Whether you’re a native New Yorker, a Chachi from the jungle of northern Ecuador or a Thai student from Ubon Ratchathani University, you probably appreciate the smell of vanilla and orange blossom. And no matter if you’re a native of Mexico City, a Mah Meri fisherman from the Malay peninsula or a Seri hunter-gatherer from the California gulf, you’re likely to wrinkle your nose at the smell of rotten onion or feet. A recently published study shows that, despite great variability among individuals, the positive and negative connotations of smells are universal.
Compared to existing research on the other senses, modern-day science still knows little about smell. Scientists can use a light wave’s length to identify its color, or ascertain a sound wave’s tone from its frequency. But even if they know a molecule’s chemical structure, they cannot predict its smell. The perception of smells depends on a complex system, affected by everything from the neurons in the nose’s olfactory mucosa to the aromas that each person has perceived since childhood. And plenty of scientists have long believed that the perception of good or bad smells fundamentally depends on each person’s cultural context.
A study published Monday in Current Biology proposes that our smell perception is universal. Researchers presented 10 molecules to 235 subjects from 10 different cultures. In addition to 20 residents of Mexico City and New York, the sample included people from traditional communities with varying levels of exposure to Western culture, from Malaysia to Ecuador. The scientists hypothesized that, if smell is shaped by culture, the results would vary widely.
The study demonstrates significant variability within each community. But, on average, everyone likes and dislikes the same smells. The 4-hydroxy-3-methoxybenzaldehyde molecule, in other words, vanillin – the primary compound of vanilla beans – was the most popular molecule in the study. The subjects also enjoyed the scent of ethyl butyrate, which gives mango and pineapple their distinctive aromas. (A synthetic version of the molecule appears in most bottled citrus juices.) Among the other popular scents were linalool, which appears in many aromatic plants, and phenethyl alcohol, which occurs in roses, carnations, orange blossom and pine.
Asifa Majid, Oxford University professor and co-author of the paper, recognizes that “several previous ethnographic studies in different communities suggested that cultures can vary widely in terms of the smells they find pleasant and unpleasant, but they had not experimentally tested with a wide range of cultures.” Her study, she adds, “fills this gap by testing it in very different settings, including hunter-gatherers and small-scale farmers, as well as people living in big cities in different parts of the world.” Majid, who carried out the study while at the University of York, concludes that “culture plays a very small role in determining how pleasant a smell is.”
Culture plays a very small role in determining how pleasant a smell isAsifa Majid, co-author of the paper
Co-author and neuroscientist Artin Arshamian, from the Karolinska Institute in Stockholm, Sweden, agrees: “We’re not saying that learning doesn’t affect smell or taste preferences. Of course they do! What we are saying is that culture has a small impact on the perception of pure smell.”
Bad smells are also universal. Participants – Malaysian, Mexican and Ecuadorian alike – ranked as their least-favorite smell isovaleric acid, which is present both in human sweat and in rancid animal and vegetable fats. Most participants also disliked diethyl disulfide, an odor found in overripe onions or rotten potatoes. Also unpopular was octanoic or caprylic acid, naturally present in palm and coconut oils and in the fat of mammal milk.
The study was limited in its use of only 10 molecules. Arshamian explains, though, that the scientists selected the odors based on a previous study, in which New Yorkers rated the pleasantness of 476 different molecules. For this study, the researchers then selected 10 smells that ranged from pleasant to unpleasant. They used the remaining 466 scents to generate a prediction model based on each smell’s molecular structure. “Thus, although we tested few odors, the model is based on a much larger set,” he says. “All 10 odors were not included when building the model, and the model was still able to predict odor classification.”
The original study, published in 2016, used 480 molecules of varying weights and complexities. It established vanillin and isovaleric acid as the most and least pleasant-smelling molecules, respectively. The researchers found that a smell’s pleasantness correlated to greater molecular complexity, except for compounds including hydrogen. Substances that contained oxygen, or that had a high number of atoms, smelled good, while any substances that included sulfur atoms smelled bad.
In 2017, Science magazine published the second part of that study. The researchers organized a crowd-sourced competition, the DREAM Olfaction Prediction Challenge, to create algorithms that could predict aromas’ sensory attributes based on their chemical makeup. Participants created machine-learning models to describe smells based on their molecular composition, using words including “sweet,” “garlic” and “sour.” The models managed to predict both the intensity and the characteristics of each aroma, as well as to match the terms that humans used to describe them.
Mexican scientist Pablo Meyer is the director of the DREAM Challenge at IBM Research and co-author of the Science study. “Most people have the idea that language does not describe odors well, but we have shown that it does. Regardless of whether you like something or not, we are capable of accurately describing olfactory sensations with words,” he says. Regarding the study by Arshamian and Majid, he highlights that “the most pronounced variations are individual, which is why you have the idea that you like or dislike smells, or they remind you of a particular smell that goes beyond culture.” So will computers soon be able to anticipate the qualities of a smell, just as they can predict a wave’s color or sound? “A good chemist would need a little bit of sociology, genetics and knowledge of AI to be able to find a universally pleasurable smell.”