The Music of Perfume

Ambitus, pitch-class entropy, and keeping it simple

Image via Tobias R.

I’ve long been fascinated by the metaphorical links between scent and music. Perfumers talk about notes and accords in a way that seems natural to musicians. The 19^th C. perfumer Septimus Piesse went so far as to match essential oils to notes on the musical scale, and a couple of German researchers in the 1920’s and 30’s offered experimental evidence along similar lines.

Some years ago I explored the topic experimentally, along with colleagues at Givaudan-Roure. After getting promising results from a pilot study with an electronic keyboard, we created a more sophisticated setup using a sine-wave generator to produce precise audio tones that were automatically equated for loudness. We found that the olfactory character of 20 fragrance materials was matched to the auditory dimension of pitch. Light-smelling scents such as neroli and bergamot were paired with higher pitched tones, while heavier scents such as birch tar and civet corresponded to lower pitched tones. Fifteen years later, a study by Anne-Sylvie Crisinel and Charles Spence confirmed these findings.

Pitch, of course, is just one dimension of musical sound. One other study examined the link between smell and the timbre of various instruments, but single-parameter studies have been the norm. No scientists have been brave enough (or foolish enough) to tackle musical expression in its full melodic complexity. Until now.

Argentine researchers Bruno Mesz and Nicolás Gorla, and Spanish scientist Manuel Zarzo have taken a novel approach. They recruited 14 experienced keyboard players and had them improvise 20 seconds of music while smelling each of 20 fragrance materials. The team had a separate panel of 15 people rate the smells on various sensory measures. That was the easy part. The hard part was statistically comparing odor qualities to the character of the resulting tunes.

The musical samples were recorded on a computer using MIDI, a digital format that allows all sort of quantifications. The simplest among them included Lowest Note, Highest Note, Mean (average) Note, and Ambitus (the difference between lowest and highest note). More complex were measures of consonance and dissonance of melodic and harmonic intervals, such as Euler’s gravis suavitatus which “measures the average dissonance of the consecutive melodic intervals of the highest voice.” The researchers also quantified the “complexity of pitch and rhythm organization” in each improvised tune, using mathematical concepts borrowed from entropy theory in physics.

As you imagine, the dataset was complex and the statistical analyses even more so. After ploughing through pages and pages of it, I think the authors’ general approach was solid and I won’t belabor it here. Instead, let’s cut to some of their conclusions.

Consistent with previous studies, mean ratings on a scale of fresh vs. warm [aka light vs. heavy] were correlated with the average pitch of the improvisation. Once again, lighter scents go with higher pitch, heavier scents with lower.

Among the novel findings: Odors with a camphoraceous quality (e.g., lavender and mint) produced non legato and/or staccato articulation in the musical samples. This would be consistent with the somewhat sharp, irritating quality of these scents.

Also novel: To the degree that scents were characterized as feminine, they “were negatively correlated with the ambitus of the improvisation,” i.e., the associated tunes had a smaller range of pitch.

These, and additional novel findings, are certainly intriguing. However, they do not (for me) lead to flashes of intuitive insight. While I admire the depth of the quantitative analysis, what I like most about this study is its basic setup. Having pianists freely improvise music in response to actual scents is a totally cool idea, but one that I would prefer to take in a totally different direction.

In my version, a pianist would improvise responses to, say, five odors. Then a panel of, say, 50 volunteers would smell odor 1, listen to the five tunes, and rate how closely each tune expresses odor 1. Repeat with odors 2 through 5. This design could be bulked up by using, say, a total of 5 pianists.

Statistical analysis would be straightforward. For example, are some players measurably better than others at improvising recognizably odor-matched tunes? Also, are some odors more easily represented in music than others? and what are the critical musical parameters? Finally, are some panelists better than others at discerning the smell:music links intended by the pianists? Having players and listeners fill out a synaesthesia questionnaire would add even more depth.

In this experiment we’d learn something about specific odor-music linkages and also about the synaesthetic ability of musicians and listeners.

The beauty of this experimental design is that it can be employed in the other direction. First, chose five distinctive keyboard samples from different composers: Mozart, Schubert, Satie, etc. Then have a few perfumers create a fragrance accord to represent each tune. Present each scent to the panel and have them rate how closely it matches each of the tunes.

Voila! We can now measure how much perfumers differ in their ability to communicate music via scent. And once again, we can rank panelists according to their ability to discern the intended music:smell linkages.

My versions of the Mesz et al. study design are stupid/simple. They are not a criticism of that study, but reflect my preference for minimally controlled setting where the stimuli and the test subjects are given room to express themselves. There are different styles of science and that’s mine.

At any rate, hats off to Mesz, Gorla and Zarzo for taking a wider view of scent and music.

P.S. Dudes, if you want to get stupid, call me!

---

Bruno Mesz, Nicolás Gorla and Manuel Zarzo. (2023) The music of perfume: Crossmodal correspondences between musical features and olfactory perception. Music Perception 41:110-131.

Kira Belkin, Robyn Martin, Sarah E. Kemp and Avery N. Gilbert. (1997) Auditory pitch as a perceptual analogue to odor quality. Psychological Science 8:340-342.