Classical singing study reveals the secret behind a beautiful singing voice

A ringing voice can impress a singing judge without simply being the loudest voice onstage. In a simulated classical vocal competition, higher scores were associated with vibrato, resonance and expressiveness. The findings help explain what experts hear when they evaluate classical singing, while leaving much of their judgment beyond a handful of acoustic measurements.

The research, published in Journal of Voice, comes from Haruka Kondo, Sotaro Kondoh and Shinya Fujii at Keio University in Japan. Kondo studies vocal science and performs as an operatic soprano. Her experience competing gives the question a practical edge: an overall score says little about which aspects of a performance a singer should develop.

Descriptions of voice quality often depend on words such as “warm,” “focused” or “ringing.” Those descriptions can carry different meanings for different listeners. The researchers wanted to connect expert impressions with defined vocal attributes and measurable sound features, giving singers and teachers a more specific way to discuss what they hear.

“Our study bridges the gap between artistic intuition and science and does not aim to replace the expert ear,” Kondo said. “It helps translate what experienced listeners hear into a shared, measurable language.” The experiment examined associations with scores, rather than testing whether deliberately changing a particular feature improves a singer’s performance.

Twenty-two Japanese classical singers participated, including 10 men and 12 women. All were studying at a music university or had graduated from one. Each sang “Caro mio ben” in E-flat major, without instrumental accompaniment, allowing the team to compare live performances of the same relatively accessible piece instead of contrasting different repertoire.

Every singer received 10 minutes to warm up and then performed once, in a single take. Five professional singers who also taught voice listened from the auditorium of a 720-seat hall. They sat in the same row, with three empty seats between judges, and completed their scoring sheets after each performance ended.

There was no general audience, only judges and recording staff. That arrangement brought the assessment closer to a performance setting than a soundproof booth, although it remained a simulated competition. The team could examine expert listening in a hall while keeping the song, recording arrangement and performance procedure consistent across participants.

Judges assigned an overall score out of 100 and rated six attributes on seven-point scales. These covered vibrato, resonance, timbre, diction, intonation and expressiveness. The researchers defined each term beforehand, distinguishing vocal projection and richness from tonal color, clear pronunciation, accurate pitch and the emotional shaping of phrases.

Three judge-rated attributes showed clear positive associations with overall scores: vibrato, resonance and expressiveness. Expressiveness included conveying emotion through music, varying dynamics and shaping phrases. The statistical model considered the attributes together and accounted for differences between judges, helping separate their overlapping relationships with the overall assessment.

Timbre, diction and intonation did not show statistically significant independent associations in that model. This does not mean pronunciation or pitch accuracy are unimportant to classical singing. The results concern differences among these trained singers performing this piece, with the other rated attributes included in the same analysis.

The distinction between resonance and timbre was particularly useful. Judges rated resonance as projection and richness, while timbre covered qualities such as brightness and warmth. A voice’s tonal color and its ability to project are related, but the study treated them as different listening judgments rather than interchangeable descriptions of a beautiful sound.

These ratings also need to be understood as expert impressions, not direct measurements of sound production. The same judges supplied the overall scores and the individual attribute ratings. Their agreement between these assessments helps describe their evaluative priorities, but does not establish that improving one isolated trait would cause a predictable increase in competition scores.

The researchers analyzed recordings for four acoustic features. Two described vibrato, the regular pitch fluctuations within a sustained note: how quickly they occurred and how wide their range was. The other measures captured relative recorded loudness and the singing power ratio, which compares prominent spectral peaks in different frequency bands.

Researchers marked the vowels in each recording and calculated acoustic features from those segments. Vibrato analysis used only intervals meeting criteria for detectable, sufficiently regular pitch oscillations. That selection matters because the resulting averages describe successfully extracted vibrato, rather than every moment of a performance. Less steady singing can be harder to summarize with these measures.

For that ratio, the team compared the strongest peak between 2 and 4 kilohertz with the strongest below 2 kilohertz. Greater relative prominence in the higher band was associated with higher overall scores. This band overlaps the singer’s formant region, associated with the ringing quality that helps trained classical voices carry without amplification.

Vibrato produced a more complicated pattern. Faster vibrato was associated with lower scores among men, while wider vibrato was associated with higher scores among women. The corresponding associations in the other group were not clear, and the experiment did not determine whether the differences reflected anatomy, voice type, aesthetic expectations or other aspects of vocal control.

Recorded loudness showed no significant association with overall scores. However, the microphone was about 20 centimeters from the singer’s mouth, while judges sat farther away. The measure described relative recording levels, not calibrated loudness at their seats, so it cannot establish that projection or volume never matters in an audition.

The acoustic model explained about 33% of score variation through its measured features, singer sex and their interactions. Including judge-level differences raised that figure to about 40%. Those results describe this dataset, rather than demonstrated accuracy for an automated scoring system applied to unfamiliar singers, different halls or new repertoire.

The small groups also limit interpretation of the vibrato differences. Some singers contributed relatively few usable vowel intervals, and unusually fast, narrow vibrato occurred alongside lower extraction rates. The authors could not exclude the influence of those cases on the statistical results.

The findings nevertheless suggest possibilities for music education. Teachers could pair listening with measurements to help students understand comments about resonance or shallow vibrato. Such feedback would supplement artistic instruction, with its usefulness needing to be tested rather than assumed from the associations found in a single session.

Kondo and her colleagues also see potential for training apps and AI voice tools that better reflect expert evaluation. The study did not build or validate those products. Clearer scoring sheets are another proposed application, although measuring selected vocal features alone would not demonstrate that competition decisions had become fairer.

The next step is to examine more works, including technically demanding arias and songs used in competitions. Their wider expressive and technical demands could change which features distinguish performances. Testing those conditions would show whether the patterns found in “Caro mio ben” remain useful when singers tackle the repertoire on which careers are often judged.

These resources explore expert judgment, vocal development and the acoustics of projection.

Toward a cognitive framework of expert singing-voice evaluation in bel canto pedagogy: a longitudinal exploratory study
An exploratory longitudinal study examines how expert teaching judgments relate to acoustic descriptions of developing voices. (Frontiers in Psychology, October 2026)

Perceived vibrato and the singing power ratio explain overall evaluations in opera singing
This earlier experiment compared expert ratings with acoustic features in recordings of 10 female classical singers. (Frontiers in Psychology, August 2025)

Singing in different performance spaces: The effect of room acoustics on vibrato and pitch inaccuracy
This study examines how performances change when classically trained singers sing in five different spaces. (The Journal of the Acoustical Society of America, June 2022)

Relating objective measurements to expert evaluation of voice quality in Western classical singing: critical perceptual parameters
This foundational research connects expert descriptions of singing quality with acoustic measurements. (Journal of Voice, June 1998)

Singing power ratio: quantitative evaluation of singing voice quality
This paper introduces a measure comparing prominent spectral peaks in higher and lower frequency bands. (Journal of Voice, September 1996)

Research findings are available online in the Journal of Voice.

The original story "Classical singing study reveals the secret behind a beautiful singing voice" is published in The Brighter Side of News.

Like these kind of feel good stories? Get The Brighter Side of News' newsletter.