The most important conclusion reached in this study is that /ɛ/ lowering as a phonological process occurs over a wide front and is far more complicated than suggested by available descriptions. This is clear from the results of the current study, which was conducted on the basis of a diverse methodology and in which a substantial and diverse set of speech-data sources was the object of study. The sources vary from personal observations during conversations, to public performances, to RSG (Radio Sonder Grense) news broadcasts and formal recordings conducted under experimental conditions.
Personal observation of the phenomenon as it occurs – or does not occur – in formal or informal speech styles over a wide front supports, in broad terms, the existing formulations of this process, particularly with regard to the standard variety of Afrikaans i.e. that the mid-low, front vowel /ɛ/ is discernibly lowered when followed by one of the consonants spelled as <k, g, l, r>. This lowering is especially clear in the speech of white, local (i.e. studying at the NWU Potchefstroom) female students; and in varying degrees over Afrikaans radio on various programmes, again most clearly in the pronunciation of young, female continuity programme hosts, particularly on RSG. On the other hand, it is mostly absent in the speech of coloured speakers of Afrikaans, locally (i.e. Potchefstroom) as well as more broadly. An important exception in the latter case is lowering before the lateral consonant [l], where it often occurs simultaneously with the diphthongisation of the vowel; here there is also often substantial palatalisation. Even in SvA (Standard variety of Afrikaans) forms one finds a diphthongised vowel in this context, although of a reduced nature with no evidence of palatalisation. Such diphthongisation, accompanied by light palatalisation, is very occasionally, in isolated cases, heard in the speech of white radio personalities.
More structured observations than the above-mentioned, in which investigations were conducted into the speech of a group of RSG news broadcasters, but during which other radio programmes were also investigated, confirmed the suspicion that /ɛ/ lowering hardly constitutes a homogenous phenomenon, and that a variety of factors need to be taken into consideration when it comes to a more precise and nuanced description. The most important of these are linguistic, geographic, ethnic and chronological factors, as well as the sex/gender of speakers.
The results of this first phase of investigation also showed that /ɛ/ lowering does not occur on a predictable basis in all cases. In short, in some types of words, often depending on the nature of the relevant conditioning consonant, there is lesser or greater internal variation. In comparison with the other three consonants mentioned above, before [r] there is much greater stability and predictability with respect to lowering.
It was also found that /ɛ/ lowering is very constant in words with /ɛ/ before /l/. In (-)stel(-), for example, no case was found in SvA where there was no lowering to [æ]. It is clear that [l], especially in the Afrikaans of so-called coloured people, has a prominent breaking effect on preceding /ɛ/, something which is either not found in SvA or is not particularly observable. One needs, therefore, to be careful to not be misled into viewing averages of F1 measurements as the only metric for determining similarities or differences between the quality of /ɛ/ across different groups. While the F1 values of /ɛ/ in, for example, wel, shows lowering for both white and coloured speakers, there is a substantial difference to be perceived on an auditory level.
The environment /_k/, as in ek (the Afrikaans pronoun I), forms a very specific category, because so much variation is discoverable here. In general, it is very clear that the lowered [æ] before [k] is seldom heard in the speech of coloured individuals and definitely not in ek; a fact which actually applies to the majority of the Afrikaans speech community. With respect to white Afrikaans, the unlowered [ɛ] pronunciation is mostly indexical of the fact that the speaker comes from the south-western parts of the country, more specifically the Western Cape. Yet it is also observable in the pronunciation of some north-eastern speakers, more specifically radio announcers from Gauteng, who have, it appears, been under the influence of certain other prominent radio personalities and who aspire to a certain kind of prestigious speech. On the other hand, the pronunciation [ɑk], with extreme lowering and retraction, is relatively sharply indexical of the fact that the speaker is white, young, mostly female, and central-north in terms of geographic provenance. This, however, still requires further thorough research.
The above-mentioned phase of research was followed by a series of empirical studies of a more structured nature. Acoustic characteristics of the relevant vowel were determined, i.e. formant frequencies, and that of F1 in particular. The results serve as the basis for precise comparisons with respect to the nature and extent of /ɛ/ lowering in the speech of speakers from a broad geographic range in South Africa and Namibia, as well as both white and coloured speakers. Two sub-studies focused specifically on the pronunciation of speakers of different age cohorts, in accordance with which a possible diachronic development of /ɛ/ lowering was identified.
The main result of these empirical studies was the existence of /ɛ/ lowering across a wide front in SvA. This lowering does not occur equally, however, in all four of the mentioned phonetic contexts. Lowering is also not completely absent in the speech of coloured speakers. Light lowering is more common in the case of /ɛ/ before the velar consonants [k] and [x], with slightly more prominent lowering found before alveolar [r]. In the case of [l], there is, as already mentioned, the presence of breaking; i.e. a diphthong, sometimes palatalised, before the lateral consonant, is clearly apparent in the speech of most coloured sub-groups. In addition, there are strong indications of the more prominent presence of /ɛ/ lowering among young speakers from the north-eastern regions of the country. It is also clear that speakers from the north-eastern parts are more inclined towards /ɛ/ lowering than those from the south-westerns areas.
An important finding of these empirical studies, and which links with what has been said before, is that /ɛ/ lowering does not occur on a uniform basis before the four consonants concerned. This was noticed especially in the speech of coloured speakers from Genadendal, Malmesbury and Kakamas, where it occurs most prominently before the two alveolar consonants, [r] and [l] (classified as [-back, +high]). The contribution of [l] to this effect is clear enough, but needs more precise specification. It is noted, for example and by way of comparison, that in Canadian English, as reported by De Decker and Mackenzie (2000), [l] constitutes the strongest influence of all linguistic factors on /ɛ/ lowering.
In this research, two empirical studies showed the clearly strengthened degree of /ɛ/ lowering in the speech of young, female speakers in comparison with older speakers. This result clearly overlaps with research on the worldwide lowering of the equivalent short front vowel in English, as can be seen from the examples provided by Hickey (2017): yes [jæs], best [bæst], west[wæst]. Hickey (2017) speculates that in Ireland young female radio and TV presenters are at the forefront of the spread of this phenomenon. In the case of Afrikaans, a similar explanation for the expansion of low [æ], especially among white, young, female speakers, is less compelling, firstly because this tendency is not as obviously apparent and widespread in the pronunciation of Afrikaans radio presenters and secondly because, on the basis of a number of quick observations among a number of local students, it is clear that this sub-group of speakers hardly ever listens to Afrikaans radio programmes, including those of RSG.
The phenomenon of /ɛ/ lowering is a clear example of movement within the vowel space, but it does not clearly form part of a more systematic pattern of chain shifting, as its equivalent appears to do in some English cases. Hickey (2017) refers to a vowel rotation in contemporary Dublin English, which includes lowering, retraction, heightening (of low, back vowels) and fronting (of the rounded back vowels). It is well known that the Afrikaans /u/ is also centralising (Wissing 2010), and that the long /a/ vowel has not only undergone rounding but has, at the same time, undergone heightening towards /ɔ/. The extreme lowering of /ɛ/ to almost [ɑ], as found typically in the pronoun ek, could, of course, be seen as a form of retraction, as is the case with what is found in the vowel of English trap (TRAP: [træp] > [trap] (Hickey 2017)).
In conclusion, the following related remark is purely by way of a suggestion: There appears to be a parallel lowering of /ɛ/, on the one hand, and the first segment of the Afrikaans diphthong /əi/ to [ʌ], or even [ɑ], on the other hand. The question of vowel chain shifts in Afrikaans is, however, a topic that requires far more attention and is an important theme in itself. It is a topic that could, in addition, benefit from recent investigations into similar phenomena in South African English by Bekker and Eley (2007) and Chevalier (2016).
Keywords: acoustic phonetics; Afrikaans of white and coloured speakers of Afrikaans; sound chance; variation; vowel lowering; vowel system.