuroscience. https://dspace.mit.edu/bitstream/handle/1721.1/148762/jocn_a_01856.pdf?sequence=2&isAllowed=y
- Bastos AM, Donoghue JA, Brincat SL, et al. (2021). Neural effects of propofol-induced unconsciousness and its reversal using thalamic stimulation. eLife. https://elifesciences.org/articles/60824v1/peer-reviews
- Suzuki M, Larkum ME (2020). General anesthesia decouples cortical pyramidal neurons. Cell. https://pubmed.ncbi.nlm.nih.gov/32084339/
- Akeju O, Brown EN (2017). Neural oscillations demonstrate that general anesthesia and sedative states are neurophysiologically distinct from sleep. Current Opinion in Neurobiology. https://pmc.ncbi.nlm.nih.gov/articles/PMC5520989/
- Cichon J, Wasilczuk AZ, Looger LL, et al. (2023). Ketamine triggers a switch in excitatory neuronal activity across neocortex. Nature Neuroscience. https://pubmed.ncbi.nlm.nih.gov/36424433/
- Lee U, Ku S, Noh G, et al. (2013). Disruption of frontal-parietal communication by ketamine, propofol, and sevoflurane. Anesthesiology. https://pubmed.ncbi.nlm.nih.gov/23695090/
Dreaming under anaesthesia and xenon
- Valli K, Radek L, Kallionpää RE, et al. (2023). Subjective experiences during dexmedetomidine- or propofol-induced unresponsiveness and non-rapid eye movement sleep in healthy male subjects. British Journal of Anaesthesia. https://pmc.ncbi.nlm.nih.gov/articles/PMC10375502/
- Leslie K, Skrzypek H, Paech MJ, et al. (2007). Dreaming during anesthesia and anesthetic depth in elective surgery patients. Anesthesiology. https://pubmed.ncbi.nlm.nih.gov/17197843/
- Franks NP, Dickinson R, de Sousa SL, et al. (1998). How does xenon produce anaesthesia? Nature. https://www.nature.com/articles/24525
- Systematic review (2023). The cellular mechanisms associated with the anesthetic and neuroprotective properties of xenon. Frontiers in Neuroscience. https://www.frontiersin.org/articles/10.3389/fnins.2023.1225191/full
Consciousness and complexity
- Sarasso S, Boly M, Napolitani M, et al. (2015). Consciousness and complexity during unresponsiveness induced by propofol, xenon, and ketamine. Current Biology. https://orbi.uliege.be/bitstream/2268/198509/1/Consciousness%20and%20Complexity%20during%20Unresponsiveness%20Induced%20by%20Propofol,%20Xenon,%20and%20Ketamine.pdf
- Casarotto S, Comanducci A, Rosanova M, et al. (2016). Stratification of unresponsive patients by an independently validated index of brain complexity. Annals of Neurology. https://pmc.ncbi.nlm.nih.gov/articles/PMC5132045/
- Schartner MM, Carhart-Harris RL, Barrett AB, et al. (2017). Increased spontaneous MEG signal diversity for psychoactive doses of ketamine, LSD and psilocybin. Scientific Reports. https://www.nature.com/articles/srep46421
- Sanders RD, Tononi G, Laureys S, Sleigh JW (2012). Unresponsiveness ≠ unconsciousness. Anesthesiology. https://centerforsleepandconsciousness.psychiatry.wisc.edu/wp-publications/rn257/
- Sanders RD, Gaskell A, Raz A, et al. (2017). Incidence of connected consciousness after tracheal intubation. Anesthesiology. https://pubmed.ncbi.nlm.nih.gov/27984262/
Clinical monitoring and history
- StatPearls: Bispectral Index. https://www.ncbi.nlm.nih.gov/books/NBK539809/
- Mion G, Villevieille T (2013). Ketamine pharmacology: an update. CNS Neuroscience and Therapeutics. https://onlinelibrary.wiley.com/doi/10.1111/cns.12099
- Li L, Vlisides PE (2016). Ketamine: 50 years of modulating the mind. Frontiers in Human Neuroscience. https://pmc.ncbi.nlm.nih.gov/articles/PMC5126726/
- The Pharmacologist. Ketamine: the second act. https://thepharmacologist.org/ketamine-the-second-act/
Philosophy and contemplative texts
- Siclari F, Baird B, Perogamvros L, et al. (2017). The neural correlates of dreaming. Nature Neuroscience. https://pmc.ncbi.nlm.nih.gov/articles/PMC5462120/
- Commentary: The neural correlates of dreaming have not been identified yet. HAL open archive. https://hal.science/hal-04098032/document
- Windt JM. Dreams and dreaming. Stanford Encyclopedia of Philosophy (on Dennett 1976). https://plato.stanford.edu/entries/dreams-dreaming/
- Māṇḍūkya Upaniṣad, verse 4, with Śaṅkara's commentary. Wisdom Library. https://www.wisdomlib.org/hinduism/book/mandukya-upanishad-karika-bhashya/d/doc143595.html
Cite this work
Mahapatra, R. R. Off or elsewhere: why propofol switches experience off while ketamine sends it somewhere else. Mind Frontier paper U0868. rajeshrmahapatra.com. DOI: to be assigned.
@misc{mahapatra_mf_U0868,
author = {Mahapatra, Rajesh Ranjan},
title = {Off or elsewhere: why propofol switches experience off while ketamine sends it somewhere else},
howpublished = {Mind Frontier paper U0868, \url{https://rajeshrmahapatra.com/papers/mind-frontier/U0868-propofol-versus-ketamine}},
note = {DOI: to be assigned}
}