About My Latin Reconstruction

My practice adheres to the current reconstruction of ancient pronunciation, more or less. However, I do make a few alterations and further assumptions, based on what I believe to have been most likely.

Syllabification

Wherever possible, I syllabify consonants with the following, rather than the preceding vowel. This applies across word-boundaries. So a sentence like "fruges effuderit" is syllabified as fru.ge.sef.fu.de.rit. When a plosive follows a liquid, I syllabify them with the following vowel as in "pa.trem."

Word-accent and Syllable weight

Current evidence suggests that Roman speech (like modern English) favored actual word-stress. Moreover, the evidence suggests that Latin verse was recited with natural word and sentence stresses, without any special attempt to give artificial prominence to heavy syllables or long vowels, thereby precluding any clear pattern from emerging in the first four feet of hexameters. (In the last two feet of hexameters, Roman poets such as Virgil very often strove to harmonize word-stress and syllable weight and avoid having a stress fall on a light syllable.) This is because Roman metrics were inherited from Greek (a language with a pitch-accent which allowed heavy and light syllables to stand in somewhat starker contrast in speech.)

Since the poet and his/her expected audience were not only fluent in Greek but also trained in the principles of Greek verse, the audience could enjoy the underlying quantitative patterns and the interplay between them and the natural stress of the Latin words, in much the way an English reader of John Donne's sonnets could appreciate the counterpoint between the ideal rhythm of iambic pentameter and the actual, slightly more dynamic word-stress of Donne's lines.

However, to the untrained hearer of such recitations, no pattern was actually evident except in the last two feet of a hexameter. The structure of the lines must therefore have seemed somewhat chaotic to him. An example of this can be found in some of the Carmina Epigraphica, composed by rather unlearned persons, where the first part of the line displays nothing short of metrical chaos and only the last two feet reveal that the line is supposed to be a hexameter.

Moreover, it seems that an unnatural "scanning" reading, where heavy syllables were stressed artificially and stress was deleted from light syllables that normally had it, was practiced in schools as a way of teaching meter, as indicated by Ausonius' De studio puerili

....tu flexu et acumine vocis
innumeros numeros doctis accentibus effers

This state of affairs in classical times leaves me, the reciter, with a difficult problem. I can recite the poems as the Romans would have but, since the listener isn't schooled in Greek metrics, much of the effect will be lost on him/her. Otherwise, I could try and exaggerate the heavy syllables like a Roman schoolboy, hoping to give the listener at least some experience of the meter.

In my recitations, I have opted for a middle-ground. I have considerably lightened my word-accent, in addition to shortening light syllables a bit more than a Roman might have, so as to give heavy syllables extra prominence. Perhaps in doing this, I am not too far off from what a Roman reciter might have naturally done when attempting to ensure that he adhered to the quantitative schemes.


The Consonants

F: For /f/ I include a voiceless bilabial fricative in more or less free variation with the labiodental.

L: For the alveolar lateral approximant /l/ I allow a velarized and non-velarized allophone in complementary distribution, with the velarized allophone occurring as the first element of a consonant cluster and at the end of all words.

H: I'm assuming that the fact that /h/ does not count metrically in Latin represents something about the way it was realized in the educated speech of living Romans. While this might lead one to conclude that Romans just didn't pronounce their H sounds, a bunch of things, including orthographic evidence from the Romance languages (along with Catullus 84, in which the poet ridicules someone for over-use of the H sound) suggests otherwise. So here's how I treated the /h/ in my readings: When a word beginning in /h/ follows a word ending in a consonant, I do not realize /h/ as a separate consonant if I can help it. If the consonant is a stop (as in rapit hora) I simply aspirate the stop. If it is voiced, I devoice it. Only at the beginning of lines and in pausal form do I actually pronounce a full-fledged /h/ sound. (And this only represents educated urban speech. Even in Varro's time words like hortus were beginning to turn into ortus.) This is the /h/ I believe Catullus referred to in his poem. Careful readers of the poem will note that he seems to think of the aspirated ch sound of chommoda as similar to the h of hionios.

M: The /m/ sound is one of the most difficult to reconstruct. Word-final syllables ending in -m are regularly elided in verse when the next word begins with a vowel. This suggests that there was something about the M sound in Latin that was different from all the other consonantal sounds. Ancient Romans themselves said as much. For example, Quintilian says

Atqui eadem illa littera, quotiens ultima est et vocalem verbi sequentis ita contingit ut in eam transire possit, etiamsi scribitur, tamen parum exprimitur, ut multum ille et quantum erat adeo ut pæne cuiusdam novæ litteræ sonum reddat. Neque enim eximitur sed obscuratur et tantum aliqua inter duas vocales velut nota est ne ipsæ coeant.
"But the selfsame letter [M], when it is in word-final position before a word-initial vowel so as to be joined with it, is written but hardly articulated, as is the case with multum ille and quantum erat, so as to virtually produce the sound of a new letter, since it is not actually dropped but merely suppressed, and is no more than a kind of mark to stop them from conjoining"

And Varius Longus says more or less the same, only he doesn't limit himself to -m before a word-initial vowel:

Quibusdam litteris deficimus, quas tamen sonus enuntiationis arcessit, ut cum dicimus virtutem et virum fortem consulem Scipionem, pervenisse fere ad aures peregrinam litteram invenies
We lack certain letters nonetheless demanded by pronunciation. For example, when we say virtutem and virum fortem consulem Scipionem you will find a veritable foreign letter has come to your ears.

Most modern scholars believe this sound to have been a long nasal vowel, so that the final syllable of a word like virtutem was pronounced like the -en of the french word bien. It is also possible that -m did not disappear entirely, but was instead an extremely weak voiceless bilabial approximant [ɸ ̞] which also caused the preceding vowel to become nasal but which was often elided entirely. The lack of word-final -f in classical Latin suggests that this could easily have happened.

Whatever sound -m represented, it certainly wasn't the English M sound you learn in Latin class.

In my readings, I go for a nasalization and lengthening, partly because this is what most scholars consider to have happened, but also because voiceless bilabial approximants are an extremely rare occurrence cross-linguistically.

N: Medial -n- before fricatives (as in the words dicens and infans lost its consonantal value early in the classical period, causing the preceding vowel to lengthen and be nasalized. Eventually the nasalization was dropped as well, causing words like consul to be pronounced as if they were cōsul. Evidence from the Romance languages (as in Italian mesa from Lat me(n)sa) suggests that this was how most Romans talked. However, prescriptive practice among the upper classes caused -n- to be re-inserted where it was once lost. The lengthened quantity of the preceding vowel, however, remained, causing words like violēns with a genitive violĕntis. Whether this re-inserted -n- simply caused a nasalization of the vowel, or whether it actually added a consonant, is anyone's guess. I choose to realize it as a consonant.

P, C, T: Voiceless stops as in Italian or French. Some scholars who know infinitely more than I ever will, such as William Allen, suggest that these letters were pronounced with some aspiration as in English. I disagree. Greek transliterations of these sounds suggest no aspiration.

Ph, Ch, Th: Aspirated stops. Technically, these sounds are quite marginal to Latin phonology, and appear mainly in Greek borrowings such as Delphicus, Theatrum, Chorus. However, they do seem to have gained traction in the language as is demonstrated by a number of native Latin words: Pulcher, Triumphus etc.


Vowels

Æ: For /ae/ I do not pronounce [aj]. Rather, I pronounce the second element of the diphthong as an ultrashort /e/.

Short E, I, U and O: In keeping with modern reconstructions, I pronounce the short mid vowels as open-mid: /e/ and /o/ ([ɛ] and [ɔ], respectively), and the close vowels as near-close: /i/ and /u/ ([ɪ] and [ʊ].)

Special note on short E:

Short E, before R, seems to have had a more open quality as attested by inscriptional evidence such as passar and cacaris as well as forms noted by Probus such as ansar. Certain developments in the Romance languages (Latin per as against Spanish para and French par) seem to suggest that this trend was productive through the Vulgar Latin period.

W.S. Allen seems to think that "because the degree of opening is not known, there is no point in attempting to reproduce it." This, however, seems rather specious, since the only reasonable possibility seems to be a sound in the vicinity of the near-open vowel [æ], though not quite as open as in the English "Bad." More like the long ā of Standard Arabic kitāb.