In a blog post I made a long time ago, I mentioned a paper that detailed the various dialects of Taishanese (namely 台城 Taicheng、端芬 Duanfen、荻海 Dihai、新昌 Xinchang). It’s likely the most thorough comparison of the dialects of Taishanese I’ve ever seen, but I’ve noticed that a few of the sounds have changed since when it was published (in 1950!). This post will list some of the changes the Taicheng (台城) dialect has experienced since this paper was published.

The reason I say some is because it turned out there’s quite a lot of changes that have happened in just 67 years, and this blog post has been stuck in limbo for far longer than it should have.

Disclaimer: I speak the Duanfen dialect, though for this post I will discuss the Taicheng dialect. The reason for this is that whereas the Duanfen dialect today hasn’t been documented by anyone, the Taicheng dialect in the present era has three pretty good references:

• (a) Stephen Li,
• (b) 台山方音字典, and
• (c) 漢字古今音資料庫

These will be used as sources for the Taicheng dialect as spoken today.

Shift from tʃ → ts; tʃʰ → tsʰ; ʃ → s

[ts], [tsʰ] and [s] is used instead of [tʃ], [tʃʰ] and [ʃ] according to all three sources.

It could be related to a similar sound shift in Cantonese that affected the same sets of characters, where [tɕ], [ts] → [ts]; [tɕʰ], [tsʰ] → [tsʰ]; and [ʃ], [s] → [s].

While in Guangfu Cantonese (henceforth referred to as Cantonese) this change was a merger, in Taishanese no merger resulted from this pronunciation shift. This is because historical Cantonese [ts] is pronounced as [t] in Taishanese, [tsʰ] is [tʰ], and [s] is [ɬ] (except for a few exceptions, of course). These changes can be demonstrated with a pronunciation comparison of some common characters:

initial ʷ → v; ʷɔ, ɔ → ᵘɔ

All cases of [ʷ] appearing at the start of a syllable has changed into [v]. This means that 黃 is now pronounced as [vɔŋ²²].

In 1950 (see p17 of the paper), one of the two Taicheng speakers pronounced the 戈 final as [ʷɔ] when after [k], while the other speakers sampled (except the Duanfen speaker) pronounced the final as [ɔ]. It seems that nowadays in Taicheng at least, all cases of [ɔ] (ie. in the 歌 final, division 1 戈 final, and 2nd division 魚 final) have shifted to [(ᵘ)ɔ] (a), [uɔ] (b), or [ᵘɔ] (c).

This results in characters such as 個 [kɔ³³] and 過 [kᵘɔ³³] having merged into [kᵘɔ³³].

Note that this shift does not cause any mergers with initial [ʷ], as initial [ʷ] changes to [v], resulting in 和 [ʷɔ²²] becoming [vuɔ²²] or [vᵘɔ²²].

i → ei shift

There is a shift in Taishanese that affects all but the t͡ʃʰ, t͡ʃ, and j initials (now t͡sʰ, t͡s and j). This is a parallel to a similar vowel shift that has no doubt been established a long time ago in Cantonese. Just like how 畀 had a shift from [bi³⁵] to [bei³⁵] in Cantonese, the vowel shift changes the pronunciation of this character in Taishanese from [i⁵⁵] to [ei⁵⁵].

Stephen Li (a) gave this vowel shift the lofty title of Taishanese Vowel Shift and he has this to say:

It is not that all the [i] sounds have disappeared – you can still hear them today as the shift is on-going and the process is not yet completed. Sometimes you can hear both [i] and [ei] from the same speaker depends on the the (sic) particular word uttered or whether the syllable is stressed.

His comment about the ongoing nature of the vowel shift is interesting as (b) treats the pronunciation as if this vowel shift is already complete, with the exception of 耳 which is listed as both ngei— and ngi—.

(c) is a bit split on whether to use ei or i — I can’t spot any pattern to it, which backs up Stephen’s statement on the vowel shift being ongoing.

Syllabic ŋ, m → m

Syllabic ŋ occurs in both 1950s Taishanese and Guangzhou Cantonese when a character has the 疑 initial (Middle Chinese ŋ) and the 模 final (Middle Chinese uo). It was kept distinct from syllabic m, which only occurs in the character 唔.

However, in Hong Kong Cantonese as well as present-day Taishanese, syllabic ŋ has merged with syllabic m, resulting in 吳 [ŋ²²] being pronounced identically to 唔 [m²²].

(a) has the merger complete, listing both syllabic ŋ and syllabic m as m. (b) and (c) still makes this distinction by listing both ŋ and m for syllabic ŋ. (Note that (c) does not have 唔 in its database.)