​Can We Really Smell One Trillion Different Scents?

A paper published in Science magazine this past year made international headlines when it answered a simple question—how many odors can we smell?—with a massive figure, finding that our noses can distinguish as many as one trillion individual odors. The report was widely lauded by the media: A testament, amidst all the depressing stuff hitting our newsfeeds on the daily, to our evolutionarily endowed awesomeness.

Or, just maybe, it's a testament to our predilection for buying into splashy, paradigm-shattering claims about just how great we are?

In a pre-print released several days ago on Cornell University's open archive arXiv, Caltech neuroscientist Markus Meister claims the trillion odors figure is off by "astronomical factors." By re-running the authors' own experiments, Meister comes up with far more pedestrian estimate for the number of smells the average person can discriminate: Ten.

Now, as you probably guessed, you can smell more than 10 scents, and that's the point. According to Meister, the paper's optimistic claims about the human nose are based on flawed mathematical logic. Using that logic, Meister showed that either 10 or a trillion scents could be calculated as the correct answer for the human scent limit.

It's a smelly, complicated world out there. Most of the scents we encounter in nature are actually amalgamations of dozens to hundreds of different odiferous compounds— the "scent" of a rose is composed of over 275 unique molecules. And while we have clearly defined boundaries for human vision (390-750 nanometer wavelengths on the electromagnetic spectrum) and hearing (20-20,000 hertz), it's proven much harder to gauge the limits of our sense of smell.

That's exactly what the researchers behind the original Science paper set out to do. They first devised an experiment to determine the resolution of our sense of smell. How far apart do two smells have to be for the human nose to register them as different?

To find out, the authors made a bunch of different mixtures of 10, 20 or 30 odiferous compounds, drawn from a collection of 128 odorants that included familiar scents like orange and spearmint. The mixtures varied in their degree of similarity to one another. The researchers then asked 26 volunteers to perform a sniff test on three vials—two with identical substances, and one with a different mixture—and identify the outlier. Each subject performed this outlier test hundreds of times.

From these 26 subjects, the researchers found that, on the average, mixtures that contained more than half the same odorants tended to smell the same, while mixtures containing less than half the same odorants tended to smell different.

Extrapolating from here to the total number of combinations possible—given just the 128 molecules used in the experiment—the researchers concluded the average person is able to distinguish at least a trillion unique compounds. Moreover, a trillion probably represents a conservative estimate, because of the many more than 128 odiferous molecules that exist in the real world.

This astonishing number means that our ability to discriminate smells kicks the crap out of our abilities to discriminate color (2.3—7.5 million23:1%3C52::AID-COL8%3E3.0.CO;2-2/abstract)) or musical tones (roughly 350,000). But Meister argues that those numbers are off.

"None of this is true," Meister writes in his rebuttal. "The authors were misled by failures in a mathematical method they designed. As a result, their claims have no basis."

I'd encourage anyone interested in Meister's approach and arguments to read his paper in detail. He declined to comment for this piece, stating that, in his view, the authors of the trillion-odors paper did a disservice to science by pushing their ideas into the public sphere before the scientific community had adequately vetted them.

To understand why the model could have been off, Meister offers the example of a hypothetical bacterium that can discriminate only three odors: "Yum," "yuck," and "meh". He follows the paper's procedure, creating combinations of 30 different scents pulled out of 128 possible, with varying degrees of overlap between different mixtures.

He makes one important addition: Each of the 128 possible odors is randomly assigned a value of +1 or -1. Meister then decides that, in any given mixture, if the sum of the + and – 1's is less than –2, the bacteria smells a "yuck". Mixtures that fall between –2 and 2 are registered as "meh," and mixtures with sums greater than 2 smell like "yum."

Following the researchers' procedure, Meister finds that his bacterium can, on average, discriminate odors ("yuck", "meh, or "yum"), when they share ten or fewer of the 30 components. According to the researchers' math, Meister finds that there are over 100 million combinations of 30, pulled from 128, that his bacterium can discriminate. But by design, this bacterium can only register three odors.

Another issue Meister raises is the dimensionality of smell. While color vision varies along three fundamental dimensions—red, green and blue—we don't actually know the number of dimensions that exist for olfaction. Meister shows that if odor is only allowed to vary along a single dimension, the odor discrimination data presented in the Science paper is consistent with just ten odors.

It's important to note that Meister is not saying humans can only discriminate ten odors, or even that the trillion-odor figure is necessarily wrong. He's saying that the researchers' experimental results are equally consistent with ten odors, a trillion odors, or anything in between. The paper, Meister argues, failed to reject alternative models that make very different predictions.

Meister goes on to show how the paper's same logic, if applied to color vision, would lead to the (incorrect) conclusion that humans can discriminate a near-infinite number of colors. This would suggest that the paper fails to pass the positive control test, because the method used cannot consistently reproduce numbers we know to be true from other lines of scientific inquiry.

Personally, I'm not too surprised by the realization that we probably can't perceive a trillion odors, but then, I've never much prided myself on my sense of smell. Give me a green Jolly Rancher, and I'll enjoy its self-proclaimed appley-ness in blissful ignorance.