Researchers find that we can distinguish more than one trillion odors. Now, is that much or not?
Out of a set of 128 basic odors (I only depicted 100 boxes here), 10 (as shown here), 20 or 30 were mixed so that test persons could compare the smell of different mixtures. Two entirely different sets of flavor compounds were easy to distinguish (left), but if a substantial part of the compounds were identical (violet squares) it quickly becomes difficult to sense a difference (middle) or at about 90% overlap mixtures can not even be distinguished at all (right). Figure inspired by figures made in Bushdid et al. (Science, 2014).
Yesterday’s edition of Science contained an interesting study on the sensitivity of the human sense of smell (article itself is not freely available but the podcast is, as well as covering articles in New Scientist or Science News). The authors took a set of 128 well-known and well-defined flavors and prepared mixtures from 10, 20, or 30 of these substances. Test persons were then given three samples to compare by their smell whereof two were identical and the third had a different composition (same number of flavors, but slightly to entirely different flavor composition). The test persons were asked to detect the deviating sample.
Most people were easily able to tell the difference between two entirely different solutions1, but the higher the number of identical compounds got, the more often people failed to identify the deviating sample. If more than 50-70% of the ingredients were the same, smelling the difference apparently becomes very difficult2. Solutions that were identical to around 90% of their ingredients became entirely indistinguishable for the human sense of smell.
Are one trillion smells a lot?
If you do some math you come to a number of at least one trillion flavor combinations that most humans could discriminate against each other. Actually the number “1 trillion” is based on a very conservative guess3 and still this low estimate of how many aromas we can smell is much much bigger than what our visual sense (around 2-7.5 millions colors we can distinguish) and our auditory sense can do (around 340000 tones). That’s why the journal praises this number so loudly, as well as other media4.
Yet this comparison of numbers is a little odd. The sense of smell is often termed “multi-dimensional” simply because it functions based on hundreds of different receptors. Also it doesn’t aim at detecting one type of stimulus as the visual sense (=detecting photons/light waves) or the auditory sense (=detecting sound waves), but can respond to uncountable flavor molecules. So to me mixing 10, 20, or 30 basic flavor ingredients should not be compared to playing one tone or to looking at one color. I find that such complex flavor cocktails rather compare to images or even movies, as well as to more complex melodies, tunes or chords composed of many tones5. And if I imagine such a comparison, our human sense of smell suddenly looks far less impressive. Trillion or not. What I actually find most interesting is that roughly half of us are apparently not able to detect the difference in smell between two substances even when 50% of their flavor components are entirely different! And not even the most gifted of us will be able to smell a difference if you would take out an average of 10-20% of all flavor notes in –say wine, perfume, or chocolate6.
- C. Bushdid, M.O. Magnasco, L.B. Vosshall, A. Keller. Science 21 March 2014: Vol. 343 no. 6177 pp. 1370-1372 (not open access)
- Review article on sense of smell (not open access), saying that
“Humans can detect and discriminate countless odorants, but can identify few by name.”:
- Another article on a plos blog on the difficulty to identify smells by name:
- Which means that both solutions contained a non-overlapping set of flavor compounds [↩]
- The precise numbers depend on how you measure the threshold of sensing the difference, and also depends on the number of compounds. The authors report that (only!) about half of the people could sense the difference between solutions with around 50% overlap. [↩]
- Because the authors calculate this number only based on the set of 128 flavor compounds, but in reality there will be many more “basic” flavor compounds. And the authors estimate this based on mixtures of 30 ingredients, but tests with higher numbers of components could lead to even higher numbers, too. [↩]
- Some covering articles appeared in New Scientist or Science News. [↩]
- The last author also mentions this in the Science podcast. [↩]
- I made this example to make things less abstract. It is of course a little more complex, because what I say only hold true if you would remove or exchange flavors that are not too dominant. After all the study of Bushdid et al. uses a balanced set of flavors, meaning that they are all adjusted to contribute similar intensity when added. [↩]