On Romanization of Greek


1 Introduction

As any regular reader of SARTRIX will be aware, the entanglements of ancient Greek and Latin literature matter to me deeply. I do not see “Greece” and “Rome” as separate cultures that merely influenced one another, but as overlapping parts of one – internally multilingual and multicultural – Mediterranean society. This is why it earnestly rends my heart to see many Anglophone hellenophiles openly hostile to anything Roman or Latin, invoking “authenticity” to justify an ahistorical logic of purity.

On this page, more specifically, I will make my case that the notion of an “authentic” Romanization¹ for ancient Greek words – and the attendant assumption that other Romanizations (namely “Latinizations”) are inauthentic or compromised – is incoherent, and that there are other, more complex factors which must guide our usage. In addition to this theoretical argument, I will also (gods willing) provide practical advice on how to parse Romanized Greek, especially in the absence of an agreed-upon standard of spelling or pronunciation.

1: Romanization refers to the adaptation of a word from a different script or spoken language into the Roman alphabet, making them usable in written English.

2 The problem with authentic “Romanization”

Let us take the name of Semele’s son as an example. The conventional spelling of his name in English is ‘Dionysus’, pronounced /ˌdaɪ.əˈnaɪ.səs/ (International Phonetic Alphabet) or /die-uh-NIE-suhs/ (English Phonetic Spelling). More recently, however, the name is also being written as ‘Dionysos’ and ‘Dionusos’, and these spellings are touted as more correct. Now, to be clear, I do not take a normative stance on which of these three should be used; rather, I oppose the normative claim that the more novel spellings are superior.

Traditional Anglicization

But first, an explanation of how we got the traditional Anglicization to begin with. Initially, English only used Latin theonyms for Ancient Mediterranean deities, since there was virtually no Greek literacy in Britain. So, the son of Semele was known as Liber Pater or Bacchus (the latter being a byname borrowed into Latin from Greek). But in the 18th century, as it seems, his Greek proper name also began to be employed, and just as with the thousands of other Greek names that had been adopted into English via Latin over the previous centuries, it was borrowed in a Latinized form, ‘Dionysus’. A strict letter-for-letter borrowing (a transliteration) would have yielded ‘Dionysos’, but this would have flown in the face of all precedent and regularity (e.g., it would have distorted the extremely close connection between ‘Dionysus’ and ‘Dionysius’, a name that had been borrowed earlier).

As for pronunciation, this followed not the Greek of the time, much less any ancient form of the language, but rather the rules for the traditional English pronunciation (TEP) of Latin (off-site link), which accommodated it to English phonetics. Both the ‘i’ and the ‘y’ (not distinguished in medieval Latin and traditional English pronunciation) were treated as long and read as /aɪ/. A penultimate syllable with a long vowel receives stress in Latin, so that the preceding and following syllables were left unstressed, and their vowels reduced to the indistinct /ə/, while the initial syllable received a secondary stress. Hence, /ˌdaɪ.əˈnaɪ.səs/, with /ˈ/ marking primary stress, /ˌ/ secondary.

‘Dionysus’ in traditional English pronunciation (TEP)

This has two advantages. Firstly, so long as one knows that the ‘y’ is long, this is the only pronunciation that the general rules of TEP allow. Secondly, it is consistent with English phonemics, and so can be picked up and reproduced accurately by anyone who speaks the English language fluently.

Novel Romanizations

As more and more languages were Romanized, and as linguistics grew more sophisticated, scholars gradually came to place more of an emphasis on also reproducing the Greek orthography closely in Romanization. At its simplest, this only means replacing the ending ‘-us’ with ‘-os’, while retaining the accessible TEP pronunciation, /ˌdaɪ.əˈnaɪ.səs/. At the other end, vowel length and accents are also marked, rendering ‘Diónūsos’ or ‘Diónȳsos’, and a closer approximation of ancient Greek – or, more recently,¹ modern Greek – pronunciation is attempted (but not so often accomplished).

This opens up several difficulties, however. Firstly, conventional, Latinizing spellings and pronunciations are fairly consistent and their rules only have to be learned once, while more recent Romanization schemes are quite diverse, and I can attest from experience that it is challenging to maintain a consistent standard in writing. This also makes it hard for the reader to recognize the import of a given spelling, and thus to identify the same word, or related words, across different contexts. It further opens up a chasm between words that have long been Anglicized (usually according to TEP rules) and words newly Romanized, such as ‘bacchanals’ and ‘Bákkhos’ – where the traditional ‘Bacchus’ makes the connection immediately apparent.

More problematic than this, I think, is the unmooring of pronunciation from ordinary English phonemics. Simply put, an English-speaker cannot accurately pronounce Greek either according to its ancient or modern phonology, unless they spend not inconsiderable time and effort to learn. This means that even the most accurate Romanization cannot produce an “authentic” pronunciation, except for those speakers who already know Greek to some degree. The mediating role of the TEP (which is already an integral part of the English language, even if not everyone consciously understands its rules) falls aways, but is not replaced by anything else.

Especially for monolingual English-speakers, the novel Romanizations thus obscure the difference between a borrowing (which accommodates the word to your language’s pre-existing rules) and code-switching, where you pronounce and use the word according to the rules of the original language, briefly slipping into Greek and then back into English. By treating traditional Anglicizations as borrowings, and novel Romanizations as direct representations of Greek, a kind of pseudo-code-switching is produced, which in reality is still nothing other than borrowing, but presents itself as more “authentic”. It is this false sense of authenticity, rather than any given Romanization scheme, that is the problem.

1: More recently in the Anglophone world, that is. Most modern Greeks have naturally always favored the modern Greek pronunciation.

Unnecessary confusion

Let me give another example. The word ‘Nymph’ has been part of the English vocabulary since the Middle Ages. Beginning from Greek νύμφη, it was borrowed into Latin as nympha, became nimphe in Old French, and was taken over into Middle English with the same spelling, nimphe. In both French and English, the final –e was subsequently lost in pronunciation, though retained in French writing. Both languages also later changed the i to y, to more closely reflect the word’s Greek and Latin origins, although this orthographic difference no longer changes pronunciation. In brief, the form ‘Nymph’ reflects a continuous tradition of speakers reaching back from now to ancient Greek – just as does the modern Greek νύφη, which could be transcribed as nífi (alongside which exists nímfi, the modern pronunciation of the ancient form).

Here are my best approximations of the pronunciation in these various languages and periods, beginning with the early Ionic Greek (e.g., Homer, Herodotus) and Koine Greek, as used in the Roman period. Classical Attic (e.g., Demosthenes, Plato) has an intermediate pronunciation, /ˈnym.pʰɛː/, with the first syllable as in Koine and the second as in early Ionic.

νύμφη in ancient Greek (early Ionic), /ˈnum.pʰɛː/
νύμφη in ancient Greek (late Koine), /ˈnym.feː/
nympha in Latin, /ˈnym.fa/
nimphe in Old French and Middle English, /ˈnim.fə/
‘Nymph’ in modern English, /ˈnɪmf/
νύμφη in modern Greek, /ˈnim.fi/

In light of this history of constant change, it generally seems most prudent to me to use the established English form, ‘Nymph’, which has a clear pronunciation and whose grammatical functioning is entirely regular: ‘a Nymph, several Nymphs’.

The alternative is to use some Romanization like ‘Numphe’, which poses a whole series of challenges. For one, the transcription ‘u’ for Greek υ implies an ahistorical ‘oo’ (/uː/) or ‘yoo’ (/juː/) sound, or at best a short ‘u’ (/ʊ/) as in ‘butch’ – the latter would be close to early Ionic, but clashes with the intuitive pronunciation of ‘ph’ as /f/, which only came into use much later, after υ had switched to /y/ (and later /i/).

Further, it is not self-explanatory to English-speakers how the final ‘e’ should be pronounced; it might be silent, or an ‘ee’ or an ‘ey’. In modern Greek, it is indeed an ‘ee’ (/i/), and so it was in late Koine; but at earlier times, it was a sound that simply does not exist in most English dialects, a long /eː/ or /ɛː/. This is often approximated as the diphthong ‘ay’ (/eɪ/), but it is only to English-speakers that they actually sound the same.

In short, ‘Numphe’ may be a more direct representation of Greek spelling, but if it is pronounced using the sound inventory of English – as /NOOM-fay/ (IPA: /ˈnuːm.feɪ/) or the like –, it is entirely as far removed from historical pronunciation as ‘Nymph’, while lacking the advantages of being well-integrated into English, and tracing back to Greek through a continuous history.

Besides, the pseudo-code-switching creates grammatical confusion. The plural of ‘Nymph’, as everyone knows, is ‘Nymphs’. But the plural of ‘Numphe’ has to be memorized. Without knowing a little Greek, it is impossible to guess that it is ‘Numphai’ (νύμφαι). Nor, if one encounters either form without prior acquaintance, is it self-explanatory which is the singular and which is the plural. Hence, Anglophone hellenophiles often use one form in place of the other, or erroneously append entirely different endings.

I want to be clear that I am not discouraging the use of the spelling ‘Numphe’, the plural ‘Numphai’, or even the pronunciation /NOOM-fay/. I personally do not like them, but that is simply my preference. What is wrong is to think that their use (or the use of any other convention) confers any kind of special authenticity that other forms do not. Not even studying ancient Greek deeply does this, because there is no one historical pronunciation that is superior to the others. There is no shortcut, and there can be none, because there is not a single goal or standard by which we are measured.

3 How to read Romanized Greek

The following is not a pronunciation guide. As I said, there are multiple valid ways to pronounce Greek that diverge greatly from each other, and many of them can only be acquired gradually. Rather, this section is meant as a guide to parsing ancient Greek, of recognizing phonemes (abstract sounds) independently of any specific realization. For instance, it is very valuable to understand that Greek υ (Ypsilon) is a unique vowel, even if you cannot pronounce it, because that way, you can see the underlying commonality between spellings like ‘Numphe’ and ‘Nymph’.

Vowel values outside of English

Perhaps the most fundamental rule of parsing ancient Greek is that there are no silent letters. Do not, for instance, let anyone tell you that the pronunciation of the Athenian month Gamelion is “game-lion”. The word has four vowels, and each belongs to a separate syllable: ga-me-li-on.

The second most important thing to observe is that ancient and modern Greek vowels are transcribed according to so-called continental pronunciation. Knowing the continental values of Roman vowel letters is generally useful because virtually every other language using the alphabet uses them; only English underwent the so-called Great Vowel Shift. In modern English, as a consequence, the same vowel sign can have very different values.

Sometimes, a vowel letter in English will stand for a reduced vowel, such as the ‘u’ in ‘Dionysus’, which represents no specific ‘u’ sound but an indistinct /ə/, which can also be written in various other ways. These reduced vowels have at most a marginal role to play in Greek phonology. Rather, you should understand each vowel letter – conceptually, at least, if not necessarily in your pronunciation – as representing a distinct vowel sound. An unstressed ‘u’ is still an /u/, an unstressed ‘i’ is still an /i/, and so on.

Further, the Great Vowel Shift in English has led to a situation where sounds may be coordinated as the short and long version of each other while actually being quite different in pronunciation. Take, for instance, the short ‘u’ in ‘but’ (/ʌ/), and the long ‘u’ in ‘boot’ (/uː/); in most dialects, these are really very distinct sounds. Or take the short ‘i’ of ‘in’ (/ɪ/), and the long ‘i’ of ‘mine’ (/aɪ/). The latter is not even technically a long vowel, but a diphthong, a sequence of two vocalic sounds. In ancient Greek, by contrast, the long and short pronunciation of a vowel are precisely the same vowel sound, either spoken briefly or more drawn out. (There are some complications to this, but as a general rule it holds true.)

So, let us get back to the example of Gamelion. It has four distinct vowels, of which two (the ‘e’ and ‘o’) are long, with a stress on the last syllable: Gamēliṓn. In Koine Greek, this would have been pronounced more or less as follows:

Gamēliṓn (Γαμηλιών) in Koine Greek

This happens to contain four of the five continental vowels (a e i o u), so let us take these in order:

A, E, I, O and U in continental pronunciation

Ancient Greek does not entirely hold to this schema of five vowels that can be either short or long (as we will see), but it is still a reasonable point of departure for this language as well as many others (Spanish, Italian, German, etc.).

Now, this does not mean that you must actually pronounce the vowels in a continental manner, which is quite challenging for people who only speak English. By all means adapt the word to the traditional English pronunciation, or key the sounds to English in some other way. But conceptualize it as being composed of short and long versions of the basic vowels.

Gamelion in TEP, /gəˈmiː.li.ən/
Gamelion in more novel Anglicization, /gəˌmeɪ.liˈɔːn/

Note, however, that vowel length is something that you cannot guess from the spelling of the word. Double consonants, unlike in English, do not indicate that the preceding vowel is short, but are actually pronounced as long (‘geminated’); and Greek often has long vowels where they would be impossible in English. Of course, there is no strict necessity to observe vowel length today; modern Greek does not do so either.

The vowel system of ancient Greek

While the continental vowel system explained above captures most nuances of, for instance, Latin vowel phonemics, except for relatively minor features, in Greek, there are some important complications. Firstly, there are seven rather than five vowel letters; secondly, one of these is Ypsilon; and thirdly, there are two digraphs – sounds written using two letters – which have been pronounced as monophthongs (simple vowels rather than diphthongs) for most of the language’s history. These three things are all interconnected.


Accents and aspiration

Like vowel length, accentuation is usually omitted from Romanization, even in the novel schemas. This is a little odd, since the TEP has fairly simple rules for where the accent or word stress goes, but Greek itself has highly irregular stress, meaning that clear indication of the accent is indispensible, even if including the accents can be a little unwieldy, or overwhelming for beginners.

But the rules are really quite simple: the vast majority of diacritics used in Romanization marks either length or accent (stress) or both. The macron (ā) always marks length; the circumflex (â) either length or length and accent, depending on the system; and the acute (á) marks accent irrespective of length. The grave (à) effectively stands in for the acute in certain situations. Everything beyond this takes us outside the realm of what most English-speakers can pronounce, so let us leave it at that.

Separate from these diacritic marks is the diaeresis, which works just as it does in English, e.g., ‘naïve’. It marks that, in this case, ‘a’ and ‘i’ are pronounced separately rather than as a diphthong. The name Laïs (or Laḯs), for instance, represents the disyllabic /laˈis/ rather than monosyllabic /laɪs/.

Related to accentuation is aspiration, that is to say the consonant spelled ‘h’ in Romanization. In the Greek alphabet, it is not written, but only indicated by an accent if it is at the beginning of a word, and not at all if it is anywhere within the word. This hardly matters when reading Romanization, but it is useful to be aware of when one compares a Romanized spelling to the Greek.



Digraphs and gemination


Diphthongs and iota sub-/adscriptum


4 Conclusion

I apologize if the above was overly long and technical; I tried my best to explain everything as clearly as was possible without simplifying any of the issues. No doubt a table of corresponding pronunciations for each Greek letter would be more handy, but it would not address any of the problems that I see in people’s engagement with the language.