2 The nature of the data

The present description of BE is based primarily on my observations of spontaneous BE use made during my joint work on the description of Bena L1 together with Mark Van de Velde and complemented by a number of recordings where BE is used spontaneously but was not the primary object of the recording. Given the lack of a substantial and varied recorded spoken corpus of BE that would allow for a variationist type of analysis, I provide a more detailed account of the nature of the data discussed.

The data considered in this article are the best that are currently available and, due to ongoing security issues, the best that are likely to be available for some time to come. The data were collected over a six-month period of fieldwork on Bena in Nigeria. Because of the ongoing security issues in northeastern Nigeria, I was able to work in the Bena community in situ in the village of Dumne for somewhat less than a month. During my stay in the community, I had the occasion to observe spontaneous BE use by dozens of speakers of different ages (ranging from teenagers to elders) and of different educational and professional backgrounds (ranging from secondary school students, farmers, university students to teachers, government officials and a former senator). Due to cultural restrictions, the speakers were almost all male. Therefore, we cannot exclude the possibility that the phenomenon is restricted to male speech. The situations of BE use were mostly natural conversations with Bena L1 speakers involving me and my colleague, some more formal events, as well as several dozens of hours of interview sessions whose subject was the description of Bena L1.

The remaining part of the fieldwork took place in southwestern Nigeria, far from the Bena community; there were three different consultants, all males in their early 30s from the village of Dumne, who had completed their secondary education and two of whom had also spent some time at university. My colleague and I shared an apartment with the consultants, which resulted in many hours of informal interaction with them in English every day, in addition to the interview sessions on the description of Bena L1.

My colleague and I observed that most speakers produced occasional examples of pre-pausal t epenthesis. We had the impression that in the speech of speakers with better overall English proficiency and university-level education, pre-pausal t epenthesis was less common, and for some such speakers possibly even absent altogether. Our three main consultants, whose BE speech I was exposed to for several months, produced such examples on a daily basis. The one consultant who did not have university-level education was most prolific in this respect. At the same time, my attempts to elicit examples of the epenthesis with our three main consultants were not successful. Although, as I argue in section 5.3, there is a reasonable explanation for the problems I had with the elicitation, the existence of a sufficiently large and varied corpus of spontaneous BE speech would be needed to quantitatively substantiate my qualitative observations. Unfortunately, the Bena-speaking area is currently inaccessible for security reasons.

Neither I nor my colleague had been aware of the pre-pausal s epenthesis and the phrase-internal t or s epenthesis before I started searching our Bena recordings for examples of t epenthesis in BE. That the phrase-internal t and s epenthesis remained unnoticed is likely to be due a combination of factors. First, it is rare, as I found only a few examples in the recordings and I observed it directly in spontaneous speech on only one occasion (several repetitions of example (2a)). Second, its phrase-internal position may negatively affect its perceptual prominence, both because the consonants in this position have shorter duration, especially when compared to the position before a pause, and because they run the risk of being masked by a following consonant. As to our lack of awareness of the s epenthesis, both in the pre-pausal and the phrase-internal position, another factor is that it is too easy to filter out occurrences of an epenthetic [s] as morphological mistakes, for instance by assuming that plural -s was used inappropriately, as is not uncommon in NE L2. For all these reasons, the discussion in the rest of the article will be mostly confined to pre-pausal t epenthesis.

My corpus of recordings contains 72 hours of interview sessions on the description of Bena L1. However, only a small part represents spontaneous BE use, for instance when the consultants volunteered an explanation in English. For the moment, I have been able to locate around 30 minutes of stretches of spontaneous BE use in these recordings which resulted in 14 cases of epenthesis: 10 before a pause (6 for [t] and 4 for [s]) and 4 phrase-internally (1 for [t] and 3 for [s]). Although in absolute terms these numbers are rather small, they suggest a relatively high rate of epenthesis, with one occurrence every 2–3 minutes of spontaneous speech. The number of non-recorded cases of at least pre-pausal t epenthesis informally observed by my colleague and me is several times higher.

The examples of t epenthesis in written BE come from a number of email exchanges with our Bena consultants and from a few pieces of written English that the consultants have produced on a number of occasions, such as writing down the English translations of Bena words and sentences during elicitation sessions on Bena L1. I have not taken into consideration written English produced using any text-processing software as it is more likely to be distorted by the availability of a spell-checker. As I have not collected the written pieces produced by the consultants consistently, I can give only rough estimates for the written corpus. In total, the written data do not exceed a couple of thousand words. I would estimate that I encountered up to 10 occurrences of clause-final t epenthesis and no examples of other types of epenthesis.

3 Bena English (L2) consonants

This section provides an overview of the consonant systems of BE as L2. As expected for an L2 system, there is much variation in how well the reference native pronunciation of L2 forms is approximated, both between speakers with different levels of proficiency and within speakers depending on various sociolinguistic parameters involved in a given speech situation. Therefore, the system presented here necessarily represents an approximation of what I judge to be the typical speech patterns of a BE speaker with an average level of L2 proficiency, such as the Bena consultants I worked with on Bena L1. The description is based on my observations and recordings of spontaneous BE use.

The consonants found in word-initial position after a pause or the final vowel of a preceding word are summarised in table 1. In general, as in Bena L1, the aspiration of voiceless stops in BE is moderate,Footnote ⁶ but more extreme realisations closer to the native English pattern are also possible. Utterance-initially, vowels are often pre-glottalised as in Bena L1. In careful speech, this prevocalic glottalisation may be preserved in phrase-internal position. It is possible that the sound classified as voiceless labial-velar approximant [ʍ] in table 1, as in what?, may be more accurately transcribed with an initial non-labialised phase as [hʍ] or [h]. The phonological status of the glottal stop [ʔ] is not clear. I do not know whether [ʒ] occurs in this position.

Table 1.

Word-initial consonants in BE (before a vowel and after a pause or the final vowel of a preceding word)

Table 2 summarises the word-final consonants as found before a pause. Like other varieties of NE, BE is generally non-rhotic.Footnote ⁷ Even though Bena L1 does have a velar nasal, in BE the etymological velar nasal tends to be replaced by the alveolar nasal stop [n], as is typical for NE in general (see Gut Reference Gut and Mesthrie2008: 45). However, in BE the preservation of [ŋ] seems to be largely dependent on stress. Thus, [ŋ] is consistently replaced by [n] in BE in words such as evening, morning, nothing, looking, where the relevant syllable is not stressed (similarly to general English ‘g-dropping’ in unstressed syllables), while [ŋ] is often preserved in words such as in king, young, thing or belong, where the relevant syllable is stressed. BE is characterised by pre-pausal devoicing, since only voiceless stops, fricatives and affricates occur word-finally before a pause, resembling the Bena L1 pattern. Unlike in Bena but like in more standard varieties of NE, as well as British RP English (Buizza & Plug Reference Buizza and Plug2012), pre-pausal stops normally have audible release which may be strong enough to qualify as aspiration (figure 1a), and pre-pausally, occasionally even as a slight affrication (figure 1b). At times, pre-pausal oral stops are glottalised as in Bena but unlike in more standard varieties of NE, viz. [], [], []. These glottalised pre-pausal stops may have an audible release, as illustrated in figure 2, and sometimes this release is nasal. As becomes apparent from the comparison of figures 1a, b and figure 2, the aspirated and affricated pronunciations are characterised by a much shorter closure duration, similar to what one finds in more standard varieties of NE, whereas the closure duration of the glottalised variant is comparable to what one finds for pre-pausal glottalised stops in Bena.

Table 2.

Word-final consonants in BE before a pause

Figure 1.

(a) right with aspirated [tʰ] (b) (re)gulate with affricated [t^s]

Figure 2.

is that with an audible release of []

Pre-pausal continuants, nasals and the lateral in BE appear to be realised largely as they are in Bena. Thus, as in Bena, they may be considerably lengthened, like [f] of have in figure 3.

Figure 3.

[f] of have

With nasals and [l] the fall in intensity is not continuous from the vowel into the consonant but tends to occur in one or two steps composed of a brief period of slightly rising or flat intensity curve and a subsequent significant downtrend, as illustrated in figures 4 and 5. In other words, although strictly speaking nasals and [l] are not devoiced, the overall decreasing trend in their acoustic energy that sets in very early may, from the acoustic point of view, be considered analogous to the full lack of voicing observed with the stops and fricatives.

Figure 4.

girl [geəl]

Figure 5.

demon [demɔn]

Furthermore, like other consonants, pre-pausal continuants, nasals and the lateral are occasionally followed by a glottal closure and an audible release, as illustrated in figures 6 and 7, where the presence of glottalisation is also supported in the acoustic record by a more abrupt final fall in the intensity curve without the second step observed in figures 4 and 5. Such glottalisation of pre-pausal continuants, nasals and the lateral is also found in Bena, where it seems to be more common.

Figure 6.

people [pʰipʰʊlʔ]

Figure 7.

skin [skʰinʔ]

Many types of word-final consonant clusters, but not word-initial or word-internal ones, are regularly simplified in BE irrespective of the phonological context to the right of the word in question, largely in the same way as in more standard varieties of NE (see Simo Bobda Reference Simo Bobda2007; Gut Reference Gut2009a, Reference Gut2009b; Ugorji Reference Ugorji2010) and Nigerian Pidgin English (see Faraclas Reference Faraclas1996). One strategy involves insertion of a non-etymological vowel, mostly for clusters involving a lateral approximant and a non-coronal consonant, as in [pʰipʰʊlʔ] ‘people’ in figure 6 (assuming that ‘people’ has the structure /pipl/). In clusters with a final stop homorganic to the first consonant or with more than two consonants involved, consonant deletion is the regular strategy. It is normally the final consonant that drops out, as in spen(t) in (1g) and husban(d) in (1h). Other types of clusters normally remain intact, e.g. clusters involving non-homorganic fricatives and stops, as in [aks] ‘ask’, [riks] ‘risk(s)’ (also showing a metathesis of the etymological sk sequence), [waps] ‘wasp’ (also showing a metathesis of the etymological sp sequence), [teks] ‘text(s)’, left, and clusters with a nasal, lateral or stop followed by a homorganic fricative (viz. [s] or [ʃ]) or affricate (viz. [ʧ]), as in else, prince Footnote ⁸ or AIDS.

Table 3 summarises the word-final consonants in contexts other than before a pause. We find in table 3 the same consonants as in the pre-pausal context in table 2 with the additional possibility of voicing for the stops and fricatives (I do not know whether the voiced affricate [ʤ] and the voiced fricative [ʒ] are possible in this position).

Table 3.

Word-final consonants in BE not before a pause

4 Bena (L1) consonants

This section provides an overview of the consonant systems of Bena as L1 (see also Van de Velde & Idiatov Reference Van de Velde, Idiatov and Kaji2017). One consonant, the labiodental flap [ⱱ], occurs only word-internally and therefore does not show up in tables 4–5.

Table 4.

Word-initial consonants in Bena (before a vowel and after a pause or the final vowel of a preceding word)

Table 5.

Word-final consonants in Bena before a pause

The consonants found in word-initial position after a pause or the final vowel of a preceding word are summarised in table 4. The aspiration of voiceless stops is moderate. Utterance-initially, vowels are often pre-glottalised. In careful speech, this prevocalic glottalisation may be preserved in phrase-internal position. Utterance-initially, all voiced stops, with the exception of the labial-velar stop but including implosives, and especially voiced fricatives may be partially or fully devoiced. The approximants and nasals are not strictly speaking devoiced but can start with a period of significantly reduced acoustic energy, which in the acoustic record looks largely like the mirror image of what happens word-finally before a pause and has been described in section 3 for BE. As a result, the nasal central approximants [] and [] may sound like nasal occlusives [ɲ] and [] respectively.

Table 5 summarises the word-final consonants found before a pause. Bena is characterised by pre-pausal devoicing. Furthermore, pre-pausal glottalisation is regular with stops. The glottal closure may have a weak audible release. If the oral closure is bilabial, the audible release is nasal. After oral stops and nasals with alveolar or velar closure, it can be nasal or non-nasal. In some words ending with an alveolar or velar oral stop that typically occur clause-finally and therefore also utterance-finally, such as ideophones and adverbs, the oral closure, especially the velar one, may be audibly released just before the glottal closure resulting in a weakly ejective sound, as in [átʰáátʰákʼ] ‘bitter’ in figure 8.

Figure 8.

[átʰáátʰákʼ] ‘bitter’ with a weakly ejective release of a pre-pausal glottalised velar stop

Furthermore, occasionally, pre-pausal glottalisation is lacking with these words, as in [sākʰāt⁽ʰ⁾] ‘quickly’ figure 9, or at least reduced in duration, as in [sākʰā] ‘quickly’ figure 10. The burst of the audible release of the oral closure is then usually weak but may be followed by noisy airflow that sometimes sounds like slight aspiration.

Figure 9.

[sākʰāt⁽ʰ⁾] ‘quickly’

Figure 10.

[gàw sākʰā] ‘come quickly!’

Pre-pausal glottalisation also occurs with continuants, nasals, [l] and vowels, as illustrated in figures 11 and 12 for [sʔ] and [nʔ] respectively. However, given that an audible release may be lacking, it is difficult to judge how common it is without articulatory data.

Figure 11.

[(bí)tʰùsʔ] (a proper name) with an audible release of [sʔ]

Figure 12.

[ɗēnʔ] ‘tomorrow’ with an audible release of [nʔ]

Similarly to what was described in section 2 for BE pre-pausal sonorants [m], [n] and [l], in Bena even in the absence of an audibly released glottal closure, before a pause these sonorants tend to show the same early decreasing trend in their intensity that from the acoustic point of view may be considered analogous to the full lack of voicing observed with other consonants.

Consonant lengthening is another important phonetic property associated with the pre-pausal position.Footnote ⁹ With continuants, nasals and the lateral such a lengthening is apparent in figures 11 and 12 and is also illustrated in table 6 by comparing two instances of [nn] ‘1sg.ipfv’, one phrase-internal, before a following [s] and one before a pause in a word-by-word repetition of the same utterance. A continuant, nasal and the lateral before a pause can be almost twice as long as the same consonant elsewhere. Thus, in table 6 the pre-pausal [n] has the duration of 190 ms, as opposed to the duration of [n] of 105 ms, 115 ms and 123 ms in the other positions.

Table 6.

Phrase-internal vs pre-pausal instance of [nn] ‘1sg.ipfv’

Similarly, (audibly released) pre-pausal glottalised stops have a considerably longer closure than their non-glottalised counterparts and the stops of the same place of articulation elsewhere. Thus, the closure duration of a non-glottalised [t⁽ʰ⁾] in figure 9 is 146 ms as opposed to 179 ms of a (partially) glottalised [] in figure 10 and 234 ms of a more clearly glottalised (and probably weakly ejective) [] in figure 13.

Figure 13.

[pàrà] ‘already’

As I argue in section 6, the phonetic properties of Bena L1 described above, such as pre-pausal glottalisation and lengthening of consonants, allow us to better understand the details of the hypercorrection scenario for the emergence of the BE pattern.

Table 7 summarises the word-final consonants in contexts other than before a pause.

Table 7.

Word-final consonants in Bena not before a pause

There are no (pre-pausal or phrase-internal) word-final consonant clusters in Bena, although consonant clusters are possible across word boundaries and within words, both across syllable boundaries and in syllable onsets. In their citation form, most words end in an open syllable. Final closed syllables are more common in minor part-of-speech classes, such as adverbs. Utterance-internally, the final vowels are often reduced and in some constructions deleted. For instance, this is what happens with the final vowel of the head N1 in a genitive construction N1 N2. When the final vowel is deleted, the result is a word-final closed syllable. At least in the case of the genitive construction, the resulting final consonant is optionally glottalised as in the pre-pausal context.

5 The phonetics and phonology of word-final post-coronal t and s epenthesis in BE

5.2 t and s epenthesis: phonetics

The audible release of a pre-pausal epenthetic [t] is generally weak without any accompanying noisy airflow, as illustrated in figure 14. In this respect, it is comparable to Bena pre-pausal [] rather than [t⁽ʰ⁾]. The audible release can also be nasal, as in three days[] in figure 15, like Bena pre-pausal [].

Figure 14.

horse [hɔst]

Figure 15.

three days [t^əɾi des]

Acoustically, the difference between the audibly released pre-pausal epenthetic [t] in BE and the audibly released pre-pausal glottalised [] in Bena appears to be primarily that of closure duration. Thus, the closure duration of BE epenthetic [t] is much shorter (e.g. 105 ms in figure 14 and 72 ms in figure 15) and comparable to that of a non-glottalised [t⁽ʰ⁾] in Bena or BE, suggesting that it is equally not accompanied by a glottal closure or only marginally so.

Within a word-final consonant cluster, the audible release of the word-final -t that has not been deleted before a pause ranges from a relatively strong one, [t⁽ʰ⁾], similar to a word-final post-vocalic -t (see section 3), to a very weak one, without any accompanying noisy airflow, similar to a pre-pausal epenthetic [t], as illustrated in figure 16 with a pre-pausal instance of the word husband where the release burst in the lexical cluster [nt] is very weak.

Figure 16.

husband [hɔzant]

I observed no obvious acoustic differences between pre-pausal epenthetic [s], phrase-internal word-final epenthetic [t] and [s] and their non-epenthetic counterparts in these contexts in terms of duration, intensity, burst quality and VOT.

Several studies dealing with the phonetic implementation of epenthetic stops as compared to that of their non-epenthetic counterparts in consonant clusters, such as Fourakis & Port (Reference Fourakis and Port1986) for American English, Warner & Weber (Reference Warner and Weber2001) for Dutch and Recasens (Reference Recasens2012) for Valencian Catalan, have argued that the two kinds of stops are produced differently. Such studies also argued that differences in production go along with differences in perception. When no differences in production were observed, no differences in perception were found either, as in Lee (Reference Lee1991) and Yoo & Blankenship (Reference Yoo and Blankenship2003) for American English. At the same time, differences in perception may not translate straightforwardly into differences in categorisation. Thus, in Warner & Weber's (Reference Warner and Weber2001) study epenthetic stops were perceptible to the speakers ‘as tokens of the stop phoneme, a great deal of time’, even though they were still perceived ‘far less often’ and ‘far more slowly’ than their non-epenthetic counterparts. The lack of obvious acoustic differences between epenthetic and non-epenthetic [t] and [s] in BE makes it unlikely that [t] and [s] from the two sources are discriminable in perception. Consequently, it is also unlikely that they differ in their categorisation, especially given that the non-epenthetic counterparts of the epenthetic [t] and [s] in BE optionally undergo deletion themselves. In other words, from a BE learner's perspective the epenthetic consonants are likely to be categorised as an optional part of the phonological representation like their non-epenthetic counterparts that optionally undergo deletion. The main difference is then probably in the lower probability of realisation associated with the epenthetic consonants.

6 The role of hypercorrection in the emergence of the t and s epenthesis in BE

As I have argued in section 5, from the synchronic perspective, it is inadequate to analyse the word-final post-coronal t and s epenthesis in BE as a type of hypercorrection. At the same time, as I pointed out in section 1, it is clear that some kind of compensatory behaviour for the BE tendency for word-final consonant cluster simplification, which we may refer to as hypercorrection, must have played a role in the development of the word-final post-coronal t and s epenthesis in BE. The tendency for word-final consonant cluster simplification is itself most likely related to the absence of word-final consonant clusters in Bena L1. Furthermore, comparative evidence from other varieties of English with word-final consonant epenthesis also points to hypercorrection as the major force driving the development of such epenthesis. The high type frequency of word-final clusters with a final coronal stop as the original phonotactic pattern of English further strengthened by the constraint on voicing of final stops in BE (and several other English varieties) also explains the preference for [t] as the epenthetic consonant.

However, an unconstrained phonological hypercorrection scenario has some trouble explaining the s epenthesis, since the deletion of a word-final post-coronal [s] is not so common in BE (see section 3). Another potentially problematic issue for a simple phonological hypercorrection scenario is the fact that t and s epenthesis in BE is found exclusively (in my dataset) after alveolar coronals, while there does not seem to be such a constraint in other English varieties with a word-final t epenthesis. Even more troubling is the fact that t and s epenthesis occurs mostly before a pause, while the presence or absence of a pause is largely irrelevant for the word-final cluster simplification in BE. Finally, a simple phonological hypercorrection scenario cannot explain why a productive t and s epenthesis is present in BE but absent in the three major varieties of Nigerian English (viz. Yoruba, Igbo and Hausa NE) and Nigerian Pidgin English, which are better studied, even though they have an equally strong tendency for word-final consonant cluster simplification.

To better account for the emergence of the s epenthesis in BE, we need to enrich the phonological hypercorrection scenario by acknowledging the possibility that morphological hypercorrection has also played a role in the development of the BE pattern. Thus, in BE the nominal morpheme -s and especially the verbal inflectional morpheme -s are often lacking even when phonotactically they would be rather uncontroversial, for instance after a vowel, as the third-person singular present suffix -s in (2d), while the clusters where the etymological word-final [s] is lexical often remain intact. The frequent omission of the past marker -ed of the weak verbs might have similarly contributed to the development of the t epenthesis but its role would have obviously been much less crucial. Note that on its own a morphological hypercorrection scenario is largely insufficient, since t and s epenthesis shows no particular part-of-speech constraints and the relevant verbal and nominal morphemes may be also lacking even when their appearance would not have created a word-final cluster.

To account for the other issues that a hypercorrection analysis raises, I argue that we need to call on phonetic properties of Bena L1 such as pre-pausal glottalisation and lengthening of consonants. As described in sections 3 and 4, the pre-pausal glottalisation is realised either as (for all consonants) a glottal closure with an audible release or (for nasals and [l] only) as a fall in the intensity of the sound. The pre-pausal lengthening of consonants, which at least for stops is also correlated with the presence of glottalisation, suggests a sustained articulatory effort in the realisation of consonants. The particular relevance of these phonetic properties of Bena L1 is due to their impact on the quality of approximation in BE of the durational and release properties of coronal consonants of more standard varieties of NE.

The hypercorrection scenario needs to be elaborated somewhat differently for t and s epenthesis in different contexts. However, all the scenarios share the assumption that pre-pausal glottalisation and consonant lengthening are occasionally transferred from Bena into BE. What may happen then is that a correction for the shorter audible duration of pre-pausal consonants of NE (as compared to those of Bena) is attempted at the same time, as suggested by the observation that on the whole, the final consonants in BE approximate the durational properties of those of NE relatively well. I hypothesise that such a correction is achieved primarily by compressing the glottalisation over the final part of the consonant.

The subsequent steps for t and s epenthesis in different contexts are, I argue, as follows. In the scenario for the emergence of pre-pausal t epenthesis after the coronal [s], [n] and [l], the transferred glottal closure is released too quickly, while the oral articulation of the preceding coronal is still largely sustained, which may occasionally result in a percept of an audibly released [t].Footnote ¹³ Against the background of the tendency for phonological hypercorrection, the latter has good chances to become phonologised as an optional realisation of a given word. The tendency for phonological hypercorrection may be expected to induce a positive bias in the perception of such glottalised coronals as being followed by an audibly released [t]. As a result, the occasional percepts of this emergent pre-pausal [t] and the occasional cases of final [t] due to phonological hypercorrection can concur for a set of speakers to bring about the reanalysis of the pre-pausal [t] epenthesis as being due to an optional phonological rule. In the scenario for the emergence of the pre-pausal s epenthesis after [t], the transferred glottal closure is released just a moment after the oral closure sounding as a weak ejective and which may occasionally result in a percept of a sibilant following [t]Footnote ¹⁴ that later became phonologised as an optional realisation of a given word. Besides the ejective source, occasional pre-pausal affricated realisations [t^s] mentioned in section 3 may also contribute to the emergence of a percept of a sibilant following pre-pausal [t]. In the scenario for the emergence of the pre-pausal s epenthesis after coronals, the transferred glottal closure is occasionally undershot when compressed over the final part of the coronal. In combination with the sustained articulatory effort in the production of the coronal, this may result in a creaky or largely devoiced period in the final part of the coronal, as in [l] or [l]. This final devoiced or non-modally voiced part of the coronal may in turn be perceived as a period of friction in the coronal area and ultimately equated with the voiceless coronal fricative [s]. The phonologisation of this optional epenthetic [s] as being due to an optional phonological rule has occurred against the background of the tendency for the morphological hypercorrection. Finally, in the case of the phrase-internal word-final post-coronal t and s epenthesis, two scenarios are conceivable. Given that the few examples of such epenthesis that I recorded occur before a vowel or an approximant and that in Bena L1 utterance-initial prevocalic glottalisation may be preserved in phrase-internal position in careful speech style, the first scenario would invoke glottalisation as in the scenarios outlined above for the pre-pausal contexts but the glottalisation would come from the word-initial vowel that follows. Alternatively, the phrase-internal epenthesis may be subsidiary to the phonologisation of the word-final epenthetic consonants in the pre-pausal context. The fact that the phrase-internal epenthesis is much less frequent than the pre-pausal epenthesis fits both scenarios. However, the fact that the phrase-internal epenthesis was found only before a vowel or an approximant is accounted for in a much more straightforward way by the first scenario.

The account involving a transfer of pre-pausal glottalisation and consonant lengthening from Bena L1 has several advantages over an unconstrained phonological hypercorrection scenario. Thus, it allows us to explain why the epenthesis in BE is largely pre-pausal, while the presence or absence of a pause is largely irrelevant for the word-final cluster simplification in BE. As outlined in the scenarios above, the transfer of pre-pausal glottalisation and consonant lengthening creates favourable conditions for the production of what may be perceived as a homorganic coronal stop or fricative, and which would thus be ripe for reanalysis as due to a phonological rule. Furthermore, it also contributes to the explanation of the fact that t and s epenthesis is found exclusively after alveolar coronals, as only in this environment can we end up with a percept of an existing BE stop or fricative.Footnote ¹⁵ For instance, when accompanied by an audible release, the pre-pausal glottalisation of a final velar stop cannot produce a percept of any existing BE consonant, as at this place of articulation there are no other consonants available in BE (see table 2). Equally, when the pre-pausal glottalisation of a final bilabial nasal is accompanied by an audible release, the latter is always nasal, which cannot result in a percept of any existing BE consonant. Finally, at the post-alveolar place of articulation no stops are available in BE, only fricatives and affricates.

In addition, the account involving a transfer of pre-pausal glottalisation and consonant lengthening from Bena L1 has the potential to explain the apparent absence of a productive t and s epenthesis in the three major varieties of Nigerian English, viz. Yoruba, Igbo and Hausa NE, and Nigerian Pidgin English. This absence is remarkable because the tendency for word-final consonant cluster simplification in these varieties of NE and Nigerian Pidgin English is at least as strong as that of BE. Of the three respective L1s, Hausa resembles Bena the most in its phonological structure. However, it differs from Bena in at least two important aspects. First, ‘words in the indigenous Hausa vocabulary all end in a vowel . . . with minor exceptions’, such as ‘ideophones and a few apocopated or clipped forms’, and while [n] and [r] occur ‘word finally in their function as gender-sensitive articles . . . word-final obstruents [are] prohibited’ (in the native Hausa vocabulary) (Newman Reference Newman2000: 317, 319). It is English that is responsible for ‘introducing large numbers of words with final consonants’ into Hausa. Second, although Hausa like Bena has pre-pausal glottalisation, in Hausa it is confined to pre-pausal vowels, and except when the vowel is phonologically short, its appearance is conditioned lexically (see Newman & van Heuven Reference Newman and Van Heuven1981). Finally, word-final consonant lengthening has not been reported for Hausa either. Yoruba and Igbo L1 lack both word-final consonants and pre-pausal glottalisation. Given that the possibility of word-final consonants in the native L1 lexicon, pre-pausal glottalisation and consonant lengthening can be argued to create favourable conditions for the production of what may be perceived as, and subsequently phonologised as, a homorganic coronal stop or fricative in L2 English, the lack of these features in Hausa, Yoruba and Igbo may explain why the available potential for phonological hypercorrection has not developed into a productive word-final t and s epenthesis in the respective L2 Englishes.

Finally, the comparison of the more standard varieties of NE and Nigerian Pidgin English with BE arguably suggests that the t and s epenthesis in BE emerged during the period when more British-like pronunciation of English, in particular as regards the faithful reproduction of word-final consonant clusters, was prestigious in Nigeria, that is during or shortly after the colonial period. Thus, as I have noted above, in more formal situations of English use, BE speakers appear to approximate the pronunciation of more standard varieties of NE in terms of cluster simplification tendencies in reducing both the frequency of the t and s epenthesis and of correct productions of etymological word-final clusters with t and s. This makes it unlikely that the more standard varieties of NE in their current form served as the trigger for the development of the hypercorrection process that originally contributed to the emergence of the epenthesis in BE. Therefore, the hypercorrection must have been primarily triggered by contact with a more British-like pronunciation at the time when it enjoyed more prestige in Nigeria, since nowadays British-like pronunciation ‘does not have a high social prestige in Nigeria and is ridiculed as affected and arrogant’ (Gut Reference Gut and Mesthrie2008: 39). In addition, it is not inconceivable that in the absence of direct contact with British-like pronunciation, the familiarity with written English at school has been contributing to the development of the hypercorrection.

7 Word-final consonant epenthesis in other varieties of English

Besides BE, cases of word-final consonant epenthesis have been reported for a number of New Englishes as well as for several of the so-called Inner Circle Englishes. Before proceeding further, note that I will not consider cases such as the development of excrescent final -t in a set of English prepositions and adverbs originally formed by means of the genitive -s, such as against, amidst, amongst, etc. Neither will I consider the few other sporadic examples of ‘the addition of non-etymological <-t>’ in the history of English cited by Minkova (Reference Minkova2014: 150), such as behest, hoist and tuft, which do not form any kind of coherent group at all. According to the OED online (2014), the likely source of the excrescent final -t in against was its frequent collocation with a subsequent te, variant of the, perhaps reinforced by association with superlatives in -st. Unlike the word-final post-coronal t and s epenthesis in BE, this excrescent final -t in the history of English is confined to a limited number of words of a certain morphosyntactic class and morphological structure. Furthermore, it has developed phrase-internally, whereas the BE pattern has developed pre-pausally.