I have had it figured out a long while ago -- /b/ and /m/ are allophones in Hokkien -- the former occurring only before non-nasalised vowels, and the latter after nasalised ones. In other words, one can analyse one of the pairs (usually /m/) as a surface realisation of the other under specific circumstances.
vijayjohn wrote:Wiktionary says this is also pronounced biān in Hokkien.
bian would be the literary reading for 麵, while mi (underlying analysis: biⁿ) would be colloquial one. The character 明 is read mia (for underlying biaⁿ) in the colloquial reading, and beng in the literary one.
The same applies to the pairs /g/ and /ŋ/ (henceforth ng in transliteration) and /l/ and /n/; the nasal in each pair being the surface/phonetic realisation before nasalised rhymes.
Sounds straightforward enough ... until one encounters nasalisation occuring in literary readings (where nasalisation is
not supposed to happen) or colloquial readings with no historical evidence of a nasal coda.
It seems that /ŋ/ is a regular outcome in characters from Qieyun syllables ŋɑ (我) and ŋuo (吾五), hardly occuring elsewhere. Nasalisation sporadically occurs elsewhere -- this deserves further scrutiny.
Ciangciu/Zhangzhou 漳州 sporadically adds nasalisation to characters read without nasalisation in the other two major Hokkien accents: 奴艾. A plausible explanation is extra-dialectal inmixture from Sinitic varieties which have fully retained nasal consonants, especially Teochew.
火 has the literary reading hɔⁿ (note the nasalisation) in all three accents -- suggesting the possibility of 'ghost nasals' occurring in syllables not deriving from Old Chinese syllables with nasal initials.