We can deduce facts about how the Romans pronounced Latin sounds by:

* descriptions of the sounds by the Roman grammarians

* wordplay, onomatopoeia, spelling errors in graffiti, etc.

* the way in which sounds from other languages (particularly Greek) were adapted into Latin

* the way in which sounds of Latin were adapted into other languages

* comparative reconstruction of the Romance languages

* evidence from metrical poetry, which provides particular clues about consonant and vowel length

A very readable book on is W. Sidney Allen's _Vox Latina_.

[edit: meant to reply to immediately following sister of this post.]

I'd never thought of using misspellings from graffiti as a pronunciation clue, but it seems so obvious.

I wonder how anthropological linguists would someday treat modern rap and hip-hop. In German, for example, it's not uncommon for the artist to slightly or not-so-slightly mispronounce words to make them fit a rhyme. Would that help or hinder a study?

I see some hand-written signs by native Spanish speakers in my neighborhood that interchange b and v (and sometimes k and c, and sometimes c and s), which would make for good evidence for a future historical linguist that these were homophones or near homophones in certain contexts.

For example, I think I've seen "besino" written by a native speaker for "vecino" 'neighbor'.

> Would that help or hinder a study?

I imagine it would help. If evidence for that mispronunciation happened consistently, it would be taken as a dialect, class, geographical or cultural thing (or subculture). If it happened consistently over a very short space of time (and there was surviving evidence) it could be confused as a blip, or fashion.

Example:

I'm constantly perplexed why 'where' is spelt as 'were' by so many, but I imagine there are a sufficient number of English-speakers' accents (or a sufficient amount of speakers of a single accent to be significant) where the two words sound similar enough to spell them out on a keyboard as the same.

Firstly, the way Italians pronounce Latin is nothing like what it should should :) In the main, c is hard (Cicero is KIKero not CHIchero), v is to be pronounced as the English w. Those are the main points that irk me.

The best approximation we have is what's called the "restored" pronunciation. If you're interested in all the nuances the best book for this in my opinion is Vox Latina by Allen. Alternatively, just listen to anything by Evan der Miller on his YouTube channel: http://www.youtube.com/user/evan1965. His Latin pronunciation is the restored Classical and extremely good.

If you took Latin in Highschool and want to get back into it you should check out a book course called Lingua Latina per se Illustrata. It introduces Latin through Latin and also contains MP3 recordings for the first 31 chapters in first book, all in very good restored Latin pronunciation by the author Hans Ørberg. (there's a torrent of all the Lingua Latina resources)

And there's an interesting tie in to the article WRT the Orberg's Familia Romana. A quick glance at the article and a text search indicates the author is snubbing Orberg for some reason, as Familia Romana is much more famous in that field for that technique of teaching than the folks discussed in the article. I like that book and occasionally digest a chapter or two. Its like solving a puzzle.

I am pretty sure there are no legal ebooks of familia romana, its full of illustrations and the ebook transition would likely be painful. Too bad.

No idea why they would snub Orberg. It's not like what they are doing is totally unique. Terentius Tunberg runs the Conventiculum Latinum, "an annual summer immersion workshop in spoken Latin held on the campus of the University of Kentucky" [1]. Not in Rome, but still full immersion.

[1] http://en.wikipedia.org/wiki/Terence_Tunberg

In my experience folks who've studied with Reginald Foster tend to admire Ørberg quite a bit, but they might think Ørberg's books are too basic for their own students (since I guess people coming to Paideia will have studied Latin for several years first) and/or study things that are not explicitly didactic works.

I've heard some academics who specialize in the phonology. I always thought it would sound like Italian, but the reconstruction actually sounded a lot more like French.

It's hard to tell with dead-languages; for example we say "chi" in english, but most grammars list χ as aspirated.

The pronunciation of Latin is extremely well studied and mostly known, including changes throughout its history.

Unfortunately it's quite far away from the pronunciation I've learned in school.

(and what Americans usually use seems to be even farther away)