What Did Ancient Languages Sound Like?

Nicholas Swift

Students who enroll in their first Latin course – or maybe find themselves placed there quite against their will – begin their studies with a survey of the alphabet. They are often surprised to find that lowercase letters were unknown to the Romans and are even astonished that ancient texts made little use of punctuation and word division. Quickly, however, they move on to more practical matters, and learn that the letter C is always “hard” (like English K) and that the letter V is pronounced not like English V but English W. This process is more involved in an introductory Greek course, but not as daunting as it first seems.

Once those basics are in place, students spend their time acquiring new vocabulary and navigating a labyrinth of inflectional endings in order to improve their written translations. They might wonder why, outside school, their priest seems to violate what they learned about the letter V in Latin class, or why their Greek professor pronounces the letter Φ as a fricative f-sound when the textbook explains otherwise. But few people, even those who go on to become professional Classicists, spend much time thinking about pronunciation.

Detail from a red-figure pelike from the late 6th century BC (ARV2 1594.48, in the Hermitage Museum, St Petersburg, Russia) containing the most elaborate ‘bubble inscription’: the figure on the left says, ἰδο<ὺ> χελιδών, “look, a swallow”, the figure in the center says, νὴ τὸν Ἡρακλέα, “yes, by Herakles!”, and the figure on the right says, αὑτηί, “there it is.”

How exactly do we know what ancient languages sounded like? We have no audio recordings after all. The question is not so simple. Consider what English sounds like. Not only did it sound different in the 16th century than it does today, but it sounds different in New Zealand than it does in Ireland, and different in Brooklyn than in Atlanta or almost any other US city. Two friends from the same small town might pronounce a word such as aunt differently, and even the same person might sound different when reciting a poem in class than they do when hanging out with friends on the weekend. This was true for ancient languages as well, even if the many of the finer details are lost to us.

A map of the various dialects (regional variations) of Ancient Greek.

The Linguistic Evidence

Once we temper our expectations and narrow our focus, however, the problem turns out to be a fascinating forensic exercise. In the case of Classical languages like Greek and Latin and Sanskrit, we have an unbroken tradition which provides the sort of foundation lacking for many other ancient languages, such as Hittite or Egyptian, which were lost completely and then deciphered by modern scholars. Because we write English with a form of the Latin alphabet, there is information encoded in a familiar writing system, and such information about pronunciation was passed down from generation to generation in both formal and informal educational contexts.

Occasionally ancient writers described what their native language sounded like. This type of linguistic analysis is far more sophisticated in the Sanskrit tradition, but we find important information among Greek and Roman authors too. Consider the sound represented by the letter rho in ancient Greek. Plato says the following about it: ἑώρα γὰρ οἶμαι τὴν γλῶτταν ἐν τούτῳ ἥκιστα μένουσαν, μάλιστα δὲ σειομένην, “for he saw, I think, that the tongue in [rho] is not still at all, but shaking rapidly” (Cratylus 426e). And Dionysius of Halicarnassus says: τὸ δὲ ρ τῆς γλώττης ἄκρας ἀπορριπιζούσης τὸ πνεῦμα καὶ πρὸς τὸν οὐρανὸν ἐγγὺς τῶν ὀδόντων ἀνισταμένης, “and rho [is pronounced] by the tip of the tongue blowing out the breath and rising to the palate near the teeth” (On Composition 14). Based on these descriptions most scholars think it was a trilled r-sound like we find in Italian and Spanish.

The various points of articulation for consonants.

We can augment that base of knowledge with tools from modern linguistics. The study of phonetics provides insight about the mechanics of sound production in language. We realize, for instance, that it is common across languages for word-final nasal consonants (M and N) to weaken and disappear, such as we find in Latin, or for the first of two consecutive aspirated consonants to lose its aspiration, such as we find in Greek and Sanskrit.[1]

Even more important for our purposes is the field of historical linguistics. Through ingenious comparative work, scholars have been able to reconstruct in great detail features of the Proto-Indo-European language, the language from which all Indo-European languages descend, despite the fact that no written evidence of that language survives, because it was spoken at a time before the technology of writing had developed. Likewise, even if we had no written evidence for Latin, we could learn a great deal about how it sounded and functioned just by comparing the Romance languages, such as Spanish, French, Italian, and Romanian, which are descendants of Latin.

The families of ‘Indo-European’ languages.

Clues from Metrical Analysis

The study of meter is revealing. When students today want to know whether a particular vowel is long or short, they can consult their dictionary. But how did those lexicographers know in the first place? Take the word ἴξαλος, which is found only once in Greek literature – in Homer’s Iliad at 4.105, where it describes a wild goat. We’re not completely sure what the word means – some have suggested “full-grown” and others “leaping” – but we can tell from the meter that the alpha is short. On the other hand, it’s unclear whether the iota is long or short: its syllable is scanned long in either case because of the following double consonant xi (ξ).

Metrical analysis provides another detail about the pronunciation of Latin. Those familiar with Latin scansion will know that often a word’s final vowel will be elided (that is, it will not be pronounced) if the following word begins with a vowel. This can also happen, however, if the word ends in M, which suggests that this final consonant was very weakly pronounced. This is supported by cross-linguistic observations, as mentioned above, and further corroborated by later Latin documents in which it is sometimes not written at all.

A rhapsode (Greek oral poet) on a red-figure neck-amphora from the early 5th century BC (ARV2 183.15, in the British Museum, London, UK) attributed to the ‘Kleophrades Painter’. He sings the beginning of a hexameter ὧδε ποτ’ ἐν Τίρυνθι, “once upon a time in Tiryns”.

The Frozen Sounds of Writing

The study of writing systems and their various orthographies (conventions of spelling) provides us with a wealth of information. By analyzing the choices that Greek speakers made in adapting a Semitic script for their own language, we can infer some phonetic details. In addition to kappa (Κ), for example, early Greek scripts often employed qoppa (Ϙ) for another k-sound – an allophone[2] – which appears only before back vowels, such as o- and u-sounds.[3] Granting this sound its own symbol suggests that it was fairly distinctive in the spoken language.

Scholars in the 2nd century BC had already developed symbols to help readers with the pitch accents found in earlier literature. This probably coincides with a change in the language to a stress-based accent linked to the loss of vowel-length distinctions. Modern students routinely learn the simple rules for writing accents, but few people attempt to recover the pitch accents when they read aloud. Nevertheless, for those interested in the history of the language, these symbols are valuable.

The opening lines of Sophocles’ Antigone as presented (50r) in the oldest surviving copy (c. 970) of the play, part of one of the world’s most valuable books (Laurentian Library, Florence, Plut. 32.9).

Precious clues are found in the spelling mistakes that we find in graffiti, personal letters, school texts, and other such documents, whose authors often had only a rudimentary literary education. We’re all familiar with the tendency for native English speakers to confuse they’re, there, and their in writing, or our and are, because they can sound alike in the spoken language, despite looking very different on the page. Such evidence is difficult to interpret in ancient texts, however, because deviations might be merely a slip of the pen or a deliberate archaism, rather than a reflection of the contemporary pronunciation. Some aspects of pronunciation can exist for a long time in speech before stumbling into a written text, yet careful analysis has much to tell us about developments in Greek and Latin, especially the vowel systems.

We often learn about pronunciation when words are borrowed between languages with different writing systems. We know, for example, that the Greek letter phi (Φ) came eventually to be pronounced as a fricative f-sound, as in Modern Greek, and we have some evidence that this change was already under way in the 1st century AD, where we find spellings such as Dafne for Δάφνη in Pompeiian graffiti. Earlier Romans could easily have written filosofia instead of philosophia, but they didn’t. Transliteration of the Greek letter Φ into Latin as P or PH instead of F, therefore, indicates that it was pronounced as an aspirated stop rather than a fricative at that stage.

Greek poetry painted in the early 4th century AD on a school wall in Trimithis (Amheida), Egypt (photo by Paola Davoli). The lacunose text says: “To my schoolpupils: My clever boys, drink from the waters of the Pierian spring until [the end?]. To the same: Work hard for me, and manly toils… which for Hercules… for the wisdom of all…”

So Why Bother?

Here we’ve only scratched the surface of a complicated question. But you might be wondering, why bother? Archaeologists and historians can do their work without ever giving it much thought, and even philologists can spend entire careers working on interesting problems where details of pronunciation are inconsequential. Still, I think that anyone who loves Greek and Latin literature, whether amateur or professional, will enjoy a keener appreciation of a highly aural artform. The difficult task of reproducing the pitch accents in Greek might be a step too far for most people, but consider the following line from Homer’s Iliad 1.345:

ὣς φάτο Πάτροκλος δὲ φίλῳ ἐπεπείθεθ᾽ ἑταίρῳ

It develops beautifully from sauntering labial alliteration (of p and ph sounds) to rapid-fire dental consonance (of th and t sounds), but if we pronounce Φ and Θ as fricatives here:

rather than aspirated stops:

then the distinction between Φ and Π and between Θ and Τ becomes so great that the sound effect is lost completely. Homer, for one, would have cared.

Nicholas Swift is a painter and independent scholar based in New York. He taught Latin and Greek at Nichols School in Buffalo, NY for several years. His publications include “The Origin of the Greek Alphabet” in Brill’s Encyclopedia of Ancient Greek Language and Linguistics (2014).

Further Reading

The two short manuals Vox Graeca (3rd ed., Cambridge UP, 1987) and Vox Latina (2nd ed., Cambridge UP, 1989) by Sidney Allen provide excellent summaries of the evidence for the pronunciation of Greek and Latin. The Cambridge Encyclopedia of the World’s Ancient Languages (Cambridge UP, 2004)provides complete linguistic overviews, including phonological details, for all ancient languages.

The Blackwell Companion to the Ancient Greek Language (Wiley-Blackwell, Chichester/Malden, MA, 2010) and Companion to the Latin Language (Wiley-Blackwell, Chichester/Malden, MA, 2011) offer a variety of relevant articles. Greek: A History of the Language and Its Speakers by Geoffrey Horrocks (2nd ed., Wiley-Blackwell, Chichester/Malden, MA, 2010) and The Blackwell History of the Latin Language by James Clackson and Geoffrey Horrocks (Wiley-Blackwell, Chichester/Malden, MA, 2011) offer more detailed accounts of the development of those languages. For the intrepid student of Latin, the two massive works of J.N. Adams, Social Variation and the Latin Language (Cambridge UP, 2013) and The Regional Diversification of Latin (Cambridge UP, 2007), treat those subjects in masterful detail, while such a Greek student might attempt The Prosody of Greek Speech by A.M Devine and Laurence D. Stephens (Oxford UP, 1994).


1 When this happens in consecutive syllables it is known as Grassman’s Law; for example, the reduplicated perfect of φύω (“I grow”) is πέφυκα (“I have grown”) instead of *φέφυκα. It also seems to happen when the sounds are contiguous, since we find the spelling Σαπφώ as well as Σαφφώ for the poet we typically spell Sappho. Where two aspirated consonants are written together, however, there is disagreement about how they were pronounced. It’s worth noting that φθ was usually transliterated into Latin as pth, and χθ as cth. I think it’s likely that the first consonant was not aspirated despite the spelling, but possible that it varied among speakers.
2 An allophone is a variant sound for a single phoneme, or, to put it another way, it is a variant sound that does not contribute to a distinction in meaning. Sometimes they are distinguished in writing, but often they aren’t.
3 Vowels can be classified and charted based on their position in the mouth. If you pronounce in sequence the vowels ου (oo), ω (ō), α (a or ā, rhyming with car), η (ē, similar in sound to hair), and ι (i or ee), you should feel them moving forward from the back of the mouth to the front.