How Smell Rewrites Hormones, Health, and Behavior: A Reading of Noam Sobel's Research

The Hardware Architecture of Human Smell: An Underestimated Precision System

In an episode of Huberman Lab Essentials, Andrew Huberman, professor of neurobiology at Stanford School of Medicine, had an in-depth conversation about chemical senses with Dr. Noam Sobel of the Weizmann Institute of Science. Sobel has long studied human olfaction and chemical signals, and his lab is one of the few olfaction research teams that studies humans rather than mice or zebrafish.

Sobel first laid out the anatomical pathway of the olfactory system. Airborne molecules travel about six to seven centimeters up the nasal cavity and interact with roughly seven million receptors on the olfactory epithelium. These receptors fall into about three hundred and fifty subtypes—meaning a considerable proportion of the human genome is dedicated to encoding olfactory receptors. Odor molecules are transduced here into action potentials, which travel via the olfactory nerve—passing through the thinnest part of the skull—to the olfactory bulb.

The olfactory bulb is organized in what Sobel calls "the most extreme case of convergence in the mammalian nervous system": all receptors of the same subtype converge onto the same glomerulus. The textbook view holds that the olfactory bulb thus forms a map reflecting receptor identity—coffee activates receptor types 1, 3, and 7, and the corresponding glomeruli "light up." But Sobel stresses that this map does not exist beyond the olfactory bulb, and how the rest of the brain encodes smell remains murky.

The olfactory epithelium is a mucosal tissue located at the roof of the nasal cavity, a few square centimeters in area, studded with olfactory sensory neurons. These neurons are among the few that directly contact the external environment and continuously regenerate—they turn over roughly every 30 to 60 days, with axons re-projecting to the olfactory bulb. Each olfactory neuron expresses only one olfactory receptor subtype; this "one neuron, one receptor" rule is achieved through a gene-regulated random selection mechanism, so that about 350 receptor genes suffice to cover recognition of a vast number of odor molecules. The receptor genes themselves constitute the largest multigene family in the human genome, accounting for about 2% to 3% of all genes—a substantial portion of which have accumulated pseudogenes over evolution, one of the molecular reasons human olfaction is often mistakenly deemed "degenerate."

After a Handshake, Why Does the Hand Go to the Nose?

Sobel repeatedly emphasizes a frequently overlooked fact: human olfactory ability is "exquisite"—we simply do not use it in the conspicuous way dogs do. His lab designed a clever experiment to capture this covert behavior.

Participants were brought into the lab to wait, unaware they were being filmed, and baseline rates of face-touching and nose-touching were recorded over three minutes—Sobel notes pointedly that this baseline is not low at all. Then a stranger entered the room and, in half the conditions, shook hands with the participant. After the handshake, hand-to-nose behavior increased "unimaginably" significantly. Nasal airflow measurements further confirmed that people were not merely bringing their hands near their noses—they were sniffing them.

More intriguingly, participants sniffed not only the hand that shook but also the other hand. Sobel believes this reflects a self-versus-other comparison mechanism, a function that many self-sampling behaviors may serve.

Sniffing behavior is typically regarded in ethology as an overt indicator of chemical sampling. Human odor sampling falls into two categories: active sniffing—short, forceful inhalations that quickly draw odor molecules to the olfactory epithelium—and sniff-respiration coupling, the unconscious nasal twitches accompanying everyday breathing. Measuring nasal airflow rather than relying on video alone allowed the experiment to distinguish "bringing the hand near the nose" from "actually sniffing." Notably, self-sampling (sniffing one's own hands, armpits, or clothing) occurs at high frequency in humans; hypothesized functions include monitoring one's own chemical state, providing a reference for social comparison, and situational self-soothing.

Odor Similarity Predicts Friendship—Not Merely a Byproduct of It

Inbal Ravrebi of Sobel's lab studied an everyday phenomenon: the so-called "instant connection" of friendship. Her hypothesis was that people with similar body odor are more likely to become friends.

The team recruited same-sex, non-romantic close friend pairs from across Israel, had subjects use fragrance-free personal care products for several weeks and sleep alone in T-shirts for two nights, then extracted body odor. Using an electronic nose (a metal-oxide sensor array), they found that close friend pairs had significantly higher odor similarity than random pairings, and human sniffers' judgments agreed.

To rule out reverse causation—friends frequent the same restaurants, so they smell alike—Ravrebi recruited complete strangers, measured their body odor first, then had them engage in close-range social interaction called the "mirror game." One prediction failed—odor similarity did not make people cooperate better in the game—but another held: after the interaction, people judged those with more similar odors as more likely to become friends and more likable. The electronic nose could predict friendship inclination, indicating that odor similarity contributes to friendship formation rather than merely being a byproduct of it.

Fear Can Be Smelled, and Tears Are Chemical Signals

Sobel avoids the word "pheromone" in his papers, preferring "chemical signal." Humans do emit chemical signals that influence others' behavior, and the "scent of fear" has been a research hotspot for the past decade. Studies from Denise Chen, Haviland-Jones, and Sobel's lab show that people can identify fear from body odor collected from others in a fearful state, and consequently increase their own autonomic arousal—fear is chemically contagious.

Emotional tears are another clue. Darwin devoted an entire chapter to tears in The Expression of the Emotions in Man and Animals but found no functional explanation. After collecting emotional tears, Sobel's team found that tears are completely odorless and undetectable—yet when men sniffed women's tears, free testosterone dropped by about 14% within roughly twenty to thirty minutes. Sobel relays that endocrine colleagues consider this an effect size even drugs struggle to achieve so quickly. About two years later, an independent Korean team fully replicated the result. fMRI showed that sniffing tears markedly reduced activity in the hypothalamus and fusiform gyrus.

Even rarer is the reverse validation: about two years after this human study was published, Nature ran a rodent study finding that mouse pup tears reduce aggression in adult males in an odor-dependent manner, and identified the key component.

The term "pheromone" is sensitive in human research because it was originally strictly defined in insects and rodents as molecules detected by the vomeronasal organ that trigger fixed behavioral or endocrine responses. The human vomeronasal organ exists during embryogenesis but degenerates after birth and lacks functional receptor genes, so many researchers avoid the term. Switching to "chemical signal" shifts the focus from fixed stimulus-response patterns to broader chemical communication that can modulate emotional and physiological states. Emotional tears differ in composition from reflex tears (such as when cutting onions)—the former contain more proteins and signaling molecules, providing a biochemical basis for "tears carrying social information."

Hexadecanal: An Infant's Chemical Tactic

Eva Michor of Sobel's lab originally studied chemical signals of aggression, collecting sweat from mixed martial arts clubs, with little success. The turning point came from German researcher Heinz Breer—he discovered in mice that hexadecanal promotes social buffering and that its receptor is highly conserved across mammalian evolution, so he sent the molecule to Sobel's team.

Tested with the Taylor aggression paradigm, the results were surprising: hexadecanal significantly reduced aggression in men, with an effect size rare in aggression research; but in women, it equally significantly increased aggression. The experiment was repeated with fMRI, and the result replicated again.

Michor's explanation: hexadecanal is one of the most abundant semi-volatile substances on infant heads. For mammalian pups, paternal aggression is often directed at them, while maternal aggression is often protective. A molecule that makes fathers gentler and mothers more protective is a double benefit for the pup. Infants chemically modulate the behavior of both men and women around them.

Digitizing Smell: From Violet to Electronic-Nose Diagnosis

The latter half of the interview turned to work from Sobel's lab published in Nature. Led by Aron Ravia and Kobi Snits, the study established an algorithmic framework that predicts the perceptual similarity of any two molecular mixtures with high accuracy, and used it to generate "olfactory metamers"—mixtures with completely non-overlapping molecular compositions that smell identical. The paper even generated a metamer of Chanel No. 5 containing none of its ingredients.

This is regarded as infrastructure for digitizing smell. The team has perceptual odor maps for four thousand molecules; any mixture can be projected onto this map, measuring its distance from coordinates like "lemon" or "tomato." In collaboration with Jonathan Williams of the Max Planck Institute, the team measured odor molecules in Germany using gas chromatography-mass spectrometry, transmitted them over IP, and reconstructed them with a mixing device. The proof of concept tested four odors: rose failed—most people smelled bubblegum; anise failed—judged as cherry; gasoline was slightly better than chance; and violet was correctly identified by 15 of 16 people—the first odor transmitted over IP was violet.

Sobel remains measured: it is roughly equivalent to 1956 black-and-white television, still far from distinguishing Cabernet Sauvignon from Merlot. But he describes the endpoint: when olfactory digitization reaches today's 4K audiovisual level, medical diagnosis will be the biggest application—every disease is a specific metabolic process, metabolism produces metabolites, and metabolites have odors. By then, an electronic nose in the bathroom could provide continuous health screening. He stresses repeatedly, however, that this is not the near future.

Gas chromatography-mass spectrometry (GC-MS) is the gold standard for odor analysis: gas chromatography separates volatile components in a mixture, and mass spectrometry identifies each component by molecular-weight fingerprint, quantitatively revealing "which odor molecules are in a space and how much of each." But GC-MS only gives chemical composition and cannot directly tell what something smells like. The methodological key of Sobel's team is building a "chemical composition–perceptual similarity" mapping model and validating it with metamers—that is, two chemically completely different mixtures that the model predicts to be perceptually identical; if human sniffers indeed cannot distinguish them, the model has captured the encoding logic of the olfactory system rather than mere molecular structural similarity.

What This Conversation Changed

Sobel's work moves olfaction off its "secondary sense" pedestal: sniffing hands after handshakes, odor-similar friendships, hormonal signals in tears, hexadecanal on infant heads—all point to the same conclusion: humans are animals highly dependent on chemical information, only this system operates below the threshold of consciousness. For AI and sensing technology, digitizing smell offers a direction not yet fully cultivated: it is not simply stacking sensors but algorithmically modeling perceptual similarity. Sobel's metamer research shows that smell can be compressed into coordinates—perhaps a more feasible path than replicating the molecules themselves.