The Science of Scent

A Of the human senses, smell has been the one philosophers were happiest to dismiss. Kant thought it the least rewarding of the senses and doubted it was worth cultivating; Darwin judged it of extremely slight service to modern humans. That verdict hardened into a scientific commonplace, and its origins are more accidental than most people assume. A recent review traced the belief to nineteenth-century comparative anatomy, and in particular to the observation that the olfactory bulb occupies a smaller fraction of the human brain than of a rat's, a claim about proportion that was quietly converted into a claim about ability. Tested directly, people do better than the reputation suggests. In laboratory trials humans outperform dogs on some odours, follow a scent trail across grass when made to crawl, and detect certain compounds at concentrations that instruments struggle to register.

B What happens in the nose was obscure until 1991, when Linda Buck and Richard Axel identified the enormous family of genes that codes for odour receptors, work for which they later shared a Nobel Prize. Humans have around four hundred working receptor types, far fewer than a mouse's eleven hundred or so, and the number alone would suggest a modest sense. The arrangement, however, is more powerful than the count implies. Each receptor responds to a range of molecules rather than to one, and each molecule activates a set of receptors rather than a single detector, so an odour is registered as a combinatorial pattern across many channels. A relatively small number of receptors can therefore encode an enormous number of distinct smells, in the way that a limited alphabet supports an unlimited vocabulary.

C This is also why the field's central practical problem remains unsolved. A chemist can look at a molecule and calculate its boiling point, its colour and its behaviour in solution, but cannot look at its structure and say with confidence what it will smell like. Molecules that differ by a single atom may smell entirely unrelated, while compounds with nothing structurally in common can be indistinguishable to the nose. The classic demonstration is carvone: its two mirror-image forms, chemically identical in almost every respect, are perceived as caraway and spearmint respectively. Perception also varies between people. A small change in one receptor gene makes coriander taste of soap to a substantial minority; another leaves some people unable to smell a compound in sweat that others find overpowering. There is, in other words, no single objective description of what a substance smells like.

D Smell reaches the brain by an unusually direct route. Signals from the olfactory bulb pass to the amygdala and hippocampus, structures central to emotion and memory, without the full relay through the thalamus that vision and hearing undergo. The consequences are familiar to anyone who has been ambushed by a smell from childhood. Memories triggered by odours are, in controlled studies, older than those triggered by words or pictures, more often drawn from the first decade of life, and reported as markedly more emotional, though not in fact more accurate. Almost none of this is innate. Preferences are learned by association, which is why the smell of wintergreen is pleasant to Americans, who meet it in sweets, and unpleasant to older Britons, who met it in medicine.

E Perfumery turned these facts into a craft long before anyone understood them. A fragrance is built in layers according to how quickly its components evaporate: light top notes that announce the scent and fade within minutes, heart notes that carry it for hours, and heavy base notes and fixatives that hold the whole structure to the skin for a day. For centuries the fixatives came from animals, and the industry was transformed when synthetic molecules began to replace them. The synthesis of coumarin, with its smell of new-mown hay, produced in 1882 the first commercially successful fragrance built around a laboratory ingredient, and the palette has expanded ever since. Formulas are almost never patented, since a patent would publish them; they are kept as trade secrets inside a handful of firms that supply most of the world's perfumes, soaps and detergents.

F The gap between craft and prediction is now closing from an unexpected direction. Machine-learning models trained on thousands of molecules rated by panels of human sniffers have begun to predict odour descriptions from chemical structure with accuracy approaching that of an average person, and the maps they build group molecules by how they smell rather than by how they are built. The medical stakes have risen too. A declining sense of smell is one of the earliest measurable signs of Parkinson's and Alzheimer's disease, often appearing years before diagnosis, and the sudden losses caused by viral infection have made anosmia a public concern rather than a curiosity. The sense that Kant thought expendable is turning out to be both measurable and diagnostic.