In a dialogue attacking divination and astrology, Cicero tells the following story: when Marcus Crassus was about to set out on the expedition to Parthia (which was to end in his and his army's destruction), a man on the quayside was selling figs from Caunus (south Turkey) and therefore shouting Cauneas ('cow-nay-arse')! Cicero comments that Crassus should have listened to the omen and realised he was shouting cave ne eas, 'beware lest you go!', 'don't go!'

Arf arf. But there is an important linguistic point to that side-splitter. It would make no sense unless (i) ne eas were run together into neas and (ii) av and au ('ow') sounded roughly similar. And indeed they do: they both sound like au. In other words - and there is much more evidence to this effect - Latin 'v' was not pronounced like English 'v'. It was, in fact, pronounced more like a semi-vowel, e.g. English 'w' as in 'wet'. Modern Latin texts print it as 'u'.

This simple fact produced outrage when I published my twenty-part QED: Learn Latin series in the Daily Telegraph some ten years ago. Many readers told me that no one knew how Latin was pronounced, but then proceeded to assert that Latin 'v' was pronounced as English 'v', not 'w', the main argument being that Julius Caesar could never have said anything as unmasculine as 'waynee, weedee, weekee'. But linguistics are, I fear, no respecter of Daily Telegraph readers.

The evidence for the pronunciation of the individual words in classical Latin is, in fact, impressive and wide-ranging. The rest of this piece will be devoted to giving some examples. Each is, I can promise, backed up by more evidence to the same effect (the standard book on the subject is W S Allen, Vox Latina, second edition, Cambridge, 1989).

The Greeks, whose culture (Horace tells us) 'took the Romans captive', often transcribed Latin names into Greek. Assuming we know how Greek was pronounced, we can expect that their transcription will tell us something about Latin. Thus Cicero was transcribed as Kikerwn, not Siserwn (the -wn being a Greek ending), i.e. Latin 'c' was pronounced hard. (The vogue for pronouncing Latin 'c' as in 'church' is the result of an effort by Pope Pius X in 1912 to impose a standard pronunciation. So - how we should pronounce the Latin of the Mass?) Greeks also transcribed Valerius as Oualrios, again suggesting that 'v' was pronounced as a semi-vowel.

Ancient grammarians are full of useful information. One says that 't' and 'd' should be distinguished by the position of the tongue: 't' with the tongue against the back of the teeth, but 'd' with the tongue against the ridge of the gum. Another describes 'b' and 'p' as a 'sound exploded from the lips', and hints that the difference is one of muscular tension. In both cases, they seem to be describing something like the sounds those letters represent for us. A grammarian tells us that 'r' was trilled or rolled in Latin, a pronunciation supported by the early satirist Lucilius who describes it as resembling the 'growling of a dog'. But enough of dog Latin.

The spelling of words, especially of illiterate inscriptions, can be suggestive. We find in pace 'in peace' written im pace, and in balneo 'in the bath' im balneo. Presumably Romans slurred 'n' to 'm' before a 'p' or 'b'. Even more surprising, we find ignes 'fires' written ingnes, and there is other evidence to suggest that 'gn' was pronounced 'ngn'. So magnus 'large' sounded roughly like English 'hangnail'. consul was often written cosul, and when Romans abbreviated it, they wrote cos., not con. or cons. So 'n' was probably not pronounced before 's'. No surprise, then, that Greek writes Hortensius as Hortsios; while we are told that aristocratic Romans like Cicero actually preferred to drop n before s and say e.g. foresia ('public matters') not forensia. Compare also Italian 'bride', sposa, from Latin sponsa ('spouse').

But they also preferred to keep 'h'. This letter, which Roman grammarians describe as a 'breath', seems to have been on the way out from quite early on (e.g. mi for mihi 'to me'), but then came to be misapplied: inscriptions produce e.g. hire for ire 'to go' alongside ic for hic 'this'. All very Cockney. Catullus writes a poem mocking Arrius who hyper-corrects, e.g. Ionios 'Ionian' into Hionios, and there was scholarly controversy whether e.g. 'sand' should be written harena or arena. The aristocrats had to get it right. Controversy over spelling and pronunciation is a very old game indeed.

Myriad problems remain, of course - from different accents to the articulation of a complete utterance. But the pronunciation of the individual sounds is pretty secure.