ScentSearch
Search scents →
ScentSearch
Search scents →
A literature summary, not applied expertise. Molecule to receptor to pattern to perception — and where the research stops.
You smell something. Your brain does not receive a label saying vanilla. It receives a pattern of activity across a set of receptors, and has to work out what that pattern means.
This page follows that process — molecule, receptor, pattern, perception — and stops where the research stops, which turns out to be sooner and more often than the popular version suggests.
I am not a neuroscientist. The other pages on this site apply training I actually have: thermal systems, heat transfer, evaporation. This one does not. It is a structured reading of primary papers, each claim traced to the study that made it, and it was reviewed for accuracy by someone with a doctorate in neuroscience.
Where a claim is contested, the page says so rather than picking the tidier version.
The obvious model for smell is a lock and key: a rose molecule fits a rose receptor, and you smell rose.
That is not how it works, and the reason is arithmetic. Humans have several hundred working olfactory receptors and can distinguish far more smells than that.
Linda Buck and Richard Axel identified the first large family of candidate odorant receptor genes in 1991,[1] and won the Nobel Prize in Physiology or Medicine in 2004. They also established something unexpected: each olfactory neuron expresses one receptor type. One neuron, one receptor.
The arithmetic came later. Working with Hirono and Sato in 1999, Buck established that odour coding is combinatorial — each odorant activates several different receptors, and each receptor responds to several different odorants.[2]
So identity is not a key. It is a pattern — which receptors fired, and how strongly. A few hundred receptors generate an enormous number of possible patterns, the same way a few dozen letters generate an enormous number of words.
That single fact explains most of what follows. It is also why a fragrance cannot be reduced to a list of the materials in it: what you perceive is the pattern the whole mixture produces, not the sum of its parts announced one at a time.
The receptor genes come in a large family, and a lot of that family is dead.
Malnic, Godfrey and Buck searched the human genome in 2004 and found 339 intact olfactory receptor genes and 297 pseudogenes — sequences that retain recognisable similarity to olfactory receptor genes but carry changes consistent with loss of function. They fall into 172 subfamilies.[3]
So in that census, roughly half the olfactory receptor genes in the human genome no longer work. Not damaged in you specifically — across the species.
⚠️ Those are the counts from that 2004 analysis, not a settled inventory. Estimates have shifted with better sequencing; the Nobel presentation the same year used a figure of about 350 receptors.[6]
That is interesting on its own. What makes it matter is that the working half is not identical between people.
This is the clearest demonstration that perception varies at the level of the hardware, and it involves one gene.
Androstenone is a steroid derived from testosterone. Ask people what it smells like and you get three different answers: offensive — sweaty, urinous — or pleasant and floral, or nothing at all.
Not three descriptions of one experience. Three experiences.
Keller, Zhuang, Chi, Vosshall and Matsunami traced it to a single receptor. OR7D4 responds to androstenone and to a closely related steroid, and did not respond to a panel of 64 other odours they tested it against. A common variant of the gene carries two single-nucleotide changes that severely impair how the receptor functions.[4]
People with two working copies were more sensitive to androstenone and found it more unpleasant. People carrying the impaired variant were less sensitive and less bothered by it.
A later study, testing androstenone in cooked pork with a Norwegian cohort, found the genotype accounted for around 40% of the variance in intensity ratings in that sample.[5]
One gene. Two point mutations. Sweat, or flowers, or nothing.
This is a concrete example of the biological basis of specific anosmia — and it is why two people can stand over the same bottle and honestly disagree about what is in it. Not merely a difference of vocabulary or taste. In this case, at least part of the difference comes from the receptors doing the detecting.
⚠️ OR7D4 is the best-characterised case, not the only one. It is the example because the link from gene to receptor function to reported perception was established end to end, which is unusual. How much of ordinary disagreement about fragrance works this way is not known.
This one is worth the space, because it is a case study in how a number becomes a fact.
About 10,000. It appears in textbooks, in press coverage, and on the Nobel Prize committee's own 2004 release.[6]
It has never been empirically validated. That is not a critic's characterisation — it is what the 2014 paper that tried to replace it says in its own abstract: the literature typically claims 10,000, and the number has never been tested.[7] Meister, reviewing the same claim, treats it as largely anecdotal.[9]
In 2014, Bushdid, Magnasco, Vosshall and Keller published in Science: humans can discriminate more than one trillion olfactory stimuli. Subjects compared mixtures of 10, 20 and 30 components drawn from a library of 128 molecules; the researchers extrapolated from how often people told the mixtures apart.[7]
A hundred-million-fold increase on the old figure, in a top journal, from Rockefeller. It was covered everywhere.
Gerkin and Castro (eLife, 2015) showed the estimate was extraordinarily sensitive to choices that should not have mattered.[8] In their words, it "varies systematically and sensitively (over tens of orders of magnitude, in both directions), for very modest changes in incidental experimental and analysis parameters."
Meister (eLife, 2015) went further, and his demonstration is the one worth remembering.[9]
The experiment discriminated 148 pairs of odours successfully. Everything above that came from extrapolation. Meister shows the data "are equally consistent with a trillion discriminable odors and with just 10 (and anything in between)."
Then he built a model organism with three possible percepts, by construction — it can only register "yum", "meh" and "yuck" — and ran it through the identical test. The method concluded it could discriminate roughly 9 × 1011 odours. Around a trillion, from a system with three.
He also applied the same method to colour vision, where the answer is known to be on the order of a million, and got more than 1027.
In a posted reply, the authors conceded specific points: that the threshold sensitivity is real and would hold for any similar dataset, and — the substantive one — that the assumption of a high-dimensional olfactory perceptual space should have been stated explicitly in the paper.[10]
They defended the assumption itself, arguing that the result holds above a certain dimensionality and that olfaction is unlikely to follow the low-dimensional rules that make colour vision tractable.
Science published an erratum in February 2024, formally indexing an earlier correction.[11]
⚠️ An erratum is a correction, not a withdrawal, and it should not be read as the rebuttals winning. This is a live disagreement with concessions on both sides.
Unknown.
The old number was never established. The trillion estimate is not established by that dataset. And the honest position — the one Gerkin and Castro put in their title — is that the number of olfactory stimuli humans can discriminate is still unknown.
That is a better fact than either figure.
You will read that smell is the only sense that bypasses the thalamus. The popular version is roughly right and stated too absolutely.
What holds: for vision, hearing and touch, the thalamus is the relay that sensory information passes through on its way to primary sensory cortex. Olfaction is the exception — its primary pathway does not.[13]
What the popular version misses: olfactory information does reach the thalamus. The mediodorsal thalamic nucleus receives direct input from primary olfactory areas including piriform cortex, and connects densely with orbitofrontal cortex.[12]
So there are two routes to the neocortex: a mostly direct one from piriform cortex, and an indirect one through the mediodorsal nucleus, which carries a smaller share of the fibres.
What that route does is an open question. Lesion, imaging and electrophysiological work suggests involvement in odour perception, discrimination, learning and attention, and the researchers reviewing it say plainly that many important questions remain unanswered.[12]
Olfactory pathways also reach the hypothalamus, which is the brain's endocrine control centre. That connection is real, and section 7 covers what it does and does not establish.
The accurate version is more interesting than the myth. Smell's route to cortex is genuinely unusual. It is not a clean bypass, and what the second route contributes is still being worked out.
Before going further it is worth separating two systems that get conflated constantly, including by everyone selling fragrance.
The intranasal trigeminal system is a third chemical sense, alongside smell and taste.[14] It is somatosensory — the same nerve that handles touch and temperature in your face — and it detects chemicals directly. Its primary job is guarding the airway: it can reflexively stop you inhaling something dangerous.
It produces cooling, tingling, burning, and the perception of airflow.
⚠️ These are not two separate populations of molecules. The great majority of airborne chemicals we detect with our noses produce both an olfactory and a trigeminal response, at least at higher concentrations.[15] So it is less that some things are trigeminal and others olfactory, and more that most fragrance materials are doing both at once, in a ratio that shifts with concentration.
| Receptor | Familiar activator | Typical sensation |
|---|---|---|
| TRPM8 | menthol, eucalyptol; cool temperatures | cooling, freshness |
| TRPA1 | mustard oil, wasabi (allyl isothiocyanate) | irritation, burning; also implicated in cold sensing |
| TRPV1 | capsaicin | burning heat |
Those plant compounds were the tools used to find the channels rather than incidental activators of them. Caterina and colleagues cloned TRPV1 in 1997 by screening for a capsaicin response, and found that the same channel is opened by noxious heat.[16] Menthol and mustard oil served the same role for TRPM8 and TRPA1.
Two consequences worth knowing.
The feeling of a clear nose is a cold receptor firing. Nasal airflow is perceived through TRPM8, activated by mucosal cooling. Which is why menthol makes breathing feel easier without changing how much air is actually moving.
And menthol changes character with dose — though not in the way the tidy version suggests.
At ordinary concentrations menthol activates TRPM8 and reads as cooling. At high concentrations people report it as irritating rather than simply cool, and that psychophysical shift is well established.
But the receptor story is not "a second channel switches on." Menthol acts on TRPA1 bimodally: submicromolar to low-micromolar concentrations activate the channel, while higher concentrations cause reversible block, producing a bell-shaped dose-response curve.[17] The same molecule, acting on two channels, in opposite directions depending on how much of it there is.
⚠️ That work used heterologously expressed channels and rodent tissue. The perceptual shift in humans is real; attributing it neatly to one channel switching on is not supported, and an earlier version of this page did exactly that.
And the two systems converge centrally. Trigeminal stimuli can activate the piriform cortex — the region usually described as primary olfactory cortex.[14] So they are not simply parallel channels reporting separately.
When a fragrance is described as warm, fresh, sharp or spicy, part of what is being described is not olfaction. A 2025 review in Chemical Senses puts it directly: descriptors such as warm, fresh or spicy reflect chemosensory input from the trigeminal nerve, which adds thermal and tactile dimensions to odour perception.[18]
That is why certain notes feel physical rather than smelled.
This is the part of olfaction with the widest gap between what is sold and what is established.
A pheromone is a chemical signal identified by working backwards from a behaviour: find the secretion that produces the effect, isolate the molecule, then confirm the synthetic version reproduces it. That is the standard used for the pheromones we are confident about in other animals.
Four steroid molecules are marketed as human pheromones: androstenone, androstenol, androstadienone and estratetraenol. None was identified that way.
Tristram Wyatt's 2015 review in Proceedings of the Royal Society B is the paper to read.[19] His position is careful. Humans are mammals, so pheromones are possible and perhaps probable. But there is no robust bioassay-led evidence for any of those four, and in the absence of a sound reason to test them, positive results should be treated as likely false positives.
Androstadienone and estratetraenol entered the scientific literature through a 1991 conference paper sponsored by the EROX Corporation, which was simultaneously patenting them as putative human pheromones. The paper reported testing five molecules supplied by the company, with no stated justification for those five. Wyatt, having traced every subsequent mention back to that symposium or a 2000 follow-up, describes it as if the molecules were plucked from the air.
Roughly 60 studies have since claimed significant results for them. Wyatt treats the whole body as a candidate example of the reproducibility crisis: small samples, positive publication bias, and no negative results appearing.[20]
In many mammals, pheromones are detected by the vomeronasal organ, a structure separate from the main olfactory system.
TRPC2 — the gene essential for vomeronasal signal transduction — is a pseudogene in humans.[21] Because it is expressed only in the vomeronasal organ, its loss of function serves as a molecular clock: analysis across 15 primate species indicates the organ was probably functional in the common ancestor of New World and Old World monkeys, and became vestigial in the common ancestor of Old World monkeys and apes. Independently confirmed the same year.[22]
The vomeronasal receptor repertoire was extensively pseudogenised alongside it. The mouse genome carries roughly 140 potentially functional V1R pheromone receptor genes. The human genome, in the 2003 analysis, has five that retain open reading frames.[22]
⚠️ That figure comes from the genome assembly available at the time; later assemblies suggest fewer may be intact, and the five were not shown to retain physiological function in any case. The direction is not in doubt — the size of the remnant is approximate.
⚠️ Both 2003 papers propose that trichromatic colour vision may have replaced pheromone signalling. That explanation was challenged the following year and is not settled — state the broken gene, not the reason for it.
This does not prove humans have no chemical communication. It proves the mouse-style channel is gone, and that any human signal would have to run through the main olfactory system.
Androstenone is one of the four. It is also the molecule in section 3 — the one where a single receptor variant determines whether you smell sweat, flowers, or nothing.
So the same compound is simultaneously the best-documented case of genetically variable odour perception in humans, and a steroid sold as a sex pheromone on evidence its own reviewers describe as absent.
This is the claim people usually reach for when pheromones are questioned, and it has a famous founding study and a poor replication record.
Wedekind's 1995 "sweaty t-shirt" study reported that women rated men's body odour as more pleasant when their MHC differed — the gene family governing immune recognition — with the effect reversing in women taking oral contraceptives.[23]
The aggregate does not support it. Havlíček, Winternitz and Roberts combined effect sizes across genomic, relationship-satisfaction, odour-preference and mate-choice studies and found no overall significant effect of MHC similarity on human mate selection.[24] A separate meta-analysis across primates found a trend toward MHC-diverse mates but no consistent dissimilarity preference, with effect sizes that are small in both cases.[25] Replications failed or contradicted in 2017, 2018 and 2020, and average statistical power across the field runs 24–57%.
⚠️ Two different claims get mixed together here. Heterozygosity means carrying two different alleles yourself. Dissimilarity means your partner's differing from yours. The evidence is modestly better for the first and null for the second, and casual explanations routinely conflate them.
In mice, MHC-based odour signalling is solid. In humans, that we choose partners by MHC-dissimilar scent has not been established.
Something better, and it is demonstrated rather than inferred.
Olsson and colleagues activated the innate immune system in healthy volunteers with an endotoxin injection, had them wear t-shirts for four hours, and gave the shirts to forty separate raters.[26]
Within hours, the endotoxin group's body odour was rated more intense, more unpleasant and less healthy — and the effect was statistically mediated by the individual's own level of immune activation. The greater the cytokine response, the worse the sweat smelled.
The detail that makes it interesting: chemical assay found no difference in the overall quantity of odorous compounds. What changed was the composition. The specific compounds remain unidentified.
A 2023 replication at a lower dose found the same direction and the same qualitative-not-quantitative pattern.[27]
⚠️ Eight donors in the original study. The authors themselves note the replication's lower dose made for less than ideal conditions.
A meta-analysis of 26 studies covering 1,652 participants found that people can communicate fear, stress and anxiety through body odour, at a small-to-moderate effect size.[28]
What makes this unusual for the field: the authors used p-curve analysis specifically to test whether the literature reflects a real effect or p-hacking. It came out diagnostic of a true effect, with no evidence of publication bias.
⚠️ The authors are explicit that they assessed statistical robustness, not the internal validity of the individual studies.
So: human chemical communication is measurable, and no human pheromone has been identified. Those are different claims, and holding both is more accurate than either "pheromones are real" or "pheromones aren't a thing."
Smell and the hormonal system share an origin, and the evidence is developmental.
GnRH neurons — the cells that initiate puberty — originate in the nasal placode and migrate along olfactory axons into the forebrain and hypothalamus during fetal development.[29] When that migration fails, the result is anosmia and failure of puberty together, from a single developmental fault. That is Kallmann syndrome.
The neurons that control the reproductive endocrine axis begin development in the same embryonic nasal region as the olfactory system. Not the same system — closely intertwined pathways with a shared origin.
⚠️ In families carrying Kallmann mutations the two phenotypes do not always occur together, so the link is not absolute. And a shared developmental origin does not mean that smelling something changes your hormones as an adult — those are different claims, and the second is not established.
The nose is not a camera. What it reports depends on the state of the person using it.
Researchers collected odour-evoked fMRI responses alongside plasma ghrelin, insulin and leptin from 25 people after a standardised meal.[30]
Higher insulin and lower ghrelin — the fed state — were associated with lower odour intensity ratings and reduced responses in the anterior olfactory nucleus, olfactory tubercle, and hypothalamus.
Put plainly: after you eat, things smell less intense, and the hypothalamus responds to them less.
The direction matters. These are hormones changing the smelling, not a smell changing the hormones. Tong and colleagues found that systemic ghrelin infusion in humans significantly increased sniff magnitude — to food odours, non-food odours, and to plain air — while leaving pleasantness ratings unchanged.[31] The authors describe it as a specific effect on odour detection rather than on the hedonic value of odours.
A sensitivity effect, not a liking effect. Which is a useful distinction: being hungry does not make things smell better. It makes them smell more.
Olfactory adaptation: prolonged exposure reduces perceived intensity, and it recovers once exposure ends.
This is one common reason a perfume seems to disappear. The odorant has not necessarily gone anywhere — your response to it has changed. That is a different thing from specific anosmia, where you never registered the material at all.
Ask someone who has just walked in.
Recovery appears to be a matter of time away from the odorant rather than anything you do. In a study that measured how long people could smell an odorant before it vanished and how it returned afterwards, time to complete desensitisation rose with concentration, and longer pauses produced a stronger subsequent sensation independent of which odorant had been used — the authors concluded that recovery "may be a uniform process."[38]
The bowl of coffee beans on a fragrance counter is the best-known answer to that question, and it is a borrowed one. Cleansing between samples is real practice in flavour work, where it is done in the mouth — and that is the problem, because the mouth and the nose are not the same organ and do not recover the same way.
In the mouth it works. Tested across beer, makgeolli, wine, soju, whisky and cognac, ice cream was the most effective cleanser of the six trialled, with water crackers close behind; water, warm water and pectin solution were the least effective.[39] So the underlying idea is not folklore. It is a technique that does what it claims, in the sense it was developed for.
In the nose it has been tested once, and nothing reached significance. Sixty-three people smelled fragrances repeatedly, then sniffed coffee beans, lemon slices or plain air before being asked which of four fragrances was new to them. They were right 57% of the time after plain air, 62% after coffee — and 86% after lemon.[40]
⚠️ None of those differences was statistically significant, and the lemon figure is the reason to be careful in both directions. Coffee did not beat plain air, which is the finding usually quoted. But lemon scored highest of the three, and with roughly twenty people per condition the study could not have detected a moderate effect if one existed. It does not show that lemon works. It equally does not show that lemon doesn't. The authors called it exploratory, and quoting the coffee number without the lemon one turns a null result into a conclusion it cannot carry.
What the adaptation research above suggests is why a sniff of anything is an unpromising intervention in the first place. If recovery runs on its own clock and behaves the same way whichever odorant caused the fatigue,[38] there is no odour you could introduce that would hurry it. A cleanser in the mouth has something to do — dilute, coat, physically clear residue. In the nose the thing that needs to change is in your receptors, not in the air.
The practical version: if a fragrance has stopped registering, leave it. Step outside, smell something else for a few minutes, come back. That is not a trick, and nothing sold at the counter beats it.
There are two routes to the olfactory receptors. Orthonasal — through the nostrils, when you sniff something. Retronasal — from the mouth, as volatile compounds travel up the back of the nasal passage while you eat.
Rozin demonstrated in 1982 that these behave as two different senses.[32] He trained participants to recognise odours orthonasally, then presented the same odours retronasally. Recognition fell drastically. His example is Limburger cheese: repulsive at the nose, pleasant in the mouth. Same molecules, different route, different experience.
Later imaging work found the two routes produce different neural responses.[33]
Which is why food goes bland when you have a cold. Your taste buds are working normally. What has stopped is aroma reaching the olfactory receptors from your mouth.
A large part of what people experience as flavour comes from aroma arriving by the retronasal route — though flavour is a multisensory construct, and taste, texture, temperature and trigeminal sensation are all part of it too.
Everyone has had this happen, and the research on it is real. It is also narrower than the version you will read.
Odour-evoked autobiographical memories tend to be:
A review by Hackländer, Janssen and Bermeitinger covering roughly 35 years of work found support for all three.[34]
There is small-sample neuroimaging evidence too. Herz and colleagues found that when a personally meaningful memory was cued by an odour rather than the same cue in another form, amygdala activation during recall correlated with how emotionally potent the memory felt. The study had five participants.[35]
And in a clinical intracranial-stimulation study, researchers triggered odour-evoked autobiographical memories by stimulating the amygdala directly.[36]
"Smell memories are more vivid." The same review treats this one as considerably more controversial than the others, and it remains an open question.[34]
"Smell is the strongest memory trigger." Explicitly contested, and it depends entirely on what "strongest" means.
Those two are the claims that appear in nearly every article about scent and memory.
The two best-supported findings — older and more emotional — are not the two that get repeated. The two that get repeated are the two the literature is least settled on.
Rasch, Büchel, Gais and Born presented an odour as context while participants learned a task, then re-presented it while they slept.[37]
Re-exposure during slow-wave sleep improved retention of hippocampus-dependent declarative memories, and fMRI showed hippocampal activation in response.
The controls are what make it a finding:
| Condition | Result |
|---|---|
| Odour during slow-wave sleep | Improved memory |
| Same odour, procedural memory | No effect |
| Same odour during REM sleep | No effect |
| Same odour while awake | No effect |
| Odour omitted during learning | No effect |
The effect requires the odour to have been present at encoding, and requires that specific sleep stage.
⚠️ This is targeted memory reactivation under laboratory conditions. It is not evidence that scenting a pillow improves learning — the effect depends on the same odour being present when the material was learned, and on delivery during a particular stage of sleep.
A molecule does not contain your grandmother's kitchen. Your brain built that link, from your history, and someone else's brain built a different one from theirs.
Which is why the emotional weight of a scent is real, personal, and not a property of the material — a point the how to choose a scent page develops.
Four things follow, and they are all practical.
Contradictory reviews can both be accurate. If one person finds a fragrance overwhelming and another barely notices it, that may be a difference in receptors rather than in judgement. OR7D4 is the proof that this happens at the level of the hardware.
Some of what you are describing is not smell. Cooling, sharpness and tingling are trigeminal. That is why those qualities feel physical, and why they behave differently from the rest of a composition.
Your own state is part of the measurement. Hungry or fed changes perceived intensity. Sitting in a scent for twenty minutes changes whether you notice it. Testing a fragrance says something about the fragrance and about the moment you tested it in.
And your associations are evidence about you. They are real and worth trusting for your own choices. They are not a property of the material, and they travel badly to other people — which matters when you are buying a gift.
⚠️ What this page does not cover. Smell can be lost or reduced, temporarily or permanently. Recovery, olfactory training, and the use of smell tests as markers for other conditions are medical questions, and this page does not address them.