Some dialects of Hakka have a splitting in the Qu tone. How can we distinguish it, without asking a speaker of a Hakka dialect with the splitting to look over a list of characters? Cantonese is a Chinese dialect which shows ying-yang splitting in the Qu tone. I have compared characters in the Qu tone in Meixian like Hakka dialects and then compare it with Cantonese. Those of Cantonese tone 3 corresponding to a YinQu tone are used to sort the characters of characters which we suppose correspond with YinQu in Hakka. All other Qu characters of Hakka are supposed to belong to the YangQu tone. Whether the latter assumptions are correct, remains to be seen.
On the Hakka Forum in Asiawind, in relation to the Hakka Pinyin thread,
http://www.asiawind.com/forums/read.php?f=1&i=7514&t=7514
I wrote:
Romanisations are contained in angled brackets < and >, and IPA are contained in square brackets [ and ].People who speak the six tone dialects which is similar to Meixian and my Hong Kong dialect can understand this as there being two types of tone 4 (hi4 sang1 or Qu Tone in Chinese linguistics), one higher and one lower. So, Git Lung will can sort all sixtone dialect tone four into two types, which one type he already associates with tone four, and the tone seven is basically a different pitched tone four, if you know what I mean.
The split in the tone four or Qu tone into upper and lower Qu is down to the voicing of the initials during the Middle Chinese stage of historical Chinese phonology. Voiced initials tended to become lower register tones, and unvoiced initials tended to become upper register initials. Consequently, if you want to analyse the upper and lower registers of the Qu tone, you need to look at the character's reconstructed MC pronunciation and compare it to modern Hakka.
One of the things about Hakka today is that there are fewer voiced initials than in MC. For example [b] [d] and [g] no longer exist in modern Hakka. They have become voiceless aspirated stops [p'], [t'] and [k'] corresponding to Hakka pinyin <p><t><k>. Voiceless unaspirated stops [p], [t] and [k] correspond to Hakka pinyin <b><d><g>
Other voiced initials include [ŋ] <ng>, [m] <m>, [n] <n>, [l]<l>, [j] <y>, [v] <v> etc. Now, Git Lung, try and read the characters which have tone four in the list of characters from Dr. Lau's input, and tell us, with the voiced initial ones, they are mostly in which tone, and have a look at those which are aspirated beginning with <p>, <t> and <k> in the romanisation, and see which is which. You'll find a fairly even distribution I am guessing.