The stop in syllables that end in a homorganic nasal-plus-stop cluster (in English, these clusters are /mb, mp, nd, nt, ŋɡ, ŋk/) is often elided. Word-final /-mb/ and /-ŋɡ/ never occur in Modern English, for example, although their dumb Middle English spellings hang around.
Final /-nd/ does occur, though not always, but it's frequently neutralized with /-nt/, especially after a stressed vowel and preceding an unstressed one. As in twenty.
This interacts with the neutralization of /d/ and /t/ in the same environment; in American English, both go to [ɾ], as in betting and bedding, which differ -- if at all -- only in the allophonic vowel length of stressed /ɛ/ in the first syllable of bedding.
Upshot: In American English, /'twəni/ is the normal pronunciation, /'twɛni/ is somewhat more formal and careful, and /'twɛnti/ is fastidiously careful.
I believe that all verbs that end in -ide (/aɪd/) regularly produce adjectives ending in -isive (/aɪsɪv/) where the diphthong is preserved.
- cicatrisive
- collisive
- decisive
- derisive
- divisive
- incisive
- indecisive
- precisive
- previsive
- subdivisive
- subrisive
- undecisive
In contrast, verbs ending in -mit /mɪt/ regularly produce adjectives ending in -issive /mɪsɪv/, with a double s to keep the i vowel “short”.
- admissive
- commissive
- dismissive
- emissive
- fissive
- insubmissive
- intermissive
- intromissive
- irremissive
- manumissive
- missive
- omissive
- permissive
- photoemissive
- promissive
- remissive
- retransmissive
- submissive
- transmissive
- unsubmissive
Both the OED and the OALD agree with this.
I’m thinking therefore that the “other” pronunciation is less broadly accepted, perhaps a regionalism or sort of hypercorrection due to interference with divisible /dɪˈvɪzɪbəl/.
It certainly sounds odd to my own ear.
Best Answer
The pronunciation of "bald" as /bɔ:ld/ is older. What seems to have happened historically for some speakers of British English is phonemic shortening of the sound /ɔ:/ to /ɒ/ in some cases when it comes before a consonant cluster /lC/ (where C stands for any consonant). This change, and the resulting variation in pronunciation, is described in the following post on John Wells's phonetic blog: scolding water. Here is the relevant paragraph:
I also found this Livejournal post about it: scolding slurry, which suggests that it is connected to the loss of the LOT-CLOTH split in most modern forms of standard British English. Apparently, another environment where shortening occured was in words like austere (listed as having the CLOTH vowel /ɒ/ by the British Library). This shortening is further discussed in this article by Piotr Gąsiorowski: The History of [ɔː]: Is There Regular Orthographically Conditioned Sound Change?; Gąsiorowski mentions a few more words with exceptional /ɒ/ such as sausage, laurel and cauliflower. I'm not sure if any American dialects that maintain a distinction between the vowels of COT and CAUGHT show the effects of such shortening in salt and related words; it seems to be mainly a British phenomenon.
It appears that your dialect shows shortening of /ɔ:l/ before voiced consonants as well, such as /d/. But this does not apply before the secondary cluster /ld/ formed when verbs that end in /ɔ:l/ are followed by the past suffix /d/. (There are attested sound changes with conditions like this in some other dialects of English, such as the Scottish vowel length rule or the distinction between words like freeze/frees and bruise/brews in Geordie.)
So for you not only halt, vault, false, alternative /hɒlt/ /vɒlt/ /fɒls/ /ɒlt.../, but also bald, scald /bɒld/ /scɒld/ have the same vowel as golf, doll, dolled /gɒlf/ /dɒl/ /dɒld/. But walk, maul, mauled /wɔ:k/ /mɔ:l/ /mɔ:ld/ have the same /ɔ:/ as torn, sort: they were not shortened because the first one lacks a pronounced /l/, the second one lacks a consonant after the /l/, and the third one has a consonant cluster only across morpheme boundaries.
The Oxford English dictionary (OED) lists alternative pronunciations with /ɒl/ instead of /ɔ:l/ for halt, vault, false, fault; but not for bald and scald. This fits with what Wells says about most speakers only having shortening before clusters of l + a voiceless consonant. However, there also seems to be some word-by-word variation, as in the case of Wells's mother.
Unfortunately, there's not a huge amount of evidence for the outcome of words with /ɔ:l/ followed by a voiced consonant, because I can't find any other common words that end with auld/ald. But here are some test words with this sound in the middle of a word; I'd be interested in knowing how you (or any other British English speakers) pronounce them:
This analysis is just based on your described pronunciations. I have no personal experience with this; I'm American with the COT-CAUGHT merger, so I have /ɑ/ in all of these words.