For example we have the verb nārrō, and in the dictionary both the "a" and the "o" are marked with a macron.
That one is actually very much contended, and I suspect it originates from some odd etymology that is not accepted anymore. It's pretty regrettable it appears in Hans Ørberg's LLPSI (which is generally excellent when it comes to this topic, "hidden quantities"), as well as the Elementary Lewis, which is apparently where people are seeing it supposedly has ā. Neither Ernout's etymological dictionary, nor De Vaan's, nor the Gaffiot 2016 dictionary have nārrō, but instead have narrō. The etymology is taken to be *gnārāō (so a cognate of gnārus and ignārus), which via the lītera ~ littera sound change produced narrō narrāre, after a stage of being *gnārō *gnārāre.
I thought the n in words like īnfāns and mēnsa represented a nasalized vowel rather than a nasal consonant, much like the singular accusative -m.
A bit unlikely, since even Old Latin inscriptions show COSOL. Allen says nasalization was a likely intermediate step before arriving at things like COSOL, but this frequent absence of -n in Old Latin and other things such as inconsistent grammarian advice on -ns vs. -s (quotie(n)s) suggest that there might've been an informal pronunciation without a trace except the long vowel ("mēsa" for MENSA), and a formal one with a fully restored [n] (or [m] before [f], mentioning republican inscriptions like EIMFERIS for īnferīs). It also matters that ancient grammarians (not named) apparently go out of their way to criticize hypercorrected THENSAURUS, OCCANSIO for thēsaurus, occāsiō.
(Btw, amusingly, he also cites Greek spellings like κλημης for Clēmēns(!), and a comment by Velius Longus (2nd c. AD) that Cicero was apparently known to have said FORESIA/HORTESIA (i.e. forēsia/hortēsia) for forēnsia, hortēnsia, something that people would probably not have said if he still had the -n- as nasalization...)
There is more evidence that Godmy didn't mention btw, in favour of vowels there being long, such as Romance outcomes (mēnsa, tēnsum/am > Old Spanish la mesa, teso/a, instead of *miesa, *tieso/a < *mɛsa *tɛsu/a < *mĕ(n)sam *tĕ(n)sum), inscriptions with an
i longum (I
N SPECTACULIS for "īn spectāculīs"), and comments by Cicero (1st c. BC), Probus (1st c. AD) and Pompeius Festus (2nd c. AD) on some words with -ns.
Who knows. AFAIK m and n are not confused in inscriptions.
Allen mentions a few like EIMFERIS, IM FRONTE for īnferīs and īn fronte, but he thinks these just reflect a phonetic articulation influenced by the [f] on the restored [n] sound ("restored" because in more informal speech at least īnferīs may have been basically "īferīs").
It's an example of -n being lost without any nasalization. What about Latin? We know that words were spelled with -m (though not always); there is vague contemporary evidence that its pronunciation was kind of unclear; in poetry -Vm is elided; -m is lost in Romance languages. Does this logically imply existence of nasal vowels? I think not. It's a plausible scenario, but a Dutch-like scenario is not ruled out, and there is no way to know for sure.
Well, regarding word-final -m (a different topic from -Vns- -Vnf-), Allen's argument seems to rely entirely on a passage from Quintilian, and something reported by Velius Longus (2nd c. AD) about the grammarian Verrius Flaccus (late 1st c. BC, which some other unnamed people, "nōnnūllī", also did). Quintilian's passage in Institutiones 9.4.40 says:
Atque eadem illa littera, quotiens ultima est et vocalem verbi sequentis ita contingit, ut in eam transire possit, etiam si scribitur tamen parum exprimitur, ut "multum ille" et "quantum erat", adeo ut paene cuiusdam novae litterae sonum reddat. Neque enim exprimitur sed obscuratur.
And this same letter [the letter M], when it is word-final touching a vowel in the next word, as if it was to continue into it, is hardly pronounced even if it is written. For example, in "multum ille" and "quantum erat", to the point it almost represents the sound of a new letter.
This "adeo ut paene cuiusdam novae litterae sonum reddat" wording suggests the -m may have still been sounded out in some fashion, so still not completely gone, yet not a full [m] either as in most other positions.
For Verrius Flaccus, Allen seems to interpret his use of a half-M (looking like Λ) for a final -m before a vowel-initial word as possibly reflecting it was pronounced differently. But I don't even think that is how the relevant passage in Velius Longus' De Orthographia 80.17 should probably be interpreted...
Nonnulli circa synaliphas quoque observandam talem scriptionem existimaverunt, sicut Verrius Flaccus, ut, ubicumque prima vox m littera finiretur, sequens a vocali inciperet, m non tota, sed pars illius prior tantum scriberetur, ut appareret exprimi non debere. Est etiam ubi vocales subducebantur, si id aut decor compositionis aut metri necessitas exigebat, ut "adeo in teneris consuescere multum est".
Some people, like [the grammarian] Verrius Flaccus, have thought of observing synaloephas [i.e. the pronunciation of the transition of two words as one syllable] writing them so that if a word ends in -m and the next starts with a vowel, they don't write the M whole, but only a part of it, so that it's clear it shouldn't be pronounced. This is where vowels were taken away if the grace of a composition or the needs of metre required it, as in "ade|ō‿in tene|rīs cōn|su̯ēscere | multum‿est".
This seems to be talking about elision in prose clausulae and poetic metre rather, which probably had their own rather artificial conventions anyway... so that this half-M "Λ" would simply be a marker of elision, instead of representing something about pronunciation (unlike Emperor Claudius'
proposed new letters).
(Allen also mentions Priscian (late 5th - early 6th c. AD) saying "obscurum in extremitate dictionum sonat" (
it sounds "obscure" at the end of utterances, Institutiones Grammaticae 2.29 in Keller's Grammatici Latini numeration), but I would not give much credit to anything a 5th century grammarian would say, except in terms of the formal pronunciation of that time and place.
Late Latin grammarians vary a lot on what they say about final -m before a vowel, in fact, some even seem to prescribe something like a nasalized vowel + full [m], quam asper [kwãm asper], with no elision, being more concerned about clearly separating words for rhetorical reasons in prose speeches than what we would normally call correct pronunciation, say, so that "quam asper" is not heard as "qua masper". See Anna Zago's fairly recent article "Mytacism in Latin Grammarians" (2018), Journal of Latin Linguistics vol. 17, iss. 1, DOI: 10.1515/joll-2018-0002 (scihubbable) for an extended discussion.)