Listening to the Smell of Stinky Bugs
Is sonification a distraction or a source of insight?
I’m always on the lookout for studies that break new ground in terms of methods or results. They’re way more fun than slogging through a paper that's just an incremental brick in the wall of science.
So I was psyched to find a paper with the jarringly offbeat title “Chemical composition: Hearing insect defensive volatiles,” which appears in Patterns, a new data science journal from Cell Press.
It took me while to get my head around the study’s premise. Apparently there is a thriving subfield of data processing and analysis called sonification—think of it as the auditory equivalent of visualization. You take a complex data set and assign a sound feature to each of its parameters. Then you listen to the data-based composition and hope to extract novel insights.
It’s definitely a weird idea, but what the hell—the proof of the pudding is in the eating.
The pudding on offer is by two Belgians interested in the chemical composition of insect defensive volatiles and their effectiveness at repelling predators. The insects are various species of sawfly larva, which squirt out a stinky mixture of chemicals from abdominal exocrine glands. These are used to discourage attacks by insectivorous insects, e.g., worker ants of the species Crematogaster scutellaris.
The larval fluids are a mixture of a few (or many) chemical ingredients—the exact composition varies by species. The Belgians measured the bioactivity (anti-ant effectiveness) of each component as well as the whole mixture using traditional behavioral tests with actual ants. The results showed that defensive efficacy was positively correlated with the number of ingredients in the glandular fluid, as well as the size of the larva’s glands.
But here’s the cool part of the study: the researchers created sonified versions of the glandular fluid and of its individual ingredients. They did this by assigning presets on a sound synthesizer to each chemical class in the gland fluid (i.e., aliphatics, aromatics, and terpenes). Then they added sound parameters for eight chemical features of a given molecule (e.g., sound pitch was negatively calibrated to the number of carbon atoms). The result was a sound file for each individual component molecule. To sonify a complete mixture of glandular fluid, they combined the files of all the individual components, using relative loudness to indicate a molecule’s abundance within the mixture. (In other words, if formic acid was 60% of the mixture and acetic acid 10%, the formic acid sound element was much louder than the acetic acid.)
The weirdest part of the study was how the researchers chose to evaluate human repellency: they played the sound files and measured how far each panelist backed away from the loudspeaker. Repellency was calculated as the percentage of maximum backing-off. It turns out the test panelists were repelled by the sonified versions of the bug chemicals, and the biggest factor determining their response was overall loudness. The Belgians claim success for sonification because the results corresponded to the ant tests:
Repellence of single volatiles, as well as mixtures of volatiles, against predators were significantly correlated with the repulsiveness of their respective auditory translates against humans, who mainly reacted to sound pressure.
Sadly, both the method and the result leave me cold.
First off, the sound elements used to represent the chemicals were chosen sorta kinda semi-arbitrarily (I found the description of the selection process rather confusing). We are not told whether the audio parameters themselves were pleasant or unpleasant. If the latter, then it comes as no surprise that the human panelists backed away from the loudspeaker. In any case, why was no attempt made to evaluate the sound parameters before selecting them? Plus: Why use digital effects on a sound synthesizer? Why not a musical instrument?
Secondly, the choice of “backing away from the loudspeaker” as a human metric is baffling. Why insist on physical distance measure? Was it to maintain an exact equivalence with the ant behavior measures? A simple sensory evaluation (“rate this sound from pleasant to unpleasant”) would have been better. They also could have inquired about the pleasantness of different aspect of the sound samples (loudness, timbre, pitch, etc.).
On the whole, I’d rate this study a swing and a miss. But the idea of sonification of smells remains an intriguing one. (To be accurate, this study sonified chemical parameters, not smells.)
P.S. As always, feel free to comment!
Jean-Luc Boevé, & Rudi Giot. (2021). Chemical composition: Hearing insect defensive volatiles. Patterns 2:100352.