Two Ways of Smelling
If you focus on molecules you'll miss the whole scent
Welcome to all the new recent subscribers! If you like what you read, consider sharing it with your friends. I write whenever something piques my interest, not according to a schedule. That said, if there’s a topic you would like me to address, mention it in the comments and I’ll consider it.
I’ve been working my way through Iain McGilchrist’s magisterial tome The Matter With Things: Our Brains, Our Delusions, and the Unmaking of the World. It’s not an easy book to summarize: its scope is vast and McGilchrist’s depth of knowledge in neuropsychology, neurology, and philosophy is extraordinary. His main concern is how the two hemispheres of our brain differ in how they apprehend the world, and how they interact with one another to create our experience and understanding of it.
McGilchrist’s basic thesis is that the left hemisphere focuses on manipulating objects in the world; it works to reduce experience into sharply defined categories and verbal formulations. It is intolerant of ambiguity. Meanwhile, the right hemisphere has a wider focus: it observes patterns and interactions as they emerge. It is not bothered by ambiguity. The left hemisphere sees only individual parts and pieces; the right hemisphere sees the whole picture. Sometimes we need the LH’s reductionist perspective and sometimes we need the RH’s global view of the gestalt.
Some aspects of odor perception are lateralized across the hemispheres, but the differences are small compared to say, speech production and language comprehension (strongly LH). What intrigues me is how the LH and RH cognitive styles are reflected in how we experience smell in the world and how we understand it scientifically.
Odors are all around us, all the time. They are part of the flow of experience. When we walk into an odor plume by accident, we may choose to fan it away or we may seek out the source. Our experience of a given smell changes constantly, not only as its concentration rises or falls, but as our nose and brain adapt to it. (We automatically tune out background odors to be ready for the next incoming odor.) We interpret a given scent differently depending on context: “The baby’s diaper is full” versus “The boiled cabbage is ready to serve.”
The on-going experience of smells in the lived moment is eminently a right hemisphere, gestalt-style mode of perception. It connects us to our surroundings and informs our motivations and decisions in a holistic way. Importantly, smell experience is not analytical. We do not routinely seek to identify individual ingredients. It is usually “The garbage needs to go out,” not “There is a rancid note that could be today’s fish or yesterday’s raw chicken scraps.”
Of course there are times when we do smell analytically: “Should I add more garlic to the melted butter?” “What is that odd smell coming from the engine compartment?”
Contrast our lived experience of smells with our scientific approach to them. Since the rise of organic chemistry in the 19th century we have sought to assign the sources of smell to individual molecules. Naming and categorizing the smell of individual molecules has become the overriding goal of objective, reductionistic, sensory science. To understand an odor scientifically is to pluck it from the flow of ongoing experience, name it, categorize it, and specify its molecular composition. This is the left hemisphere approach par excellence.
This narrow-beam reductionism reaches its peak in the current obsession with AI models that “predict” a molecule’s smell from its chemical structure. This is the ultimate LH take on the world—it removes smell from the smeller and from its place in the world.
Obsessed with identifying “the” smell of a single molecule, AI models ignore the big picture—they don’t even try to deal with complex mixtures, much less their ever-changing flow through time.
A new paper by Xu, Zou, and Firestein reviews how the mammalian olfactory system perceives odor mixtures. There are ~350 human olfactory receptors and each responds to one or more odor molecules to a greater or lesser degree. Conversely, a given odor molecule can activate a few or many receptors, again to varying degrees. Linking patterns of receptor activation to specific odor perceptions has proven to be difficult and time consuming.
Zu et al. point out that the problem is even bigger than that. Odor molecule A might activate receptors x, y, and z, but it also might block receptors q, r, and s, and enhance the response of receptors e, f, and g. A complex odor mixture sets off a farrago of activations, inhibitions, and amplifications across the sensory nerves of the nose. The olfactory code (if there even is one) is not going to be simple.
That is, the olfactory system is not a chemical analyzer that builds a perception out of molecular parts. Perceptions are created in a holistic manner . . .
While Zu et al. politely concede a role for machine learning and AI approaches to smell, they propose that modeling of odor perception should take account of the biological architecture and functioning of the living nose. In other words, we should stop treating odors as chemometric abstractions (LH) and bring the smeller back into the picture (RH).
Let’s not miss the full, ever-evolving bouquet by over-focusing on chemical features and ingredient names.
Lu Xu, Dong-Jing Zou and Stuart Firestein. (2023). Odor mixtures: A chord with silent notes. Frontiers in Ecology and Evolution 11:1135486.
***Stuart Firestein is a longtime friend and colleague of mine.