[The “INDSCAL” Multidimensional Scaling Analysis in Vowel Perception]
A recent model of multidimensional scaling, INDSCAL, was used in an attempt to investigate some aspects of ihe perception or vowels. Two major features make the INDSCAL method original: first, the external structure it provides has to be interpreted without any rotation of the axes; second, configurations can be obtained, that are specific to each subject or to each value of an experimental variable, depending on the inputmatrices. — In a first experiment, seven French vowels [a], [ε], [e], [i], [ɔ], [o], and [u] synthesised at seven different values of fundamental frequency mixed with the same vowels uttered by male and female, were presented for identification to eight subjects. INDSCAL analysis yielded a two-dimensional configuration that was found to be very similar to the physical plane F1–F2. Furthermore, the two perceptual axes match perfectly well with the two first formants. A second experiment, this time with natural speech, was designed to test the relevance of these results using other experimental variables. One period samples of ten vowels [a], [ɔ], [o], [u], [y], [i], [ε], [e], [ø] and [oe] pronounced by ten native French speakers were presented to ten listeners. From the several analyses of these data, it was concluded that F1 and F2 remain the perceptually relevant dimensions throughout the whole identification process. Comparison between individual speakers’ and listeners’ differences revealed normalisation processes which arc specific 10 each perceptual dimension. An explanation in terms of reciprocal interactions perception-production is proposed. This particular approach of INDSCAL is most attractive in that the dimensions emerge as being meaningful for one particular stage of the process and therefore leads to a functional description of it.
How to Cite:
Beeckmans, R., 1978. L’étude de la Perception Des Voyelles Par la Méthode D’Échelle Multidimensionnelle «INDSCAL». Psychologica Belgica, 18(2), pp.125–150. DOI: http://doi.org/10.5334/pb.629