'No one knows how ancient languages were pronounced'

This one sounds like it should be a no-brainer. We don't have any audio recordings from ancient Rome, after all. On an online forum earlier this month, in answer to the question 'What knowledge is considered lost by mankind?', one user answers

The actual pronunciation of Latin. And ancient Greek, for that matter, but Latin fascinates me because for all the Latinists out there, the best we're getting is a guess at what it probably might maybe have sort of sounded like.

and then goes on to cite a source.

The web page linked there starts off with a couple of caveats -- evidence is 'difficult to interpret', ancient sources 'lack clarity' -- but as it goes on, it makes it look like exactly the opposite is the case. What's going on?

Socrates (the late Tony Steedman) tries to speak Classical Greek to Bill S. Preston, Esquire (Alex Winter) and Ted 'Theodore' Logan (Keanu Reeves). Vowels are a mix of Modern Greek and something else, consonants are British, and he doesn't even make an attempt at the pre-Hellenistic pitch accent. But certainly not laughable. (Bill and Ted's Excellent Adventure, 1989)

There's a trick to this topic. The naysayers, the folks who insist that no one knows how these languages were pronounced -- at least the ones who know what they're talking about -- what they're talking about is a much more strictly defined kind of knowledge than most people are thinking of when they wonder how Latin and Greek were pronounced. What your average Joe is usually wondering about is the phonetics and phonology of Latin and Greek. And those are very well understood indeed.

What's the difference? We're not talking about audio recordings, or idiosyncrasies of speech. Phonetics doesn't refer to the actual noises you'd hear if you took recording equipment back in a time machine: audio data on intonation, pace, and so on truly is gone forever. Phonetics instead refers to a kind of abstraction of the sounds, like the instructions your English dictionary gives on how to pronounce words; and phonology refers to the relationships phonetic elements have with each other and with the words that they're in.

The naysayers' claim is equivalent to saying: if your only contact with English is through texts, you can't ever know how English is pronounced. In some strict senses there's some validity to that: there are many kinds of mannerisms of speaking that you can't extrapolate from written texts alone. But that's not what most people are thinking about when they ask how an ancient language was pronounced. What they're really after is phonetic values. And as I said, that's something that's very well understood.

Even if you've never heard someone speaking English, it's perfectly possible to know that the th in thirty typically represents a dental fricative, but that when spoken with an Irish Republic accent the same word usually has an alveolar stop (as in tin) or, in some parts of the country, a dental stop (a sound that doesn't exist in most other forms of English or other Germanic languages). We don't have audio recordings of ancient Greek, we don't know much about intonation or pace, and the variance within a given dialect is anyone's guess. But we know perfectly well 'how it was pronounced' in this more colloquial sense.

Well, we know a lot more than is often imagined, anyway. There are areas of uncertainty. But they're relatively limited, and they're probably not the kinds of uncertainty that you'd expect. Here are a few examples.

(1) The letter v in Latin. Depending on the Latin teacher you had, you may be aware of the dogma that the Romans pronounced the letter v like English w: so for example the Caesarian motto veni, vidi, vici ('I came, I saw, I conquered') was pronounced weeh-ny, weedy, weaky. It was the consonantal version of u: that's why the Romans didn't have a separate letter u. One letter represented both the w and u sounds, as i represented both the y and i sounds. But in Romance languages and in modern Church Latin, v is pronounced just like English v (so veeny, etc.). So which is right? Are the linguists all at sea? Well, no. The linguists know perfectly well that they're both right: languages change over time. At one time v was pronounced like w; later on, w sounds shifted to v, so the phonetic value of the letter shifted too. There's nothing mysterious or unknowable about that.

Jesus (Jim Caviezel) chats with Pilate (Hristo Shopov): probably the best on-screen attempt at Latin, but that's not saying much (and it's vanishingly unlikely that Jesus knew any Latin). The rhythms and vowels of the minor Roman characters, played mostly by Italians, carry verisimilitude even if they're not always borne out by evidence. But again, much too modern: phonetic values are all those of modern Church Latin. Some sounds we only start to see evidence for in the 1st century; some, not until centuries later. And worst of all, Jesus' lines are chocker with glottal stops -- Latin was much more continuous than Germanic languages like English. (The Passion of the Christ, 2004)

But we don't know very well the timeline of the shift. It seems to have been a gradual thing. We get our earliest evidence of v representing /v/ in the 1st century CE (spelling confusion between v and b; a 2nd century grammarian says v is pronounced with some exhalation, implying a fricative). But v was still representing /w/ in some contexts right up to the 6th century (e.g. the name Valerius being transliterated into Greek as Οὐαλέριος; Consentius writes of the word veni being pronounced as though it had three syllables). So it wasn't like flipping a switch. How exactly did the shift play out? What were the key moments? We don't know the answer to that as well as we'd like. Is that the same thing as saying that our state of knowledge is at best 'a guess at what it probably might maybe have sort of sounded like'? No, that's just dishonest.

(2) β in Greek. This is similar to the first example above: we know the letter β (beta) had two different pronunciations in different periods, but the timeline of the shift is unclear. In Classical Greek beta β had the phonetic value of an unaspirated English b (as in gib board), but in later Greek it was a fricative v (as in sieve), as it still is in Modern Greek. This is a pretty typical shift that you can see in lots of languages: in Latin, too, there was confusion between b and v after v acquired the value /v/. But our timeframe for when the pronunciation shifted from b to v is at least a 300-year window: we know it was pronounced b in the 2nd century BCE, thanks to testimony from two grammarians, and we know it was pronounced v in the 2nd-3rd centuries CE, when we start to find Latin v being transliterated into Greek as β. We know the switchover happened somewhere in between, but we don't know where or precisely when. Sometimes you'll hear this kind of uncertainty exaggerated to the point of saying that we don't know how β was pronounced at all. And that's nonsense.

(3) ττ in Attic Greek. In Attic -- that is, the dialect spoken in Athens up until the late fourth century BCE -- the letter τ (tau) had the phonetic value of an unaspirated t sound, like the first t in coattail. However, we don't know how the consonant combination ττ was pronounced in Attic. On the face of it it looks like ττ ought to represent a double t sound. But it isn't that simple. Where Attic has ττ, other dialects consistently have σσ (which looks like it ought to be unvoiced ss; not as in tosser, but as in toss sir), and ττ/σσ are used in words whose prehistoric forms had several different phonologies. It could be that both ττ and σσ represented affricates, but separate ones; it could be that only Attic ττ did, and represented a /t͡s/ sound (as in N.Z. tea) or /tt͡s/ (Italian grazia), or some other possibilities, depending on what word it appeared in. The fact that Attic spelling is so consistent suggests that all the variants might have been simplified, most obviously to /t͡ʃ/ (chin), or maybe even to exactly what it looks like, /tt/. But we can't be certain. This is one area where we are genuinely in the dark.

'Let it go' in Classical Greek, performed by the University of Auckland Classical Society (May 2014). Pronunciation is student Greek, which is to say a mixture of Erasmian with smatterings of Modern Greek ... but great fun anyway.

So, how do we know what we know about the phonetics of these languages? Like many things that relate to knowledge, it's not simple. You wouldn't expect to extrapolate the Big Bang theory from a single thing in the sky that you can see with a pair of binoculars; you don't reconstruct the phonetics of dead languages in a single easy-to-sum-up step. It's complicated and difficult, but that doesn't automatically mean that it's uncertain. There are several kinds of evidence. Here are a few:

• Direct testimony from ancient sources. For example, the 1st century CE Roman grammarian Nigidius Figulus tells us that the letter v in Latin was enunciated with protruded lips, which tells us straightaway that it was a bilabial continuant (/w/); an anecdote in Cicero (1st century BCE) relies on wordplay between consonantal v and the vowel u. Dionysius Thrax and Dionysius of Halicarnassus both tell us that Greek β, γ, and δ were 'intermediate' between aspirated and unaspirated consonants, and they don't call them continuants (ἡμίφωνα). And so on.

• Transliteration. For example, Cicero (1st century BCE) tells us that Greek βινεῖ was pronounced like Latin bini: this tells us that, at that time, (a) Greek β and Latin b were both pronounced the same way (but that might have been either /b/ or /v/); and (b) Greek ει was pronounced /i/.

• Variant spellings. When we find Byzantine writers translating the Latin names Valerius and Vergilius as Οὐαλέριος and Βεργίλιος, that tells us (a) that Latin v could represent both /v/ and /w/ at the time, and (b) Greek β was pronounced /v/. Another Greek example: in Classical Attic ει, ι, υ, and οι all represented different sounds, but at some point they all acquired the Modern Greek value /i/. From variant spellings we know that ει and ι combined with one another and represented the same sound by Cicero's time, and υ and οι combined with one another from around the 4th century CE onwards; but the two pairs represented distinct sounds until at least the 10th century. This indicates that between 500 and 900 CE, υ and οι both represented the older pronunciation of υ, /y/ (as in French une, German über), while ει and ι represented the unrounded form of the vowel, /i/ (as in see).

• Verse. Ancient verse forms relied on certain phonetic properties of their text, so we can often infer those phonetic properties from the texts. For example, Homeric Greek contains formulas that rely on the consonant /w/, but is generally indifferent about whether or not to observe the presence of /w/. So we can safely deduce that in the Ionic dialect of the early-to-mid 7th century, /w/ had no moraic value, which in turn tends to imply that it wasn't pronounced at all.

• Common linguistic shifts. For example, we know that /v/ and /b/ are interchangeable in many languages, and we can see shifts between them in both Latin and Greek. Another example: we know that Indo-European had different 'flavours' of laryngeal consonants, and that one of these, h₂, regularly turned into /a/; but in the Ionic dialect, we see words from h₂ roots written with both α and η, and we know that at a later time these were pronounced /a/ and /e:/ respectively; from this we deduce that η represented a vowel that had once been intermediate between /a/ and /e/, namely /æ/. (And in fact we also have evidence that η still represented /æ/ in the sub-dialect spoken on the island of Naxos up until the 6th century BCE.) And so on.

Anyway, yes, on the whole we have a very, very good idea of how Greek and Latin were 'pronounced', on the understanding that this refers to phonetics and not to anything more nuanced. Yes, Julius Caesar pronounced his own name yoo-li-uss ka-eece-ar; yes, Cicero was kee-ke-raw. We can make a good estimate of pitch contours in Greek when it still had a pitch accent (like modern Norwegian; not a tonal language like Chinese).

Would a time traveller in Rome, armed with a modern rulebook about phonetics and phonology, be able to communicate with an ancient Roman? Absolutely yes. Getting to grips with the accent and pace would be challenging, but neither of them would find the other's pronunciation at all ambiguous, barring personal and local idiosyncrasies.

Would the time traveller stick out like a sore thumb as an obvious foreigner? Also yes.

---

Further reading

• Allen, W. Sidney 1968. Vox Graeca: a guide to the pronunciation of Classical Greek. Cambridge University Press.
• Allen, W. Sidney 1975 [1965]. Vox Latina: a guide to the pronunciation of Classical Latin, 2nd edition. Cambridge University Press.
• Horrocks, Geoffrey 2010. Greek. A history of the language and its speakers. Wiley-Blackwell.