Did an internal m nasalize the preceding vowel?

Question

We know that the final m was not a full consonant in classical Latin, but denoted nasalization and elongation of the preceding vowel. See this or this old question for more details. Was this effect limited to word-final position, or did an internal m nasalize the preceding vowel as well?

For example, how were sumus, summus, amplecti, sumpsisti pronounced? I would assume m had a fully consonantal nature between two vowels to avoid hiatus, but that does not rule out nasalization. There seems to be more freedom as the first consonant of a consonant cluster. For example, perhaps summus could be [sũ:mus] instead of the pronunciation [sum:us] I was taught. Sometimes the vowel before a consonant cluster beginning with m is already long (eg. sūmpsistī), which makes the combination of nasalization and elongation sound unlikely. The vowel before a final m is always short as far as I know — when m is pronounced as a consonant.

Answer 1

Sampson 1999 summarizes research on nasalization of vowels in Latin:

"there is every reason for believing that in the history of Latin significant vowel nasality, allophonic and perhaps even phonemic, may well have been found at different times, in different places, and with different sociolinguisic significance" (p. 42).

Researchers don't entirely agree on its phonetic environment though.

For example, Sen 2015 claims that "[v]owels preceding nasals in Latin nasalized" (p. 72). Clackson 2008 writes that nasalized vowels in Latin were inherently long and occurred in two positions only: word-final and before a nasal+continuant (p. 77).

There are some dissenting voices out there. For instance, Marotta 1999 argues that "the evidence for postulating long nasalized vowels in Latin does not seem to be strong enough" (p. 289).

Since there are no audio recordings of Classical Latin (obviously!), all we can assume is that vowels before a nasal stop were most likely nasalized.

I wouldn't worry about making a special effort to pronounce vowels nasalized before an m since it should happen automatically in such contexts. It'd be hard not to utter a nasalized vowel before a nasal consonant anyway.

Answer 2

A pronunciation like [sũːmʊs] for summus in Classical Latin seems rather unlikely considering that the Italian reflex is sommo [somːo] and not [suːmo].

The letter M in Latin corresponded to the nasal consonant [m] in any position but word-final (and excluding compound words, which I will discuss in the section below)*. So sumus, summus, amplecti, sumpsisti are something like [ˈsʊmʊs], [ˈsʊmːʊs], [amˈplɛktiː], [suːmpˈsɪstiː]. It seems plausible that the vowels preceding the nasal consonant [m] in these words were pronounced with some amount of phonetic nasalization, but that kind of "nasal vowel" would not be phonologically distinct from a regular Latin vowel phoneme, as the presence of nasalization here would be entirely predictable from the phonological context (as is the case for vowel nasalization in present-day English, which isn't normally marked in phonetic transcriptions). I don't know of any actual evidence that these vowels were nasalized.

The possible existence of nasal(ized) vowels in word-final position, or before voiceless fricatives, is notable because we have evidence that the nasal consonants that originally conditioned the nasality of the vowels may have been lost in these two contexts, with "compensatory" lengthening of the vowel. We have clear evidence of the lengthening of vowels before NF and NS, and fairly clear evidence of the possible loss of the consonantal value of N before S, and of M in word-final position before a vowel. The existence of distinctively nasal long vowels in Latin (as opposed to non-nasal long vowels, or vowel + nasal consonant sequences) is mostly an inference, not something that we have clear evidence for (and so, as Alex B.'s answer mentions, it is a bit controversial).

*The pronunciation of M inside compounds might be a bit unclear in some cases

The pronunciation of certain compound words where the first element ends in m and the second element starts with a vowel, e.g. circumeo, is apparently doubtful. Andras Cser's "Aspects of the Phonology and Morphology of Classical Latin" (2016) has a section on circum- (8.2.3.4.). In some forms of circumeo, it was common to not write the letter M (e.g. circuit), which could be interpreted as an indication that nasality was totally lost in these forms.

An M that was originally in word-final position is assumed to have been pronounced as a nasal consonant before a plosive or nasal consonant, with assimilation to the place of the following consonant. So [m] before [p b m], [n] before [t d n], and [ŋ] before [k g]. For some words, we have evidence for these changes in the form of a spelling with N: thus eam + -dem > eandem.

This kind of place assimilation is thought to have applied even when the consonant remained written as M. So before the enclitic -que, M was probably pronounced [ŋ] (making the form "uirumque" something like [wɪˈrʊŋkwɛ]), and before the enclitic -ne, M was probably pronounced as [n] (making iamne [ˈjanːɛ] and mātremne [maːˈtrɛnːɛ]).

Answer 3

Yes, as Lindsay and others have long noted. See the following passage from Victorinus:

Clari in studiis viri, qui aliquid de orthographia scripsere, omnes fere aiunt inter M et N literas mediam vocem, quae non abhorreat ab utraque littera, sed neutram proprie exprimat, tam nobis deesse quam Graecis: nam cim illi σάμβυξ, scribant, nec M exprimere nec N; sed haec ambiguitas in his fortasse vocabulis sit, ut in 'Ampelo,' 'Lycambe.'

Lindsay further notes variants in inscriptions like sentemtiam and Decenber, which show nasalization of both N and M before certain consonants.

Answer 4

As Sumelic and Alex B have noted, it's almost certain that vowels became a bit nasalized before nasal consonants. This is somewhat of a universal in phonetics. Different languages nasalize the vowels to different extents (e.g. English has a lot more nasalization than Italian), but some amount of nasalization is a given.

However…

In three particular environments, there's evidence that the nasal consonant disappeared entirely, leaving the nasalized vowel on its own. Those environments seem to be -Vm, -Vns-, and -Vnf-, where V stands for "any vowel".

There's some solid Classical evidence for this: final -Vm elides in poetry, for example, while no other final consonant does, and the word cōnsul is generally abbreviated COS. But there's also evidence from the Romance languages, where we see e.g. -Vm turning into plain -V in inflectional endings, and vowels before -ns- and -nf always acting long. This doesn't happen in other environments, where e.g. the n in erunt and the m in amat show through clearly.

P.S. In Romance, final nasals notably don't disappear in one-syllable words, so sum and rem keep their ms. This is also the only time they're ever in stressed syllables. I don't know if that reflects any Classical reality or just a later change.