The anatomy of onomatopoeia

Abstract

Virtually every human faculty engage with imitation. One of the most natural and unexplored objects for the study of the mimetic elements in language is the onomatopoeia, as it implies an imitative-driven transformation of a sound of nature into a word. Notably, simple sounds are transformed into complex strings of vowels and consonants, making difficult to identify what is acoustically preserved in this operation. In this work we propose a definition for vocal imitation by which sounds are transformed into the speech elements that minimize their spectral difference within the constraints of the vocal system. In order to test this definition, we use a computational model that allows recovering anatomical features of the vocal system from experimental sound data. We explore the vocal configurations that best reproduce non-speech sounds, like striking blows on a door or the sharp sounds generated by pressing on light switches or computer mouse buttons. From the anatomical point of view, the configurations obtained are readily associated with co-articulated consonants, and we show perceptual evidence that these consonants are positively associated with the original sounds. Moreover, the pairs vowel-consonant that compose these co-articulations correspond to the most stable syllables found in the knock and click onomatopoeias across languages, suggesting a mechanism by which vocal imitation naturally embeds single sounds into more complex speech structures. Other mimetic forces received extensive attention by the scientific community, such as cross-modal associations between speech and visual categories. The present approach helps building a global view of the mimetic forces acting on language and opens a new venue for a quantitative study of word formation in terms of vocal imitation.

Figure 1. Sketch of the vocal model.

The figure in the middle represents the concatenation of tubes that approximate the vocal tract. The upper panel represents, from left to right, the voiced source spectrum of fundamental frequency , the vocal tract transfer function for a tube of about 17.5 cm and the multiplication of both, corresponding to the resulting voiced sound. In the lower panel, a colored noise sound source characterizing the turbulent flow at the exit of the constriction at the section of the vocal tract and the resulting fricative sound, filtered by the vocal tract.

Figure 2. Anatomy of vowels.

Each point in the graph corresponds to a vowel sample ( ms) taken from normal speech recordings of 20 Spanish speakers of different age and sex. We performed a Fast Fourier Transform to the time series to get the vowel spectrum and plot the first two formants and . The points naturally cluster into five groups, associated with the Spanish vowels . The figures defined by the black lines are vocal tract shapes taken from a corpus of MRI-based anatomical data reported in . In each case, we selected from the corpus the vowels that were closer, from a phonetic point of view, to the Spanish vowels: . MRI-based data consists of 44 area functions taken from equally spaced slices of vocal tract shapes , . The shapes drawn here correspond to the solid of revolution of radius . On the other hand, the gray shapes are the reconstructed vocal tracts from our model (see Methods).

Figure 3. Co-articulated fricatives.

From top to bottom, reconstructed vocal tract configurations of co-articulated fricatives , , , and (gray shapes) and their associated MRI vowel data (black contours). The obtained shapes are a combination of the preceding vowel and a constriction at the velar level (located around half the vocal tract length), indicated by the watermark. These vocal tract configurations along with the source parameters are: , , , , generate sounds having the spectra in black, to be compared with the experimental spectra, in gray.

Figure 4. Anatomy of onomatopoeias.

We compare sound time series, spectra and anatomy of the click (panel a) and knock (panel b) onomatopoeias and their corresponding sounds. As evident from the time series for the knock and click words (upper insets), the occlusive consonants are naturally isolated from the rest of the speech sounds during the pronunciation of the onomatopoeias in normal speech. However, co-articulation strongly affects their spectral content (medium insets): the occlusive consonants and consist of superimposing a velar constriction on a vocal tract that globally resembles the vowels in click and in knock (lower insets). The figures to the right within the frame represent the best vocal tracts imitating the click and knock sounds as retrieved by our model, without anatomical restrictions. To the right, outside the frame, we show the area functions for the occlusive consonants (black) and (gray) for the click (dotted) and (black) and (light gray) for the knock (gray). In the bottom panel we show the first two components of the PCA for the co-articulated consonants and best imitations: ; ; ; ; ; knock =  and click = . The distances between the knock vocal tract and the coarticulated consonants are:  = 0.90;  = 0.82;  = 1.00;  = 0.26;  = 0.38. The distances between the click vocal tract and the coarticulated consonants are:  = 0.95;  = 0.26;  = 0.21;  = 1.07;  = 0.50.

Figure 5. Associations between co-articulated consonants, knocks and clicks.

We evaluate the similitude of sounds with respect to the knock (solid line) and click (dotted line) sounds. Participants graded the audio files using a scale from 1 (poor or no association) to 10 (perfect identification). The left panel summarizes the responses of 20 participants to 7 synthetic sounds: the 5 co-articulated , using the parameters of (figure 3 and table 2) modulated by an experimental envelope (see Methods). The other 2 sounds were generated using the best vocal tracts for the knock and click sounds, modulated by the same envelope (points in light gray). The stronger associations with the click and knock sounds are and respectively. The best vocal tracts performed better than the consonants. In the right panel, we show the results of the experiment for 20 subjects using experimental isolated fricatives . The trend is the same as before, but grades are systematically higher.