Physical models of the vocal tract sound different with the same shape but different temporal characteristics and vice versa

Two physical models of the human vocal tract have been successfully developed for producing English /l/ and /r/ and Japanese /r/. The first model was originally designed for the alveolar lateral approximant and the retroflex approximant, while the second model was designed to produce English “bunched /r/.” With these models, we observed that different configurations of the vocal tract can produce similar sounds in addition to retroflex and bunched /r/. We also observed that the models produce the target sounds when the articulators move with certain patterns of temporal change. However, moving the same articulators with the same movement but different temporal patterns produced less intelligible and/or different sounds. Thus, using the two physical models, we tested and confirmed that different configurations of the vocal tract using similar temporal changes yielded similar sounds, while the same configuration of the vocal tract with different temporal changes yielded different sounds.