The Problem

From time to time the paper The Error of Erasmus and Un-Greek Pronunciations of Greek, by Chrys Caragounis, is cited to argue for the remarkable notion that Ancient and Modern Greek are pronounced essentially the same. The article is flawed in many ways, but it keeps popping up. There used to be a lengthy article available online, written by an actual linguist, detailing the many problems with Caragounis' arguments. Unfortunately, that article has disappeared without a trace. Since I can no longer refer to that I need someplace to direct people so I don't have to make the same arguments endlessly myself. I haven't aimed for the same thoroughness as that other article.

It is not my goal to explain the Erasmian or reconstructed pronunciations here. For that you should read Allen or Horrocks (see the references below). Nor am I claiming that it's illegitimate to pronounce Ancient Greek in the Modern way — I don't worry myself about how others pronounce their Homer or Plato. I do, however, get cranky when people assert that Ancient and Modern Greek are pronounced the same and that I must pronounce Ibycus like Cavafis — the notion is completely unsupportable.

ἀπὸ τῶν πρώτων

The fundamental error of Caragounis is his evident belief in the exceptionalism of the Greek language. It is, we are expected to believe, completely unlike any other language in human history known to us, in that it alone has been immune to sound change for nearly two and a half millennia. Why this should be so is never explained. He freely admits that the lexicon and grammar have changed over time, and, further, posits a great many sound changes in vowels between about the 6th and the 3rd centuries BC.

2. EI = I. EI interchanges with I since the VI-Vth century B.C., indicating both that it was sounded monophthongally, i.e. as one sound, and that it was sounded as ‘I’, or something very similar to it.

4. UI = I. The I of this diphthong very early had come to be swallowed by or contracted with the U, and the diphthong was pronounced as a simple U (see above). This phenomenon is clearly witnessed since the Vth c. B.C.

8. Ο, ΟΥ and Ω. The letter Ο interchanges with ΟΥ very frequently from the VIth century to the IIIrd century B.C.

The mechanism by which a language that experienced a massive reconfiguration of its sound system in a brief, 300 year span should suddenly seize up and remain essentially static needs to be explained.

He is prepared to make impressionistic judgements in lieu of argumentation:

The impossibility of pronouncing the diphthongs in diaeresis (i.e. each vowel distinctly) becomes obvious also from a word such as Εὐαοῖοι (see IGA 110, 2, early VIth c. B.C.). This word, which consists of seven vowels, pronounced in the Erasmian way, would give the comical sound: `E-u-a-o-i-o-i' - as if it were an exercise in vowel mnemonics.

What Caragounis, or anyone else for that matter, finds comical is quite irrelevant to the question. Nor does it matter what he finds difficult to pronounce:

Here one should bear in mind that Greek, basically a polysyllabic, vowel-loving language, avoids the concentration of unnecessary, difficult-to-pronounce consonants so characteristic of German, cf. e.g. Nietzsche and other words with six or even seven consecutive consonants. Greek pronunciation cannot be determined by what is possible or acceptable in other languages.

The entire science of linguistics rests on comparisons between languages. There is simply no reason to exempt Greek. I will leave it to Germans to comment on their unnecessary and difficult consonant clusters.


Caragounis consistently conflates the Erasmian and the reconstructed pronunciations. This is quite misleading. The standard academic pronunciation in common use these days is called "Erasmian" but isn't necessarily much like the proposals Erasmus published in De recta latini graecique sermonis pronuntione, nor is it anything at all like the reconstructed pronunciation found in Allen's Vox Graeca. Here are some examples with the academic Erasmian pronunciation as practiced in the U.S., the reconstructed pronunciation and the Modern Greek pronunciation (it is somewhat confusing here that the International Phonetic Alphabet draws so much of its inventory from Modern Greek):

Letter Erasmian Reconstructed Modern
β /b/ /b/ /v/
γ /g/ /g/ /ɣ/
δ /d/ /d/ /ð/
ζ /dz/ /zd/ /z/
θ /θ/ /th/ /θ/
φ /f/ /ph/ /f/
χ /x/ /kh/ /x/
αι /ai/ /ai/ /ε/
ε /ε/ /e/ /ε/
η /e/ /ε:/ /i/
οι /oi/ /oi/ /i/
etc., etc.

The academic Erasmian, then, is a sort of hybrid of Modern Greek and Latin. No one argues this is correct or accurate.

In a document from his web site (Erasmianism in New Garb: The Chimera of the ‘Reconstructed’ Pronunciation of Greek, updated June 14, 2007) he badly misrepresents the work that has gone into the reconstructed pronunciation:

Those who make the preposterous claim of being able to tell us exactly how Greek was pronounced in a particular place and at a particular time, in the face of utter lack of evidence for such reconstructions, simply cannot be taken seriously. Their concoction is a Chimera. It is self-illusion and it leads others astry.

If anyone working on the reconstructed pronunciation actually claimed to know exactly how Ancient Greek was pronounced, I would agree with Caragounis. But of course no one makes such a claim. To say the earth is flat is false. To say the earth is a sphere is also false — it is slightly deformed by rotation. Nonetheless, it is a lot less false to say the earth is a sphere than it is to say it is flat. In the same way, using Modern Greek as a good representation of, say, Homer is false. The reconstructed pronunciation is doubtless incorrect in some details, but it's a lot less wrong. This notion ought not be troublesome to any scholar. It is, further, completely untrue that there is an "utter lack of evidence" for the reconstructions. There's all sorts of evidence: the phonotactics of the language itself; inscriptions, especially spelling variations; foreign borrowings into Greek and Greek borrwings into other languages; descriptions by ancient Greek grammarians; and historical linguistics permits comparison to sister languages.

In The Error of Erasmus he claims that "more recent Erasmians avoid the inscriptions (particularly the earlier ones)." This is also false. In Vox Graeca Allen constantly references inscriptional evidence.

Nor does he seem to fully understand what exactly the reconstructed pronunciation is. Here he is on the circumflex accent:

However, the form of the circumflex only indicated that it was the result of the contraction of two vowels, one ὀξυνόμενον the other βαρυνόμενον, but it had no rising and falling tone in pronunciation — an impossibility in actual speech — for once the contraction had taken place there was but one position in the mouth and one dominant accent, the acute.

I trust he does not mean that pitch doesn't change during speech and that people speak in a monotone. But that means he has somehow come to the stunning idea that the reconstructed pronunciation posits a simultaneous rising and falling pitch for the circumflex, which of course would be impossible. But the circumflex marks a pitch contour — up then down — not Tuvan throat singing.

Going back to the Εὐαοῖοι example, neither the academic Erasmian nor the reconstructed pronunciation match his seven-syllable rendition, "E-u-a-o-i-o-i". The reconstructed version of this is /ew-wa-oj-joi/.

Factual Errors

Throughout his article Caragounis makes basic errors of fact when talking about linguistics. The word "impossible" comes up rather too often, one example of which we have already encountered in his discussion of the circumflex. In discussing aspirates:

Finally, the fact that the preposition ἐκ does not change before κ, τ, π, but before θ, φ, χ it actually often becomes ἐχ (e.g. ἐχ Θετταλίας, ἐχ θητῶν, ἐχ φυλῆς, ἐχ Χαλκίδος), which would be impossible to pronounce as ek+h-K+h-αλκίδος etc.

It is not, in fact, impossible to produce clusters of aspirated stops with aspiration articulated for both consonants. Several Indic languages do so, for example, as do Georgian and Armenian. You can hear sound samples of it at the UCLA Phonetics Lab Archive (the Armenian corpus has an example right away).

Phonology, the central concern of his article, has an extensive vocabulary which allows one to describe vocal sounds with considerable precision. Yet nowhere does Caragounis use this technical vocabulary to describe the sounds under discussion, favoring instead idiosyncratic terms without precise meaning. Iota, ι (high, front, unrounded, i.e. [i]), is described as "thin" and the development of an iota-like pronunciation of ypsilon, υ (high, front, rounded, i.e. [y]), as "thinning down." It's not clear if this "thinning down" represents the change of [u] to [y] (fronting) or of [y] to [i] (unrounding) or the complete [u] to [i] development. He describes the development of αυ and ευ to (Modern Greek) [af] an [ef] as "labialization (the pronunciation with the lips, i.e. as consonants) of these diphthongs." Labialization in normal linguistic terms refers to a quality of consonants. The term for the process he's describing is fricativization. Further, the diphthongs in question match his own definition of labialization — pronunciation "with the lips" — while still vowels.

He claims that "all Indo-European languages are based on stress accent." Even trivial research will show this to be false. We have every reason to believe Vedic Sanskrit used a pitch accent, and several Slavic and Baltic languages have such accent systems today (Lithuanian, Croatian/Serbian). Further, he does not understand what pitch accent means: "it is commonly assumed that ancient Greek accent was musical pitch-accent, not stress-accent, as though the Greeks always sung and never used ordinary speech." The pitch accent system is no more musical than is the tonal system of a language like Chinese or Yoruba. Most languages use tone contours to indicate something, such as emphasis or type of statement. No one imagines that the rising pitch of a question in English is singing, and similarly there is no reason to compare a pitch accent to music.

The circumflex accent is the cause of numerous difficulties:

Each syllable of a Greek word is accented. However, polysyllabic words stress one and only one of the syllables above all others. This dominant accent (stress) was called acute (ὀξεῖα) and was indicated by the mark ('), while all other syllables received the mark of the grave (`) (βαρεῖα): e.g. ΚᾺΤᾺΞῚῺΘΈΝΤῈΣ. The third mark to come into being was the circumflex (^) (i.e. περισπωμένη) placed on contracted vowels and explained as the combination of the acute with the grave (ˆ), i.e. the percussion or stress and its absence on two adjoining vowels prior to their contraction: e.g. νόὸς > νοοῦς.

Any long vowel or diphthong may receive the circumflex accent according to the rules of Greek accentuation, not just contracted vowels, as he appears to be saying here: τῆς, δῶρον, πνεῦμα, κτλ. Later he says that "in the rules governing accentuation, the circumflex functions exactly like the acute: cf. ἐκ τῆς γῆς ἐστιν with ἀληθής ἐστιν." This is not true. For enclitics, when the accent of the previous word is on the penultimate syllable the behavior with circumflex and acute is different: παῖδές μου vs. φίλος μου (Smyth §183).


Most people will learn of the reconstructed pronunciation from the third edition of Sydney Allen's Vox Graeca: The Pronunciation of Classical Greek. It has an introductory chapter explaining some phonetics which will be of use to those with without much lingistics training.

A more detailed work is Geoffrey Horrocks' Greek: A History of the Language and its Speakers. It discusses in great detail the sound changes in Greek by time period. A second edition was recently released (March 2010).

The Wikipedia article on Ancient Greek Phonology is quite good (as of August 3rd, 2008).

Document History