Vowel hiatus and non-diphthong vowel pairs (compared to Romance languages)

Question

Classical Latin's 6 major diphthongs are clear-cut, phonologically speaking. We know ae is pronounced as one phoneme, such as in [ˈsae̯.pɛ], "saepe." However, we often come across words that have 2 vowels next to each other that aren't a diphthong (ex. Italia, puella). As a Romance speaker, I am curious how this would have been pronounced in spoken Latin.

Spanish and Italian distinguish between "strong" and "weak" vowels. /a e o/ are strong and /i u/ are weak. Combining a strong and weak vowel makes a diphthong. For example, Italia is pronounced as /iˈta.lja/, not /iˈta.li.a/. The vowel combination /ia/ is pronounced like a diphthong /ja/, not /i.a/. This phenomenon reduces hiatus in the spoken language. Spanish uses the acute accent to indicate when hiatus does occur, like in /panadeˈɾia/ (panadería).

I have not yet found any sources clarifying this topic in Latin. In fact, most Latin transcriptions imply that hiatus is standard. The word Italia is transcribed as /iˈta.li.a/. and puella as /puˈel.la/. Was this standard pronunciation in Latin, or more of a formal ideal? I could imagine Romans (especially commoners) diphthongizing these vowels in quick speech, and this could carry over into the Romance languages. Any references from ancient scholars would be extremely helpful.

Reference Guides:
Diphthongs and Hiatus in Italian
Diphthongs and Hiatus in Spanish

Answer 1

I think poetry is the biggest data source indicating that hiatus was usual in Latin for i e u + vowel.

In the stage of the language that was ancestral to the Romance languages, both i and e were reduced when unstressed to a glide [j] before a following vowel. However, the gliding process doesn’t seem to have worked quite the same as in modern Spanish/Italian. My understanding is that Spanish and Italian syllabify the glide along with the preceding consonant as an onset (as in your example [iˈta.lja]), so the preceding syllable is “open” (ends in a vowel). In Proto-Romance, the preceding syllable seems to have been closed in words like folia (the ancestor of Standard Italian foglia, which has a long palatal consonant) which could imply a syllabification like [ˈfɔl.ja] or possibly [ˈfɔl.lja]. See also the geminate b in Italian rabbia vs the single b in Latin rabies.

I read some paper about the syllabification issue that I will try to cite when I find it. It's not the source that I remember learning about it from, but I see that it is mentioned in the chapter "Syllable, segment and prosody" by Michele Loporcaro in The Cambridge History of the Romance Languages: Volume 1, Structures, edited by Martin Maiden, John Charles Smith, and Adam Ledgeway. Loporcaro gives the following examples and analysis:

(57)        Pt.   Sp.  Fr.   Prv.   Rm.    It.
     sapiat saiba sepa sache sapcha sapcha sappia

While only Italian preserves gemination, Gallo- and Rhaeto-Romance have glide hardening, pointing to a syllabification -C.j- that is confirmed by the non-application of /a/-palatalization in French. The Spanish and Portuguese examples in (57), on the other hand, show in this case the application of a conflicting process, viz. -Cj/w- > -j/wC- metathesis, as further reflected in the outcomes of e.g. SAPUIT > OSp. sope, BASIUM > Sp. beso (see Rini 1991).

(pages 101-102)

Replacement like this of /i./ in hiatus with /.j/ (causing the preceding syllable to scan long from the presence of a coda consonant) is occasionally seen in poetry. Likewise, in poetry, U before another vowel was sometimes replaced with the consonant V or vice versa, as described in Scott Brown’s answer here: How were vowels u and i discerned from consonants v and j?