Resolving the bouba-kiki effect enigma by rooting iconic sound symbolism in physical properties of round and spiky objects

Abstract

The "bouba-kiki effect", where "bouba" is perceived round and "kiki" spiky, remains a puzzling enigma. We solve it by combining mathematical findings largely unknown in the field, with computational models and novel experimental evidence. We reveal that this effect relies on two acoustic cues: spectral balance and temporal continuity. We demonstrate that it is not speech-specific but rather rooted in physical properties of objects, creating audiovisual regularities in the environment. Round items are mathematically bound to produce, when hitting or rolling on a surface, lower-frequency spectra and more continuous sounds than same-size spiky objects. Finally, we show that adults are sensitive to such regularities. Hence, intuitive physics impacts language perception and possibly language acquisition and evolution too.

Figure 1

Schematic representation of the phenomenological model used for predicting round (over spiky) score based on the properties of each stimulus s (t) for each experiment. First, a spectrogram is generated thanks to a classical gammatone filter-bank inspired from auditory filtering. Then, two independent analyses are applied. Spectral analysis (top pipeline) provides the Balance index defined as the difference between summed energy in low and high spectral components around a boundary b (Eq. 1). Temporal analysis (bottom pipeline) provides the Continuity index based on the ratio between the minimum and maximum energy through time within the corresponding stimulus (Eq. 2). Finally, the round (over spiky) score is linearly predicted from the combination of Balance and Continuity values. For more details, see "Methods" and Supplementary Information.

Figure 2

Model round score predictions (Model, y axis) as a function of Experimental round score data (Data, x axis, on the same scale as in the original studies, see Table 2 for more details) for the 10 experiments. For each speech stimulus (displayed by empty circles) in each experiment, the round score prediction in the model is obtained for the boundary b value providing the best fit for the corresponding experiment (cf. "Methods"). Linear regression lines and corresponding explained variance r² between experimental and predicted round scores are added to each display.

Figure 3

Stimuli manipulation and results from the Noise Band Experiment. (a) Top. Spectral content of the non-speech stimuli in the Noise Band Experiment made of bands of white noise, centered at values varying from 300 Hz to 1,200 Hz (Central frequency). Bottom. Amplitude modification of the stimuli to manipulate temporal continuity. Non-speech stimuli were 500 ms long and modulated by an envelope containing a dip from 225 to 275 ms, at an amplitude taking 4 possible values at 0, 0.1, 0.5 and 1 relative to the maximum value (Dip amplitude: 0 meaning a silent dip, and 1 meaning no dip at all). The amplitude was constant from 0 to 225 ms and from 275 to 500 ms. (b) Experimental results for the non-speech stimuli. The percentage of round (over spiky) score is averaged across participants as a function of the central frequency of the noise band and the size of the envelope modulation dip. Error bars are standard error from the mean.

Figure 4

Stimuli and results from the Beating Toys Experiment. (a) Images of the visual stimuli (children’s plastic toy reproductions of fruits and vegetables). (b) Typical frequency spectrum resulting from hitting the base of a round (apple) or the tip of a spiky (watermelon slice) object on a hard surface. The spectral balance is lower for the round as compared to the spiky object. (c). Percentage of round (over spiky) score averaged across participants as a function of each stimulus object and Toy Shape Category (from left to right, Round: Apple, Lemon, Potato, Strawberry, Tomato; Spiky: Banana (tip), Carrot (tip), Cucumber (tip), Fennel (tip), Watermelon slice (tip)). Error bars are standard error from the mean.

Figure 5

Stimuli and results from the Rolling Balls Experiment. (a). Picture of the two 3D balls (one round, one spiky) propelled by a flick on a plane shelf used in the forced-choice task. The balls were made with a 3D printer. (b) Examples of spectrograms of two original sounds produced by the round and the spiky balls rolling on a shelf. The acoustic envelope, drawn in black, is smoother for the round ball. Darker areas represent more intense low- (for the round ball) or high- (for the spiky ball) frequency bands. (c) Experimental round score averaged across participants for the Rolling Balls Experiment. These scores are represented as a function of Acoustic Envelope (Smooth—circles vs. Sharp—triangles) and Spectral Balance (Low-Frequency, LF, in blue vs. high-frequency, HF, in red) for the original and the synthetically modified sound stimuli. Big circles and triangles show overall means. Error bars are standard errors.