SoundSketcher experiments

Between Bouba and Kiki lies Grougi

A crossmodal experiment where listeners placed 30 sound stimuli onto a morphing shape continuum — from a smooth circle, through rough undulating contours, to sharp star-like spikes — revealing systematic and acoustically predictable sound–shape correspondences.

27listeners
30sounds
3shape categories
Try the task See results

How the data was collected

Building on the Bouba/Kiki effect, this study extended the paradigm from words to musical timbres. Participants adjusted a single continuous slider to find the visual shape that best matched each sound they heard.

1. Bouba/Kiki framework

The study extends the Bouba/Kiki paradigm — the well-known crossmodal mapping between sounds and shapes — to musical timbres, focusing on the perceptual middle ground between smooth and sharp: roughness.

2. Thirty audio stimuli

Thirty short audio stimuli (1.5–3 s) were drawn from sound libraries and designed to span smooth, rough, and sharp timbral qualities equally — 10 sounds per category.

3. Shape-morphing interface

A browser-based slider morphed a circle into increasingly jagged, star-like forms. Participants set it to wherever the image best represented the sound's perceptual quality. Slider values were hidden from participants.

4. Acoustic modelling

A multinomial regression model with Vassilakis' roughness and Zwicker's sharpness as predictors reached 80% cross-validation accuracy, showing these acoustic features reliably drive the visual categorisation.

Try the task

Pick a sound, play it, and drag the slider until the shape matches what you hear. Then compare your choice with the 27 listeners of the study.

Select a sound to start.
Smooth Sharp
Shape value: 0

Key findings

Strong, non-random associations emerged between the three shape categories and the 30 audio stimuli. Acoustic roughness and sharpness models effectively predicted participants' assignments, supporting a perceptually grounded mapping between sound and image.

χ² = 478.89

Chi-square across 30 stimuli × 3 categories, p < 0.0001

V = 0.544

Cramér's V — relatively strong systematic structure

80%

mean cross-validation classification accuracy (5-fold)

27/30

sounds with significant uncorrected stimulus–category associations

Publication

CMMR 2025 · London

Between Bouba and Kiki lies Grougi: Modelling the Rough Middle Ground of Sound–Shape Correspondences

Asterios Zacharakis, Konstantinos Velenis & Emilios Cambouropoulos

CMMR 2025 Semantic ratings (Exp. 2)

Crossmodal correspondences between audition and vision have been extensively studied in relation to pitch and loudness, but timbre remains more challenging to capture empirically and model acoustically. Building on the well-known Bouba/Kiki effect, this study extends the framework to musical timbres and investigates visual forms spanning a continuum from full roundness to high angularity. Results revealed strong, non-random associations between sounds and three shape categories (smooth, rough, and sharp), with acoustic modelling showing that established roughness and sharpness models effectively predicted participants' assignments.

Cite this work (BibTeX) 
@inproceedings{Zacharakis2025Grougi,
  author    = {Zacharakis, Asterios and Velenis, Konstantinos and Cambouropoulos, Emilios},
  title     = {Between {B}ouba and {K}iki lies {G}rougi: Modelling the Rough Middle Ground of Sound--Shape Correspondences},
  booktitle = {Proceedings of the 17th International Symposium on Computer Music Multidisciplinary Research ({CMMR})},
  year      = {2025},
  address   = {London, United Kingdom}
}