Towards a feature-based L2 pronunciation teaching approach: variations in Hong Kong English and ELF intelligibility

5.1.1 Consonants

Our analysis yielded variations in HKE features in terms of (1) their prominence in the whole dataset (percentages of specific feature production among all instances; see Figure 1 for the consonantal features) and (2) the percentages of speakers who produced (at least one instance of) these features (Table 3). For the case of HKE consonants, both analyses yielded a similar pattern. This indicates that most of the HKE speakers in our data (over 90 %), to some extent, produced features such as devoiced /z/, /l/-vocalisation/deletion, /ð/ substitution, variation/deletion of final /d/, /t/, /p/ and /k/, and /v/ substitution but not necessarily every time. This may also reflect their degree of difficulties in production, i.e., the higher percentages of the features in the figure/table (/z/, dark /l/, and /ð/), the more unlikely (or challenging) the features are to be produced. Some less frequently occurring features (e.g., /n/-/l/ conflation, the substitution of /r/ by [w]) were more likely found among the Band 3 students, who presumably have a relatively lower academic or English proficiency level. However, the reason for the prominence of /r/ substitution among students at University Y is unclear. Alternatively, /θ/ substitutions were less commonly produced by the students at both universities (particularly University Y), probably due to their greater awareness of this feature as they progressed through the education system. Details of each feature according to their frequency of occurrence are discussed below (see Appendices 2–6).

Figure 1:

Prominence of HKE consonants.

/z/: The devoicing of voiced consonants is a typical feature of HKE (Hung 2000; Setter et al. 2010). The most common feature is the realisation of the voiced /z/ (94.1%) as a voiceless [s] (‘these’), and this feature is consistent in any position of a word (initial/medial/final).

Dark /l/: /l/-vocalisation or deletion is another frequently-occurring HKE feature (90.6 %). In the former case, the dark /l/ was replaced by a vowel (‘will’ [ʊ]), whereas in the latter case, it was often deleted when positioned after a back vowel (‘all’) (see also Deterding et al. 2008; Hung 2000). Vocalisation of dark /l/ is also common in other L1/L2 English varieties (Deterding and Kirkpatrick 2006).

TH: The L1 pronunciation of ‘th’ (in spelling) mainly consists of /θ/ (‘think’) and /ð/ (‘they’) depending on the words and consonant positions. Consistent with previous research (Bolton and Kwok 1990; Deterding et al. 2008; Hansen Edwards 2018; Hung 2000), it is common that HKE speakers substituted voiced /ð/ with [d] (‘those’) (90.3 %). Substitution of voiceless /θ/ with [f] was less prominent (‘think’) in the overall data (45.5 %), but it was occasionally produced by many HKE speakers (72.5 %). This may indicate their awareness of the L1 pronunciation, but they pronounced it inconsistently. Substitution of voiceless /θ/ for the ‘th’ sound in words beginning is common in the ASEAN region, but it may be pronounced differently in different varieties ([t] in Singapore/Malaysia English, [s] in China English) (Deterding and Kirkpatrick 2006).

Final consonants: Word final consonants such as final /p/, /d/, /t/, and /k/ were often pronounced differently by HKE speakers from those of L1 varieties (e.g., RP) (Setter et al. 2010). Specifically, these final consonants were frequently found to have been unreleased (‘shop’, food’, ‘not’, ‘back’) or omitted (‘about’), and in very few cases, /d/ was substituted by [t] (‘workload’). It is likely that the speakers were aware of these final sounds but were unable to or did not pronounce them as they are absent in the Cantonese (their L1) phonology. Overall, variations or omissions of final /p/ (85.5 %) were more prominent HKE features than /d/ (65.6 %), /t/ (64.7 %) and /k/ (57.8 %), but they are present to some extent in most participants’ speech (>90 %).

/v/: Another prominent example of devoiced consonants is /v/ (78.2 %), which was produced by most participants (94.7 %). Hung (2000) hypothesised that this phoneme is virtually non-existent in HKE and is phonetically realised as [w] at the onset of a stressed syllable but as [f] at the beginning of an unstressed syllable. His hypothesis is generally supported in our data in that /v/ in words such as ‘view’ was pronounced as [w], whereas it was produced as /f/ in words such as ‘heavy’. Nonetheless, there were also cases where /v/ at the onset of a stressed syllable was substituted with [f] (rather than [w]). In addition, /v/ was frequently replaced by /f/ at the coda of syllables. The variations in the production of /v/ may be related to the participants’ English proficiency level or awareness of the L1 pronunciation. The pronunciation of [f] is also arguably more similar to that of /v/ than [w].

/r/, /n/ and /l/: Although substitutions of /r/ by /w/ (5.6 %) (‘really’), /n/ by /l/ (1.0 %) (‘nice’) and /l/ by /n/ (0.3 %) (‘lesson’) in onset syllables were reported to be common HKE features (Bolton and Kwok 1990; Hung 2000; Setter et al. 2010), they only accounted for relatively small proportions in the data (cf., Deterding et al. 2008). There were relatively more participants who featured /r/-/w/ substitution (34.5 %). In Cantonese, /r/ is absent and may be replaced by a similar existing consonant /w/. Despite the presence of /n/ and /l/ in Cantonese, they are sometimes interchangeable for some Hong Kong speakers when they speak Cantonese.

5.1.2 Initial consonant clusters

Our analysis has identified a list of HKE initial consonant clusters (Figure 2), yet they only contributed to a small proportion of the data (cf., Deterding et al. 2008; Setter et al. 2010). They included simplifications or variations of /tr/ (25.0 %) (‘train’), /pr/ (23.0 %) (‘pressure’), /pl/ (22.2 %), (‘plan’), /kl/ (20.3 %) (‘class’), /fr/ (16.1 %) (‘friend’), and /br/ (12.5 %) (‘breakfast’). Indeed, most of these features were only present in less than 25 % of the sample population (Table 4), except for /pr/ (36.6 %). All these HKE initial consonant clusters involved /l/ or /r/ as the second sound, which was sometimes omitted or pronounced incompletely. Some participants in the data also simplified /tr/ to [tʃ]. Among the six participant groups, several categories of consonant cluster simplification were more common among the Band 3 students (e.g., /fr/, /kl/, /tr/) and professionals with diverse backgrounds (e.g., /pr/, /fr/, /pl/).

Figure 2:

Prominence of HKE initial consonant clusters.

5.1.3 Final consonant cluster

Final consonant clusters do not exist in Hong Kong speakers’ L1 and are often simplified. As this simplification also frequently occurs in connected speech in an utterance, our analysis of consonants and consonant clusters took account of pauses, and vowels and consonants in the previous/following words. As can be seen in Figure 3, many frequently occurring final consonant clusters in our HKE data belonged to this category. These features also appeared in most participants’ speech production (Table 5).

Figure 3:

Prominence of HKE final consonant clusters.

The symbol ‘+C’ in the figures indicates that the final consonant clusters were followed by another consonant (including that at the beginning of the following word). The speakers often omitted the middle consonant(s) in this cluster of three (or more) consonants. They include /sk/+C (93.3 %) (‘ask you’), /nd/+C (89.0 %) (‘stand behind’), /st/+C (88.0 %) (‘first step’), /ŋk/+C (78.5 %) (‘think so’), /nt/+C (77.2 %) (‘different people’), /kt/+C (67.8 %) (‘products’), /ft/+C (62.5 %) (‘raft building’). Two prominent HKE final consonant clusters that do not belong to this category (i.e., with less than 3 consonants) are simplifications of /nd/ to [n] (88.0 %) (‘friend’) and /ŋk/ to [ŋ] (50.2 %) (‘think’), where the final consonant was often missing (cf., Deterding et al. 2008). There is a lack of a clear pattern in the differences among the six participant groups, although certain final consonant clusters (e.g., ‘/kt/ + C’, /ŋk/) appear to be less commonly found among the Band 3 students. A possible explanation is the relatively low occurrence of these features in the sub-datasets.

5.1.4 Monophthongs

Figure 4 displays the HKE monophthongs identified, which are consistent with previous studies on HKE phonology (Bolton and Kwok 1990; Deterding et al. 2008; Hung 2000; Setter et al. 2010). Interestingly, although most of them have a relatively low overall percentage, they were present in most participants’ speech (Table 6). Almost all the participants occasionally produced /æ/-/e/, /ɪ/-/i/ and /ɒ/-/ɔ/ mergers and stressing of unstressed syllables (>90 %). This may indicate their awareness of these features according to L1 norms. Interestingly, /u/-/ʊ/ mergers were more commonly found among the university students and professionals for unknown reasons.

Figure 4:

Prominence of HKE monophthongs.

Mergers of /æ/ and /e/: This is the only feature in this category with over 50 % frequency of occurrences (55.5 %). In the data, there is a much higher tendency that /æ/ was shifted towards the more closed and fronted /e/ (‘rational’) (3036/3041 instances of this HKE feature) than vice versa (‘better’). This is probably because only /e/ (but not /æ/) is present in Cantonese.

Mergers of /ɪ/ and /i/: The study considered audible differences in vowels and, thus, only focused on the contrast between /ɪ/ and /i/ regarding vowel quality. Although this is a commonly recorded HKE feature in the literature, it has a relatively low frequency of occurrences in the data (23.9 %). The majority of the /ɪ/-/i/ mergers were a shift from /ɪ/ to /i/ (‘this’) (2374/2384) rather than one shifting from /i/ to /ɪ/ or to a sound between /ɪ/ and /i/ (‘need’).

Mergers of /ɒ/ and /ɔ/ and of /u/ and /ʊ/: As compared to the previous two monophthong features, merging between /ɒ/ and /ɔ/ (‘top’, ‘important’) (14.6 %) and between /u/ and /ʊ/ (‘to’, ‘good’) (5.1 %) was found to be less prominent among the HKE speakers. Among all the HKE monophthongs, there were relatively fewer participants featuring /u/-/ʊ/ mergers (65.8 %).

Stressing of unstressed syllable /ə/: Most L1 English varieties have a stress-based rhythm consisting of stressed and unstressed syllables, and the unstressed syllables tend to be realised as schwa (/ə/). In contrast, L2 varieties in many places, such as Hong Kong and Singapore, are perceived to adopt a syllable-based rhythm, in which these unstressed syllables are often stressed (Deterding 2010, 2013; Setter et al. 2010). Our findings generally confirm the existence of this HKE feature: this feature is present in all participants’ speech, but /ə/ is not always stressed in the unstressed syllables (24.7 %). The HKE continuum may reflect different degrees of syllable-based and stress-based rhythm depending on individuals. When /ə/ was stressed, it was sometimes pronounced as full vowels such as [ɔ] (‘condition’), [ɜ] (‘teenagers’), [a] (‘popular’), [ʌ] (‘museum’), [u] (‘today’), [oʊ] (‘provide’) and [e] (‘adventure’).

5.1.5 Diphthongs

Different forms of diphthong reduction/variation were identified in our analysis, and they often occurred when followed by certain consonants (cf., Setter et al. 2010) (Figure 5). In order of their prominence in the data, these diphthong-consonant combinations included /eɪl/ (100 %), /eɪk/ (97.1 %), /aɪk/ (95.3 %), /oʊn/ (85.6 %), /eɪn/ (74.4 %), /aɪt/ (70.9 %), /oʊm/ (64.9 %), /aʊt/ (55.8 %), /eɪm/ (35.4 %), /ɔɪn/ (25.3 %) and /aɪm/ (22.1 %). This pattern is similar to the proportion of speakers producing them (Table 7). Clear inter-group variation patterns could not be identified for most diphthong categories. One observable pattern, however, is the absence or relatively low occurrence of certain HKE diphthongs (e.g., /eɪn/, /oʊn/, /oʊm/, /eɪm/, /ɔɪn/, /aɪm/) among the students at both universities (except for /oʊm/ among University Y students). Similarly, none of the professionals simplified /oʊm/ in the data, while the simplification of /eɪm/ was more common among the Band 3 students.

Figure 5:

Prominence of HKE diphthongs.

Reduction of /eɪ/ to /ɪ/ or /e/: Reduction of diphthongs in syllables ending in [l], [k] and [n] were more common than those ending in [m]. The variation depends on the diphthong-consonant combination in specific words. For instance, /eɪ/ was often pronounced as [ɪ] in syllables closed by [k] (‘take’) and, in those ending with [n], [eɪn] was replaced by /ɪŋ/ (‘change’). In other /eɪ/ syllables with [l] and [m] endings, /eɪ/ was reduced to /e/ (‘female’, ‘game’). Particularly, the dark /l/ in /eɪl/ was always vocalised (‘fail’ [eʊ]). /eɪ/ in open syllables was retained (‘may’).

Variation of /aɪ/ and /aʊ/: When pronouncing the diphthongs /aɪ/ and /aʊ/, the HKE speakers frequently replaced /a/ with [ʌ] if the diphthongs were followed by final voiceless oral stops such as [k] (‘like’) and [t] (‘quite’). There were few instances where /aɪm/ was reduced to [am] (‘time’).

Variation of /oʊ/ and /ɔɪ/: Similar variation/reduction patterns were also identified for HKE speakers’ production of /oʊ/ and /ɔɪ/. For example, the second vowel in these diphthongs was sometimes omitted when followed by [m] (‘homework’ [om]) and [n] (‘point’ [ɔn]). Alternatively, /oʊ/ followed by /n/ tended to be reduced and modified to [oŋ] (‘phone’).

5.2.1 Consonants

Intelligibility research has suggested that all consonant sounds are important, except for ‘th’ substitutions and ‘dark’ /l/ (Deterding 2010, 2013; Deterding and Kirkpatrick 2006; Jenkins 2000). These less important HKE features are common in ELF interactions both in Europe and the ASEAN region (including Hong Kong). It is also suggested that initial consonants and those with a higher FL are more important for maintaining understanding (Brown 1988; Levis 2018; Munro and Derwing 2006; Sewell 2017). In contrast, medial (individual) consonants between vowels and devoicing of final consonants are less important (e.g., /z/-/s/ in HKE); many L1 speakers also do not fully voice stops at word ends in any environment (Deterding 2013; Levis 2018).

From this perspective, the majority of the prominent HKE consonants (deletion/vocalisation of dark /l/, substitution of /ð/ by [d], substitution of /θ/ by [f]) do not necessarily impede intelligibility (Table 8). This accords with previous findings that HKE is generally intelligible to international audiences (Kirkpatrick et al. 2008). Although /z/ was frequently devoiced to [s] in our HKE data, this sound is often located in the final position (rather than initial) due to the low frequency of words beginning with /z/. Subsequently, priorities in teaching HKE should be given to consonantal contrasts with a higher FL (/f/-/v/, /n/-/l/) and initial consonants (/z/, /r/, /v/, /n/, /l/). Nevertheless, some of these features are not that common among HKE speakers (initial /r/, /n/, /l/). Deterding (2013) reported instances of miscommunication caused by missing final consonants (/s/, /z/, /v/, /t/, /d/, /p/, /k/), but most of them accompanied other contributory factors. As HKE speakers tend to vary (rather than delete) these consonants in word-final positions, the extent to which these variations (e.g., unreleased plosives) affect intelligibility may require further investigation.

While Levis (2018) argues that aspiration following initial voiceless /p/, /t/ and /k/ are crucial for intelligibility (also Jenkins 2000, 2002), these features are often found in other Southeast Asian English varieties but rarely in HKE. In Deterding (2013), however, there were few instances of misunderstanding caused by a lack of aspiration on initial voiceless plosive among ELF speakers in Southeast Asia. Deterding proposes that this feature could be an emerging norm in the region.

5.2.2 Consonant clusters

Initial consonant clusters are especially important for intelligibility, particularly if the words begin with a stressed syllable (Deterding 2013; Jenkins 2000, 2002; Jeong et al. 2020; Levis 2018) (Table 9). In our speech data, several initial consonant clusters were sometimes simplified (/tr/, /pr/, /pl/, /kl/, /fr/, /br/) by the HKE speakers, and should, hence, be the foci in the English classroom. All of these clusters consist of either /r/ or /l/ in the second position, which, according to Deterding (2013), is a major cause of misunderstanding in ELF interactions in Southeast Asia.

Some final consonant clusters are crucial for intelligibility, especially when they are the placement of inflectional endings (–s/’s and –ed endings), but medial consonant clusters tend to be less important because they can be easily re-syllabified into different syllables (‘nt’ in ‘painting’) (Levis 2018). Nevertheless, it is permissible that some complex final consonant clusters are simplified according to L1 English rules of syllable structure (Deterding 2010, 2013; Jenkins 2000, 2002). One example is the omission of the medial consonant in clusters of three consonants (‘scripts’), and this also applies to those in connected speech when a final two-consonant cluster is followed by another word beginning with a consonant. This feature of connected speech was frequently found in our HKE data. Deliberate attempts to pronounce the medial consonant may not be necessary as they may make the word less intelligible or affect speech fluency. Deterding (2010) argues that the deletion of final consonants in final clusters (/t/, /d/) is common in Southeast Asia (and L1 varieties) and may potentially be part of the regional standard (also Hansen Edwards 2016a). This may apply to the simplified final consonant clusters [nd] in HKE.

5.2.3 Vowels

Table 10 summarises areas of HKE vowels that are more or less important in pronunciation teaching according to previous intelligibility findings and the concept of FL. There have been inconsistent, inconclusive findings about the impact of vowel variations on intelligibility. While Jenkins (2000, 2002) suggests that vowel length contracts are particularly important, there is little evidence that this would cause misunderstanding in Deterding (2013). Alternatively, Brown (1988) has ranked different FL vowel contrasts according to their importance. Given the inconsistent and inconclusive evidence, it seems safer at this stage to prioritise teaching features with the potential to cause misunderstanding. For example, as the vowel pairs /e-æ/ and /iː-ɪ/ have a higher FL than /uː-ʊ/ and /ɔː-ɒ/ and have a higher frequency of occurrences in the HKE data, they should have a higher priority in learning/teaching. Levis (2018) also argues that distinctions in which vowels have merged in certain varieties (/ɔː-ɒ/) are less important for intelligibility. In fact, the pedagogical foci of most of these distinctions in HKE can be on the contrasts between long and short vowels (e.g., between ‘live’ and ‘leave’), which are particularly crucial according to ELF intelligibility findings. According to Walker (2010), the importance of long-short vowel contrasts also applies to diphthongs, such as those in HKE that are often reduced (e.g., /eɪk/>[ɪk], /eɪn/>[ɪŋ], /oʊn/>[oŋ], /oʊm/>[om], /eɪm/>[em], /ɔɪn/>[ɔn], /aɪm/>[am]). Since it is generally agreed that most L2 regional qualities (if consistently pronounced) do not affect intelligibility (Deterding 2013; Jenkins 2000), several HKE diphthongal features in our data are likely acceptable (e.g., /eɪl/>[eʊ], /aɪk/>[ʌɪk], /aɪt/>[ʌɪt], /aʊt/>[ʌʊt]).

Another area of vowel production concerns the pronunciation of unstressed syllables. Deterding (2010) suggests that stressing unstressed syllables in di-/multi-syllabic words (/ə/ in ‘condition’) is common in many L2 English varieties and may enhance intelligibility in international communication. From a teaching and learning perspective, while L2 learners may need to learn to recognise the presence of schwa as a marker of unstressed syllables (particularly if the prospective interlocutors are L1 speakers), they need not produce schwa in all places (Levis 2018). When producing an utterance, what is important for L2 speakers is to be able to identify key words (or syllables) in the stream of speech and to lengthen the stressed syllables.