How smells shape human hormones, social behavior, and the future of digital olfaction
Andrew Huberman interviews Dr Noam Sobel, a leading researcher in human chemosensation, covering the science of smell, chemical signaling between people, and efforts to digitize odor.
Summary
Andrew Huberman speaks with Dr Noam Sobel about the remarkable and often overlooked role that smell plays in human behavior, hormones, and social bonding. Sobel explains that humans are far more olfactory than commonly assumed, presenting research showing that people unconsciously sniff their hands after handshakes to chemically sample strangers. He describes studies demonstrating that body odor similarity predicts friendship formation, that the molecule hexadecanal emitted by infant heads reduces aggression in men while increasing it in women, and that odorless emotional tears cause a measurable 14% drop in free testosterone in men within 20–30 minutes. Sobel also describes his lab's work toward digitizing smell, including the first successful transmission of an odor — violets — over an internet connection, and outlines a long-term vision for electronic-nose medical diagnostics.
Key Takeaways
FULL TRANSCRIPT
The architecture of the human olfactory system
Andrew Huberman: Welcome to Huberman Lab Essentials, where we revisit past episodes for the most potent and actionable science-based tools for mental health, physical health, and performance. I'm Andrew Huberman, and I'm a professor of neurobiology and ophthalmology at Stanford School of Medicine. And now for my discussion with Dr Noam Sobel. Dr. Sobel, welcome.
Dr Noam Sobel: Thank you.
Andrew Huberman: I must say I am extremely excited for this conversation. I've been a huge fan of your work, and today I'd love for you to share with us all about the amazing landscape of chemosensation, in particular olfaction, our sense of smell, and some related perceptual abilities — or subconscious abilities — including pheromones, etc. To get everybody on the same page, I'd like to start off by asking: what are the major components of our ability to smell? Obviously, or I like to think, it involves the nose at some level.
Dr Noam Sobel: It does.
Andrew Huberman: If you could give us the top contour — or even deep contour if you like — of the parts list and the various roles they play.
Dr Noam Sobel: Primary olfaction is referred to as orthonasal olfaction — that is, through our nose. We sniff, and sniffing is a big thing. I have a sense we might talk about that a lot today in all sorts of contexts. So we sniff in through our nose, and to answer your general question about the organization of the system: molecules, airborne molecules, travel up our nose a distance in the human of about 6 or 7 centimeters, to about here, where they interact with — I will use the word "sheet" of receptors, but "sheet" is a bit misleading. It's not a sheet; it's very convoluted. We have about 7 million such receptors lining a structure known as the olfactory epithelium. This is the sensory surface of the olfactory system. Again, about 6 or 7 million receptors in the human, and in the human, probably about 350 different kinds. That's amazing — it means a meaningful percentage of your genome is devoted just to this, just to the kinds of olfactory receptor subtypes you have in your nose.
So odorants interact with these receptors in our epithelium, where they undergo what is referred to as transduction. That is, the odorants dock at a receptor and turn into a neural signal. This neural signal — in fact, action potentials, not gradient potentials of any kind — propagates via the olfactory nerve through the thinnest part of our skull, which is perforated. It has a lot of holes. The nerve goes through those holes and synapses at the first target in the brain, which is the olfactory bulb. The pattern of that synapsing follows what's referred to as the most extreme case of convergence in the mammalian nervous system. More specifically, what happens is that all the receptors of one subtype converge to one location in the bulb. This location is referred to as a glomerulus, or in plural, glomeruli. That may be a slight oversimplification — it's in fact two glomeruli.
The textbook view of how things work is that every such receptor subtype is responsive to a small subset of different molecular shapes, sometimes referred to as odoritopes — the molecular aspects of the odorant. Each odoritope will activate a different subset of receptors. Then, because of this convergence, you end up having on the bulb a kind of map reflecting receptor identity. So let's say this coffee activates receptors of type 1, 3, and 7. The glomeruli of receptors 1, 3, and 7 will light up, quote unquote, when I smell the coffee. And if you could take a snapshot of that, theoretically, you would have the map of coffee. This is sort of the textbook view of how the system works.
Then information goes from the bulb to several targets in the brain. What is referred to as primary olfactory cortex is piriform cortex and entorhinal cortex. This is on the ventral surface of the brain, the lower portion of our temporal lobe, and information goes there directly — but it also goes directly to the amygdala. It probably goes directly to the hypothalamus. It may go directly to the cerebellum. It goes all over the brain. Information projects widely from there, and as far as people understand, the map that may exist on the bulb doesn't exist in the rest of the brain, and the understanding of how coding occurs in the rest of the brain is murky.
Sometimes when I give public lectures to non-olfaction audiences, I reiterate this: humans have an utterly remarkable sense of smell. We look at human behavior a lot. We try to look at it through our nose and think about what people are doing. If I was a dog, what would I think of this? When dogs interact, they visibly sniff each other. It's very obvious — they walk up to each other and sniff each other. And yet humans don't typically walk up to a stranger and carefully sniff them. We're sort of obliged to sniff our babies — that's considered almost something you're supposed to do. And it's not culturally taboo to sniff our loved ones. But we don't sniff strangers. Or do we?
Handshaking as covert chemical sampling
We're finding more and more mechanisms where we do this. One example: we started looking at handshaking. There's really poor understanding of how this behavior evolved. Where did this thing come from? If you look for the Wikipedia version, they'll tell you it's to show that you're not holding a weapon in your hand. But there's really no good evidence for that. We started looking at people handshaking, and we noticed that people will shake hands and then do this — touching their hand to their face.
Andrew Huberman: For those of you listening and not watching, Noam is taking his hand and wiping it on his face.
Dr Noam Sobel: Yes, touching myself. These things that we do all the time.
Andrew Huberman: After a handshake.
Dr Noam Sobel: Well, first of all, we do them all the time just as a baseline — and we'll get to that in a second. But these behaviors that you could easily not notice — we asked whether that's a real thing. This was a study led by Idan Frumin in our lab at the time. We brought participants into our lab, sat them in the experiment room, and told them the experiment would start soon and they should wait. They didn't know what they were coming for. Unbeknownst to them, they were already being videoed, and they would sit there for 3 minutes. During those 3 minutes we could later quantify how much, just by baseline, they touch their nose, their forehead, their chin — how many times their hand reaches their face. And by the way, that baseline is not low.
Then 3 minutes later, an experimenter would walk into the room and deliver a consistent text: "We're still setting up our equipment in the other room, so just wait here and we'll be right back with you." In half of the cases it included a handshake. This was a new experimenter — not the one who had put them in the room — so it was the first time they met. They would say hello, put out their hand, and shake — or not. So you had handshake and no-handshake conditions, and then you could quantify the behavior of the hand going to the nose after the handshake. And there was a remarkable increase in the hand going to the nose after a handshake. It's unreal — the hand goes to the nose.
We did a few controls to verify that this is an olfactory behavior. One is that, unbeknownst to participants, we measured nasal airflow. People not only bring their hand to their nose — they sniff it. And this is perfectly timed. So they're sniffing their hands. This was an olfactory behavior; we're quite confident of that. And there's an interesting thing going on here too, because people didn't only smell the hand that shook. They also smelled the other hand. We think there's something going on here comparing self to other, and we think a lot of self-sampling might reflect that.
Body odor similarity and friendship formation
There's a study we published by Inbal Raviv in our lab. Inbal came with this basic interest in the phenomenon loosely referred to as click friendships — people you meet and you click right away. Everybody has experienced this to some extent. The question is: what was there? Was it because you looked the same? Was it because you had the same sports team you liked? Or is there something deeper? Inbal's theory was that a similarity in body odor may contribute to this — that people who smell the same will click in some way.
To address that, she recruited click friends from all over Israel. She posted all over social media to identify pairs of friends — same-sex, non-romantic dyads whose friendship started as a click. Then she sampled their body odor. We have a protocol for this: they're given odorless shampoo and soap to use for three weeks or so, and then they sleep two nights in a T-shirt, sleeping alone. We then extract the body odor from the T-shirt.
She first asked whether click friends are more similar in their body odor than you would expect by chance. She tested this with a device we call an electronic nose — a very poor effort to mimic what the mammalian nose does. Basically, it's a bunch of sensors that respond to airborne molecules, in this case sensors referred to as MOX sensors, metal oxide-covered sensors. She found that click friends are indeed more similar to each other than you would expect by chance, by random dyads. This was a significant difference. After she found that a device could do this, she had other participants do it too — people smelling the click friends versus non-click friends — and they judged them as being more similar to each other than not.
Now, you might wonder whether this is causal or not, because maybe click friends go to the same restaurant all the time or live in the same neighborhood, and that's why they smell the same. To address causality, she recruited total strangers, first smelled them with the electronic nose, and then engaged them in a social interaction called the mirror game. In the mirror game, one person moves their hands and the other person — right up close, so they can smell each other — has to move their hands with the other person.
One prediction panned out, but another didn't. She predicted that people who smelled more similar to each other would be better at the mirror game — that is, they would follow each other better. That did not pan out. However, the interaction was completely nonverbal; they were not allowed to speak with each other. She did an entire round robin, so everybody played with everybody else — an insane experiment to run. At the end, each person rated each other person as to how much they thought they would want to be their friend, and on a bunch of other ratings: how nice they think they are, how affectionate, and so on. All of this was predicted by the electronic nose. People who smelled more similar to each other thought the other person was more likely to be their friend, more likely to be a nice person, etc. We could actually predict friendship using the electronic nose. So this is not a result of friendship — it plays into the causal elements of building friendship.
People are constantly smelling themselves, constantly smelling others.
Chemical signaling, fear, and the smell of emotion
Andrew Huberman: I want to touch on some other so-called pheromone effects. Whether or not it's a classic pheromone effect, or whether it's olfaction or something else, this is chemosensory signaling between individuals.
Dr Noam Sobel: It all becomes semantics, right?
Andrew Huberman: Like nerdy inside ball.
Dr Noam Sobel: Yeah, it's all semantics. In our publications, we don't use the term "pheromone," because it would not help us and would probably only hurt us. We talk about chemo signals. Humans definitely emit chemo signals from their body, and these chemo signals influence other humans and influence their behavior.
The flavor of the month for the past ten years in this field is what's referred to as the smell of fear. Humans emit a particular body odor when they're in a state of fear, and this body odor influences other humans, in effect increasing their autonomic arousal, their sympathetic state. So in effect, you could say that fear is contagious — the smell of fear is contagious. Culturally, we've known for ages that dogs can smell fear in humans, but that was only really shown about a year and a half ago in a study. It was always said, but it wasn't really shown effectively until then — that dogs indeed can smell human fear. And humans can smell human fear. Several labs, starting from Denise Chen and Haviland Jones, and then in our lab and others: if you collect body odor from people in a state of fear and collect body odor from the same people when they're not in a state of fear, other people can determine which is the state of fear, and this influences their behavior.
This just highlights again that we're incredibly olfactory animals.
Andrew Huberman: We had a guest on this podcast who I'm sure you're familiar with — Charles Zucker, a professor at Columbia who has done incredible work in vision, olfaction, and thirst sensing. I talked a little bit about this: that there's something in the breath of romantic partners that's hopefully appetitive, not aversive. As well as in children — he was talking about the smell of his grandchild's nape of the neck and how he misses that smell, because when he thinks about missing his grandchild, it's that smell that's associated with that feeling.
Hexadecanal: the chemical signal from infant heads
Dr Noam Sobel: Hexadecanal.
Andrew Huberman: Hexadecanal?
Dr Noam Sobel: Yes. This is a study run by Eva Mishor, who was a graduate student in our lab. Eva was interested in aggression. When she started off, we said, "Okay, let's do chemo signaling of aggression." She was actually going to MMA clubs and collecting body odors. We had all sorts of ideas going, and she worked on that for quite a bit. It never went anywhere really.
At the same time, Heinz Breer was studying in his lab a molecule, hexadecanal, that was a chemo signal in mice promoting social buffering. When lots of mice are together, they feel good about being in a group — that's social buffering — and it's promoted by hexadecanal, which they emit in their feces. Breer and his colleague Strotmann discovered the receptor for this, and then discovered that the receptor is very highly conserved throughout mammalian evolution. They therefore hypothesized that maybe this is a universal mammalian signal — which is unusual, because in chemo signaling, you typically think of things as being very species-specific. He approached us and said, "Look, you've got to study this stuff in humans," because he knows us as the human people. We go to these olfaction conferences where lots of people study mice and zebrafish and whatnot, and we're the human group. Eventually, he just FedExed us hexadecanal.
Eva was not going anywhere with her aggression studies using sweat from human participants, and yet she had built the entire paradigm to study human aggression. There are standard paradigms — this one is known as the TAP, the Taylor Aggression Paradigm. What's nice about it is that it allows you to quantify aggression. It's very well validated, studied all over, a very standard protocol.
So we brought in participants and had them play the TAP either under exposure to hexadecanal or a control. I'll first tell you the result with men, which is that hexadecanal consistently reduced aggression. People were less aggressive under hexadecanal. The effect size was quite meaningful compared to the effects seen in aggression research — really, really strong effects, unusually strong. So hexadecanal lowered aggression in men, and we were like, "Cool, this is sort of what we were hoping to see, consistent with the hypothesis and consistent with what it seems to do in mice." But then we looked at the data from women, and hexadecanal increased aggression, equally significantly.
Andrew Huberman: Is this thought to be something related to maternal protectiveness?
Dr Noam Sobel: We're getting there. So she went and did the entire experiment again, this time inside the fMRI magnet so that we could also track brain activity while this was happening. First of all, it replicated. Once again, hexadecanal made men less aggressive and women more aggressive. And yet, I was telling everybody, "But this makes no sense to me." Then Eva's insight — which of course afterwards is like, "Duh" — was: "No, there's a place where this makes perfect sense, and that is if you're a mammalian offspring." Because paternal aggression is often directed at offspring. There's infanticide all over, and sadly, there's male aggression towards human children as well. And maternal aggression is often protective. So if you're an offspring, if you have a molecule that will make your mother more aggressive and your father less aggressive, both of those are good for you. You're winning.
The bottom line of all this is that hexadecanal is the most abundant semi-volatile compound in baby heads. Tons of it is coming out of baby heads. Babies are conducting chemical warfare — reducing aggression in their fathers or males around them, and increasing aggression in their mothers or females around them. Both of those things are good for them.
Emotional tears, testosterone, and chemical protection
Andrew Huberman: What is the relationship between odors and hormones, and in particular crying? Tears are a bodily liquid — emotional tears — that we emit in emotional situations where nonverbal communication is critical. The most influential text in emotion research, I think to this day, is Darwin's book The Expression of the Emotions in Man and Animals. And a chapter of it is devoted to tears. Darwin systematically went through the expressions of emotions and for each one found their functional antecedent and explained everything very nicely. And then he got stuck with tears, because tears are an obviously emotional expression and he could not find a functional antecedent. He ended up saying, "Yeah, this is an epiphenomenon."
Dr Noam Sobel: So we thought, "Well, maybe the function is a chemical signal." With that in mind, we harvested emotional tears. We collected tears and exposed participants to them. First of all, the tears are completely odorless. You cannot detect them at all. And yet when you sniff them, you have a pronounced reduction in testosterone within about 20 minutes, half an hour.
Andrew Huberman: This is men and women smelling women's tears?
Dr Noam Sobel: Men smelling women's tears, but not perceiving any odor.
Andrew Huberman: Nothing. Just sniffing them.
Dr Noam Sobel: And you have about a 14% drop in free testosterone. People who study testosterone — which is not me, but they tell me — say this is a really strong effect. It's hard to even pharmacologically get an effect like that that fast.
Andrew Huberman: Yeah, years ago I spent time studying endocrine effects of this sort, and it's a tremendous, precise effect.
Dr Noam Sobel: One of the concerns we had, because of the effort required to run this study, was that nobody would ever try to replicate it. To our joy, about two years later, an independent group from South Korea — Oh et al., who I don't know at all — replicated the testosterone effect to a T. Same numbers. So it lowers testosterone.
We then also looked, using MRI, at the effect on brain activity and saw a pronounced dampening — a lowering of activity under an arousing state — both in the hypothalamus and in the fusiform gyrus, amongst other things. We don't know why, but it was pronounced.
Usually in our chemo signaling work, we look at what happens in rodents and see if the same thing is happening in humans. This was a rare case where after we did this work, more or less identical effects were discovered in rodents. A paper published in Nature two years later found that mouse pup tears lower aggression in male adult mice towards them, in a smell-dependent way. They also actually found the actual component in tears — the tear pheromone that lowers aggression.
So you can think of tears as a kind of chemical blanket — you're covering yourself up to protect against aggression.
Digitizing smell: olfactory metamers and the first odor transmitted over the internet
Andrew Huberman: Smell being such a rich source of behavioral, hormonal, and other sorts of deep information — what is your experience with generating smells in silico, in computers? What we're really alluding to here is the idea that you'll put into a search engine "blueberry pancakes recipe," and not only will you get photos of those blueberry pancakes and a recipe, but you will get the hopefully validated odor of those pancakes coming at you in real time through the computer.
Dr Noam Sobel: Google had an April Fools spoof. They put out this video of Google Smell, with all these classic sales images of holding up your phone to a rose and it generating rose — all these things. So Google is now trying to do that, to digitize smell. There are other companies trying to do this as well. And recently in our lab, we've had — and I hope this doesn't come across as overly arrogant — a sort of mini breakthrough on this front.
Andrew Huberman: To call something a mini breakthrough is far from arrogant.
Dr Noam Sobel: This is a paper led by Aaron Ravia from our lab, with Kobi Snitz also a major contributor. A paper published in Nature. In this paper we develop an algorithmic framework where we can predict the perceptual similarity of any two molecular mixtures with very, very high accuracy. If you give me two molecular mixtures, I can predict how similar you will smell them to be. Not only could we predict that, but we could design it. We can generate mixtures with known similarities.
The result that was highlighted — and you'll appreciate this coming from vision — is that using our algorithmic solution, we generated olfactory metamers. We created mixtures completely non-overlapping in their molecular structure that smell exactly the same. Now, if you went to a classic perfumer and told them you can generate two mixtures with zero molecules in common that smell exactly the same, they would tell you no. And yet we did, and anybody can recreate them. This is actually simple. In the paper we do things like generate a metamer for Chanel No. 5 with no component from Chanel No. 5 in it.
In a way, we've generated the infrastructure for digitizing smell, because what our algorithmic framework predicts is similarity. But in a way, that's enough. We have a map of 4,000 molecules. For each one, we know their perceived smell. Now, you can make up any mixture you want for me. I can project it into that map and measure its pairwise distance from all the points in the map. If it falls on lemon, then what you generated smells like lemon. If it falls on tomato, then what you generated smells like tomato. We can now predict the odor of any molecular mixture — we can say how it's going to smell.
What we can then do is find a set of components, which we call odor primaries, that can be used to mix any odor that you can perceive. That's what we're working on now. This is in collaboration with the lab of Jonathan Williams at Max Planck in Munich. Jonathan Williams is an atmospheric chemist, but he's really good at using GCMS — tools that measure molecules. Jonathan Williams measured odorants in Germany, transmitted the information to us over IP, we fed that into our algorithmic framework, and recreated it from a device that mixes primaries.
We tried to do four different odorants in our proof-of-concept test. One of them was rose, and we failed at recreating rose. We did recreate something that had a percept, but most people perceived it as bubblegum. The second one we tried was anise, and we failed at recreating anise — most people said it was cherry, which is not very far, but it failed. The third was gasoline, and we were slightly but significantly better than chance at recreating gasoline. And the fourth was violets. Fifteen of sixteen people said violets. So the first odor ever transmitted over IP is violets.
Don't hold your breath for this to be on your table tomorrow. All we have in hand is this very initial proof-of-concept, but we're on the path. We're talking now of trying to achieve the olfactory equivalent of circa 1956 black-and-white TV. I'm not dreaming of being able to transmit to you the difference between a Cabernet or a Merlot. But if I can generate something that's vaguely wine, that will be an amazing success from my perspective.
Jump ahead in your imagination to 4K odor transmission, and then medical diagnostics is where you want to be talking about. You can almost say that every disease will have an odor. Every disease is a specific metabolic process. Metabolic processes have metabolites, metabolites have a smell. When olfaction is digitized and at high resolution — which in our hands it's not going to be; we're talking, in my retirement maybe I'll read about this one day — but when olfactory digitization is brought to the equivalent of 4K vision and audition that you have now, it will be in medical diagnosis. You will have an electronic nose in your bathroom. Each one of us will have one in the toilet, and it will be doing diagnostics all the time. That's where it's going to go. But again, not anywhere in the very close future.
Closing reflections
Andrew Huberman: I want to say thank you for your time today. This was a tremendously interesting conversation. We touched on so many things — hormones, smells, the architecture of the olfactory system. I know that people listening to this are realizing — but I'm going to say it anyway — what an incredible gift you've given us. As an expert in this field, giving us this tour of the work that you and others, who you credit so generously, have done, and the spirit that you bring to it. Whatever drives that spirit — as the great late Ben Barres used to say — keep going, because we are all benefiting tremendously. And I also just want to say that for people listening to this, the spirit of science is one of — as you mentioned, there's complex politics and all these things — but it's absolutely clear that you delight in the work you do. I delight in it. I'm grateful for it. I'm grateful for your time today. On behalf of me and the many, many people listening to this, I just want to extend a huge debt of gratitude. Thank you so much.
Dr Noam Sobel: Thank you so much for those very warm words.