Pronunciation model: Hong Kong English

View the key for Hong Kong English here.

Hong Kong has three official languages: Cantonese (the ‘usual language’ for 89.5% of the population according to based on 2011 census), English (3.5%), and Putonghua (Modern spoken Mandarin, 1.4%). Yet for historical reasons, English remains an official language and widely used in governmental, legal, professional, and business environments. Most speakers are at least bilingual in Cantonese and simple English, while trilingualism with Putonghua (with variable fluency) is growing since sovereignty of Hong Kong returned to China in 1997. There remains debate as to whether Hong Kong English, being spoken primarily (but not exclusively) by first-language Cantonese speakers, is a distinct variety or a learner interlanguage (for more on this discussion, readers are directed to chapter six of Setter, Wong & Chan 2010).

The two primary sources used in compiling the Hong Kong English [HKE] model for the OED are Setter, Wong & Chan’s (2010) overview of HKE, and Cummings & Wolf’s (2011) dictionary (DHKE), with support from a compiled feature list by the Hong Kong Institute of Education (2014). The HKE transcription is intentionally broad, as the variety has no fixed or established phonology. Considerable variation is found, and some patterns in the OED model are necessarily extrapolated from other models or estimated where data is lacking.


Note that the vowels in NEAR, SQUARE and CURE may be produced as two syllables (disyllabically) by some speakers when no consonant follows (‘open’ syllables), for example here as /ˈhi.ə/. This also demonstrates another feature, namely that the vowels of NEAR, FACE, PRIDE, VOICE, MOUTH and GOAT could legitimately be transcribed with ‘closer’ close points:

/iə/, /ei/, /ai/, /ɔi/, /au/ and /ou/ respectively.
This would be potentially more reflective of the Cantonese influence, but the OED follows the DHKE and Setter, Wong & Chan in using the more open symbols. The presence of ‘full’ vowels in unstressed syllables (as in the first syllable of absorb) is assessed on a case-by-case basis (although the default is as in AGO).

The OED HKE model is non-rhotic and does not have flapping of /t/, although Deterding, Wong & Kirkpatrick (2008) and Setter, Wong & Chan (2010) note a significant minority of HKE speakers who prefer to adopt North American features. Stress structure is relatively under-studied at present, but significantly differs from British placement in a range of respects, most notably with particular suffixes (such as -ative, which attracts stress to the a unlike in British) and compound forms (in which main stress falls on the final part, rather than following British patterns, hence banana TREE rather than baNAna tree). Word stress is further complicated by the use of lexical tone (rather than lexical stress) in Cantonese, and there is some evidence that HKE may mesh the two (see the summary of Luke 2008 in section 2.14 of Setter, Wong & Chan 2010). Durational differences between levels of stress are not as strongly maintained as in British English, and HKE has fewer ‘weak’ syllables with:

/ə/ or similar
than it does ‘unstressed’ syllables (with a full vowel). The Cantonese influence on HKE seemingly results in a less strongly stress-timed rhythm (potentially appearing more ‘syllable-timed’, with less distinct durations between stressed and unstressed syllables). This is shown in the OED by a flattened stress structure, with all syllables transcribed as having at least secondary stress marks (unless they have a clearly reduced /ə/ vowel) and with double-primaries permitted on multi-word entries.

As mentioned above, many features are extremely variable from speaker to speaker, and the OED model is attempting to provide a largely consistent set of pronunciations amidst considerable inconsistency. In this respect, it is important to note that the pronunciations given are descriptive rather than prescriptive, and indicative rather than comprehensive. For example, the HKE model reflects the common (but not unlikely universal) absence of assimilations, elisions, and linking typical of British English (e.g. handbag as /hɛndbɛɡ/ without the bracketed /(d)/, and a lack of /j/-coalescence in most words with /tj/, /dj/, /sj/, /zj/).

HKE is subject to a broad two-consonant limit in syllable final clusters, with a reduction from even two consonants to one being common. /t/ and /d/ are particularly vulnerable to this, and word-final stop consonants may be deleted when preceded by nasal or vocalized lateral sounds (this is not an explicit feature described by the sources, but the evidence consulted so far is consistent with this approach). /i/ is often added to avoid final consonant clusters necessitated by inflectional suffixes (e.g. equipped /ikwipit/ – note the precise realisation of the final two vowels may not be as close as /i/ but is represented as such at the phonemic level). A specific syllable-initial cluster feature is that both /tr/ and /tw/ are pronounced as /tʃw/.

A number of British English contrasts are represented differently in HKE in the OED, even though they are only common features rather than universals. /z/ and /ʒ/ are transcribed as their voiceless counterparts, while the dental fricative /ð/ is usually /d/ (there are documented instances of it becoming /z/, as in this, but for consistency the OED retains /d/). The voiceless dental fricative /θ/ is more commonly documented as becoming /f/ rather than /t/ (as in thanks, but again for consistency the OED retains /f/). /v/ may become /w/ or /f/ (as in vine or even; this being reviewed as HKE entries are added to the OED). Word-finally, the stop consonants (/p/ and /b/, /t/ and /d/, /k/ and /ɡ/) have neutralized voicing contrasts, and all are represented as the voiceless one of their pair. Affricates /tʃ/ and /dʒ/ are often subject to this neutralization but there is greater variability and these remain represented separately in the OED. Word-final stop sounds have no audible release and are often glottalized, but this is reflected only in the spoken pronunciations (not the transcriptions).

Syllable-final /l/ is vocalized to /o/, even in final clusters (e.g. childhood). Although vocalization is an allophonic (phonetic variant) pronunciation not usually represented in the OED pronunciations, its consistency is a highly distinctive feature of HKE and is represented for this particular variety. The /l/-/r/ substitutions which are sometimes noted are not reflected here, nor the /l/-/n/ merger (in one study occurring in fewer than 10% of tokens).

Another feature of HKE is that /s/ can become /ʃ/ in some words, while /ʃ/ may become /s/, with resulting loss of contrast. These occur too frequently to ignore, but generalisation proposals (e.g. that /s/ moves back to /ʃ/ before back rounded vowels) do not currently appear to capture the variation observed. The default choice remains as in the British pronunciations but are assessed on a case-by-case basis, forming the area of least consistency in the HKE model.

A number of other features have been considered, but their variability has been deemed too great for inclusion at the present time. These are features such as: initial consonant cluster simplification (which may involve a reduction to a single consonant, or the insertion of an extra syllable with the addition of an /ə/ or /i/ vowel, or they may not be simplified); two patterns noted for two-quality (diphthong) vowels followed by consonants (they may preserve the diphthong and lose the consonant, make the diphthong into a monophthong and produce the consonant, or produce both; the OED follows the latter as it is the least collapsed form); and inserted final consonants (word-specific; they require a higher degree of evidence to support their inclusion than more common features).