Proposals for a representation of sounds based on their main acoustic-perceptual properties, Tones and Features

摘要

This paper proposes a representation of sounds based on their main acoustico-perceptual properties", The articulatory description of the phonemes is extremely useful, but sometimes difficult to achieve in sufficient detail. Phonetic transcription using IPA has proven useful, but in practice, transcription raises fundamental issues since the choice of IPA symbols depends in no small part on the transcribers' native language and on the instructions that they received during their training. In addition to the set of symbols for vowels and consonants, the IPA proposes diacritics to transcribe some differences between similar sounds; however, in order to describe fine differences such as that between the vowels transcribed as /u/ in French, English and Spanish, there is a clear need of well-established references for comparison. The cardinal vowels devised by Daniel Jones can be used as references but there are disturbing discrepancies between Daniel Jones's production, and the rendering of the same vowels by Peter Ladefoged. The specification of the phoneme in terms of distinctive features does not always reflect the acoustic-perceptual similarity between the sounds. Acoustic description, based on observations of the data on modelling is a welcome addition to a description in terms of articulation and in terms of distinctive features. Acoustic description is sometimes done in the literature, but it is often incomplete. The habit of using only the two first formants to represent vowels still persists, but it is not entirely justified, at least for front vowels. The correlation between front/back constriction and F2, on the one hand, and high-low and F1, on the other is overestimated, whereas the role of the lips (in determining F2 of the back vowels) and the relative amplitude of the formants (which plays a role for contrasting oral and nasal vowels) is often neglected. The lack of information on F3 and F4 makes it difficult to determine the position of a vowel relative to the cardinal vowels. The proposed notation system is based on the combination of acoustic and perceptual properties of sounds,and relies on Jakobons, 's, Fant's and Halle's Preliminaries to Speech Analysis, Jakobson, Fant and Halle , on Stevens's quantal theory, Fant's acoustic theory of speech production and Maeda's articulatory model and our own experience in spectrogram reading in different languages. It intends to integrate the articulatory, aerodynamic, acoustic and perceptual properties of sounds that have previously been put forward as explaining "natural" phonetic changes. The method originally based on articulatory modelling consists in placing five vowels in the maximally stretched three-dimensional vowel space: these five vowels are proposed as references for language comparison or to transcribe detailed variations in the realisation of the phonemes. Since these vowels are meant to serve as references, they are compared with Daniel Jones' (1918) cardinal vowels and their rendition by Peter Ladefoged . The same type of notation can used for enhancing the similarities between vowels and consonants and describing the effect of coarticulation. In conclusion, some of the cardinal vowels correspond to quantal regions as described by Ken Stevens. They are good candidates to be defined and used as reference vowels. The quantal vowels as described in Stevens's Quantal theory (with converging formants) may not correspond precisely to the most frequent vowels in the world's languages, but this does not detract from their usefulness as references in the description of vowel systems.