The Shape of Smell

A perfumer sits in front of a wall of bottles — a few thousand raw materials, each a world. From those, a formula draws maybe twenty or forty, in proportions that matter to a fraction of a percent. Do the arithmetic and the number of possible perfumes is not large; it is monstrous, larger than the count of atoms in the visible universe. And yet a working perfumer navigates it the way you find your kitchen in the dark. How?

An impossible inventory

Start with the size of the room. Choosing a handful of materials from a palette of thousands is a combinatorial act, and combinatorics grows at a speed the mind was not built to feel. The instrument below does the multiplication so you can watch it happen. Pick a palette and a formula length, and read the exponent.

This is the fact people reach for when they call perfumery infinite. It is true, and it is also a trap. The number counts combinations, not experiences. The nose does not visit every point in that space, because most of the space does not smell like anything at all.

Where the space collapses

Here is the strange result. In 2012 a team in Noam Sobel's lab mixed odorants together — ten, twenty, thirty components at a time, drawn at random — and asked people to compare them. As the number of components rose, the mixtures stopped smelling distinct. By around thirty, almost any random blend smelled like almost any other: a single, generic, unplaceable haze. They named it olfactory white. Add enough of everything and you get nothing — the beige of smell.

Drag the slider. One material is vivid. Pile on more, chosen carelessly, and watch distinction drain out.

The lesson is not that complexity is bad; the house's whole catalogue argues otherwise. The lesson is that the perceptually distinct region of smell is vastly smaller than the combinatorial one. Coherence is rare. The interesting fragrances do not sit scattered through the cloud; they lie along a thin, structured ridge inside it — a manifold, in the language of geometry — and getting onto that ridge, and staying on it while you move, is most of what craft is.

Ten directions

If the meaningful part of smell is small, it should also be low-dimensional — describable by a few independent axes rather than thousands. It seems to be. When researchers take large tables of how people describe odours and factor them mathematically, the cloud of descriptors keeps collapsing onto roughly ten directions. And one axis dominates the rest: not floral or woody, but simply pleasant against foul. More of how we perceive a smell sorts along pleasantness than along anything else — and, remarkably, that axis correlates with the physical structure of the molecule. Beauty, here, is partly written into chemistry.

A reading of those axes, in the house's own materials:

The cartographer

Put the three findings together and a picture forms. The combinatorial space is astronomical. The perceptually distinct part of it is a thin, curved sliver — step off it and you fall into olfactory white. And that sliver is organised by only a handful of axes, with pleasantness running through it like a meridian. The perfumer's gift is not the ability to imagine everything. It is a learned sense of where the ridge runs — which moves keep you on coherent ground, and which step you off the edge into haze.

This is also, quietly, what a house system is for. A matrix of styles and intentions is a set of coordinates laid over the manifold — a way of saying this house works these regions, in these directions, with these landmarks. Taxonomy is not the enemy of creativity here. On a map this strange, it is the only way to keep your bearings, and to invite someone else to find the same place you did.

Smell feels, in the moment, like the least mathematical of the senses — the one that goes straight past language to memory and the body. But under it there is a shape, low and curved and surprisingly small, and a great deal of the art is the long apprenticeship of learning, by nose, to feel it.