The Taste of Smell

This week we take a trip down research lane…

Taste and smell are the most important, least understood senses — or are they even distinct senses? About 80% of flavor comes from smell, not taste.

Try this Jelly Bean Test: put several very different jelly beans in a bag, close your eyes and seal your nose. Most people are shocked at the lack of distinct taste. Delicious black licorice tastes the same as the disgusting coconut!

Taste and smell exist in the most primitive life forms. Chemoreception, detecting molecules in the environment, is arguably the oldest sense, predating multicellular life. Bacteria detect and move toward nutrients and away from toxins.

Smell and taste are harder to describe, recall, or introspect on than vision. Language for smell is notoriously impoverished (we mostly describe smells by their source: “it smells like…”.) The line in Shakespeare’s Romeo and Juliet “A rose by any other name would smell as sweet,” doesn’t use a word for a smell. (Sweet in Elizabethan English would have meant “pleasant.”) Bart Simpson had one response: “Not if you called ‘em Stenchblossoms.”

There’s neuroanatomy behind the language gap; smell processing is physically located in brain regions that predate and are poorly connected to language. Vision and touch route through the thalamus to the cortex — they get processed in dedicated cortical areas that are tightly integrated with language regions. You can describe a shape, a color, a texture because the processing pathway runs near or through areas that connect to verbal description. Smell is the exception. It’s the only sense that routes directly to the limbic system without a thalamic relay. These are ancient, pre-verbal structures.

If smell and taste are poorly represented in language, then any system trained on language would have impoverished taste/smell simulation. In the movie the Matrix, machines recreated the simulation from human knowledge; mostly visual, verbal, textual. The “chemosensory” dimension is underrepresented in the training data, as the AI people would say.

Matrix thinks about Qualia

Mouse: “Because you have to wonder: how do the machines know what Tastee Wheat tasted like? Maybe they got it wrong. Maybe what I think Tastee Wheat tasted like actually tasted like oatmeal, or tuna fish.”

The Chemistry of Taste and Smell

Taste and smell are nothing like the other senses; molecular interaction with receptor cells, not energy. Our biological heritage is so much more sophisticated than electrical technology; when it does work, it’s because scientists did an enormous amount of work to find one specific molecule and build a chemical reaction that can produce an electrical signal. You can digitize a waveform (light or sound) on your computer, but digitizing the over 800 volatile compounds in coffee takes a laboratory. This will be a short chapter.

Taste

Taste itself is limited to detecting useful chemical properties. The five tastes are not random but highly useful signals. Imagine being a starving toddler walking through the jungle, putting things in your mouth:

Umami highlights how blurry our language is around taste. In 1908, Kikunae Ikeda identified glutamate as the active compound in Dashi, the essential Japanese savory fish stock made from kelp and dried fish flakes. He named the flavor umami, from “Umai” (savory-delicious) “Mi” (taste). It wasn’t widely accepted in the West until much later when glutamate receptors in the mouth were identified. People understood umami to be real, but perhaps not distinct enough to be a primary taste.

Umami signaling is a more general protein-cue, perhaps telling our primate ancestors something like “keep eating this.” Umami subtly drives many cuisines, one of the reasons people love mushrooms as an ingredient: they have little flavor, even less nutrition, but lots of glutamate. By the time you’re old enough to read this, you’re past the primary audience for umami. Glutamate is extremely abundant in human breast milk; getting babies to eat was always job one.

Smell

In 2001, DigiScents raised $20M to build the “iSmell” device, bringing smells from the internet to your PC. They attempted to define smells as a 64 or 128-parameter digital file, with inkjet-printer-type chemical cartridges. It died within a year — a classic dot-com bust. It’s unclear whether it was poor management, technological challenges, or simple lack of interest in what websites smelled like.

The nose has approximately 400 receptor types, each tuned to a class of volatile (meaning reactive) odorant. The brain reads all of them at once, a chord, not a single note. Coffee, roses, or even gasoline are not one thing, but hundreds. Scientists can make a generic ‘coffee’ smell with around 6 compounds, but people today are looking for a “light roast shade grown Madagascar single farm varietal.”

The human sense of smell is bidirectional; the way we perceive smells changes according to the direction of our breath. In experiments, people often fail to recognize the same odor when sniffed or from inside their mouth. The brain treats both directions as different kinds of sensory events, even if they hit the same receptors.

Dogs have over 220 million olfactory receptors (compared with 5–10 million in humans), a 40% greater area of the brain dedicated to smell, and the ability to smell 1000–10,000 times better than humans, and in stereo.

The gap between “smelling” and “detecting a specific molecule” is enormous — and worth being honest about. An e-nose would need hundreds of sensors plus pattern-recognition trained on known combinations; the sensors would need to refresh and recalibrate in normal use. The field is generations of research away from that.

The sophisticated tools that do exist aren’t smelling either. GC-MS (gas chromatography–mass spectrometry) — the standard instrument used to analyze food and perfume — separates compounds in a column, then identifies each one by its mass. Airport security swabs use ion mobility spectrometry (IMS): molecules are ionized and measured by how fast they drift through a tube in an electric field. Both are remarkable instruments. Neither is smell.

After billions of dollars of research, the nose still has a job only it can do.

The Best Burger I Never Saw

Many years ago, while consulting for International Flavors & Fragrances (IFF), I was taken on a product tour in an unassuming building off the highway in New Jersey. I can still remember the shock of sniffing a test tube to vividly smell a freshly grilled (not fried) burger, with lettuce, tomatoes and ketchup, on a toasted bun.

In case you want to recreate this at home, here is a recipe off the internet.

Grilled beef patty (Meaty, roasted, fatty, savory, slightly sulfurous) 2-ethyl-3,5-dimethylpyrazine, 2,5-dimethylpyrazine, methional, 3-methylsulfanylpropanal, bis(2-methyl-3-furyl)disulfide, 2-furfurylthiol, octanal, nonanal, 1-octen-3-ol

Toasted bun (Toasty, baked, caramel, sweet-bready) Furaneol, maltol, ethyl maltol, 2-acetylfuran, 2,5-dimethyl-4-hydroxy-3(2H)-furanone

Lettuce (Fresh-cut, green, leafy) cis-3-hexenal, trans-2-hexenal, hexanal, cis-3-hexen-1-ol

Tomato slice (Juicy, green-fruity, slightly leafy) hexanal, cis-3-hexenal, trans-2-hexenal, (E,Z)-2,6-nonadienal, hexanol

Ketchup (Sweet tomato, cooked, tangy, caramelized) Furaneol, maltol-like notes, hexanal, cis-3-hexenal, trans-2-hexenal, 2-acetylfuran

(Please do not try this at home.)