The perception of vowel sounds

Abstract

This thesis discusses what is meant by "phonetic quality" in so far as vowel sounds are concerned. The first chapter points out that phoneticians implicitly describe speech sounds in two different ways firstly in terns of phonetic quality which is in practice the attributes of a sound -which convey information by virtue of being part of a socio-linguistic code; and secondly in terms of their personal quality, which consists of the attributes that convey information only about the speaker considered as an individual.The second chapter discusses, the historical development of our present descriptions of the phonetic quality of vowels. It is suggested that not only is there very little evidence for the traditional articulatory descriptions, but also that the acoustic descriptions of vowels in tens of their formants have never been shown to be adequate for all vowels.The third chapter describes the collection and spectrographic analysis of a large number of sets of cardinal vowels spoken by twelve phoneticians. It appears that no precise .statements about the acoustic correlates of phonetic quality can be made; but it is probable that vowels such as [e, ɛ, a,ɑ] are best specified in terms of the relation between the pitches of their first two formants and the pitches of the first two formants. of other vowels of this general type spoken by the same speaker. This is not true of vowels such as [i, u, o, ɔ]. These vowels often cannot be conveniently analysed as having two or three formants; nor, even when a procrustean formant specification is possible, are they adequately specified in terms of formant frequencies (or pitches). An illustration is given of a pair of vowels spoken by one speaker in which the frequencies of the first three formants are almost identical, but which are nevertheless heard as having quite different qualities (viz. [ɔ] and [o]) , probably because their formant bandwidths are very different.In chapter four it is shown that in so far as a formant frequency specification is adequate, it is the relative and not the n. salute values of the formant frequencies which convey linguistic information. In an experiment using synthetic speech it was found that the same test sound was identified in different way when it was preceded by different versions of an introductory sentence.Chapter five discusses the psychological mechanism responsible for this process. It is suggested that previous theories accounting for similar processes are not fully adequate, in that they incorrectly consider a subject's adaptation tJ a series of stimuli to be independent of the response which he makes to the stimuli.The sixth chapter discusses an experiment in which 18 phoneticians specified the vowels in ten words in a language which was unknown to them in terms of a standardised cardinal vowel diagram.. It is shown that among. phoneticians trained in the use of the cardinal vowel system, there is often a high degree of agreement in the dpecification- of a vowel which has á degree of lip rounding' similar to that of the nearest cardinal vowel; but many disagreements occur in the specifications of vowels with less familiar lip positions. Nevertheless it is concluded that, in view of the results detailed in the previous chapters, at present the best method of specifying many vowels is in terms of the traditional cardinal vowel system,