How Smell Secretly Shapes Hormones and Behavior

Shake a stranger's hand, and watch what happens next. Within moments, many people lift that same hand toward their face and quietly sniff it, usually without any awareness that they have done so. Olfaction researcher Noam Sobel calls this an olfactory behavior, and he has the measurements to prove it. His lab filmed volunteers sitting alone in a waiting room, then sent in an experimenter who either shook their hand or didn't before leaving again. The volunteers who got the handshake touched their nose and sniffed their hand far more often than the ones who didn't (05:46, 07:51). Some even sniffed the hand that never touched anyone, as if comparing their own scent to the stranger's.

That small, invisible gesture is the thread running through Sobel's conversation with Stanford neurobiologist Andrew Huberman: humans behave like a nose-first species while insisting we are not. The hardware supports that claim. Odor molecules travel roughly six or seven centimeters up the nose before reaching the olfactory epithelium, a folded tissue packed with about seven million receptors across some 350 different subtypes (01:14, 01:40). Each receptor subtype converges onto a single relay point in the brain called a glomerulus, so a cup of coffee produces a specific pattern on the olfactory bulb, in effect a map of that smell (02:34, 03:25). From there, signals travel directly to the amygdala and the hypothalamus, brain regions tied to emotion and hormones, skipping the layover in the thalamus that vision and hearing require (03:50, 04:09). Smell has a backstage pass most senses don't get.

What a Nose Can Predict

Sobel's lab has used that backstage pass to explain something people usually credit to conversation and shared interests: the instant click that turns strangers into friends. His graduate student Inbal Ravreby recruited same-sex friend pairs across Israel whose friendship had started as a click, had them use odorless shampoo for three weeks, then sleep two nights in the same T-shirt to collect their body odor (10:19, 11:12). An electronic nose, a device whose metal-oxide sensors read airborne molecules the way a biological nose does, found that click friends smelled more alike than randomly paired strangers (11:42). Human sniffers agreed.

The sharper test came next. Could scent predict friendship before it existed? Ravreby brought in strangers, measured their odor with the electronic nose, then paired them in a silent mirror game where they copied each other's hand movements at close range (12:35). Odor similarity did not predict how well they mirrored each other. But after a full round of pairings, each person rated how much they liked every other participant, and those ratings tracked the electronic nose's similarity scores closely.

"People who smell more similar to each other think that the other person is more likely to be their friend." (Noam Sobel, 13:34)

That is a strong claim: a machine reading of body chemistry anticipating a social judgment before anyone speaks. It doesn't make smell destiny. It means scent contributes to a process people assume runs entirely through words.

A related channel runs through fear. Several labs, including Sobel's, have shown that body odor collected from people in a state of fear changes the physiology of whoever smells it afterward, raising their sympathetic arousal, their fight-or-flight state (14:38, 15:21). Dogs can smell human fear too, something long assumed but only confirmed experimentally in the last couple of years (15:00).

The sharpest single example of chemical signaling in the conversation involves one molecule, hexadecanal. German researcher Heinz Breer had found it calmed groups of mice through what's called social buffering, and traced its receptor across a long stretch of mammalian evolution, suggesting it might be a signal shared across species (16:32). Sobel's then student Eva Michor tested that idea in humans using the Tyler Aggression Paradigm, a standard lab measure of aggressive behavior (17:50). Under hexadecanal, men became reliably less aggressive, with an effect size unusually large for aggression research (18:44). Women tested the same way became more aggressive, by a similarly strong margin (19:08). The result repeated inside an MRI scanner. The explanation arrived once Michor and Sobel asked who actually emits the molecule: hexadecanal is the single most abundant semi-volatile compound released from the heads of human infants (20:13).

"Babies are conducting chemical warfare, right? They're reducing aggression in their fathers or males around them and increasing aggression in their mothers or females around them, and both of those things are good for them." (Noam Sobel, 20:13)

Paternal aggression toward infants is a documented risk across mammals; maternal aggression, by contrast, tends to protect offspring. A molecule that turns one down and the other up, radiating from a baby's head while both parents hold it close, is doing exactly the job natural selection would favor.

Tears carry a stranger version of the same logic, because they are the one emotional secretion that carries no detectable smell at all (22:57). Sobel's lab collected women's emotional tears and had men sniff them. The men perceived nothing, yet their free testosterone fell by about 14 percent within 20 to 30 minutes (22:57, 23:24). Brain scans taken during exposure showed reduced activity in the hypothalamus and in the fusiform gyrus, a region involved in face recognition (24:15). A South Korean team led by a researcher named Oh independently replicated the testosterone finding two years later with nearly identical numbers (23:49), and a later Nature paper found that tears from mouse pups lower aggression in adult male mice through smell as well (25:04). Charles Darwin, writing in his 1872 book on the expression of emotion, found a plausible evolutionary function for nearly every emotional display he studied, except tears, which he filed away as an unexplained byproduct (22:11). A chemical signal that disarms aggression nearby is a reasonable candidate for the function he was missing.

Teaching a Computer to Smell

If smell carries this much information, the obvious next question is whether it can be captured, measured, and sent somewhere else, the way a photograph captures light or a recording captures sound. Sobel's lab has made a start. A paper led by lab members Aharon Ravia and Kobi Snitz built an algorithm that predicts how similar any two molecular mixtures will smell to a person, with high accuracy (27:10). The team then used that algorithm to design olfactory metamers: mixtures that share no molecules in common yet smell identical, including a version of the perfume Chanel No. 5 built from entirely different ingredients than the original (27:41).

"We measured mixtures completely non-overlapping in their molecular structure, but they smell exactly the same." (Noam Sobel, 27:41)

That similarity framework doubles as the start of a map. The lab has charted 4,000 molecules by their perceived smell, so any new mixture can be placed on that map and matched to its nearest neighbor, lemon, tomato, or anything else (30:00). The next step, in collaboration with atmospheric chemist Jonathan Williams at Germany's Max Planck Institute, is finding a small set of base molecules, odor primaries, that can be combined to recreate any smell, the olfactory equivalent of red, green, and blue light (30:27). In an early test, Williams measured real odorants in Germany and sent the data over the internet to Sobel's lab, which rebuilt them from primaries on the other end. Attempts at rose and anise failed, read by participants as bubblegum and cherry. Gasoline came through weakly but correctly. Violets succeeded outright: 15 of 16 participants named it correctly, making violets the first odor transmitted over the internet (31:22, 31:51).

Sobel is careful not to oversell where this stands. He compares it to the earliest days of black-and-white television, not anything close to high resolution (32:20). But he is willing to say where high-resolution smell would eventually matter most: medicine. Every disease produces its own metabolic signature, and metabolites have odors (32:43). An electronic nose sensitive enough to read those signatures, sitting quietly in a bathroom, could in principle flag disease long before symptoms do (33:10). That is decades away by his own admission. What the handshake study, the friendship study, the hexadecanal work, and the tears all point to in the meantime is simpler and already true: people are reading each other's chemistry constantly, acting on it, and almost never noticing that they do.