Sadness in speech: minor thirds?
« previous post | next post »
In "Poem in the key of what" (10/9/2006), I blogged about a paper by Maartje Schreuder, Laura van Eerten and Dicky Gilbers, "Speaking in major and minor keys". Its abstract:
In music the difference between sad and cheerful melodies is often indicated as a difference between a minor and a major key. In order to investigate whether the same difference can be found in language, we analyzed intonation contours in emotional speech. We made cluster analyses in order to find out which fundamental frequencies were most present in the contours. Furthermore, we analyzed the musical scores of sad and cheerful speech as well. In the pitch contours of all speakers we found intervals of three semitones in sad passages and intervals of four semitones in cheerful passages. We therefore conclude that emotional speech melody, just as musical melody, involves major and minor modalities.
The idea behind this paper is that the pitch contours of speech naturally express the same sorts of melodic intervals that occur in music. This is an old idea, prominent already in Paṇini's work two and a half millennia ago, but Schreuder et al. have a new idea about how to look for the phenomenon. While it's clear that musical intervals are part of the stylized forms of speech that we call "chanting", I've always been skeptical that well-defined intervals (in the sense of small-integer ratios of pitch values) play a role in unchanted speech. I'll explain some reasons for my skepticism later in this post. However, it would be fun to be wrong on this one.
My first problem was the idea that the contours of spoken intonation stem involve well-defined musical tones and intervals, either in the speaker's intent or the hearer's perception. When I ask someone with perfect pitch or good relative pitch to tell me (for example) the musical interval between the pitches of "good" and "noon" in a normal performance of "Good afternoon!", their reaction is generally uncertainty or puzzlement, as if I'd asked them to judge the color of sincerity. And I'm also unconvinced that tonal intervals in music have universal emotional associations across cultures, but never mind that for now.
In the 2006 post, I did confirm the claim that F0 histograms of spoken intonation contours are often multimodal, as the F0 histograms of chanting or a capella singing always are. But the nature of those distributions, and their relationship to musical intervals and scales, remained unclear to me.
I recently stumbled on a 2010 paper by Meagan Curtis and Jamshed Bharucha, "The minor third communicates sadness in speech, mirroring its use in music". Its abstract:
There is a long history of attempts to explain why music is perceived as expressing emotion. The relationship between pitches serves as an important cue for conveying emotion in music. The musical interval referred to as the minor third is generally thought to convey sadness. We reveal that the minor third also occurs in the pitch contour of speech conveying sadness. Bisyllabic speech samples conveying four emotions were recorded by 9 actresses. Acoustic analyses revealed that the relationship between the 2 salient pitches of the sad speech samples tended to approximate a minor third. Participants rated the speech samples for perceived emotion, and the use of numerous acoustic parameters as cues for emotional identification was modeled using regression analysis. The minor third was the most reliable cue for identifying sadness. Additional participants rated musical intervals for emotion, and their ratings verified the historical association between the musical minor third and sadness. These findings support the theory that human vocal expressions and music share an acoustic code for communicating sadness.
Oddly, Curtis and Bharucha don't mention the work by Schreuder et al.
Curtis and Bharucha used "bisyllabic speech samples conveying four emotions recorded by 9 actresses", while Schreuder et al. used data from "the performances of five professional readers reading passages from A.A. Milne’s Winnie the Pooh in Dutch", taking Eeyore to be sad and Tigger to be happy.
Unfortunately Curtis and Bharucha's footnote promising "Examples of speech samples" at a Tufts URL no longer works, and Schreuder et al. didn't offer data samples at all.
Since I've recently been working on audio samples from poetry readings from PennSound, I thought I'd try comparing the F0 histograms of some poems differing in emotional valence, taken from a set of 125 readings by John Richetti. Just for fun.
I selected five poems with (apparent) positive valence and five with (apparent) negative valence.
Here are the poems with negative valence (according to my perception of their texts), with links to the texts, to John's readings, and to the F0 histograms plotted in semitones. Click on the histogram images for larger versions, and you'll see that the first three out of five indeed have histogram modes separated by minor thirds:
| reading |  |
|
| A Narrow Fellow in the Grass | reading |  |
| Spring and Fall | reading |  |
| Lucifer in Starlight | reading |  |
| Dolor | reading |  |
And here are the five poems whose texts seemed to me to express positive feelings. None of the five histograms exhibit modes separated by minor thirds:
| Lullaby | reading |  |
| She Walks in Beauty | reading |  |
| These I, Singing in Spring | reading |  |
| The Daffodils | reading |  |
| A Psalm of Life | reading |  |
So this crude test sort of somewhat supports the "sadness in speech is expressed by minor thirds" thesis. Of course John Richetti may have chosen to express other emotions in those poems, or no strong emotions at all. And there are other uncontrolled co-variates, like line and verse lengths, overall poem length, differences in syntactic patterns, and so on.
Some day I'll try something similar on a better-controlled dataset, like this one — and look for better methods for evaluating the hypothesis (or other related ones).
Rick Rubenstein said,
April 23, 2026 @ 4:44 pm
I find it almost unbelievable that the universality (or otherwise) of emotional responses to basic intervals in music is still an open question. Even given that musicology and music psychology are (I assume) underfunded fields, and finding sufficiently large pools of subjects who have never been exposed to Western music (or non-Western music which has itself been influenced by Western music) must be quite challenging — if we can't pin down this most basic foundational question, how can we expect to establish anything?
Julian Hook said,
April 23, 2026 @ 4:55 pm
Another apparent problem in the first abstract is that the authors seem to conflate major thirds with major keys (or "modalities") and minor thirds with minor keys. In fact, both major and minor thirds occur in similar numbers in music in both major and minor keys. People may think that a major chord sounds more cheerful than a minor chord—but a major chord contains one major third and one minor third, and so does a minor chord. Using the notes of a major scale, you can form four different minor thirds and only three different major thirds.
Jonathan Smith said,
April 23, 2026 @ 5:04 pm
Well til today I guess I assumed car horns ("honk!") were intended to be either minor or major thirds… but internet (www.losdoggies.com) says the idea is 360 cents-ish i.e. "rather somewhere in between, a unique Car Horn Third, that evokes the spectrum of triadic emotions." IDK how that makes me feel…
Yuval said,
April 24, 2026 @ 2:08 am
One might argue (?) that poetry reading is already on the road to singing, though.
Andreas Johansson said,
April 24, 2026 @ 5:32 am
I found myself wondering if the choice of samples from actresses in particular was wise – one might suspect that thespians are unusually prone to convey emotion in conventionalized ways.
Daniel Deutsch said,
April 24, 2026 @ 5:33 am
Following on Julian Hook’s point, imagine a cheerful, chant-like “HEL-lo!” on the pitches G-E. I would hear it as part of a C major triad, although that is not the only possible interpretation. The “Good Morning” song from “Singin’ in the Rain” has emphatic minor and major thirds. They are both components of the major triad, which is doing the happy work.
Roscoe said,
April 24, 2026 @ 9:14 am
That is indeed how the word is set to music in “Hello!”, the opening number of “The Book of Mormon,” where it mimics the sound of ringing doorbells. (Perhaps not coincidentally, it’s the exact opposite of the two-note “hello” heard in “Hello, Dolly!”).
Alex Temple said,
April 24, 2026 @ 2:59 pm
I wish scientists would consult music scholars before doing studies like this. In addition to the issue that Julian Hook pointed out above, even the idea that "major mode is happy, minor mode is sad" is wildly oversimplistic. Luis Fonsi's "Despacito," Michael Jackson's "Billie Jean" and Benny Goodman's "Sing, Sing, Sing (With a Swing)" are all in minor and few people would call them sad. On the flipside, there are plenty of sad songs/pieces in major keys, including Everly Brothers' "All I Have to Do is Dream," the Disemberment Plan's "Come Home," and Big Star's "Holocaust." Chopin's Etude in E Major, Op. 10 No. 3, is even nicknamed "Tristesse"!
Viseguy said,
April 24, 2026 @ 3:51 pm
Off-topic and on a personal note, John Richetti was one of my Core Curriculum teachers at Columbia in the late '60s. His pear-shaped delivery never failed to hold one's attention. So nice to see his photo and read his bio.
Philip Taylor said,
April 24, 2026 @ 4:59 pm
Sorry, what exactly (or even approximately) is a "pear-shaped delivery" in this context, Viseguy ? If it were cricket that we were discussing, I might be able to hazard a guess, but in the context of speech I would have no idea of even where to start in trying to understand the phrase.
AntC said,
April 24, 2026 @ 6:40 pm
@myl I'm also unconvinced that tonal intervals in music have universal emotional associations across cultures, (and as others comment)
A counterpoint might be to compare speech contours in cultures whose music has non-Western tunings. (Of course ideally before the whole world became swamped in Western pop slop.)
Javanese Gamelan has two major tuning systems, pentatonic and heptatonic, neither comparable to Western major/minor. There's a definite emotional feel for me from each; but each piece is in either sléndro or pélog [**], no 'modulation'.
[**] to get something like a Western scale, you need to combine notes from both, although it still doesn't sound right. I've seen this done to produce 'Happy Birthday to You'.
Viseguy said,
April 25, 2026 @ 12:00 am
@Philip Taylor:
From the OED:
pear-shaped,
. . .
2. Of a musical or (more often) vocal tone: rich, mellow, sonorous.
Philip Taylor said,
April 25, 2026 @ 4:09 am
Good Lord. Thank you, Viseguy — I never knew of that meaning.
AntC said,
April 25, 2026 @ 5:25 pm
I hunted for the 'pear-shaped' of musical tone to derive from the pear-shaped body of a lute or oud. But that seems to be fanciful etymology.
Viseguy said,
April 26, 2026 @ 3:12 pm
@AntC: Nice try. It is indeed a rather odd metaphor when applied to tone, musical or vocal. When the word popped into my head, I wasn't sure it meant what I thought it meant, so I looked it up before posting here. Glad I did.
Jichang Lulu said,
April 27, 2026 @ 4:48 pm
Satie's Trois [actually seven] morceaux en forme de poire are arguably so shaped, but I doubt he had the English idiom in mind.
Robert Coren said,
April 28, 2026 @ 8:45 am
Satie tended to be creative, not to say whimsical, in his choice of titles, but my understanding is that this particular title was intended as a response to critics who complained that his pieces didn't have any form.