Romanising Shataukok Hakka Sounds

The following is used merely as a guide to pronunciation. Agreeing a set of sound values such as this helps us to see the dialect written down. In doing so, we are in fact beginning to standardise how the language is spoken, though I do not intend that this should be the case. We note here that, within the Shataukok region itself there are variations between the local speakers also. Such minor differences in pronunciation are not great. When speakers communicate, they overlook such quirks because the jist of the conversation is realised. On paper, however, differences are more glaring. Since a single word may be spoken in alternative pronunciations, any word list given is not intended to be fixed as a standard. It is merely how the web author's dialect is recorded. The nature of langauge, afterall, is fluid and so changes, from one speaker to the next.

N.B. Please note that I have used the sounds of English wherever possible to cover the Sa Tdiu Gok Hakka sound. So, the sound 'tong' has the initial 't~' like the english 'taking' and so on.

Letter Sound Values if used as
InitialsEnd Letter(s)

ashort in 'cat' always a long value as in 'tar'

b'bigger'p can be substituted

ch'chance'never used as an ending

d'dog't an be substituted

eshort 'e' as in 'bed'long 'e' rhymes with 'hair'

f'fat' never used as an ending

g'get'k can be substituted

h'hat' Not used, except to lengthen a vowel - see also the ending ~r (Now discontinued as of Nov 1997, but may still be found in the dictionary pages)

ilike short 'i' in 'hit'always long rhymes with 'bee'

j'jam'never used as an ending

k'kite'g can be substituted

l'long'never used as an ending

m'man'soft m as in 'ham'

n'n' has become mutated into 'l' in Sa Tdiu Gok Hakka when used as an initialnasal 'n' in 'an'

ngmerged nasal 'ng' in 'singer' as ending in 'song'

oshort 'o' as in 'hot'always a long value rhymes with 'core'

p'park'b can be substitued

rnever used as an intialNot used except when untrilled 'r' in 'ore' or a soft 'h' is need to lengthen a vowel (Now discontinued as of Nov 1997, but may still be found in the dictionary pages)

s'sand'never used as an ending

sh'shout' never used as an ending

t'tap'd can be substituted

tdlike the 't' in 'star' never used as an ending

tslike 'ts' in 'its'never used an ending

ushort 'u' as in 'but'always a long value rhyming with 'loo'

v'van'never used as an ending

wThis is the same as the 'v' sound never used as an ending

ySome STdG Hakka people use y, but at this site, all y~ ('yesterday') initials are mapped onto z~ never used as an ending

z'zest'never used as an ending


There are five vowels; a, e, i, o, and u. They all have short values when followed by a consonant with the exception of 'h' and 'r'. If it is followed by any vowel, it can have a long or a short value. Addition of 'r' or 'h' acts to lengthen the vowel or may be omitted. Generally, if a word ends in a single vowel, it has a long sound.


Dipthong consonants such as 'ch', 'ng', 'sh', 'td' and 'ts' are thought of as a single letter. Letter 'r' is the only one that is never used as an intial consonant. When a consonant ends a word, it is usually soft in quality. There are many consonants that do not appear as endings. Values of soft b,d,g are interchangeable with soft p,t,k.
The intial consonant 'ng' is different from 'ng' when used as an ending. As an Initial it is hard, whereas used as an Ending it is soft.

Revision of above table

After some very useful discussions with Mr. Thomas Chan, a cantonese speaker, I have revised the above list of words and sounds to that below.

The ending -ioi occurs only in two words of Hakka I know of. One is ngioi4 (character above) in the compound ngoi4 fo3 which has something to do with the chinese medicinal art of moxibustion, and kioi4 meaning tired or weak.

I have prepared a number of Real Audio sound files for demonstating the words in read life. Please click on the table below and compare with the sound inventory above:

62,144 bytes
Endings -a-
19,608 bytes
Endings -e-
13,772 bytes
Endings -i-
53,524 bytes
Endings -o-
19,332 bytes
Endings -u-
17,384 bytes


Over the last two months of 1997, I have been in correspondence with two linguists, Thomas Chan, a student at New York State's Cornell University, and Professor Marjorie Chan at Ohio State University. They have asked me the exact nature of my 'td' value in the romanisation above. For those familiar with Mandarin, then it agrees with the PinYin 't' value. The Yale Cantonese 't' is also like my 'td'. Since I have approached this project from the point of view of a speaker of English, the values of the 26 letters of the English roman alphabet have been used wherever possible. So then, the letter 't' in STdG Hakka has already been taken to transcribe the English 't' as in the words "tank" and "title", and so the 'td' combination has been created merely as a visual aid, rather than the linguistically correct IPA forms.
Click here for a short example list of STdG Hakka 'td~' words below is the list with the translations.

tdai4 tdam1 tdang4 tdau2 tden1tdiak5 tdiu2 tdoi2 tdong2 tdu4 tdung3
big to reach to listen peach sky to kick head to carry sugar rabbit bucket

ts versus j

I have noted in the table of Sa Tdiu Gok sounds, that certain times one form of the set j- ji- and tsi- can be used interchangeably, however, this is not universal, as some of the resultant sounds end up sounding like a non-Hakka effort. Click here for an audio explanation of ts versus j.

Sa Tdiu Gok Romanisation and the Value of each Letter

The Hakka sounds may be thought of as consisting of two parts. These are the starting romanisation letter, called the initial, and then the attached ending called the final. For instance, the sound kong may be split into the intial k~ and final ~ong. The sound 'kong' embodies the sound of the word. The following is a rationalised list of the initials found in Sa Tdiu Gok Hak Ga sounds. There are 24 initials and 47 finals.

Hakka Initials

Each one of these initials can begin the sound of a word. In order to have a rationalised set of letters, the sounds of these letters are fixed when used as the initials. However, when they are used as an ending, they hold different values. So, for each letter of the romanisation, they embody more than one type of sound value, depending on its use as an initial or as an ending.

Hakka Endings

a. Sa Tdiu Gok Vowel Use
Vowel Short Initial and MiddleLonger Final
a vowel in cata in father
e vowel in setai in fairy
i vowel in fitee in meeting
o vowel in hotor in order
u vowel in butue in blue
b. Sa Tdiu Gok Ending Types
Nasal endings~m ~n~nglonger duration
Final Stops ~p (~b) ~t (~d)~k (~g)abrupt shortening of the sound
a,e,i,o,u have short sounds when used as an initial or in the middle of the sound, but have a longer duration when used at the end of a sound. We note here that when a vowel ends a sound, a resultant legato tone is produced. These are associated with tone numbers 1, 2, 3, and 4. See Table a.

m,n, and ng all have nasal qualities which allow for the lengthening of the sound. This produces somewhat of a glide, or humming of the end letter's sound. It is associated with the long or legato tones, (tones 1, 2, 3, and 4) see Table b.

On the other hand, p (b), t (d), k (g) are three sound values that creates a stop to the sound, hence shorting the sound somewhat. These lead to the production of the short or staccato tones (tones 5 and 6) see Table b.

You may be wondering why we have also have the bracketed letters. The sound is stopped short so that the full effect of the letters are not realised. This gives the ending "~p" a "~b"-like quality, ending "~b" a "~p" like quality and so one. The convention here at this site is to end staccato tones with the stops '~p', '~t' and '~k'.

This convention allows comparison to Cantonese which also has all six consonant type endings, though Cantonese initials and final vowels are slightly different. It is considered that Cantonese, also a Southern Dialect like Hakka, is more conservative in preserving the ancient vestiges of the Middle Chinese dialect of the Tang Era. However, Hakka is quite faithful to this also. Where this is not true, the following two points demonstrate the reason why.

There is a tendency for Hakka to have a ~in ending where in the Cantonese there is a ~ing ending. Also, in some words, the Cantonese ends in ~k whilst the Hakka has changed to ~t. But, we note that this is not always 100 percent the case. (M. bei C. baak; H. bet; J. hoku K. baeg E. north)

Though these two points detract from the prestige of Hakka as a faithful conserver of the ancient endings, it is noted however, that Cantonese has over time mutated the ng~ initial into y~ whereas the Hakka retains the ng~ initial feature faithfully in most cases.

Back to Hakka Links : Back to Dylan's homepage

This page was last updated on Saturday 28th February 1998. Started on Sat 29.3.97.

©Dylan W.H.S. 1996 onwards

E-mail me if you like...