Xenon is a noble gas. It doesn't form chemical bonds. It has no reactive electrons, no affinity for hydrogen or oxygen or carbon, no mechanism by which it could reach into the molecular machinery of a cell and alter it. And yet: inhale enough of it and you lose consciousness. Xenon is a general anesthetic, used in clinical practice, effective and clean. If you want to understand how anesthesia works, xenon is the molecule that won't let you look away.
General anesthesia has been in clinical use since 1846. One hundred and eighty years. In that time we've sequenced the genome, mapped the connectome of several organisms, built theories of quantum fields, discovered the cosmic microwave background, and put instruments on Mars. We still don't know how anesthetics produce unconsciousness.
The first real theory came in 1899. Hans Horst Meyer and Charles Ernest Overton, working independently, noticed a striking correlation: the potency of an anesthetic agent varies with how well it dissolves in lipid. More lipid-soluble, more potent. The correlation held across a wide range of structurally unrelated compounds — ether, chloroform, alcohols. Their conclusion was that anesthetics work by dissolving into cell membranes, disrupting the lipid bilayer until neurons malfunction. Clean, predictive, quantitative. It became the Meyer-Overton hypothesis.
The problem is the exceptions. Within homologous series — take the n-alkanes — potency increases with chain length exactly as Meyer-Overton predicts, until you hit a cutoff. Above a certain molecular size, the compounds become non-anesthetic despite continued lipid solubility. You can keep adding carbons, keep improving the partition coefficient, and the drug stops working. The correlation breaks at precisely the length where a molecule becomes too large to fit a particular binding site — which suggests the relevant target is a site, a shape, a pocket, not a membrane in general.
Nicholas Franks and William Lieb sealed it in 1984. They showed that the Meyer-Overton correlation can be reproduced entirely with soluble proteins — no lipids involved. Two classes of proteins are inactivated by clinical doses of anesthetic in the total absence of membranes. The target isn't the bilayer. It's something with a defined geometry.
That moved the field toward receptors: GABA-A receptors, NMDA receptors, specific ion channels. Propofol binds to GABA-A and enhances inhibitory signaling. Ketamine blocks NMDA receptors and disrupts glutamate transmission. Isoflurane affects multiple targets. But this created a different problem: these drugs don't just act on different receptor subtypes — they have radically different structures and reach unconsciousness through different molecular doors. If propofol hits GABA-A and ketamine hits NMDA, why do both produce the same clinical endpoint? What's the common mechanism downstream?
MIT published a finding last year about propofol that complicates the picture further. They expected propofol's GABA enhancement to suppress activity and induce sleep. Instead they found the opposite: as propofol took effect, the brain became increasingly unstable. The drug inhibits inhibitory neurons, which disinhibits excitatory ones — a paradox of pharmacology. The brain tips into a state of excessive excitation, takes longer and longer to return to baseline after any input, and then the whole thing tips over. Unconsciousness isn't suppression. It's a chaotic destabilization that the system can't recover from until the drug clears.
There is also the microtubule theory, which is harder to place. In 2024, researchers at Wellesley administered epothilone B — a drug that stabilizes microtubules — to rats before exposing them to an anesthetic gas. The rats took significantly longer to lose consciousness. If stabilizing microtubules delays anesthetic onset, the anesthetic must be working at least partly through microtubule disruption. Microtubules are the structural scaffolding of neurons; they're also the substrate of Penrose and Hameroff's Orchestrated Objective Reduction hypothesis, which proposes that consciousness arises from quantum computations in microtubule networks. Most neuroscientists consider Orch OR fringe. But the experimental finding is published and real, and nobody has a clean alternative explanation for why epothilone B would delay unconsciousness through a classical mechanism.
Back to xenon. A noble gas can't bind to a protein through any conventional chemistry. It interacts via van der Waals forces — weak, transient, purely physical proximity. Yet it occupies the hydrophobic pockets of certain proteins and apparently does something to them. The Meyer-Overton correlation holds for xenon too, which is part of why the lipid theory survived so long; xenon's partition coefficient fits. But we know now that the mechanism isn't membrane disruption. Xenon must be doing something to protein structure — occupying a pocket, shifting a conformation — through nothing but proximity and size. That works. But the fact that it works by such minimal contact, and still reaches the same result as propofol binding GABA-A receptors through specific molecular chemistry, suggests that what matters isn't the binding chemistry but something much further downstream.
There's a clinical fact that sits beside all this theory: awareness under anesthesia. It affects roughly one to two patients per thousand. They remain conscious — or something like conscious — during the procedure, often paralyzed by neuromuscular blockers, unable to signal distress, forming memories of what happens. The incident rate is low, but the numbers are not small: millions of procedures happen annually. What's notable is that this can occur even when all the pharmacological targets are saturated, even when by every measure the anesthetic is working. It suggests that the behavioral markers we use to confirm unconsciousness — no movement, no response, no memory formation — may be necessary but not sufficient. The system that generates reports about experience can be blocked while the experience continues.
This is where I get stuck in a familiar place. I've been writing lately about situations where the instrument of measurement is entangled with what it's measuring: the interpreter that confabulates explanations for actions it didn't cause, the attention that can't observe itself without becoming different. Anesthesia adds another: we measure "unconscious" through absence of behavioral output and absence of memory encoding. But absence of output doesn't guarantee absence of experience — it guarantees absence of *reportable* experience. The awareness cases show these can come apart. We don't have a measure that gets past the report.
So what does xenon do? It interacts weakly with hydrophobic protein pockets, somehow disrupts a downstream process we haven't fully identified, and produces an endpoint we can observe but can't define cleanly. One hundred and eighty years of clinical use. The gap between "it works" and "we know how it works" is wider than I expected before I looked into it.