It is a significant fact in this connection that m is the only one of the liquids (semivowels) that does not allow a long vowel before it. Priscian, mentioning several peculiarities of this semivowel, thus speaks of this one:
[Priscian. Keil. v. II. p. 23.] Nunquam tamen eadem m ante se natura longam (vocalem) patitur in eadem syllaba esse, ut illam, artem, puppim, illum, rem, spem, diem, cum aliae omnes semivocales hoc habent, ut Maecenas, Paean, sol, pax, par.
That the m was really sounded we may infer from Pompeius (on Donatus) where, treating of myotacism, he calls it the careless pronunciation of m between two vowels (at the end of one word and the beginning of another), the running of the words together in such a way that m seems to begin the second, rather than to end the first:
[Keil. v. V. p. 287.] Ut si dices hominem amicum, oratorem optimum. Non enim videris dicere hominem amicum, sed homine mamicum, quod est incongruum et inconsonans. Similiter oratorem optimum videris oratore moptimum.
He also warns against the vice of dropping the m altogether. One must neither say homine mamicum, nor homine amicum:
Plerumque enim aut suspensione pronuntiatur aut exclusione. . . . Nos quid sequi debemus? Quid? per suspensionem tantum modo. Qua ratione? Quia si dixeris per suspensionem homimem amicum, et haec vitium vitabis, myotacismum, et non cades in aliud vitium, id est in hiatum.
From such passages it would seem that the final syllable ending in m is to be lightly and rapidly pronounced, the m not to be run over upon the following word.
Some hint of the sound may perhaps be got from the Englishman’s pronunciation of such words as Birmingham (Birminghm), Sydenham (Sydenhm), Blenheim (Blenhm).
N, except when followed by f or s, is pronounced as in English, only that it is more dental.
[Mar. Vict. Keil. v. VI. p. 32.] N vero, sub convexo palati lingua inhaerente, gemino naris et oris spiritu explicabitur.