|dc.description.abstract||The main purpose of this research was to develop an integrated framework for the description, analysis and interpretation of vowels, which should be applicable first and foremost to Afrikaans, but also to other languages. This broad applicability is, to a large extent, due to the nature of articulatory and acoustic phonetics. This framework can therefore be seen as an attempt to develop a universally applicable descriptive model. Although many well-written works exist which provide detailed discussions on the two types of phonetics, a comprehensive, integrated model such as the one proposed here has not yet been developed. This article is an attempt to contribute to the development of such a framework. I focus on the description of the physical, i.e. the acoustic, characteristics of vowels, as this facet gives one the opportunity to provide a wider-ranging and more precise description, which is imperative when different vowel systems are being compared. Such comparisons include those between the related languages Afrikaans and Dutch, but also, and in particular, the comparison of a number of pronunciation varieties of Afrikaans. The latter is an important aim of this comprehensive research project, of which this article provides an overview.
In the absence of sufficient artwork, I have made ample use of Dutch works (particularly that of Rietveld and Van Heuven, 2009), which constitute a comprehensive discussion of the articulation of vowels. Furthermore, these authors personally made available to me a complete set of magnetic resonance imaging (MRI), one for each Dutch vowel. The second author also provided his sound files of his reading of the vowels in the framework h_t. This was used in the development of an online interactive vowel chart. As can be seen and heard on this web page, the manner in which vowels are formed (articulated) is illustrated graphically. The pronunciation of vowels is combined with these images. An attractive feature available on this interactive chart is the comparison of the tongue and lip stances of a specific vowel with each of the remaining eleven vowels in a graphic manner.
The acoustic parameters under consideration in this article are the temporal parameter, duration and particularly various spectral parameters. The most important parameters are those which refer to the vowel formant frequency parameters F1, F2 and F3, as well as F0, which is the fundamental frequency. The bandwidth of vowels is often neglected in vowel descriptions in general. Nevertheless, I do discuss them here with specific reference to F1. Other parameters that are also interesting and important for a comprehensive description of vowels are total intensity and the intensity that can be measured in certain spectral areas (I investigate two: those called BF1 and BF2, respectively Band Pass Filter 1 and Band Pass Filter 2). According to my knowledge, the ratio between harmonicity and noise that can be found in vowel waveforms (i.e. harmonicity-to-noise ratio, or HNR) has not yet been used in the way in which I represent it in this research. The latter may be valuable when studying emphasis on vowels, be it stress or sentence accent.
In this article, both acoustic vowel charts and line graphs are proposed to describe the acoustic characteristics of vowels in a precise and comprehensive manner. These vowel charts and line graphs were derived from acoustic information extracted from the acoustic wave form and provided in the form of tables. It is indicated that although the use of gradient-related parameters is not required in the case of the short, steady-state vowels, the opposite applies to the diphthonged vowels with gliding first and second formants. This aspect of vowels receives special attention. Various methods of approach are illustrated, among others those related to the one that I propose. I refer to the method of description of Adank et al. (2004) in particular, which simply uses two readings on the temporal line of the formants, one on the 25% mark and one on the 75% mark. Smakman (2006) refines this method by suggesting a diphthongisation index. I indicate that this method is not superior to the one suggested in the present article.
In addition to explaining the use of acoustic vowel charts and line graphs, it is demonstrated how the statistical method of discriminant function analysis can be used to determine which parameters discriminate best between two or more naturally occurring groups of vowels. It is pointed out, for example, that one should always be able to discriminate between vowels that belong to the same natural phonological class in a more precise manner when making use of the full set of parameters rather than only the two traditional vowel frequency parameters, namely F1 and F2, even if duration, the other customary phonological feature, is also brought into the equation. This is demonstrated with reference to the high middle and high back vowels (/u/ and /o/). When only F1 and F2 are used, /u/ is classified correctly for only 56%, and /o/ for 75%. This level of success is increased to 94% and 95% respectively if all parameters are used||en_US