On Sat, 8 Mar 2003, John Cowan wrote:

> Peter T. Daniels scripsit:
> > In pinyin, the four tones are indicated by the four accents macron,
> > acute, hachek, and grave respectively. If you wanted to transliterate
> > Cantonese with a similar system, there are certainly another three or
> > four familiar accent-marks that you could employ -- umlaut, circumflex,
> > dot, for instance
> Unfortunately, although the Mandarin tone accents are nicely iconic with
> respect to the standard pronunciation, no such thing can be done for
> Cantonese, where it would be necessary to associate the six (tonetic)
> tones with arbitrary marks, and there would be constant tension between
> wanting to use the same mark that the Mandarin reflex uses and wanting
> to be iconic. Still, without doubt the thing could be done.

For complete iconicity, Chao-style tone letters (which are in IPA) could
be used, at risk of increasing the typographic requirements. But with five
tone levels defined, they can be hard to read if not neatly

Yale's romanization scheme for Cantonese halved the diacritics problem by
infixing an <h> for low register tones, although its choice of diacritics
does not depict the modern contours (rather, it semi-arbitrarily assigns
the grave to ping-class tones, the acute to shang-class tones, and the
macron to qu-class tones).

I do believe I have seen less-known romanization systems that do increase
the number of diacritics, such as using doubled macrons, but I don't
recall if any attempt was made to have the diacritic suggest the tone

Somewhat iconic, but in a different way, is to transcribe the tones in
such a way that the tones in the same tone classes (ping, shang, qu, ru)
but in different registers, are related. e.g., a subscript prefixed "c"
means the yinping tone (modern 55 contour); the same but underlined means
the yangping tone (modern 21 contour). Numerous 19th and early 20th c.
schemes, like those of Samuel Wells Williams, Dyer Ball, Louis Aubazac,
Ernest J. Eitel, et al., were like this.

> What's really challenging is to come up with a tonal spelling for
> Cantonese. After a lot of work, I finally came up with a satisfying
> tonal spelling for Mandarin that doesn't deviate too far from HYPY.

There's a Hanyu Pinyin-like system for Cantonese, which is used in books
(about Cantonese) from the PRC, but despite all the other
misrepresentations and compromises made to make it resemble Hanyu Pinyin
(for Mandarin), it transcribes tones with superscripted numerals.

There's a copy of the scheme[1], to which the transcriber has added
commentary and (his) transation. Among the distortions is that the
affricate and sibilant initial phonemes are written with allographs <j>,
<q>, and <x> (rather than <z>, <c>, <c>) before a high front vowel, in
mimicry of Hanyu Pinyin (where they may be treated as phonemes in some
analyses) and the labiovelars initials are transcribed as if they were a
sequence of a velar initial and a following [u] (a misrepresentation of
the phonology). It does make some attempt to depict the initial glides
[j] and [w], rather than treating them as [i] and [u], however. One
example of compromise, which is not explained in the scheme, but which can
be seen from examples in the books that use it, is the transcription of
the final consonants as <b>, <d>, <g> for unreleased, unaspirated, and
unvoiced [p], [t], [k] (e.g., 'white' is <bag>, 'bone' is <gued>) which is
totally unlike other romanizations, and while internally consistent, there
is no aspiration distinction in non-initial position.

[1] http://www.sungwh.freeserve.co.uk/chinese/gzhhpy.htm

> Westerners who knew Chinese characterized it as "readable but ugly" --
> people always call spelling reforms "ugly", AFAIK. (Yes, I know about
> Gwoyeu Romatzyh, and a more unsystematic and messy torturing of the
> conventions of Latin script I can scarcely imagine.)

Have you seen CHAO Yuen Ren's own Gwoyeu Romatzyh-like system for
Cantonese? It's in his _Cantonese Primer_ (Cambridge, MA: Harvard
University Press, 1947), which grew out of materials used in a summer
course at Harvard in 1942. His solution to the depiction of tones is by
using different letters for the initials, depending on if they are high or
low register tones (i.e., making use of the correlation in most cases that
the high register tones are associated with unvoiced initials, the low
register tones with the voiced initials). The zhongru, or xiayinru (lower
yinru) tone, I believe, was treated by making use of the correlation in
most cases with the vowel tenseness to distinguish it from the yinru (or
shangyinru--upper yinru). If that isn't "torture" enough, Chao also
builds into it (I don't think it actually had a name, like Gwoyeh Romatzyh
did) the distinction between two sets of affricates and sibilant initials,
knowing full well that by the 1940s the distinction had been lost in
standard Cantonese--the idea was that the additional burden in distinction
would not be a problem for new learners (but admitedly one for those who
already spoke it), and give them an edge when/if later learning Yue
dialects that maintain the distinction, or even Mandarin (the z/c/s vs.

(Chao also has GR-like schemes for his own dialect of Wu, and "General
Chinese", a sort of pan-Sinitic language that incorporates the union of
their distinctions, both reprinted in his and Anwar S. Dil's _Aspects of
Chinese Sociolinguistics_ (Stanford: Stanford University Press, 1976)).

Another tonal spelling scheme, if it could be said to be one, is that used
by circa 1940s US War Department books (technical and field manuals)
teaching Mandarin and Cantonese, which made use of lower/upper case and
normal/italics to represent four tone levels, and with combinations (i.e,
within the syllable, like <paA>), up to eight tones.

> But going from four tones to six, and having to represent vowel length
> and roundness to boot, makes Cantonese a much tougher proposition.
> I never came up with anything that wasn't so ugly that I myself couldn't
> stand it, never mind anyone else.

I don't think anyone ever really attempted to use Latin script seriously
for Cantonese besides linguists and teachers, although there are
completely romanized versions of Cantonese bibles and other religious
works, mostly 19th/early 20th c., though. For the Han
character-illiterate, I suppose.

Thomas Chan