2. Free variation in English

The concept of free variation in phonology accounts for the occurrence of two or more different sounds in the same environment without constituting a meaning change and without being considered incorrect by the native speakers. The phenomenon may also be explained as a case of instability or free fluctuation in sounds (Hughes, Trudgill & Watt, Reference Hughes, Trudgill and Watt2005) as a result of which a particular sound may be variably pronounced at different times by the same speaker or differently by two speakers (Fatunsin, Reference Fatunsin2007). Free variation may occur at the segmental (e.g. schedule as /ʃeʤu:l/ or / ʃkeʤu:l/, dilemma as /dɪlemə/ or /daɪlemə/) and supra-segmental (e.g. detail as /'di:teɪl/ or /dɪ'teɪl/, finance as /'faɪnæn^ts/ or /fə'næn^ts/) levels.

At the segmental domain, free variation may manifest in phonetic or phonemic form. It may occur as an instance of phonetic realisations of a phoneme, for example, the pronunciation of the initial voiceless alveolar plosive in team as aspirated [t^hi:m] or unaspirated [t^oi:m], or a situation where certain sounds are realised as different phones from one native accents to another (e. g., RP /ɒ/ as /ɑː/ in GA). In its phonemic form, free variation may occur due to differentiation in the number of phonemes in a language (McCully, Reference McCully2009; Roach, Reference Roach2009), thereby allowing a set of phonemes in a particular language to be used interchangeably. For example, both long /e:/ and diphthong /eɪ/ are present in some northern British English accents, while the same long /e:/ is realised as/eɪ/ in BBC pronunciation.

At the suprasegmental domain, free variation may occur in the stress pattern of a particular word in isolation with no change in grammatical category (e.g. vaccine /'væksi:n/ or /væk'si:n/; complex /'kɒmpleks/ or /ֽkɑːm 'pleks/) or in connected speech (e.g. a shift in stress placement in eco'nomic, as in think about the 'economic 'value they have), yet without a meaning difference. Lastly, free variation may also manifest through the processes of assimilation, coalescence and reduction in connected speech. These may result in phonological alternations such as this shop as /ðɪs ∫ɒp/ or /ðɪ∫∫ɒp/; God bless you as /gɒd bles ju:/ or /gɒd ble∫ʊ/ and would you as /wʊd ju:/ or /wʊʤu/.

This study focuses on free variation in lexical stress patterns. Essentially, this phenomenon explains the possibility of more than one stress pattern on a particular word in its citation form (e.g. ciga'rette or 'cigarette, ap'plicable or 'applicable, 'distribute or di'stribute, 'communal or co'mmunal) without a change in meaning or grammatical category (Mompean, Reference Mompean and Henderson2010; Cruttenden, Reference Cruttenden2014). This, therefore, precludes stress variation that is due to stress shift (e.g., su'ccess versus 'success 'power) or leads to a change in grammatical class such as au'gust (adjective) and 'August (noun), 'present (adjective/noun) or pre'sent (verb).

The occurrence of lexical-stress free variation may be due, largely, to accent differences (Shitara, Reference Shitara1993; Katalin & Szilárd, Reference Katalin and Szilárd2006; Henderson, Reference Henderson2010) or what Simo Bobda (Reference Simo Bobda2010) refers to as stress variation in space. Variability in the accents of the two major native English varieties (RP and GA), globally considered as models for teaching and learning English in the second language setting, has considerably yielded lexical-stress free variation in various English varieties. Examples are advertisement pronounced as /əd'vɜ:tɪsmənt/ (RP) and /ædvə'taɪzmənt/ (GA); formidable as /fɔ:'mɪdəbəl/ (RP) and /'fɔ:rmɪdəbəl/ (GA); harass as /'hærəs/ (RP) and /hə'ræs/ (GA); and so on (Shitara, Reference Shitara1993; Simo Bobda, Reference Simo Bobda2010).

Word-stress free variation has also been shown to be conditioned by social factors, such as age, level of education and gender, among others. For example, preferences for certain word stress variants (e.g., harass, formidable, justifiable) are found to be influenced by age, gender or level of education of speakers in both British and American English varieties (Shitara, Reference Shitara1993; Wells, Reference Wells, Elenius and Branderus1995).

Another explanation for lexical-stress-free variation, according to Simo Bobda (Reference Simo Bobda2010), is variability in time. Evidence abounds of diachronic stress shift on English words over time which has yielded several competing stress patterns. For example, the lexical items ´illustrate, ´revenue, and ´abdomen with stress preference on the initial syllable in present-day British English were, prior to the 18^th century, stressed on the penultimate syllable as i´llustrate, re´venue, ab´domen. Today, both the traditional and newer forms are in free variation (Wells. Reference Wells2008, Simo Bobda, Reference Simo Bobda2010).

Generally, the concept of free variation has not received detailed attention in phonological theory, possibly due to the traditional assumption that it was insignificant to the linguistic field (Mompean, Reference Mompean and Henderson2010). Nevertheless, a few studies in the native English varieties have paid attention to the phenomenon, whose findings have improved our understanding of the concept (Henderson, Reference Henderson2010; Mompean, Reference Mompean and Henderson2010) and have provided support for the revised editions of the English Dictionaries (Shitara, Reference Shitara1993; Wells, Reference Wells1999, Reference Wells2003).

On the other hand, research into lexical-stress free variation in a second language setting such as Nigeria is yet to be seen, despite that the accent is replete with variable pronunciations under the influences of native Englishes and substrate languages. The present study, therefore, is an attempt to extend the study of free variation in word stress to a non-native variety.

3. Word stress in Nigerian English

English, in Nigeria, has graduated from just being the language of trade and of teaching the Christian ethos to that of government, diplomacy, literature, the mass media, and of inter-ethnic and international communication (Oyeleye, Reference Oyeleye2016). It is the nation's official language and, unarguably, the most influential and dominant language (Gut, Reference Gut, Kortmann, Schneider, Burridge, Mesthrie and Upton2004). The contact of English with the over 500 indigenous languages (Ebenhard, Simons & Fennig, Reference Ebenhard, Simons and Fennig2019) has, however, exposed it to substratum influence which has resulted in a domesticated variety known as Nigerian English. The existence of NigE is no longer in doubt (see Gut, Reference Gut, Kortmann, Schneider, Burridge, Mesthrie and Upton2004; Jowitt, Reference Jowitt2019). Rather, commendable efforts are geared towards its codification and standardisation (Gut, Reference Gut2012; Adegbite et al., Reference Adegbite, Udofot and Ayoola2014) which have yielded series of studies at different linguistic levels.

At the suprasegmental level of phonology, several studies have posited that NigE differs remarkably from British English (BrE) in stress assignment. These studies are agreed that, in view of the serious challenge which stress poses to most users of English as a second language, NigE speakers have developed nativised stress-placement strategies which have yielded peculiar but definable stress patterns. Some of these include backward and forward stress shift in contrast to the native-English stress patterns (Atoye, Reference Atoye1991; Simo Bobda, Reference Simo Bobda2010; Atoye et al., Reference Atoye, Soneye, Adejuwon, Faleye and Fajobi2018; Jowitt, Reference Jowitt2019) and replacement of English stress with tone (Amayo, Reference Amayo1980; Gut, Reference Gut, Bisang, Hock and Winter2007; Udofot, Reference Udofot2007; Gussenhoven & Udofot, Reference Gussenhoven and Udofot2010; Fajobi, Reference Fajobi2014).

In addition to these, however, users of English in Nigeria are also exposed to the American and British English stress patterns, which are often in free variation (Atoye et al., Reference Atoye, Soneye, Adejuwon, Faleye and Fajobi2018; Obasi, Reference Obasi2019), in view of the initial contact with the British and the subsequent political, economic and educational interactions with the United States of America (Awonusi, Reference Awonusi1994; Jowitt Reference Jowitt2019). This presupposes that NigE speakers are beneficiaries of different linguistic influences with different variants of English word stress in their repertoire.

It is against this backdrop that the study of word-stress free variation in NigE is worthwhile. The present study goes beyond mere identification and description of NigE word stress patterns which was the preoccupation of earlier studies. It is intended, as a first step, to empirically establish the pronunciation preferences of NigE speakers amongst competing word stress patterns through corpus analytical methods.

4. Methodology

The data for this study were sourced from the International Corpus of English (ICE) – Nigeria, which contains 1,010,382 words of spoken (609,586) and written (400,796) NigE usage. The ICE–Nig was compiled between 2007 and 2014 at the Universities of Augsburg and Münster (Gut, Reference Gut2014) as part of the International Corpus of English project founded by Greenbaum and associates in 1990 (Unuabonah & Gut, Reference Unuabonah and Gut2018). The text samples were supplied by male and female Nigerians with a wide range of age groups who have been educated in Nigeria through the medium of English to at least the end of secondary school. These include media practitioners, politicians, university lecturers and students, medical doctors, sport analysts and commentators, lawyers, judges, technocrats and clergy. The choice of these categories of informants as the population sample for this study is premised on their recognition as suitable candidates for the codification of NigE (Gut, Reference Gut2012).

Specifically, the study concentrates on the spoken categories of the corpus which contain formal, semi-formal and informal texts, such as conversations, unscripted speeches, broadcast interviews, classroom lessons, broadcast talks, non-broadcast talks, broadcast news, and so on. Table 1 shows the data types in the spoken part of the corpus and the total number of files and words contained in each of the texts.

Table 1:

Word count of the spoken part of the ICE–Nig corpus

(Source: Gut, Reference Gut2014)

Forty-four lexical items that are prone to stress variation in native Englishes were compiled from Shitara (Reference Shitara1993), Mompean (Reference Mompean and Henderson2010) and Wells’ (Reference Wells2008) Longman Pronunciation Dictionary (LPD). The words, in their inflectional forms (plural, present tense and past tense), were searched in ICE–Nig using AntConc corpus analysis toolkit version 3.4.4 (Anthony, Reference Anthony2014); the search returned 31 words with various rates of occurrences. In order to ensure a well balanced representation of each word, therefore, 18 items which occurred ten times and more were eventually chosen for the analysis. Table 2 contains the lexical items analysed in this study.

Table 2:

Tokens, number of occurrences and speakers

The analysis involved listening to the relevant audio files of each word in ICE–Nig by the authors to determine the particular variants produced. Both raters are Nigerian L2 speakers of English and trained phonologists. Each listened to the audio clips independently and noted the variants produced, after which their findings were compared. Cases of disagreement were resolved by jointly listening to the disputed audio clips; and where it was difficult to reach a consensus, the opinion of another trained phonologist was sought and then a decision was made. The data were subsequently analysed quantitatively by counting the tokens of occurrence of each variant and speaker and converting them to percentages. The variants with higher percentages of tokens of occurrence and higher number of speakers were taken as the preferred variants. In a situation of intra-speaker variation, where the same speaker pronounces the two variants of a lexical item, the number of times such variation occurs is indicated as a superscript above the total number of speakers (see Table 2). Wells’ (Reference Wells2008) LPD model was adopted for the transcription of native English pronunciation variants while Adegbite et al.'s (2014) phonemic model of NigE was adopted for the NigE phonemic representations.

To establish the preferred stress variant of a particular lexical item, two levels of analysis were carried out. First, the sum of each variant was calculated as a percentage of the total sum of occurrences of the lexical item in the corpus; and second, the number of speakers who produced each variant was also calculated as a percentage of the total number of speakers who produced the word. Therefore, the variants with the higher (or highest) percentage of occurrence and the higher (or highest) rate of speakers were regarded as the preferred. This is statistically represented as shown in (Figure 1):

Figure 1.

Statistical representation of preferred variants

5. Findings

A total of 18 lexical items that typically exhibit stress variation in the native English varieties were analysed in this paper. Each word appears in different inflectional forms grouped as a word family (e.g. CREATE: create, creates and created). The lexical items were grouped into three categories: disyllabic items with invariant word stress in the corpus, disyllabic items with variant word stress and trisyllabic words whose stress alternates between the initial and penultimate syllable in native Englishes. The native British and American English pronunciation forms, as well as the NigE speakers’ pronunciation variants and preferences, are tabulated and discussed accordingly. Since the inflectional forms of each item do not show stress differences, only the base form is shown on the tables as representing each word family. For example, CREATE stands for create, creates and created (see Table 3).

Table 3:

Words with invariant words stress in Nigerian English

Table 3 shows that no stress variation was exhibited in five disyllabic words: adult, detail(n,v), weekend, dictate(v) and increase(n) in the corpus. In all instances of occurrence of the lexical items, all the speakers demonstrated total preference for stress on the initial syllable of 'adult (11 tokens produced by six speakers), 'detail (76 cases produced by 41 speakers), 'weekend (25 instances produced by 18 speakers) and 'increase(n) (17 tokens produced by 12 speakers), and on the final syllable of dic'tate(v) (12 tokens produced by nine speakers). The result for 'adult in this study is similar to that of Wells’ (Reference Wells2008) LPD which indicates that 84% of BrE speakers preferred stress on the initial syllable, but different from Shitara's (Reference Shitara1993) finding which reveals that 88% of AmE respondents showed a preference for stress on the final syllable of a'dult. The score for 'detail tallies with Mompean's (Reference Mompean and Henderson2010) and Wells’ (Reference Wells2008) LPD findings respectively which show the BrE (85.5%) and AmE (75%) speakers’ predilection for stress on the initial syllable of the same word. Wells’ (Reference Wells2008) LPD entry for 'weekend shows that NigE speakers’ preference for stress on the initial syllable of the item is in tandem with the AmE prereference, as against the BrE primary variant: wee'kend. Stress on the final syllable of dic'tate(v) agrees with the BrE preferences but differs from the word-initial stress variant favoured in AmE as LPD (Wells, Reference Wells2008) reports.

The lexical items in the second category are disyllabic words that take stress either on the initial or final syllable. Table 4 shows that NigE speakers preferred stress on the initial syllable of 'dispute(n) while a'ddress(n), i'dea, re'search(v,n), fi'nance(n), con'tact(v), tran'slate, cre'ate and re'source were predominatly stressed on the final syllable. Seven (87.50%) out of eight speakers exhibited a preference for stress on the initial syllable in 21 (95.45%) of the 22 cases of 'dispute(n) found in the corpus. This differs from Mompean's (Reference Mompean and Henderson2010) and Wells’ (Reference Wells2008) LPD findings which reveal a predilection for stress on the final syllable of the same word in both BrE and AmE, which is an exception to the English noun-verb alternation stress rule and further confirms the notion that English stress is unpredictable (Simo Bobda, Reference Simo Bobda2010; Jowitt, Reference Jowitt2019).

Table 4:

Disyllabic words with variant word stress in ICE–Nig.

V1- Variant 1 (NigE main/prefered variant); V2- Variant 2 (NigE alternative variant)

In a'ddress(n), the final syllable was predominantly stressed in 37 (88.10%) out of 42 tokens by 26 (86.70%) speakers. This includes an incidence of intra-speaker variation within the same discourse by a female speaker (btr_11) who produced address with final stress twice and with initial stress once. The result compares with Wells’ (Reference Wells2008) LPD entry for the same word, which reveals that 58% of AmE speakers showed preference for stress on the final syllable and that the variant is the primary pronunciation in BrE. Table 4 also reveals that 40 (60.61%) speakers in 72 (72%) cases preferred stress on the final syllable of i'dea. This finding is similar to the LPD data (Wells, Reference Wells2008) which shows the stress variant as the BrE speakers’ only pronunciation form and the preferred variant for 86% of AmE respondents based on a preference poll. Again, a number of speakers in the corpus varied their pronunciation of idea, shifting between initial and final stress. For example, in (bdis_12), a male discussant produced /'aidia/ once and /ai'dia/ thrice, while a female lecturer (btr_07) and a male lecturer (les_01) each pronounced both variants of idea in two different instances.

Final stress was also adopted in 70 (80.46%), out of 87 cases of re'search (both the verbal and the nominal forms) by 42 (82.35%) speakers. The poll results for the same word in LPD (Wells, Reference Wells2008) reveal that 80% of BrE speakers also preferred stress on the same syllable, while 78% of AmE respondents placed stress on the initial syllable ('research). The finding also supports the possibility of distinguishing between the verb (re'search) and the noun ('research), which LPD (Wells, Reference Wells2008) reports for the native speakers, as the two tokens of the verb form in the corpus were accented on the final syllable, whereas the noun received stress on the intial syllable in majority of the cases.

Preference for final stress was also observed by 11 (84.62%) speakers in 19 (82.61%) of the 23 tokens of fi'nance(n) and by eight speakers (66.67%) in ten instances of the total cases of con'tact(v), representing 71.43%. This includes a case of intra-speaker variation which occurred in the pronunciation of 'finance and fi'nance by a news reporter in (bnews_31). However, both findings are at variance with Wells’ (Reference Wells2008) LPD entries which indicate that the initial syllable is the preferred stress position for both words in the American and British English varieties. Final stress was again favoured in 12 (92.31%) of the 13 tokens of tran'slate by 11 (91.67%) speakers and in 80 (95.24) of the 84 cases of re'source by 42 (95.45%) spearkers. In both cases, NigE speakers aligned with the BrE final stress preference against the initial stress pattern in AmE. Finally, in this category, 58 speakers (98.31%) overwhelmingly placed stress word-finally on cre'ate in 83 (96.31%) of the 86 tokens found in the corpus. A presenter also varied his pronunciation of 'creates and cre'ated with stress on the first and the second syllable respectively. The finding agrees with LPD (Wells, Reference Wells2008) which shows that 87% of American respondents preferred final stress on the same lexical item and that it is the primary pronunciation in BrE.

The third category of lexical items considered in this study comprises trisyllabic words: contribute, umbrella, revenue and vehicle which are typically stressed on the initial or penultimate syllable in native Englishes. As shown in Table 5, contribute was stressed on the intial, penultimate and final syllables by NigE speakers; however, it was final stress, contri'bute, that was overwhelmingly favoured by 25 (89.29%) of the 28 speakers in 31 (83.78%) cases of all the total incidence of the lexical item. A female lecturer (unsp_05) also exhibited internal variation in her pronunciation of the lexical item; toggling between initial stress on 'contribute and penultimate stress on contri'buted. The result is different to Mompean's (Reference Mompean and Henderson2010) finding and the LPD (Wells, Reference Wells2008) poll results which show that 78% and 73% of British respondents respectively preferred stress on the penultimate syllable (con'tribute), which is also the case in AmE.

Table 5:

Trisyllabic words with variant word stress in ICE–Nig.

Table 5 further shows that 7 (87.5%) out of 8 speakers preferred stress on the initial syllable in 7 (70%) of the 10 cases of umbrella found in the data; whereas LPD (Wells, Reference Wells2008) reveals that both American respondents (86%) and British speakers favour penultimate stress. Revenue was pronounced with intial stress in 15 (93.75%) of 16 cases by 11 (91.67%) speakers, which agrees with the main variant in British and American English (Wells, Reference Wells2008). Finally, out of the 64 tokens of vehicle found in the corpus, stress was placed on the first syllable ('vehicle) in 63 (98.44%) instances by 30 (96.77%) speakers. This trend agrees with Wells’ (Reference Wells2008) findings in LPD which marks the same stress position as the norm in BrE and the preferred for 95% of AmE speakers.