The bearcat, an animal from Southeast Asia, marks its territory with popcorn-smelling urine. Researchers, i.e. professional piss-sniffers, used gas chromatography-mass spectrometry to identify the compound 2-acetyl-1-pyrroline in the bearcat’s urine. This is the same compound that gives popcorn its smell. It forms when heat drives a reaction between the sugars and amino acids in the kernel. A similar thing happens when bread is toasted and when rice is cooked.
The researchers say that with the bearcat, the compound 2-AP is probably created when the animal’s urine combines with microorganisms living on its skin and fur. These microorganisms break down the urine in the same way human-armpit microorganisms turn our sweat into “body odor.”
Although popcorn is an unusual example of body odor, for sure, it is considered one of the ten primary categories of smell. It should be noted that this very recent method for classifying odors is pretty loose, meaning the categories are not rigidly defined, and by viewing this chart, one can see the slight alterations that are almost as acceptable. Researchers used statistical analysis to condense the categories into the following list: Fragrant, Woody/resinous, Fruity (non-citrus), Chemical, Minty/peppermint, Sweet, Popcorn, Lemon, Pungent, Decayed.
Notice here the persistent difficulty in organizing smells – one category requires its own disambiguation (fruity non-citrus), and one isn’t even a smell (sweet).
The organization of smells is an arduous task, and although a handful of people have tried, none have been successful. Using statistical semantic analysis is a relatively new approach. They began with a standard catalogue of odors, taken from the Andrew Dravniek's 1985 Atlas of Odor Character Profiles (144 odors, I believe). Upon this, they do a form of statistical analysis called non-negative matrix factorization (NMF). NMF is a dimensionality-reduction technique, which makes it handy for categorizing the perceptual space of smell. It has to do with confusing things like normalization and consensual matrices, but all we need to know here is that a huge network of smell descriptors are matched against each other to measure their similarity, both to each other, and to a baseline. Each descriptor is then given a kind of similarity score. The descriptors that are the most similar to others are then called the primary categories.
A hypothetical corpus of all possible smells is a multidimensional thing that has never been (and perhaps can never be) reduced to a small set of categories like colors or musical notes. Statistical methods such as NMF reveal that smells are not evenly spread throughout odor-space, meaning that they are not equally different from each other. Instead, they form clusters of similarity, albeit a very loose clustering. As the scientists note in their paper, “Because NMF is an iterative optimization algorithm, it may not converge to the same solution each time it is run (with random initial conditions).” -source
This method, which provides not answers but approximations, turns out to be quite appropriate. The odor lexicon is an ephemeral thing, like smell itself. It is a precognitive perception which bypasses the language centers of our brain, yielding an unpredictable set of descriptors that will change with the verbalizing person. I just wonder – in cultures where they don’t have popcorn, what would they call bearcat piss?
phys.org, April 2016
Castro JB, Ramanathan A, & Chennubhotla CS (2013). Categorical dimensions of human odor descriptor space revealed by non-negative matrix factorization. PloS one, 8 (9) PMID: 24058466.
see this chart listing of other potential primary categories