18.1 Introduction

The growing interest in New Englishes has been accompanied by an impressive number of synchronic studies on these varieties. However, diachronic investigations of postcolonial Englishes are still the exception. What studies there are mostly adopt a macro-sociolinguistic perspective and focus on the external history of (post-)colonial Englishes, with little or no reference to linguistic structure. Even Schneider's (Reference Schneider2007: 113–250) case studies, illustrating his Dynamic Model of the evolution of New Englishes and describing postcolonial Englishes at different stages of development, are based mainly on synchronic data.

One main reason for the lack of diachronic studies of the structural development of postcolonial Englishes in the Outer Circle is that in many cases authentic historical language data is either non-existent or has not yet been located and analysed by linguists.Footnote ² The situation is somewhat less problematic with regard to written texts, and some systematic diachronic projects in this area have started recently, e.g. Hoffmann, Sand and Tan's (Reference Hoffmann, Sand and Tan2012) Corpus of Historical Singapore English and Biewer et al.'s (Reference Biewer, Bernaisch, Berger and Heller2014) Diachronic Corpus of Hong Kong English, both based on the design of the Lancaster-Oslo-Bergen Corpus (1970–1978), or Brato's (forthcoming) Historical Written Corpus of Ghanaian English (2014), modelled on the International Corpus of English. However, for many Outer Circle Englishes few if any early recordings have been located and analysed so far.

Focusing on Ghanaian English (GhE), this chapter is an attempt to redress this situation with regard to West African Englishes. Adopting a quantitative-variationist approach, I will explore some ways in which early radio broadcasts and recordings of political speeches from the 1950s and 1960s, stored in the archives of the Ghana Broadcasting Corporation, can be used to reconstruct the phonological development of GhE. This complements a previous study (Huber and Schmidt Reference Huber and Schmidt2011), which used early popular music recordings as data for the investigation of the history of some GhE vowels. That study found substantial changes over the past fifty years, and it will be interesting to see whether radio broadcasts and political speeches show similar developments.

In this pilot study I will present the results of an exploratory auditory analysis of the broadcasts and speeches collected, digitised and transcribed so far. After a discussion of the potentials and problems associated with early audio recordings as data, four phonological variables that show variation in present-day GhE will be analysed: two consonant variables, (ing) and (wh), as well as two vowel variables, (nurse) and (strut). The research question is whether the variation observable today was already in place in the GhE of 50–60 years ago and whether the distribution of variants changed over time. Diachronic studies based on recordings and focusing on language structure can make a significant contribution to models such as Schneider's (Reference Schneider2003, Reference Schneider2007) evolutionary model that have mainly been based on external language history and synchronic structural data.

18.2 English and Indigenous Languages in Ghana

After a period of first contacts from the 1550s to 1570s, anglophone traders regularly frequented the Gold Coast (modern Ghana) from 1632 on. Formal colonisation started about two centuries later, when coastal chiefs yielded part of their jurisdiction to the British crown in the Bond of 1844. English was the language of colonial administration, but the percentage of English-speaking indigenous Gold Coasters remained low until English-medium schools were established in the 1880s,Footnote ³ so 1900 can be taken to be an approximate starting date for modern GhE. The Gold Coast was an exploitation colony, and the number of anglophone foreigners has always been low: the 1948 Census, the last before independence, returned 4,102 British, 126 US and 10 West Indian nationals, together a mere 0.1% of a total population of 4.5 million (Government of the Gold Coast 1950: 10, 83). The percentage of anglophone foreigners remained low also after the independence of Ghana, in 1957. In 1960, there were 7,502 UK and Irish nationals (0.1% of the total population) and 1,103 from Canada, the US and the West Indies (0.02%; Census Office 1964: 102) in a population of 6.7 million. The latest census (2010) enumerated 24.7 million inhabitants, among which were 4,493 Europeans (0.02%) and 2,714 aliens from the Americas (0.01%; Ghana Statistical Service 2013: 215, 247). The vast majority of Ghanaians speak indigenous languages as L1, but English has been Ghana's de facto official language since independence. It is used in formal contexts like the educational system, parliament, higher courts of justice or the media.

With the exception of Hausa, the approximately eighty indigenous Ghanaian languages (Lewis et al. Reference Lewis, Simons and Fennig2013) belong to the Niger-Congo family, whose Kwa genus includes the most widespread languages in the south: Akan (spoken by 47.5% of Ghana's population) with its various dialects (Akuapim, Asante, Fante etc.), Ewe (13.9%) and Ga-Dangme (7.4%). Several western Oti-Volta languages of the Gur genus are spoken in the north, of which Farefare (c. 4%), Dagbani (c. 4%) and Dagaare (c. 3.5%) are the most widespread. Twi (a dialect of Akan) is the main lingua franca in southern Ghana (cf. Huber Reference Huber, Buschfeld, Hoffmann, Huber and Kautzsch2014: 87).

18.3 The History of Broadcasting in Ghana and Material Analysed

Broadcasting in colonial Ghana started in 1935. A wired relay station (Radio ZOY) transmitted BBC Empire Service programmes and music to a small number of subscribers in Accra, the capital of the Gold Coast colony.Footnote ⁴ In the following years, the service expanded and rediffusion stations were opened in other cities. In 1940, a new broadcasting house with a transmitter was built in Accra, and programmes in the Ghanaian languages Twi, Fante, Ga, Ewe and later Hausa were introduced in the following years. 1953 saw the establishment of the Gold Coast Broadcasting System. Up to the mid-1950s, the English-language programme mainly consisted of governmental announcements and BBC re-broadcasts. The number of local productions only rose in 1956. On independence in 1957, the station was renamed Ghana Broadcasting System and later the Ghana Broadcasting Corporation (GBC), popularly known as ‘Radio Ghana’ (Buckley et al. Reference Buckley, Apenteng, Bathily and Mtimde2005: 5–6; Ghana Broadcasting Corporation 2010; Peasah Reference Peasah2009: 3–4).

In the early years, the GBC was assisted by BBC staff, as becomes clear, e.g. in a 1958 speech by the Minister for Education and Information, Kofi Baako: ‘We are grateful … to the British Broadcasting Corporation for their assistance in seconding a number of their staff to help us’.Footnote ⁵

The material analysed for this study comes from the GBC audio archives. The archives hold one of Africa's largest collections of gramophone and vinyl records, reel-to-reel tapes and cassettes. In 2008, digitisation of the material started in the music collection (‘Gramophone Archive’, www.gbcghana.com/gramophone/index.html, accessed 18 June 2014) and for that reason most of the material computerised so far is music.Footnote ⁶ However, the GBC also digitised a small number of its original analogue media from the so-called Sound Archive (spoken word) on the occasion of the 2009 centenary of the birth of Kwame Nkrumah, Ghana's first prime minister. These recordings relate to the Nkrumah era, from the year of independence 1957 to 1966, when the prime minister was ousted in a coup.

The Nkrumah centenary audio files form the basis of the present analysis.Footnote ⁷ The data available to me consists of sixty-five mp3 files of varying lengths (from 3 to 43 minutes), almost 18 hours in total, including music and ambient sound. The GBC digitisation project primarily focused on Nkrumah's public speeches, radio announcements and interviews. However, and crucially for the study of early GhE, the recordings also include other speakers, mostly correspondents and radio announcers who report on Nkrumah's travels, public appearances and politics. They also contain speeches by, and interviews of, Ghanaian national and local politicians as well as other officials.

The digitised Nkrumah recordings described above were used for the compilation of a small corpus of early educated spoken GhE, according to the following guidelines: first, only material recorded during the Nkrumah era (1957–1966) was selected for the corpus. These are the years immediately following political independence, one of the sociopolitical events that characterise the Nativisation and Exonormative Stabilisation phases in the Dynamic Model of the Evolution of New Englishes (Schneider Reference Schneider2003, Reference Schneider2007). Second, to avoid skewing the corpus too much in the direction of Nkrumah's idiolect, I excluded all of his speeches but one (Nkrumah's presentation of the Seven Year Development Plan in the National Assembly on 11 March 1964). I further removed all clearly non-Ghanaian speakers. Table 18.1 provides an overview of the number of words of individual speakers in the corpus.

This corpus of just above 10,000 words represents the GhE of around 1960 as spoken by national and local politicians, chairmen, government officials and radio reporters in a public setting. The individual contributions (reports, speeches, interviews) appear in most cases to have been scripted or were at least based on notes. The recordings thus document the formal educated spoken English of Ghana's political and cultural elite in the years immediately following political independence.

At present, these are the earliest known recordings of GhE of any substantial length and thus constitute very valuable material for the study of the variety around 1960. However, some words of caution are in order at this point. Until more material has been digitised by the GBC and transcribed for linguistic purposes, the corpus by necessity remains an unbalanced sample of early GhE, and all analyses based on it can only be exploratory: at present, all speakers in the corpus are male, individual speakers’ contributions are of unequal length, and they are for the most part scripted, unspontaneous monologues.

A crucial step in the compilation of the corpus was the decision whether a speaker was Ghanaian or not. Speakers’ names, if at all known, can provide a guideline here but have to be used with caution. While a clearly Ghanaian name is a relatively reliable indication that the speaker qualifies for inclusion in the corpus, names such as Sam(my) Morris or James Malcolm are ambiguous since it was, and is, not uncommon for Ghanaians to have English names. In some cases, biographical research helps in establishing the identity of such individuals, as in the case of J. B. Millar, the first director-general of the GBC (1954–1960), an Englishman seconded to Ghana from the BBC, who was accordingly excluded from the corpus.

Although I was able to establish the names and sometimes the socio-biographies of a number of the speakers, there are some individuals where this background information is still lacking and where further research is required. It is these anonymous speakers that represent the most serious challenge to the representativeness of the audio material used for the present study: excluding them would reduce an already small corpus in size and internal diversity and therefore run the danger of overlooking (some aspects of) variation that was actually present in early GhE. On the other hand, leaving these speakers in the corpus runs the danger of including the English of non-Ghanaians. The problem is less acute if the individual has an identifiable Ghanaian accent, but the more serious question is how to proceed with individuals who have a British or near-British accent. A British accent alone cannot not be a criterion for removing a speaker from the corpus. It may well have been the case that (some speakers’) early GhE was closer to British English (BrE) than today's GhE, so excluding these individuals would result in making early GhE more ‘Ghanaian’ than it actually was. A case in point is that of a Mrs Simpson, who, we are told, was the wife of the Solicitor-General of Ghana. In the 1957 recording she speaks English with an RP accent and unmistakably British intonation but on the other hand clearly identifies herself with Ghanaian culture, when sending her greetings to the women of Malaya:

Judging by her accent, it seems clear that Mrs Simpson acquired her English in Britain, or at least in direct contact with RP speakers, but there is no clue in the recording itself whether she was British or Ghanaian. Excluding her from the corpus on the basis of her accent alone would be tantamount to data-streamlining and making early GhE more Ghanaian than perhaps it was: note that, in addition to the fact that the norm at the time around independence was still exogenous, many prominent Ghanaians were trained abroad, notably in Britain. It was the linguistic reality of the Ghana of fifty to sixty years ago that the Educated GhE speech community included a fair number of Ghanaians with a (near-)BrE accent, and to a certain degree this is still true today. Biographical and archival research is needed in cases such as that of Mrs Simpson, in order to establish whether a speaker is Ghanaian in spite of the British accent and/or name, in which case the recording should not be excluded from the analysis.Footnote ⁹

The GBC Sound Archive's index cards can be of help in identifying speakers. They provide information on the recordings, such as the languages used, the place and date of recording, reporters’ and other speakers’ names and short summaries of the recordings. The archive also has complete scripts of some broadcasts. However, all this documentation still awaits digitisation and was not available for this study. For the present purposes it was therefore decided to leave anonymous speakers in the corpus unless there were strong reasons for doubting that they were Ghanaians. Also note that since identification based on voice quality alone is difficult because of the varying quality of recordings, it may well be that some contributions listed as coming from different anonymous reporters in Table 18.1 (Anonymous 1–10) were actually made by the same speaker. As the present study does not consider inter-speaker variation, this does not impair the results.

Apart from the considerations regarding the identity of speakers and the representativeness of the corpus, the quality of recordings themselves also poses a challenge for linguistic analyses. While in most cases transcribing the material proved possible, technical limitations of the original recordings (e.g. clipped sound), recording practices (ambient noise during recording, including studio recordings with background noise from neighbouring booths etc.), editing practices (voice-overs, e.g. simultaneous translations) or the deterioration of the analogue media before they were digitised (e.g. dull or distorted sound) have impaired the sound signal in a number of files. There are cases where the sound is so muffled or distorted that transcription and auditory analysis are hopeless tasks. In other cases, individual passages could not be transcribed, and in yet others the transcriptions can only be of a tentative nature. Tokens in such passages were ignored in the following analysis, as were tokens whose specific realisation was unclear, for example because of superimposed ambient noise.

The relatively small size and particular composition of the corpus impose certain limitations on linguistic analyses. For example, the corpus is too small to yield enough tokens for a quantitative analysis of a good number of morphosyntactic variables. Also, since the corpus mostly contains scripted monologues, i.e. material that was written to be spoken, we can expect relatively few deviations from the British norm. This is because nativisation in African Englishes proceeds more quickly on the level of phonetics/phonology than in morphosyntax, the latter being more exonormative in its orientation towards BrE, particularly in the written mode (Brato and Huber Reference Brato, Huber and Hickey2012: 181). In spite of these limitations, 10,000 words are enough for a simple variationist study of frequent variables such as phonemes. Because of the varying sound quality of the recordings, it was decided to perform an auditory analysis, as this often allows for the identification of particular realisations of variables even if the sound signal is not pristine.

18.4 The Development of Four Phonemic Variables in Ghanaian English

The following is an exploratory auditory analysis of the corpus of early educated spoken GhE. Four phonological variables were chosen for the study, based on the variation they show between more ‘British’ and ‘Ghanaian’ variants in present-day GhE: (ing), (wh), (nurse) and (strut). These variables will be investigated for the distribution of individual variants, and the results will be compared with their distribution in present-day GhE. The aim is to determine whether during the nativisation of GhE since independence we can observe an increase of Ghanaian variants at the expense of a pronunciation that is closer to Standard BrE. In this study, the corpus of early educated spoken GhE is taken to reflect the pronunciation at around 1960, while the present-day GhE pronunciation is determined on the basis of structured sociolinguistic interviews mainly conducted during a students’ excursion to Ghana in 2008.Footnote ¹⁰ To ensure comparability of the data, only tokens from the reading passage and word list modules were considered here, as they come nearest to the scripted monologues in the 1960 corpus. Table 18.2 shows the speakers’ L1, gender, age and educational background (S = secondary school leaver, U = university student/graduate) at the time of recording (2008, 2010).

18.4.1 Variable 1. (ing): [iŋ] ∼ [in, ĩ]

The realisation of /-ɪŋ/ as [-ɪn] in post-tonic -ing (e.g. in swimming) has been described for many English accents and is ‘practically universal’ in varieties of English worldwide (Schneider Reference Schneider2004: 1124).

According to Wyld (Reference Wyld1921: 289), the alveolar nasal in [-ɪn] for -ing has a long history and at one time was common in almost all varieties in England. In the 1820s, [-ɪŋ] emerged as a spelling pronunciation and developed into a prestige variant. At the time that Wyld was writing (late 1910s), the older [-ɪn] was still common in the dialects of ‘the South and South Midlands, and among large sections of speakers of Received Standard English’, but the new variant [-ɪŋ] had ‘a vogue among the educated at least as wide as the more conservative one with -n’. At the beginning of the twenty-first century, [-ɪn] has become even more restricted geographically and socially in Britain. It is today found ‘in Northern and West Midland English as a stigmatised feature’ (Upton Reference Upton2004: 1073).

With regard to present-day GhE, Huber (Reference Huber2004: 858) observes that ‘RP /-ɪŋ/ in progressives or deverbal nouns is more often than not replaced by [-ɪn], cf. morning [mɔnin], leading [lidin], the meeting [dɛ mitin]’. This is corroborated by Adjaye (Reference Adjaye2005: 193, 194), who finds that in her 1980s data from Ewe, Ga and Akan L1 speakers, /ŋ/ ‘is commonly pronounced by Ghanaians as … an alveolar nasal, [n], in … the -ing post-tonic’. Likewise, Koranteng (Reference Koranteng2006: 242) says that in her data, collected in 2001, ‘the “ing” form is in almost all instances realized as [ɪn]’. In most varieties of English today, [-ɪn] is the informal, non-standard variant, while [-ɪŋ] is the standard pronunciation. In GhE, however, ‘[-ɪn] cuts across all socioeconomic and educational levels. It is therefore not necessarily a marked or stigmatized form’ (Adjaye Reference Adjaye2005: 194).

Simo Bobda (Reference Simo Bobda2003: 19) proposes that West African [-ɪn] ‘was brought, or reinforced, by the eighteenth century African Americans who settled in Sierra Leone’ and Adjaye (Reference Adjaye2005: 194) speculates that it ‘would be fair to say that in Ghana, it was the [-ɪn] version that became established from the onset’. That [-ɪn] must already have been common in early twentieth-century West African English can be inferred from the corrective in Harman (Reference Harman1931: 127), who taught in Nigerian and Gold Coast schools in the 1920s: ‘Avoid replacing ŋ by n in such words as something, talking …’.

GhE [-ɪn] can participate in the optional process of syllable-final /n/-elision with compensatory nasalisation of the preceding vowel (Adjaye Reference Adjaye2005: 191; Huber Reference Huber2004: 857; a process already observed in 1960s GhE by Strevens Reference Strevens and Strevens1965: 114), so that meeting has the following GhE pronunciations:

1. (1) meeting [mitiŋ ∼ mitin ∼ mitɘ]

For the present study, a binary variable was constructed, contrasting [ŋ]-realisations (the British standard pronunciation) with realisations from which [ŋ] is absent (the ‘Ghanaian’ pronunciation):

1. (2) (ing): [iŋ] ∼ [in, ĩ]

Tokens in neutralising contexts were ignored, i.e. when the word following -ing had an initial velar plosive /k, g/ (e.g. wedding gifts), because this can trigger assimilation of [n] to [ŋ]. Figure 18.1 shows the ratio of the two variants in 1960 and 2008.

Figure 18.1 Realization of GhE (ing), 1960 and 2008. (χ² = 103.164, df = 1, p_two-tailed < 0.001***, p_sim < 0.001***; G = 106.028, df = 1, p_two-tailed < 0.001***).Footnote ¹¹

There was a strong (42%) and highly significant increase of the Ghanaian variant [in, ĩ] in just fifty years. While in 1960 the British standard variant clearly dominated, with almost 70% of the tokens realized as [iŋ], it has become a minority variant by the beginning of the twenty-first century, accounting for only 27% of the realisations. That is, in half a century a complete reversal has taken place in the ratio of the two variants.

Nevertheless, the data suggest that the position of the British standard variant [iŋ] is and was stronger in GhE than proposed by Adjaye (Reference Adjaye2005: 194), who appears to imply that [ŋ] has been so marginal ‘from the onset’ that it ‘does not have a phonemic status except in the speech of a few individuals’.Footnote ¹² Figure 18.1 reveals that [in, ĩ] only became dominant in post-independence Ghana. This is also corroborated by Sey's (Reference Sey1973: 152) observation that in late 1960s GhE there was variation between [-iŋ] and [-in]: ‘killing, singing, etc., may be pronounced /kilin/ and /siŋin/’ (my underlining). If it is legitimate to extrapolate to the past from Figure 18.1, the Ghanaian [in, ĩ] realization in all probability only existed as a minority or marginal variant in the educated GhE of the first half of the twentieth century.

18.4.2 Variable 2. (wh): [w] ∼ [^hw, ʍ, ^hʍ]

The pre-aspiration [^hw] and/or devoicing [ʍ, ^hʍ] of orthographic <wh> is generally in decline in varieties of English. On the whole, /ʍ/ or its variants tend to be restricted today to rural, non-standard or conservative varieties and older speakers. In the British Isles, /ʍ/ is attested in Scottish Standard English (Stuart-Smith Reference Stuart-Smith2004: 61) and in Shetland (Melchers Reference Melchers2004: 42). In Ireland, Hickey (Reference Hickey2004: 78, 81, 92–93) finds /ʍ/ in conservative Ulster Scots, Popular Dublin English, the rural south-west and west and variably in supraregional Southern Irish English. In American English, ‘dialects have retained a historically older consonant cluster with an initial velar fricative [x] before the approximant [w], so that, unlike many mainstream varieties of English, which is not homophonous with witch’ (Schneider Reference Schneider2004: 1086). For varieties of English in the Pacific and Australasia, Burridge (Reference Burridge2004: 1093) similarly remarks that ‘[a]s in other parts of the English-speaking world the distinction between /w/ and /hw/ has virtually disappeared … The /hw/ cluster is preserved only for the most conservative speakers of these varieties …’.

Pre-aspiration/devoicing of <wh> is a feature that has hitherto gone largely uncommented in African Englishes in general and GhE in particular.Footnote ¹³ To my knowledge, Gyasi (Reference Gyasi1991: 27) is the first to mention that in GhE ‘[t]he voiceless glottal fricative /h/ is clearly heard in where, when, white and wheat’. In the analysis of her 1980s data, Adjaye (Reference Adjaye2005: 208) says that ‘wh- words [are] usually pronounced [hw-] in GhE’ and Koranteng (Reference Koranteng2006: 249) finds that ‘in words with spelling “wh” … there is the use of [hw-] or [w] alternatively’.

Figure 18.2 indicates the distribution of the voiced [w] and pre-aspirated/unvoiced variants [^hw, ʍ, ^hʍ] of <wh> in the 1960 and 2008 corpora. <who-> spellings that generally lack a /w/ like who or whole were omitted from the analysis.

Figure 18.2 Realisation of GhE (wh), 1960 and 2008. (χ² = 8.849, df = 1, p_two-tailed < 0.003**, p_sim < 0.004**; G = 8.808, df = 1, p_two-tailed = 0.003**).

We can observe a very significant increase of [^hw, ʍ, ^hʍ] from about 29% in 1960 to about 49% in 2008. As with (ing), it is the ‘Ghanaian’ variant that has risen in the half-century following independence. Regarding its origin, Huber (Reference Huber2004: 861) submits that this ‘is another feature that could have its historical origin in Scottish [possibly nineteenth/twentieth century missionary] influence in the Gold Coast, reinforced by spelling pronunciation’. Again, while the voiceless/pre-aspirated variant may have been present in early twentieth-century GhE, Figure 18.2 suggests that it must have been in the minority.

18.4.3 Variable 3. (nurse): [ɜ] ∼ [ɛ]

The /ɜː/ > /ɛ/ substitution in the nurse lexical set is a characteristic that clearly distinguishes GhE from other African Englishes (Huber Reference Huber2004: 851), where /ɜː/ > /ɛ/ is more restricted: most other Englishes in Africa replace the nurse vowel in <or, our, ur, ir> spellings with /ɔ/ or /a/ (Simo Bobda Reference Simo Bobda2003: 22, 28–29), but in Ghana /ɛ/ is general for all spellings representing the nurse vowel.

For the Nigerian and Gold Coast English of the 1920s, Harman (Reference Harman1931: 57) records the following realisations of nurse: [ɛ], [ɑ], [ɒ] and [ɔ]. In the literature on GhE, /ɜː/ > /ɛ/ was first mentioned by Schachter (Reference Schachter1962: 18–19), who observed this substitution in speakers with a Twi L1 background. A decade later, Sey (Reference Sey1973: 145–147) called the phenomenon ‘extremely common’ and proposed possible BrE dialectal or American English influence. In her mid-1980s data, Adjaye (Reference Adjaye2005: 141–144) found the following variants: fronted [ɛ, ɛ̈, ɛ̝] c. 70%, central [ɜ] c. 21%, backed [ɔ, o] c. 5%, [a] c. 4%. Koranteng (Reference Koranteng2006: 161–162) has similar proportions in her 2001 data: fronted [ɛ] c. 79%, central [ɜ] c. 19%, backed [ɔ] c. 2%.

Simo Bobda (Reference Simo Bobda2000b: 190) hypothesises that ‘the systematic occurrence of /ɛ/ for RP /ɜː/ is fairly recent. It must have supplanted an earlier /ɔ/ for the graphemes <or, ur, our>, as attested by data collected from Ghanaian speakers of the old generation’. According to Simo Bobda (Reference Simo Bobda2003: 33–34), /ɔ/ is ‘fading out in Ghana as the substitute for the nurse vowel with <or, our, ur>’ and has ‘over a generation’ almost entirely been replaced by /ɛ/. He judges this replacement to be a conscious divergence occasioned by Ghanaians’ belief that they speak better English than other West Africans.

Figure 18.3 shows the ratio of the fronted [ɛ] and central [ɜ] variants of (nurse) in the 1960 and 2008 corpora.

Figure 18.3 Realisation of GhE (nurse), 1960 and 2008. (χ² = 8.541, df = 1, p_two-tailed = 0.003**, p_sim = 0.004**; G = 8.418, df = 1, p_two-tailed = 0.004**).

In the half-century following independence, there has been a moderate (c. 17%) but very significant rise of the ‘Ghanaian’ fronted variant [ɛ]. [ɔ], marginal in Adjaye's (Reference Adjaye2005) and Koranteng's (Reference Koranteng2006) data, is not attested in my corpora, neither in 1960 nor in present-day GhE. The absence of [ɔ] and rather strong presence of [ɛ] in the 1960 corpus call into question Simo Bobda's (Reference Simo Bobda2000b, Reference Simo Bobda2003) assertion that [ɛ] is a recent innovation, having emerged ‘over a generation’, i.e. from around 1970. It is true that [ɛ] was not the majority variant in 1960, but it still accounted for about 28% of the nurse tokens.Footnote ¹⁴ If the rise of fronted [ɛ] was more or less constant, the rates in Figure 18.3 suggest that it must already have existed at the beginning of the twentieth century.

18.4.4 Variable 4. (strut): [ɛ] ∼ [a] ∼ [ʌ] ∼ [ɔ]

The realization of the strut vowel is a further distinguishing feature of GhE. While other Englishes in West Africa usually replace RP /ʌ/ by /ɔ/, GhE shows a lot of variability, including fronted, central and backed variants: [ɛ] ∼ [a] ∼ [ʌ] ∼ [ɔ].

That strut can be realised as [ɛ] was first noticed around 1890 by a Cape Coast missionary, who reported that missionary school pupils pronounced butter like better and Hull like hell (Kemp Reference Kemp1898: 179). Harman (Reference Harman1931: 53–54) suggested that in 1920s West African English, [ɛ] was a hypercorrection of the common realisation of strut as [ɔ], with speakers ‘push[ing] the tongue well forward to avoid making ɔ’. Thirty years later, Schachter (Reference Schachter1962: 18) pointed out that Twi L1 speakers realise RP /ʌ/ as [ɔ]. Similarly, with reference to the GhE of the late 1960s, Sey (Reference Sey1973: 145, 147) described [ɔ] as ‘widespread’ and [ɛ] as a minority variant, ‘common in the Cape Coast area’. He did not specifically mention centralised realisations, but since he described GhE as a system ‘of tendencies rather than specific Ghanaian usage’, we can assume that [ʌ] was present, too. In addition, we can deduce from Sey's (Reference Sey1973: 147) observation that ‘/ʌ/ does not occur in L1 but the most likely substitute for it would be /a/ and not /ɔ/ or /ɛ(ː)/’ that the [a] variant must have been marginal. Gyasi (Reference Gyasi1991: 27) only mentions /ɔ/ as a possible realisation of strut. Interestingly enough, [a] has a much stronger position in Adjaye's (Reference Adjaye2005: 71–75) mid-1980s recordings: [ɛ] c. 13%, [a] c. 35%, [ʌ] c. 16%, [ɔ] c. 36%. Koranteng's (Reference Koranteng2006: 137–145) 2001 data show the following proportions: [ɛ] c. 6%, [a] c. 18%, centralised [ʌ, ä, ə] c. 55%, backed [ɔ, ɒ] c. 26%. The great variation between these figures (and the ones presented in Figure 18.4) may well be due to the fact that an auditory analysis of naturally occurring speech is always subject to perceptional differences between researchers and that it is not always easy to distinguish between [a] and [ʌ].

Figure 18.4 Realisation of GhE (strut), 1960 and 2008. (χ² = 174.002, df = 3, p_two-tailed < 0.001***, p_sim < 0.001***; G = 169.705, df = 3, p_two-tailed < 0.001***).

Simo Bobda (Reference Simo Bobda2000b: 188) observes that ‘only a generation or so ago, Ghana had /ɔ/ for cut, just, mother, done, etc. like the rest of West Africa’ and submits that the realisation of the strut vowel as [a] or [ɛ] is another change that has only recently occurred in GhE, again as a result of a conscious effort on the part of Ghanaians to dissociate their accent from that of other anglophone West Africans (Simo Bobda Reference Simo Bobda2003: 33–34; Reference Simo Bobda2000a: 258fn).

The development of the four (strut) variants in the 1960 and 2008 corpora can be seen in Figure 18.4. Tokens were identified based on the pronunciation as suggested by the OED. This includes function words like but, does, us or some, whose citation form contains an /ʌ/, because these tend to be realised with a full vowel quality in GhE rather than a schwa, even in unstressed positions.

Figure 18.4 shows that there has been a highly significant and considerable (31%) increase of [a] at the expense of central [ʌ] and back [ɔ] realisations of strut from 1960 to 2008. [ɛ] has remained marginal and, as the only variant, has seen no significant change. The findings agree with Simo Bobda's observation that [a] is spreading, but they contradict his hypothesis that it has only recently replaced older [ɔ]. In the 1960s corpus, [ɔ] already is a minority variant (21%), while 50% of the tokens are realised as [a]. This means that the rise of [a] must have started much earlier than supposed by Simo Bobda. Based on apparent time evidence, Huber (Reference Huber2004: 854) proposes that it must have begun in the 1930s. Figure 18.4 also shows that, contra Simo Bobda, [ɛ] is not a particularly new variant either. In fact, the missionary's observation, quoted above, confirms that it was already around in the late nineteenth century.Footnote ¹⁵ Note also that the RP variant [ʌ], still found in about 26% of the tokens in 1960, has all but vanished in the 2008 data (c. 3%).

18.5 Conclusion

Because of the lack of data, the actual structural development of postcolonial Englishes has not been studied much hitherto. The recordings from the GBC Sound Archive thus represent invaluable data from the period just after Ghana's independence in 1957. For the first time, the availability of a sufficient amount of real-time data makes it possible to study the diachronic phonological development of GhE in a variationist framework.

The comparison of the development of four phonological variables in spoken educated GhE at around 1960 and 2008 has shown that there have been considerable changes over the last half-century or so. In both the 1960 and the 2008 corpus, each of the variables has British (RP) variants but also alternative, more nativised Ghanaian ones. The analysis has shown that in all four variables, the RP variant has receded over time – (ing): [ɪŋ] by 42%, (wh): [w] by 20%, (nurse): [ɜ] by 17% and (strut): [ʌ] by 23%. If this is representative of the development of the GhE phonological system as a whole, it suggests that spoken educated GhE was closer (but not identical) to RP at the time of Ghana's independence and that Ghanaian variants have gradually been replacing the British ones over the past fifty to sixty years. That is, we can observe a development from a more exonormatively to a more endonormatively oriented variety. While this is what would be expected during the Nativisation Phase in the evolution of a New English (following Schneider Reference Schneider2003, Reference Schneider2007), it was also shown that all variants that are characteristic of GhE today were already in existence in 1960 (even if they were not the majority variants then), in some cases considerably earlier than assumed in the literature. Simo Bobda (Reference Simo Bobda2000b: 190, Reference Simo Bobda2003: 33–34; Reference Simo Bobda2000a: 258fn) suggested that in the case of nurse and strut, GhE has developed away from the pronunciation patterns of other West African Englishes since about 1970 as Ghanaians dissociated themselves linguistically from other anglophone West Africans. Attitudes towards other West Africans’ English may have furthered the spread of the characteristically Ghanaian variants, but as this study has shown, they were certainly not the reason for their emergence. If Simo Bobda's dissociation theory is correct, a possible scenario may have been that Ghanaians selected and promoted those variants that were most dissimilar from the dominant realisation in other West African Englishes.

All in all, the four variables have undergone considerable change, attesting to rapid phonological restructuring during the nativisation of GhE. The growing acceptance of an endonormative model as a corollary of linguistic emancipation following political independence as well as the dissociation from other West African pronunciation practices may have played a role in this. Another factor in the spread of distinctly Ghanaian variants may have been the 1951 Accelerated Development Plan for Education, which aimed at providing free basic education for all children and led to the establishment of a large number of new primary schools. The resulting staff shortage was alleviated by employing primary school leavers as teachers, dramatically decreasing the percentage of trained teachers in primary and middle schools from over 52% to 28% (Inkoom Reference Inkoom2012: 10). Together with the early transition from African languages to English as the medium of instruction, this may have led to an endonormative shift, the Ghanaianisation of pronunciation (see also Boadi Reference Boadi and Spencer1971: 56; Huber Reference Huber2004: 844).

Diachronic phonological studies of early recordings of postcolonial Englishes can provide much-needed data to test and refine evolutionary models that so far have been based mainly on external language history and synchronic structural data. Once more early spoken GhE data becomes available, a more sophisticated study of the linguistic and extralinguistic factors determining the choice of individual variants can be conducted, providing insights into possible restructuring processes within linguistic variables during nativisation.