cedh wrote: ↑09 Mar 2018 11:55An idea: Both of these numerals had not just one but two consonants after the stressed syllable in Latin, so I could easily imagine that your usual allophony rule for word-final vowels before a nasal-initial word in the same phrase would be blocked in this situation, probably keeping the vowels distinct. If you don't like to decree that offhand, you could also check what your sound changes would give if you treat {number + noun_with_initial_nasal} as a single word.
Ah, but there's the rub. I didn't mention it because I didn't want to get too into the weeds initially, but attributive numbers
are treated as part of the noun they modify. Through a series of steps, that fact leads to the current predicament.
To lay it all out: Those Latin consonants cause the initial consonant of the noun to geminate precisely because of the phonological linkage. You get nasalization because geminate consonants take up a slot in the syllable coda, and coda nasal consonants nasalize the preceding vowel. Then, because of extensive mergers in nasal vowels,
sí and
sé end up with the same vowel. Hmm, I hadn't thought about how long that chain of logic is; hopefully that was clear.
Jackk wrote: ↑09 Mar 2018 12:30Alternatively, French irregularly revived final -t in the word sept "seven" after it was dropped, probably for reasons of clarity like this.
I'm guessing the /t/ came from the prevocalic form. Though, that is something. Prevocalically,
sí becomes [siːz]. Perhaps I'll just extend the long vowel across all forms. That also blocks the gemination and nasalization I mentioned above.