Go to previous page Go to main page Go to next page

All That Glitters Is Not Gold

The principal question raised by the sceptic who is informed of a "reconstructed" ancient phonetic model is, in the words of Allen (ALLE78, p. viii):

"How do we know [what we claim to know]?"

The text within brackets is a necessary addition, since we have no direct data (i.e. sound recordings) from the era and language in question, but the "reconstruction" is based on inference from several kinds of secondary data. And it is exactly these kinds of secondary data that will be the subject of this chapter.

Allen devotes only a few lines to list six categories of data (ALLE87, p. xiii) and another couple of pages to explain that, unlike Latin which was one language and spawned many scions, Greek was already split into several dialects which left but a single descendant; he does not discuss the strengths and weaknesses of each type of evidence, but merely cites, for every letter, all available data in support of the conclusion that he has already announced (a regular practice of his). Sturtevant is certainly superior to Allen in that respect, because he took great pains to discuss the "Nature And Value Of The Evidence" in 12 pages (STUR20, pp. 1-12), providing typical examples for each category of evidence and discussing the pros and cons thereof. Chatzidakis surpassed both of them,Since Chatzidakis predates Allen and Sturtevant by six and two decades, respectively, it would be more accurate to say that he has set the standard too high for any of his successors. having devoted an entire chapter of 88 pages (CHAT02, pp. 284-371) to the general discussion of the various kinds of evidence. Unfortunately, his main concern is not to set global rules, which he intends to follow, but principally to disqualify the work of Papadimitrakopoulos (PAPA89) by pointing out his questionable conclusions from dubious evidence or via faulty logic, forgetting in the next (more specific) chapter the very same standards he has set.For example, on pp. 342-343, where he discusses the value of the ancient testimonies on pronunciation, he correctly refrains from drawing conclusions from vague descriptions in Plato's dialogs, but he takes Plato's submissions literally or as he sees fit on pp. 374-375, where he wants to prove that Η had the value of [ɛ:].

My intension, in the present chapter, is to build upon the work of Chatzidakis, evaluate the various categories of evidence, show that some may carry more weight than others and establish guidelines for their usage in drawing conclusions about the phonology of "ancient Greek".

Tradition

It should be clear from the historical overview that Greek has been continuously in use in both written and spoken form from well before our time of interest down to the present day. In agreement with Sturtevant's approach, we will consider that "The original clue to the speech-sounds of Greek - the starting-point of our knowledge of the subject - is tradition" (STUR20, p. 1).Chatzidakis seems to also have tradition at the highest rank: "Αἱ κύριαι ὁδοὶ εἰς εὕρεσιν τῆς γνησίας προφορᾶς εἶναι ἡ ζῶσα Ἑλληνικὴ γλῶσσα καὶ προφορά...|The main paths to the discovery of the genuine pronunciation are the living Greek language and pronunciation..." (CHAT98, p. 394). Here, it should be clarified that the term "tradition" always refers to the mapping of letters (or "graphemes") and sounds (or "phonemes"), which will be our primary concern (since the main issue a "reconstructionist" is expected to tackle is how to read aloud the myriads of ancient texts that survived to the present date). In other words, for determining the "pronunciation of ancient Greek" we will seek the phonetic values assigned to the letters or digraphs of the Greek alphabet.

In contrast to Latin, there is a single tradition for Greek, the one that survives in mainstream Greek (cf. ALLE87, pp. xiv-xv), which includes the individual sounds of the language as well as their correspondence with the letters of the Greek alphabet. This (i.e., the uniqueness of tradition) has been challenged only by a few Catholic fanatics, such as Blass (BLAS90, pp. 12-13) under the guidance of Psichari, who claim that there are several traditions, such as "the genuine language" (meaning the language of ordinary people) and "the language of the learned" or "according to dialects and localities"; Blass, thus, cites the phenomena of consonantisation and dissimilation as evidence that the tradition of "the language as now spoken" differs from the tradition of "the cultivated language". For a scholar that was considered an authority in linguistics, it is not very flattering to confuse the question of pronunciation (i.e., how to pronounce a written word) with that of phonotactics (i.e., that certain sound combinations are favoured over others); any modern Greek, irrespective of background and dialect spoken, would always read out the written words "νύκτα|night" and "παλαιός|old" as [ˈnikta] and [paleˈos] (i.e., in accordance with the Orthodox grapheme↔phoneme mapping rules) and never [ˈnixta] and [paˈʎos] as Blass suggests, even if these last ones were the only forms of the words that he was familiar with.

Blass also cites regional differences in, e.g., the palatalisation of velars, with κε being variably pronounced as "kye, tye, chye, che, tsye, tse".What he means with all that strange notation is not clear. I cannot identify any difference between the modern-Greek κε and the German (i.e., Blass' native) ke, which sound both as [ce] to me; the only reason why the palatilised κ (i.e., [c]) is perceived as a different sound in Greek, but not in German, appears to be the phenomenon of consonantisation (which is not to be found in German), wherein κι (in, e.g., κια, κιο, κιου) "stands for" the palatal sound ([c], i.e., as [ca], [co], [cu]), as opposed to the German practice of pronouncing palatal+[i] (cf. also ciao pronounced as [t͡ʃiao] in English and occasionally in German, as opposed to the italian pronunciation [t͡ʃao]). The only diverging pronunciation I am aware of is that of some islanders, particularly those formerly under Venetian or Italian occupation, who pronounce κε (almost) the "Italian way", i.e., [t͡ʃe] or [t͡ɕe]. This is probably the sound that Blass is trying to approximate with all that clumsy notation. Such regional differences in the pronunciation of particular letters do indeed exist. For example, in my hometown, σ(=[s]) is sometimes (as exemplified by the late former Prime Minister Konstantinos Karamanlis) pronounced [ʃ]; in Thessaloniki, λ(=[l]) is often (particularly among populations originating from Istanbul and Turkey in general) pronounced as a "dark l", i.e., [ɫ]; in Crete and other islands that were once under Venetian or Italian occupation, κι(=[ci]) and κε(=[ce]) are pronounced the "Italian way", i.e., [t͡ʃi] and [t͡ʃe] or [t͡ɕi] and [t͡ɕe]; in certain areas, such as the Peloponnese (particularly around Patras), λ(=[l]) and ν(=[n]) in the syllables λι and νι are palatalised, i.e., the syllables are pronounced as [ʎi] and [ɲi], even if they are followed by a consonant. However, none of these should be a problem for a true linguist; they would all be taken into account as part of the oral tradition, bearing the following in mind:

Under the above principles, the reasonable conclusion from the present state of oral tradition is that there is only one tradition, the Orthodox. All diverging traditions can be ascribed to external causes acting on and modifying the Orthodox tradition. It is for this reason that the Orthodox model will be adopted as long as there is no reason to doubt individual letter-sound mappings, an approach which is consistent with the Presumed-Innocent Principle.

Tradition=Orthodoxy
For a particular grapheme, the value proposed by the Orthodox model will be adopted until other evidence prove it wrong.

The same approach (authority of Orthodoxy) had been apparently advocated by Bursian, with Blass responding that "we have here two traditions", meaning the written and the oral tradition (a view shared by Chatzidakis; cf. CHAT02, pp. 286-289) and that "it is a not less recognized principle to prefer the older and the literary to the later and oral" (BLAS90, p. 12). I do not know when was the last time written tradition spoke to Blass and how it sounded to him, but I would expect that no aural information can be obtained from something as static and as... silent as the dead letter of an ancient language. Blass goes as far as proclaiming that "The present sound in any language proves nothing [!] for the earlier", as if a particular letter or combination of letters might have a certain phonetic value during one period of the language and a different, unrelated one during another period (e.g., that A could have been a consonant in Latin, despite being a vowel in Italian). The problem of assigning phonetic values to the letters (and combinations thereof) of ancient Greek is already complicated enough, so let's not make it an impossible one to solve by disregarding any actual evidence available to us nor by relying on imaginary data represented by our subjective interpretation of a graphic representation.I remember a Finnish friend of mine trying to pronounce the name of the place in Greece where he had been, Kokkari. He strived, in perfect Finnish fashion, to geminate the second K and when I informed him that the double K is pronounced exactly the same as the single K, he said he was sure that it must have "originally" been pronounced the Finnish way, otherwise why would anyone spell it that way? It is not clear whether the name goes as far back as the times when gemination is said to have existed in Greek and I wonder whether my friend is also sure that English consonants as in, e.g., "bitter", "butter", "better" etc, were "once upon a time" geminated (apparently they were never geminated, since the doubling of a consonant in English is rather a convention for indicating that the preceding vowel is "short"; cf. ALLE87, p. 12).

Tradition≠Script
There is no such thing as "written tradition".

Tradition has not been used evenly in the various works on ancient phonology. For instance, in discussing "voiceless plosives" of Latin (ALLE78, p. 12), Allen uses the fact that "The Romance languages also generally agree in lacking aspiration" as a regular argument for the voiceless character of word-initial P, T, C. Oddly enough, the similar facts that Romance languages "generally agree" in the lack of a pronunciation [ke], [ki] or even [ce], [ci], for CE, CI and in the lack of any (consonantal) value for H (cf. STUR20, pp. 70-71, Romanian H coming from Slavic and French "h aspiré" coming from Germanic) is only considered when discussing C (p. 14) or H (p. 44) as an indication that the pronunciation has changed from an a priori accepted pronunciation (namely, [ke], [ki] and [h], respectively).

Furthermore, dialectal evidence from contemporary dialects (Cretan, Cypriot, Griko, Pontic, etc.) are readily considered as direct continuations of an ancient pronunciation, merely because they happen to coincide with speculative "reconstructions" (the aforementioned a priori accepted pronunciations). One has to be very careful with such claims, because, with the exception of Tsakonian (the origins of which can, pretty safely, be traced to ancient Laconian), there is a significant probability that all other modern dialects are direct offshoots of the medieval rather than an ancient language. For example, the following testimony is relevant for the question of "preservation" of ancient elements in Cypriot: "As an observer and a writer of Byzantine history, Cyprus strikes me as one of the few places in the world where the language, religion, the character and customs readily lends itself to the Byzantine years. You can see it with the old churches, monasteries, and the grammar and pronunciation of Cypriot Greek which has a canny similarity to the medieval style of spoken Greek." This together with the already mentioned example of Blass rushing to the premature conclusion that η "retained" its "e-sound" when followed by ρ, before being pointed at an alternative explanation by Psichari, lead us to the last conclusion about tradition.

Preservation?
Claims about preservation of (assumed or alleged) ancient features in modern speech should be taken with a grain of salt.

Linguistics

This category comprises arguments based on general teachings or results of the science of linguistics.

Inevitability of Change

Orthodoxy is looked at with scepticism by many scholars because it appears to contravene the basic "law" of linguistics, that of perennial change.This is as much of a "law" as is the famous "Moore's law" (as a matter of fact, Siamakis attempts to define a Moore's law for linguistics by claiming that sound change happens every 300 years or so; SIAM88, §2,769-772). Actually, they are both at most observations about the course of events, rather than scientific theories that explain why these phenomena (language change and processing-power doubling) happen, and as such they are not inviolable. The argument is summarised by Chatzidakis (CHAT02, p. 371): "ὅπως πανταχοῦ τῆς γῆς ἐν πάσῃ γλώσσῃ παρατηροῦνται σύμπασαι αἱ ἀλλοιώσεις αὗται, οὕτως ἀνάγκη νὰ ὑποτεθῇ ὅτι ἐγένετο καὶ ἐν τῇ ἡμετέρᾳ, ἐκτὸς ἄν τις πείσῃ ἑαυτόν, ὅτι αὕτη εἶναι τέρας τι οὔτε ἐξελισσομένη οὔτε ἀλλοιουμένη κατὰ τοὺς γενικοὺς τῶν γλωσσῶν νόμους|as allover the world in every language all these changes are observed, it is necessary to assume that the same happened in ours, unless one convinces oneself that that one [=our language] is some monster, neither evolving nor changing in accordance with the general laws of languages". This is a recurrent argument, which we have already seen and addressed. The historical distribution of sound change in Greek according to the official Catholic timeline is neither uniform nor consistent with the aforementioned "law". In order to demonstrate how puzzling the official Catholic timeline is, we can divide it in the following five time periods:The approximate times for each sound change are taken from ALLE87, particularly the tables on p. 78 and on p. 86 for the vowel sounds.

  1. X/VIII BC to V BC (3-5 centuries): from the adoption of the alphabet (which was presumably phonetic) to the dawn of the "classical age", only 4-6 sound changes occurred ([u(ː)]→[y(ː)], [ej]→[eː], [ow]→[oː]→[uː], also possibly [w]→∅ mostly word initially and [aː]→[ɛː] in most cases)
  2. V BC and IV BC (2 centuries): the so-called "classical times" were a period of alleged phonological stability, with at most one sound change ([ɛːj]→[eː])
  3. III BC to III AD (6 centuries): in the "Hellenistic and Roman times" (from Alexander to Constantine the Great), the phonological model turned upside down with sixteen sound changes ([zd]→[zː], [eː]→[iː], [oːj]→[oː], [aːj]→[aː], [ɛː]→[eː]→[iː], [aj]→[ɛː], [oj]→[yː], [h]→∅, [kʰ]→[x], [tʰ]→[θ], [pʰ]→[f], [g]→[ɣ], [d]→[ð], [b]→[v], loss of vowel length)
  4. III AD to X AD (7 centuries): in the eastern "Middle Ages" (from Constantine I to Basil II), at most three (doubtful) sound changes took place ([ew]→[ev] or [ef], [aw]→[av] or [af], [y]→[i])
  5. X AD to XXI AD (10 centuries): in the "late-Byzantine", Ottoman and modern times (from Basil II to the present), virtually no sound change was observed

One easily observes that, according to this timeline, almost two thirds (61,5%) of the sound changes that occurred in Attic/Koine/Greek in the last 3 millennia are concentrated in the 6 "post-classical" centuries (period C, which represents 20% of the time span); essentially, the phonological model was turned upside down in the period that is characterised by a (fairly uniform) common language, the great Grammarians, (relatively) high literacy and above all a period proximate to the one (="classical") that served as a beacon for all the subsequent ones. Only a few minor changes are to be found in the "dark" pre-classical era (period A) and the "Middle Ages" (period D), two periods that cannot be considered favourable to linguistic stability (the first was called "The Dark Ages" and the second coincides with the break up of Latin into the various Romance languages). Finally, there is an unusually long (1000-year) period of essential stability (period E), which ironically happens to be the most turbulent time period of the Greek history in terms of political cohesion and central governance.

The first conclusion from the above is that, by all accounts, Greek phonology defies the "law" of obligatory sound change, since both Catholic and Orthodox agree that there has been essentially no sound change in the last 1000 years. Thus, the special linguistic status of this particular language should come as no surprise even to the most devout Catholic.This feature of Greek (phonological stability over a very long period) does not seem to be unheard of. For one, Icelandic "was reputed to 'have remained virtually unchanged for a millennium'" (LYON06, p. 6).

Supergreek
The immunity of Greek to the "law" of sound change is a fact and its extend transcends the "normal" limits.

This does not mean that Greek has never undergone a process of sound change, but merely that the possibility that such sound change has not occurred in the last 2000-2500 years (i.e., "Orthodoxy") should not be regarded as abnormal (since we already accept that none has occurred in the last 1000 years, a time period much longer than any phonologically stable phase of any language we know). Two questions are associated with this finding:

In examining the "crystallization of orthography" (a.k.a. "historical orthography") in English, French and Greek, Blass (BLAS90, p. 9) attributes it to the deference to a previous language in a higher stage of cultivation", such as Latin for English and classical Attic for Greek. The same (i.e., respect for a higher previous stage of the language) appears to be a satisfactory explanation also for conservatism in phonological (and morphological) change; once an alternative pronunciation (or word form) appears, the weight of this past model provided enough centripetal force to discard the new variation as rustic, vulgar or even barbaric and to pass the old standard to the next generation through education, an explanation also adopted by Menardos: "this colossal attraction of the past explains also another feature of later Greek; namely, why its evolution, in comparison with that of Latin, which was broken asunder into the modern languages, has been so slow" (MENA08, p. 18). Thus, the answer to the first question appears to comprise two factors:

  1. a respectable (past) model creating a literary tradition; and
  2. a functioning educational system that perpetuates the model.

The answer to the second question (causes of sound change in Greek) cannot be decoupled from the answer to the first. In other words, the typical explanation "change happens" is not satisfactory, since it does not explain why change has not happened in the last 1000 years. What's more, it appears that the time period with the most sound changes according to the Catholic model (period C) is the one that better suits conditions a and b above, since it is, at the same time, the period immediately succeeding the "model" (period B) and the age of the Great Grammarians. On the contrary, the period of longest phonological stability (period E) is the one that least corresponds to the two conditions, since it is the most remote one from the classical "model" and is characterised by political fragmentation and higher illiteracy.As a matter of fact, it appears that there has been considerable differentiation in the various local varieties of Greek, at least at the morphological level, particularly after the "dark" Ottoman times, as can be inferred from the post-revolution work of Vyzantios "Βαβυλωνία|Babylonia [i.e., Babel]" (VYZA36), which deals with the problem of the various (often mutually unintelligible) dialects of Greek. The fusion of these into a common language based on the classical model is certainly the work of a successful educational system. If anything, one would expect the situation to be exactly the opposite, with period C being a time of relative stability and all sound changes occurring in period E.

I have not seen any attempt to explain this paradox of the Catholic timeline (i.e., the fact that the Attic/Koine phonology underwent changes for 10-15 centuries only to stagnate in the next 10-15), except for the one put forward by Chatzidakis: "ἀπὸ τοῦ 300 κἑξ [...] ἡ ἀνάμειξις τῶν Ἑλληνικῶν φυλῶν πρὸς ἀλλήλας καὶ πρὸς ἀλλογλώσσους πλείστην καὶ τῶν φθόγγων καὶ τῶν τύπων κλπ. ἐπήνεγκε σύγχυσιν|from 300 [BC] on [...] the blending of the Greek tribes with each other and with foreign-speaking ones brought about a lot of confusion of the sounds as well as the forms" (CHAT02, p. 342). Chatzidakis speaks as if the Greek tribes were blended once (and for all) in post-classical times and remained perfectly uniform after that (which is far from the truth). His second culprit is the non-Greek population of the conquered (by Alexander) lands, who adopted Greek as their official language, but were not able to pronounce it properly; however the "barbarians" of Egypt and Syria were never "mixed" with Greeks in Greece proper (i.e., Athens, Corinth, etc.) and any influence they might have had on the phonology of the motherland must have been from afar, a scenario as unlikely as claiming that the pronunciation in the Americas may be transplanted in and (significantly) influence that of Great Britain, Spain or Portugal. All in all, despite its verisimilitude, this theory of Chatzidakis' is not very credible. The Catholic scholars who came after him did not even attempt an explanation of the paradox, but silently passed over the issue.

TO BElieve OR NOT TO BElieve?
There is something rotten in the state of Catholicism (as regards the timeline of sound changes).

I believe that one point which has been overlooked is the fact that the present (=Orthodox) phonological model is fairly stable: it is fully symmetrical with sufficient distance between the various sounds, leaving little room for variation. A further fact is that of continuous political unity (which was first provided by Alexander) combined with literacy and scholarly tradition; this can be readily seen in the example of the western Empire, where the idioms of the various medieval states, neglected by an educational system that focused exclusively on Latin (considered a different language), attained the status of full-blown language, in some cases a common language, like Italian (from Florentine) and French (from Langue d'Oïl), being engineered for the purposes of a newly created state; on the other hand, it is highly unlikely that the various Greek scholars, starting with the first Grammarians, would perpetuate a model that would significantly differ from the classical one.Here, I might have been scorned by Chatzidakis, who declares (CHAT02, p. 295) that whoever claims that sound change could not have gone unnoticed and concludes that the same pronunciation is used since the first Grammarians (a clear reference to Papadimitrakopoulos) "οὐδὲν ἄλλο ἤ δηλοῖ σαφῶς ὅτι ἀγνοεῖ παντελῶς τὸν φυσιολογικὸν τρόπον καθ’ὅν μεταβάλλονται οἱ φθόγγοι πάσης γλώσσης|does nothing more than clearly declare that he completely ignores the natural manner in which the sounds of each language change". To that, I have to point out that it is not a question of how the sounds of a language change, but of how they are preserved (i.e., how the changes are reversed) under the weight of a respectable tradition. And I may ignore the "natural manner" of sound change, but so did Chatzidakis (as well as any other linguist), since "no one has yet observed sound change" (LLOY87, p. 11, quoting Hockett)!

Bearing that in mind and examining the five periods into which the history of the Greek language was divided, it appears that period A, which does not fulfil any of the aforementioned conditions for stability, is a far better candidate for any sound changes that led from the pre-historic pronunciation to the Orthodox. In other words, the Orthodox thesis that most changes were accomplished in period A, with a sudden interruption of the "normal" course of sound change in Greek after the classical age seems more likely than the Catholic thesis that the changes occurred in period C, which favoured stability, but were absent from a period that did not.

Comparative Linguistics

A favourite starting point for many "scientific reconstructionists" is a prior state of the language. This works well for, e.g., the Romance languages, the prior state of which (=Latin) is well documented. In the case of ancient Greek though, there is no documented prior state. Some scholars pretend that such a prior state is represented by "Mycenaean", referring to the imaginative "readings" of the Linear B tablets by Chadwick and his successors. I will refrain from making any reference to this doubtful (I have already explained that I do not accept Linear B as "decipherable", much less "deciphered") predecessor, since, even if the decipherment is correct,

  1. the spelling rules prevent us from assigning specific sound values to Linear B symbols with sufficient certainty;
  2. "Mycenaean" is not supposed to be a direct ancestor of Attic;
  3. in the seven centuries between its alleged last use (XII BC) and the time period of interest (V BC), anything could have happened (in phonology); and
  4. its phonology is based on a pre-conceived (=Catholic) model and cannot be used to prove the validity of the model on which it is based!

The "decipherment" of these prehistoric tablets being a recent development, few works on ancient-Greek pronunciation actually mention "Mycenaean".Allen, for one, seems to use it only when discussing the existence of semivowels. Most scholars (notably Chatzidakis) rather prefer to rely on a much more ancient language, which they call PIE (or PIG, as already mentioned), from which they attempt to trace the sound changes that led to "ancient-Greek phonology". That all sounds very scientific, but there is a... tiny problem: PIE is not attested in any kind of written or spoken form left behind by its alleged speakers! This is implicitly acknowledged by prefixing an asterisk (*) to the alleged PIE words, a practice scorned by Papadimitrakopoulos who aptly points out that "ὁ ἀστερίσκος παρὰ τοῖς γλωσσολόγοις σημαίνει λέξιν ἤ τύπον ὑποθετικὸν καὶ μόνον ἐν τῇ διανοίᾳ ὑπάρχοντα|the asterisk in linguistics denotes a word or type that is hypothetical and exists only in the mind" (PAPA98, pp. 36-37). Despite their lack of attestation, many scholars use PIE forms as if they were something given. Their certainty about the correctness of the imaginary PIE forms comes from a two century-long tradition of "reconstruction", a procedure that combines cognates (i.e., words that may sound similar and also have related meanings) from languages thought to be related and attempts to guess the corresponding words of the "parent language". I do not wish, at this point, to lay out my objections on the validity of the (widespread and dogmatically defended) "reconstruction",I do not fully share the exaggerated view of Papadimitrakopoulos: "Εἶνε μωρία τὸ νὰ θέλῃ τις νὰ συγκρίνῃ τὸ ἔργον τῆς μαθηματικῆς τῆς θετικωτάτης τῶν ἐπιστημῶν, πρὸς τὴν ἐπιστήμην τῶν Γλωσσολόγων καὶ τοῦ κ. Χ. ἥτις κατὰ τὸ πλεῖστον αὐτῆς μέρος σύγκειται ἐξ ὑποθετικῶν καὶ φαντασιωδῶν στοιχείων, ὡς δεικνύει ὁ ἀπειροπληθὴς κόσμος τῶν άστερίσκων τῶν σελαγιζόντων ἀπό ἄκρου εἰς ἄκρον τών πλείστων γλωσσολογικῶν γραμματικῶν τῆς Ἑλληνικῆς γλώσσης|It is foolishness for someone to want to compare the work of mathematics, the most positive of sciences, to the science of the Linguists and of Mr. Ch[atzidakis], which in its largest part consists of hypothetical and imaginary elements, as indicated by the ornamentation of countless asterisks, which span most linguistic grammars of the Greek languages from one side to the other" (PAPA98, p. 74). It is legitimate for a field of science to develop theories, in order to explain certain observed phenomena. One such phenomenon is the remarkable similarity of certain words that have the same meaning, e.g., ĕgo[LAT]ἐγώ|I, fĕro[LAT]φέρω|I carry/bear and most notably the number system of the "Indo-European" languages. It is reasonable to assume that the answer to this phenomenon is a common ancestor language (but cf. HORR10, p. 16: "the view that languages develop through divergence initiated by innovation on the part of subgroups within a previously uniform parent [...] is clearly an unrealistically restricted view of language development, particularly when it is known that speakers of the different varieties involved remained in long-term social and geographical contact: Greek, for example, developed initially within the confines of the Balkan peninsula, and any assumption of clean and permanent breaks between endlessly diverging varieties is plainly implausible"). I do believe though that the "reconstructors" were too eager to reconcile marginally compatible "cognates" (or "to emphasize the little that is generally common to the whole group, and overlook the innumerable specific differences", in the words of Jannaris, JANN97, p. ix), leading to monstrous and barely pronounceable forms, e.g., *bʰn̩ɡʰ-, which are reminiscent of cough rather than actual speech. But what bothers me the most is the presentation of the (PIE) speculation as certainty (cf., e.g., STUR20, p. 156, "The rough breathing comes chiefly from Indo-European s or i̯"), whereas both the identification of cognates and the reconstructed form are at most "best guesses". but merely to explain the reasons why PIE cannot serve the objective set forth (classical-Attic pronunciation):

  1. Chicken-and-egg: A reconstructed PIE form relies on a few attested forms in one of the Indo-European languages. For example, the stem *bʰer-|to bear/carry is derived from the cognates fĕro[LAT], bhárati[SAN|Sanskrit] and φέρω. The symbol "bʰ" corresponds to the initial consonants of the three cognates, namely "f", "bh" and "φ". In order to assign a phonetic value to this symbol, so that we can use it for determining the pronunciation of the corresponding Greek letter Φ, we need to know the sound that corresponded to Latin "f", Sanskrit "bh" and... Greek "φ". But if we knew Φ's value, we wouldn't be looking for it, would we? The most one can say in this (and the other similar) case(s) is not that Φ corresponds to PIE "bʰ", but rather that in (some?) cognates it corresponds to (the position of) Sanskrit "bh" and Latin "f".
  2. Very Early Mycenaean: Like Mycenaean, PIE predates the period of interest (V BC) by several centuries. In fact, the time interval between the time that PIE was spoken and the classical age is at least as long as the interval between the classical age and today. If (according to the official Catholic narrative) the bulk of the aforementioned sound changes took place in the six post-classical centuries (a time period of relatively high scholarship), the sound changes that must have taken place in the 25-35 centuries between PIE and classical Greek (a time period characterised by illiteracy and lack of any educational system or even state) make the PIE phonology almost irrelevant for the objective at hand.
  3. Not The Right Mixture: PIE, if it ever existed, was not simply an early form of Greek. In the 19th century, Greek was treated as the normal evolution of PIE, when the "Greek tribes" broke away from the other "Indo-Europeans". However, the current view (or at least a more likely hypothesis) is that the indigenous population played a significant role in the development of the language,cf. HORR10, p. 21: "Greek in toto is the product of the consequential contact between the Indo-European dialect(s) of the incoming population and the language(s) of the indigenous populations"; also CHAD76, p. 2: "This hypothesis is that the Greek language did not exist before the twentieth century B.C., but was formed in Greece by the mixture of an indigenous population with invaders who spoke another language" and p. 3: "the mispronunciation of Greek by these aboriginals led to permanent changes in the phonetics of the language". which means that they may be responsible for the majority of the diverging features of modern-Greek phonology from the (alleged) one of PIE.

Even Chatzidakis, a fervent proponent of "Japhetic" etymologies, admits that "ἡ ἐτυμολογία ἡ ἐπὶ τῆς συγκρίσεως τῶν ἀδελφῶν γλωσσῶν καὶ της ἱστορικῆς ἐξετάσεως τῶν φθόγγων, τῶν τύπων, τῶν συντάξεων, τῶν σημασιῶν κλπ. τῆς Ἑλληνικῆς γλώσσης στηριζομένη, βοηθεῖ μὲν ἡμῖν μέγιστον εἰς ἐξεύρεσιν τῶν φθογγικῶν αὐτῆς νόμων, ἐγκαταλείπει δ’ ἡμᾶς ὅλως προκειμένου νὰ ἐξετάσωμεν τίς ἦν ἡ προφορὰ τῆς Ἑλληνικῆς ἐν δεδομένῳ τόπῳ καὶ χρόνῳ|the etymology based on the comparison of the sister languages and the historical examination of the sounds, the forms, the syntaxes, the meanings etc of the Greek language is of great help to us in discovering its sound laws, but abandons us completely [i.e., is of no use to us] when it comes to examining which was the pronunciation of Greek in a particular place and time" (CHAT02, p. 357).

Pre-Alphabetic Greek
Any pre-alphabetic stage of the language, even if taken at face value, bears little significance for the classical phonology.

One might argue that the PIE sounds may be used as a guide, a starting point for the determination of the Attic ones. In fact, many scholars, e.g., Blass, Chatzidakis and even Teodorsson, start from such a fiction (i.e., phonological model they think applied) and make their way to fact (the present-day pronunciation), on the way describing all intermediate phonological stages of the language. Here, the approach will be more pragmatic, namely from fact (what the pronunciation indisputably IS today) to fiction (what it MIGHT have been in the past up to the classical age).One may accuse me of disregarding all the great "achievements" of modern linguistics and of trying to re-invent the wheel. However, I believe I have amply demonstrated that the "wheel" that has been "invented" may be of... square rather than circular shape. Neither is the apparent consensus among the majority of scholars a reason to consider PIE a certainty, since on one hand scholars are products of a biased academic system and on the other hand there is little scientific merit in proof by majority without even an... Ecumenical synod (cf. also the view expressed about religion: "Isn't it obvious that scholarly communities are defined by certain axioms in which grad students are trained, and that they will lose standing in those communities if they depart from those axioms? [...] It simply does not matter how many scholars hold a certain opinion. If one is interested in the question one must evaluate the issues and the evidence for oneself."). One may, thus, correctly accuse me of heresy, because heretic is exactly what I am against the official linguistic religion. Any reference to evidence from comparative linguistics will be made through the "sister languages" rather than PIE (e.g., for φέρω the likely cognates fĕro[LAT] and bhárati[SAN] will be cited in lieu of the "reconstructed" *bʰer-).

Internal Evidence

I will use this term to describe evidence that relate to the usage of language itself, particularly in conjunction with its graphical representation, since the main question is the determination of the sound corresponding to each symbol (=letter).

The principal internal evidence in the case of ancient Greek is the use of the alphabet to write the language. It is assumed that each letter "originally" represented a single distinct sound and it is conjectured that phonemic identities (such as Ι=Η=Υ=[i] and Ο=Ω=[o] presently in Greek) and monophthongal digraphs (such as ΕΙ=ΟΙ=[i] and ΑΙ=[e] in current Greek) had no reason to exist. In the words of a Catholic enthusiast: "If you make a language from scratch why should you invent seven vowels for four sounds?". Even though the argument (obviously referring to the writing system, since Attic was certainly not "made from scratch") sounds reasonable,Even that assumption (perfect initial grapheme-phoneme correspondence) cannot be considered safe. I have argued elsewhere that "it is quite improbable that the ancient Greeks (the first users of a full alphabet) suddenly grasped all complexities of phonemic representation, something that eluded the users of previous writing systems". it is barely relevant for our main objective, since there is a long (dark) period of 3-5 centuries between the adoption of the alphabet as the writing system of Greek and its use in classical Attic (period A above). For similar reasons, Blass' argument "How can anyone possibly think that such an orthography was originally shaped to fit such a language?" (BLAS90, p. 11) is not directly relevant to the question of classical pronunciation, but is potentially relevant only for the time of adoption/modification of the alphabet (X-VIII BC).

Often the argument is made that the writing system "made from scratch" was the (Ionic) alphabet adopted by Athens during the archonship of Eucleides (a story mentioned by Arvanitopoulos), which came to be known as the "Greek alphabet". We have seen how Blass makes a case for the coincidence of writing and sound based on the assertion that "in Attica towards the close of the fifth century the entire system was absolutely changed" and on the claim that "Here was the opportunity in those cases, where the living sound had here and there deviated from the writing, to bring them again into harmony" (BLAS90, p. 10). However, this assessment is not correct: a) it was not "the entire system" that was changed, but merely a few orthographic conventions (specifically ΕΙ/Η, ΟΥ/Ω for long Ε, Ο and Ξ, Ψ for ΧΣ, ΦΣ) and b) it was not a case of fabrication of a new (and better) alphabet, but a mere (official) ratification of an ongoing practice, namely the use of the (already existing) Ionic alphabetThis (the fact that it was a mere case of adoption of an imperfect alphabet rather than engineering of a new alphabet) is evident from the lack of one-to-one correspondence between graphemes and phonemes in several cases, such as
a) the use of Γ for the velar nasal ([ŋ]) and the pure velar ([g] as per Catholicism),
b) the use of Σ for its usual voiceless value ([s]) and for its voiced allophone ([z]),
c) the use of Υ for its vocalic (allegedly [y]) and semi-vocalic (possibly [w]), even though the letter Ϝ (digamma) representing the latter (semi-vowel) was known from other dialects and past usage,
d) the use of the digraphs ΕΙ and ΟΥ to represent monophthongs (purportedly [eː] and [oː]), as well as the two letters in dieresis ([e.i]/[ej], [o.y]) in some cases,
e) the use of no symbol at all for indicating the so-called "spiritus asper", which was (so the catholic story goes) still pronounced.
Why did the Athenians not introduce new letters for all those cases? After all, the Greeks were not shy about introducing further letters in the original alphabet and in some cases, such as the digamma, it would be a mere case of restoration. Nothing like that happened, but the Ionic alphabet was merely ratified "as is" (was).
(maybe because it represented the current phonology better or because it could help in understanding metre better). Further similar statements made by Blass about the classical timese.g., "since the Athenians and also the other races did not yet possess any grammarians or etymologists to attach importance to a historical mode of writing, the only principle which could have weight was the phonetic" (p. 10), "Since then the ancient Greeks were not in a position to pay deference to a previous language in a higher stage of cultivation, they must consequently have originally striven to bring their writing as near as possible to the sound" (p. 9), "where there is but little reading and writing, as in Greece in the classical period and in western Europe in the Middle Ages, unless the sound is very stable and well defined, the orthography is extremely shifting" (p. 10) are actually his personal estimations or opinions (addressed, e.g., by Teodorsson) rather than fact-based assessments.

Graphemes do not speak
No safe conclusions about the classical phonology can be drawn from the usage of the alphabet for recording the language.

Another type of internal evidence is based on the study of (what Jannaris calls) phonopathy, namely phonological transformations (such as assimilation, dissimilation, elision, etc) occurring upon interaction of neighbouring sounds in a particular language, from which it is attempted to draw conclusions about the language's phonology. A typical example is Allen's thesis that in ancient Greek "there is no loss of nasal consonants before β, δ, γ as there is before the fricative σ, or as before the modern Greek sounds (e.g. acc. sing. το γάμο)" (ALLE87, p. 31), whence he concludes that the modern- and ancient-Greek values of β, δ, γ were different (similarly for φ, θ, χ on pp. 21-22). Here, Allen assumes that the mechanism governing the loss of a preceding nasal is the same in both ancient and modern Greek; hence, he concludes, the nature of the corresponding sounds must have been different. However, the assumption (and, consequently, the conclusion) is not correct, since (unlike, e.g., the palatalisation of the velars before front vowels) there is no physical reason that necessitates the loss of nasals before "fricatives", as is evident from other languages (cf. infeasible, Manfred, in front). The predilection of a certain language to avoid certain consonant clusters reflects rather the "Sprachgefühl" of its speakers and is known under the term phonotactics. The fact that different phonotactics do not imply different pronunciation is evident when considering, e.g., that Spanish, contrary to most other languages, such as English, French or Italian, does not tolerate word-initial consonant clusters starting with s (thus, Estefán, Esvetlana, escándalo),Incidentally, a Spanish friend pointed out that the very country is called España after all; however, this name may derive directly from Hispania[LAT] rather than, e.g., Spagna[ITA]. but this does not imply that Spanish s sounds any differently from English, French or Italian s. Similarly, modern Greek is generally intolerant to successions of stops; thus κτ, πτ are usually converted to χτ, φτ, but no-one has dared conclude from this fact that the modern and ancient values of κ, π, τ differed.

Phonotactics≠Phonology
A difference in phonotactics does not necessitate a difference in phonology.

Sometimes the conclusions drawn from internal evidence violate the Invariability Principle. For instance, Chatzidakis argues (CHAT02, p. 318) that the "diphthongs" ΑΥ and ΕΥ could not have been pronounced in antiquity as they are today ([av] or [af] and [ev] or [ef], respectively), "Διότι ἐν τύποις οἷον Ζεῦ, βασιλεῦ, εὖ, παῦ, ἄνευ, κττ. φαίνεται τοῦτο μὲν τελικός φθόγγος τὸ β [=[v]] ή το φ [=[f]], ὅπερ ἀνελλήνιστον, τοῦτο δὲ περισπώμενος τόνος ἐπὶ βραχέος φωνήεντος ε, ᾰ, ὅ ἀδύνατον|Because in forms like Ζεῦ, βασιλεῦ, εὖ, παῦ, ἄνευ, etc. it appears that on one hand β [=[v]] or φ [=[f]] is the final sound [of the word], which is ungreek, while on the other hand a circumflex is placed over a short vowel ε, ᾰ, which is impossible", disregarding the fact that he, the leading Greek scholar, is both pronouncing them and writing them in the same way that he calls ungreek!As a matter of fact, the modern language is even less tolerant than the ancient one as regards final consonants, since the latter allowed three more letters as final (Ξ, Ρ and Ψ), which are no longer acceptable; hence, the modern language is no less "greek" than the ancient.

And more often than not, some strange rules are devised when studying internal evidence; rules that appear to come from a linguist's own expectations or prejudices, rather than physical necessities or actual evidence. For example, Sturtevant decrees that "the product of contraction must be either identical with one of the original sounds or between them" (STUR20, p. 122), without explaining which physiological reason would impose this restriction or what mechanism underlay ancient-Greek crasis; furthermore, such a statement is not based on facts, as can be seen from the ancient-Greek contractions Ε+Ο→ΟΥ←Ο+Ο ([e]+[o]→[uː]←[o]+[o])The official storyline is that the contractions Ĕ+Ŏ and Ŏ+Ŏ resulted in Ō ([oː]), which was later shifted to the corner of the vowel triangle ([uː]); however, it appears that the only evidence in support of the claim [e]+[o]→[oː] is the grapheme used for the product of the crasis, which was represented by Ο, a character which was, at least in Attica, used for all vowel sounds on the back axes, i.e., later Ο (allegedly [o], or [o̞]), ΟΥ (allegedly [oː] or [uː]) and Ω (allegedly [ɔː]) and the only certainty (?) is that the modern value ΟΥ=[uː] dates from as far back as the classical age. Note also that the assumed value for short Ο (at least according to Allen; ALLE87, p. 62) is [o̞] (i.e., "between" [o] and [ɔ]); hence, the alleged result of the contraction [e̞]+[o̞]=Ε+Ο→[oː]←Ο+Ο=[o̞]+[o̞] is still not "identical with one of the original sounds or between them". or the development of the Latin diphthong OE (allegedly [oi], or better [oj]) in Vulgar Latin e ([e]). I refer to this kind of theories, as Lingo-Math, the linguists' attempt to establish a mathematical basis for their field.

Transcriptions

This category of evidence relates to certain uses of the alphabet to transcribe sound patterns for the value of which (we think) we have a pretty good idea. Their (un)importance is summarised by Chatzidakis with the phrase: "Τὰ λογοπαίγνια, αἱ παρηχήσεις καὶ αἱ τῶν φωνῶν τῶν ζῴων μιμητικαὶ λέξεις ἐλαχίστην ἔχουσιν ἀποδεικτικὴν δύναμιν, ἐπειδὴ διὰ τούτων σχεδὸν οὐδέποτε ἐπιτυγχάνεται πιστὴ καὶ ἀκριβὴς ὁμοιότης καὶ ἀναπαράστασις τῶν ἐκφωνουμένων φωνῶν|Word plays, alliterations and words imitative of animal voices have minimal proof potential, because faithful and accurate similarity and representation of the pronounced sounds is almost never achieved through them" (CHAT02, p. 351).

Extralingual

Natural sounds provide a good reference point and may help in deciphering the letters of the words that are used to describe them. In many cases, a word associated with a known natural sound is constructed so as to approximate the sound as closely as possible, a practice called onomatopoeia. The examples in the English language are abundant: fizz, zip, hiss, click, etc. One problem with natural sounds is that they are almost never reproducible in human language (e.g., /ˈklɪk/ is not exactly the sound produced by the mouse button), but are perceived differently by different people and are represented with crude approximation. Thus, the same sound may be represented by different words, as in the case mentioned by Wikipedia: "the sound of a clock may be tick tock in English, dī dā in Mandarin, or katchin katchin in Japanese". A careless Englishman attempting to decipher the Chinese symbols might rush to the wrong conclusion that the first symbol (corresponding to "dī") stands for "tick" and the second (corresponding to "dā") for "tock".

A particular type of natural sounds are the cries of animals. Here, the (reasonable) assumption is that animals of the same species sound more or less the same allover the world and that their cries must be perceived likewise by speakers of all languages. The present uniformity is also a sign of lack of linguistic development in animalspeak, so it is almost certain that the ancients heard (practically) the same sounds. Their occasional representation of such sounds may, thus, provide hints about the values of the letters of the representation. However, clues of this kind suffer from the same weaknesses as the evidence from the representation of the rest of the natural sounds. Primarily, the range of sounds available in any human language is too poor to express the actual sound produced by the animals. This leads to different perceptions and different representations in the various languages.

Take for example the owl. Its (present) greek name is κουκουβάγια (=[kukuˈvaʝa]), an obviously onomatopoeic reference to its cry κου-κου (=[ku-ku]). Plautus, however, makes a pun on the habit of Menaechmus' wife to use "tu" way too often and equates it with the cry of an owl (cf. STUR20, pp. 7 and 33), which means that in Latin the owl cried [tu-tu]. A further pun, this time from the movie Tinker Bell and the Lost Treasure, has the "wise owl" cry "hooowooo", which is perceived by Tinker Bell's friend as the question "who?" (=[hu(ː)]), the apparent cry of the owl in English. While there is agreement about the vowel ([u]) across all three languages, the consonant varies wildly (from [k] to [t] to [h]) and it would be precarious to draw any conclusions about the value of Latin T based on the Latin and the modern (either Greek or English) representations (which would lead us to the conclusion that, in Latin, T=[k] or T=[h], respectively).

Hebrew SheepIn a book for learning Hebrew, there is an interesting drawing (see right) for teaching the pronunciation of the vowel [e]: a sheep bleats מֶה(=[me(ː)]). This is strange, for an Englishman would swear that sheep bleat "baa" or "bah"(=[baː]). There is, thus, no coincidence between these two languages as concerns the sound that a sheep makes; both the consonant and the vowel differ. The future (English-speaking) linguist might be tempted to conclude that the Israelis of the 20th century pronounced מ like English b and ֶ (or rather ֶֶ ה) like English ah. Interestingly, modern Greeks decided to split the difference, as the most popular rendering of a sheep's cry is μπέε(=[beː]), taking the consonant from one language and the vowel from the other (however, the alternative rendering μέε, i.e., identical with Hebrew, also exists).

German DogAnd to provide a final example, Germans think (see left) their dogs bark "wau" (=[vaw], sounds like English "vow"), while modern Greeks use exclusively γαβ(=[ɣav]). And just when one is ready to at least acknowledge a coincidence of the vowel ([a]), comes the English "woof"(=[wuf] or [wʊf]) to prove beyond doubt that humans are poor judges of any kind of animal sound, even that of their "best friends".

If for representing the same sound some humans use a stop and others a continuant (t/k↔h, b↔m) or some use an open vowel and others a mid or close vowel (a↔e, a↔u), it is evident that no safe conclusions can be drawn from the graphical representations of animal cries or any other natural sound, neither in the present nor in the past.

Naturally Inappropriate
The transcription of natural sounds in the human languages is never accurate and seldom uniform; hence, inappropriate for any conclusions.

Intralingual

A different, but related class of evidence comprises the cases for which we know (or suspect) that two different representations stand for the same or similar sound. In those cases, any conclusions that can be drawn are relative, i.e., of the kind "X sounds like Y"; only if we have established the value of X or Y would we be able to determine the (absolute) value of the other.

The most helpful kind of relative evidence is the rhyme. Rhymes are very useful for determining that two optically different words (or at least their last parts from the stressed vowel onwards) are acoustically identical. We have seen that through rhyme we can ascertain how Erasmus pronounced his Greek (eruditionis↔ὄνοις, solos↔ὅλως, astrologi↔λόγοι, grammatici↔εἰκῆ, famelici↔λύκοι). However, according to Rhankabes (RHAN92, p. 19), there are only some "faint traces" of rhyme in ancient poetry and its introduction into (modern) Greek poetry is the result of the influence of the Romance nations. I am not sure what kind of "faint traces" he has in mind, but I suspect that they would not follow the strict rules that usually apply today.Rhankabes sets the following rules: "ἐπὶ ληγούσης τονιζομένης πρέπει να ὁμοιοκαταληκτῇ ὁλόκληρος ἡ τελευταία συλλαβὴ (οἷον φρι-κτὸς δὲν ὁμοιοκαταληκτεῖ μετὰ τοῦ αὐτὸς, ἀλλὰ μετὰ τοῦ ἐ-κτὸς· ἐχ-θρὸς μετὰ τοῦ νω-θρὸς, οὐχὶ μετὰ τοῦ και-ρὸς κτλ. θε-ὸς μετὰ τοῦ υἱ-ὸς οὐχὶ μετὰ τοῦ κα-λὸς.) Ἐν δὲ παροξυτόνοις καὶ προπαροξυτόνοις λέξεσιν ἡ ὁμοιοκαταληξία ἄρχεται ἀπὸ τοῦ τονιζομένου φωνήεντος (οἷον πλ-άσμα, χ-άσμα, γ-έροντα, φ-έροντα)|when the stress is on the ultima, the entire last syllable must rhyme (thus φρι-κτὸς does not rhyme with αὐτὸς, but with ἐ-κτὸς; ἐχ-θρὸς with νω-θρὸς, not with και-ρὸς etc. θε-ὸς with υἱ-ὸς not κα-λὸς.) As for the words that receive stress on the penult and the antepenult, rhyming starts from the stressed vowel (thus πλ-άσμα, χ-άσμα, γ-έροντα, φ-έροντα)" Furthermore, even today the rhyming part of two verses does not always coincide perfectly. Consider for example the popular German kid song

"Backe, backe Kuchen,
der Bäcker hat gerufen|Bake, bake a cake,
called the Baker
"

In order for the rhyme to be 100% correct, the sound of German ch (in "Kuchen|cake") and f (in "gerufen|called") should be identical, which they are not (they are pronounced [x] and [f], respectively). The rhyme is, thus, sometimes imperfect or inaccurate and does not always imply phonetic identity, but phonetic similarity would also do the trick.

A-rhyme-ia
Rhymes are very rare in ancient Greek and are never decisive.

Even less helpful are alleged cases of alliteration or word plays. The latter exploit similarly or identically sounding words or parts thereof to allude to two different meanings at the same time, for example "If Hitler got Hungary, he would have Turkey with a lot of Greece|If Hitler got hungry, he would have turkey with a lot of grease". Chatzidakis uses the term "παρηχήσεις" to refer not only to the classical alliteration, wherein the beginning (usually the first sound) of a series of words must be (nearly) identical, but also to the coincidence of large parts of two words in the same phrase, such as Diogenes' quote "κοινῇ συνοικεῖς πόρνῃ, ἤ κύνιζε οὖν ἤ πέπαυσο|you live with a common prostitute, so either behave like a dog [be cynical?] or stop" (CHAT02, p. 354). From that example, one might be tempted to conclude exact coincidence of οι and ῇ (in "κοινῇ") with ύ and ι (in "κύνιζε"), respectively. However, Chatzidakis points out (using examples from well known ecclesiastical hymns of VI AD) that one needs "οὐ πλήρη καὶ τελείαν ὁμοιότητα πάντων τῶν φθόγγων τῶν σχετιζόμενων λέξεων [...], ἀλλ' ὀλίγων τινῶν μόνον|not full and perfect similarity of all sounds of the related words, bun only of a few of them", i.e., it is enough for the majority of sounds to coincide in the two words, in order to provide the alliterative effect. However, it is true that the word play is more evident if identical pronunciation (as per modern Greek) is used.

Not A Toy
It is uncertain when an ancient intended to play with words and what extend of sound identity was required.

True alliteration (coincidence of the first sound in a series of words) is also rare in ancient Greek. One example is provided in the work "Οἰδίπους Τύραννος|Oedipus the King" of Sophocles: "τυφλὸς τά τ᾽ ὦτα τόν τε νοῦν τά τ᾽ ὄμματ᾽ εἶ|you are blind in the ears, the mind and the eyes". While it is unquestionable that an alliteration (repetition of τ) is produced, I do not see how we can be certain that it is intentional, as it seems to be an isolated (i.e., non-recurring) case in hundreds of verses. Furthermore, it is doubtful what kind of conclusions can be drawn from cases of suspected alliteration. For example, Pindar's verse "ἀλλ' ἐν ὄρφναισιν πέτρας φοίνισσα κυλινδομένα φλὸξ ἐς βαθεῖαν φέρει πόντου πλάκα σὺν πατάγῳ|but in the darkness of night the crimson flame hurls rocks down to the deep plain of the sea with a crashing roar [Svarlien's translation]" describing Aetna in action, is often quoted as an example of alliteration that implies (ex)plosive activity of the volcano. While there is unquestionably a repetition of φ in the first part and a repetition of π in the last part, the Catholics claim that the entire sentence involves alliteration of the labials π, φ and β (of "βαθεῖαν"). Such a claim, if taken at face value, could mean one thing and one thing only: that π, φ and β represented the same sound. Yet this is not what Catholics suggest, but merely that the sound of φ and β were much more "close" to that of π(=[p]) than they are today. However, since the very basic rule of alliteration (that of phonetic identity) is violated, one may ask how close should "close" be for the alliteration to work, yet the sounds to remain different. There does not seem to be an objective answer to that question, at least more objective than the aforementioned requirements for a successful rhyme.

Too Many Assumptions
In alliteration a two-tier assumption model is used: a) intentional alliteration, b) unspecified degree of similarity, yet not coincidence.

Interlingual

There are numerous cases of loans from or to other languages, wherein the same word is found in both Greek and another language. These concern mainly proper names (like Καῖσαρ↔Caesar), but also common words (like cynicus↔κυνικός) that originate in one language and are borrowed (almost) "as is" in the other, albeit spelled in the other language's alphabet according to a process called transliteration. There are two kinds of transliterations: occasional, when it concerns individual words transliterated either once or several times (like the examples already mentioned), and persistent, where a certain letter in one language is almost always replaced by a specific letter in another (e.g., Greek η corresponding to Latin e or rather ē). In the first type, one cannot exclude the possibility of spelling mistake or perpetuation of such a mistake, while the second may be a matter of convention.

More specifically, transliterations may only serve to suggest (with various degrees of probability) correspondence of two graphemes (i.e., one letter or combination of letters in one language with another in the other language), since graphemes are the actual data that we have (i.e., only the written forms have survived). This correspondence may be due to a number of factors:

  1. Coincidence of sound: This is the most straightforward explanation; if two graphemes represent the same sound, it is only reasonable that one be used to transliterate the other (e.g., Greek φ for English f in foul→φάουλ, Norfolk→Νόρφολκ, etc); however, we have seen that the coincidence of sound is not always the only reasonable assumption.
  2. Similarity of sound: If no exact sound is available in the target language, the transliterator must make do with whatever comes reasonably close (e.g., Greek σ for English sh in shoot→σουτ=[sut], Bush→Μπους=[bus], etc).
  3. Pure chance: Some transliterations, such as καταπέλτης→catapulta and Γανυμήδης→Catamitus (STUR20, p. 170 and p. 98), cannot be due to a desire for phonetic identity or similarity, but they appear to be products of a random process, possibly misunderstanding or other psychological factors; for example, Albanian q (reportedly [c]) in the syllable qi is sometimes transliterated with Greek τσ(=[t͡s]), e.g., KuqiΚούτσι, even though κ in κι(=[ci]) would be the reasonable choice, as it represents the exact same sound; another example is the use of Greek β(=[v]) not only for its English equivalent v (e.g., VictoriaΒικτωρία), but also for b (e.g., ByronΒύρων) and w (e.g., WashingtonΒασιγκτών), although the correct grapheme to use would be μπ and (γ)ου (thus, Μπάιρον and (Γ)Ουάσιγκτον), respectively.
  4. Established convention: While the previous factor mainly concerns the "occasional" type of transliteration, there may also be some consistent transliterations that do not attempt to provide the same or closest sound, but they are the results of some convention; in addition to the already mentioned consistent transliteration of German ch with the Greek χ, even though their allophones appear in different environments, one may also cite the colloquial Spanish transliteration of English j(=[d͡ʒ]) with y(=[j]) (cf. Michael Jackson→Maikel Yason, junkie→yonki, George Michael→Yors Maikel), even though Spanish ch(=[t͡ʃ]) would be a more reasonable choice, also consistent with the general Spanish tendency of sibilant devoicing (cf. LLOY87, p. 267ff.), as well as German's use of ä and y to represent Greek αι(=[e]) and υ(=[i], allegedly formerly [y]), e.g., αἱμ-→Häm-, δαίμων→Dämon, χαμαιλέων→Chamäleon, although e (at least in most dialects) and ü have the same pronunciation, respectively.
  5. Aesthetics: Some times, it appears that a certain transliteration was chosen because it merely "looked better"; for example, the choice of ph for representing the sound [f] in the names Ralph, Randolph, Rudolph (all of germanic origin) cannot be explained by any of the above; in other cases, it may be that the transliterator wished to preserve the orthography of the original language,cf. RENA49, p. 26, note 4: "le cas où les diphthongues sont représentées par deux caractères ne prouvent pas contre l’iotacisme; car on peut supposer que le transcripteur a voulu alors représenter chaque lettre en particulier plutôt que le son résultant. Ainsi le moyen âge, qui transcrit d’ordinaire οι par i, écrit aussi paroicos (transcription anglo-saxonne), etc.|the case wherein the diphthongs are represented by two characters does not prove against iotacism [i.e., their pronunciation as [i]]; because it may be assumed that the transcriber wanted to particularly represent each letter rather than the resulting sound. Thus, the Middle Ages that usually transcribe οι with i, also write paroicos (english transcription), etc. which may lead to a systematic use of the transliteration as an established convention.This is reportedly what happened with the Gothic transcriptions of Greek: "Ulfilas tried to follow the original Greek text as much as possible in his translation [and] used the same writing conventions as those of contemporary Greek".

All in all, any of the above explanations may have contributed to a certain transliteration, so that we need to consider all of them as possible causes for the evidence. Chatzidakis observes that: "συνεργάζονται κατὰ τὴν ἀναπαράστασιν τοῦ γνησίου ἔν τινι γλώσσῃ φθόγγου ὑπ’ ἀλλογλώσσων τοσαῦτα ψυχικὰ στοιχεῖα, ὥστε ἡ παραβολὴ τούτων ἀποβάλλει πᾶσαν ἀξίαν|so many psychological evidence cooperate during the representation by foreign speakers of a genuine sound in one language, that their juxtaposition removes any value", so that "διάλεκτος, ὀρθογραφικὴ συνήθεια, γλωσσικαὶ περίοδοι παριστῶσι τὸν αὐτὸν φθόγγον δι'ἄλλου γράμματος|dialect, spelling habit, linguistic periods represent the same sound with different letter[s]" (CHAT02, p. 348). This wholesale rejection of transcription evidence (which he only uses against Papadimitrakopoulos' evidence) is too strict. It is probably more reasonable to state that a certain transliteration points to all the above causes with different degrees of probability (correspondence of sound being the most probable); it is then up to further evidence to determine whether a certain cause is reasonable or has to be rejected.

Anything Goes
Transliterations may be due to aural, visual, psychological or probabilistic reasons.

There is a bigger problem though than identifying the reason behind a certain transliteration. Even if we somehow establish a certain correspondence of graphemes and are pretty certain that it is due to coincidence or similarity of sound (factors 1 and 2 above), we would still be at a loss about the actual sound these two graphemes represented. For instance, even if we assume that the aforementioned correspondence C↔Κ implies an equation C=Κ in the aural domain, we cannot use this to determine the value of C until we have established that of Κ (and vice versa); however, when relying on transliterations, we are kicking the can down the road, since we essentially transform the question "what is the sound corresponding to grapheme A in Language AA?" to "what is the sound corresponding to grapheme B in language BB?". Since languages AA and BB have to be contemporary (in order for the transliteration to be useful), it must be equally difficult to determine the pronunciation of B as it is to determine that of A.

When it comes to determining the pronunciation of Greek based on transliterations, "it is mainly a question of Latin" (CARA95, p. 162). This was actually the very first type of "clues" for the early Catholics, wherein "the pronunciation of Greek was determined almost solely with the pronunciation of Latin as the arbiter" (CARA95, note 10). But which exactly "pronunciation of Latin" was used as reference point? If the current pronunciation of Latin in the West was used, then it is unfair to take that pronunciation (which is also divergent in many aspects) for granted, but not the current Greek one (which is uniform). Since arguments based on Latin are often the main reasons for objecting to the Orthodox pronunciation, it is important to examine how we "know" the pronunciation of Latin and how certain our knowledge is. In many cases, it turns out that the establishment of a letter's phonetic value, as well as an argument based on the assumed pronunciation of Latin are characterised by such sloppiness that they actually serve against the Catholic pronunciation. For example, the correspondences β↔b and η↔ē (established though transliterations) are often used as proofs that β=[b] and η=[ɛː], as per the Catholic model, instead of the Orthodox values β=[v] and η=[i]; however, this argument disregards the facts that original Latin b sometimes corresponds to v in many Romance languages (e.g., februarius[LAT]→février[FRA], habere[LAT]→avere[ITA]) and that the Catholic specifications require that ē=[eː], not [ɛː]! Elementary investigation reveals that we can be as sure about (most of) the values of the Latin letters, as we are about those of the Greek ones.

Solving One Problem By Creating Another
In our attempt to determine the pronunciation of Greek through that of Latin, we need to investigate how we know about the latter.

While things are rather clear about Latin (the alphabet used, the falsifiability of the reconstructed pronunciation, etc), they are much more vague when it comes to transliterations to/from languages other than Latin. The good news is that they only play a marginal role in the determination of Greek phonology and are only used as secondary evidence, i.e., confirmatory of a suspected value, and they seem relevant only when the principal evidence are sound. Nevertheless, it makes sense to point out the weaknesses of using third-language phonologies.

The principal "other language" that has been studied at least as much as Greek and Latin is Sanskrit, which everyone treats it as if its phonology has been firmly established beyond any reasonable doubt. It is not clear that this is the case. In fact, Rhankabes (RHAN81, p. 17) submits that the pronunciation of Greek "aspirates" by the "Erasmians" is "nach der vermeinten aber keineswegs bewiesenen Aussprache des Sanskrit|according to the believed but in no way proven pronunciation of Sanskrit", which means that there may be considerable doubts that the assumed pronunciation is adopted, but not proven (cf. also JANN97, p. 57: "Sanskrit (whose pronunciation, however, is still more hypothetical)"). Reportedly, although "true native speakers of Classical Sanskrit do not exist, and have not existed for over two millennia", the phonology of "the Classical Sanskrit of the great grammarians" has been determined based on "the Vedas, the epics, fables, drama, legal and philosophical treatises, treatises on phonetics and grammar, commentaries on these, and commentaries on the commentaries" (ZWIC65, pp. 1-2). Essentially, out of all types of evidence just mentioned, only the "treatises on phonetics and grammar" by "the great grammarians" seem to be directly relevant, notably the grammar of the mysterious Paṇini (we actually know nothing about him, even the time he lived being a matter of pure speculation), which defines a few thousand (!) rules of the Sanskrit grammar including phonology.

Although I have not been able to find a Vox Latina/Graeca-like work on Sanskrit phonology, I suspect there are two possible major sources of "clues":

  1. Surviving Indian languages: This is a legitimate starting point (consistent with the concept of tradition, which is also our primary source for Greek phonology). However, the current pronunciation cannot be declared as "proof" of the ancient one. Even if there is near-unanimous agreement among the various offshoots of Sanskrit about a particular sound, this does not mean that the ancient sound was the same (cf. the pronunciation of H in Romance and its officially assumed value in Latin). Furthermore, we should not forget that Greek is also a surviving language and at least as well documented as the Indian languages; if the transcriptions indicate a correspondence between two sounds of Greek and Sanskrit, which do not coincide nowadays, something's gotta give and it is not necessarily the Greek pronunciation. Sometimes it is argued that, because Sanskrit was the liturgical language of Hinduism, its pronunciation was preserved; however, it would then be unfair to deny the very same principle to other liturgical languages, e.g., Vatican Latin, the validity of which has been questioned since the times of Erasmus.
  2. Statements of the grammarians: Apparently, Indian grammarians surpassed all their ancient (and sometimes modern) "colleagues".According to Allen (ALLE87, p. 30), they are responsible for the discovery of "the nature of 'voice', i.e. glottal vibration, as a distinctive feature of consonants", which escaped the Europeans until XIX AD! To what extend they managed to provide clear and unambiguous phonological descriptions of the various sounds of Sanskrit I cannot tell. I have tried in vain to search the Internet for phonological rules mentioned in the revered work of Paṇini, but the only relevant rules that I found concern phonotactics (such as sandhi rules); in fact, it appears that the very structure of Paṇini's rules (or "sutras") makes it difficult to formulate a proper phonological description. The little information that I have encountered about phonological descriptions by Indian grammarians comes from Allen. For example, he points out (ALLE87, p. 22) that "although fricatives and aspirates are not identical, they are phonetically (and often historically) related - in fact the ancient Indian phoneticians apply the same term both to the air-stream of the fricatives and to the aspirated release of the plosives"; such a statement (in addition to needing verification) allows for the interpretation that what are perceived by modern (European) linguists to be "aspirates" may as well be plain fricatives. Similarly, he submits (ALLE87, p. 16, note 9) that "Sanskrit grammarians describe the aspirated and unaspirated plosives as 'mahāprāṇa' and 'alpaprāṇa', i.e. 'having big/little breath' respectively"; again, this statement does not describe presence/absence of "aspiration" better than it would describe unobstructed/blocked airstream. Thus, the (admittedly) little evidence I have seen so far, are in no way conclusive that Sanskrit phonology can be safely determined from the statements of Indian Grammarians.

Apart from the uncertainty about the assumed Sanskrit phonology, there is a further problem related to the usefulness of Sanskrit transliterations for determining ancient-Greek phonology: there was no direct contact between Greek and Sanskrit, because the two never coincided geographically AND temporally. The only contact between Greeks and Indians occurred in III and II BC in the north-west of the subcontinent (modern Pakistan) and it is unlikely that the language of that area at that time was Sanskrit; it was probably rather some Prakrit (which were reportedly spoken between VI BC and XI AD), the phonology (and identity) of which needs to be established anew!

It's All Sanskrit To Me
The evidence from Indian transliterations and Sanskrit phonology may be more dubious and less relevant than is actually believed.

There is a host of other languages that are used for determining the pronunciation of ancient Greek: Persian, Armenian, Coptic, Gothic, Aramaic, etc. The written form of most of those languages is not contemporary with classical Greek and the evidence put forward is usually post-Christian. It is for this reason that the scholars do not rely on such evidence to establish the value of certain Greek letters, but they rather cite the evidence as confirmation of "preservation" of their (proven otherwise) "original" values. Usually, the argument is along the same lines as the dubious C↔Κ transliteration: a correspondence of a Greek letter and a foreign letter is interpreted as indication of similarity or identity of sound. The same question applies here: how do we know the phonology of those languages (better than the Greek one)? Again, there are two likely sources of information. The first is to judge from their modern scions. But this is a right denied to Greek; why should it be granted to every other language? The other possibility is that their phonology was established based on other evidence. But Greek is the best documented and most studied language of all; if any conclusions are to be transferred from one language to another, it would be more reasonable to derive the phonology of Armenian et al from Greek rather than vice versa.This appears to be the case with Gothic: "Since the Greek of that period is well documented, it is possible to reconstruct much of Gothic pronunciation from translated texts". That is to say, Gothic sounds were determined based on the Greek ones, not the other way around.

The aforementioned problems are exemplified by Allen's statement: "A Demotic Egyptian text of the 2 c. A.D. containing some Greek transliterations shows that Greek φ and χ there represented Egyptian ph and kh, and not the fricatives f and ḫ" (ALLE87, p. 25). In this argument, the sounds represented by the Greek letters φ and χ are the variables to be determined and the Egyptian sounds the constants. To begin with, there is a hidden premise that is often used in similar linguistic works and usually goes unnoticed: neither the Egyptians nor any other of the ancient peoples (but the Romans) used the Latin alphabet. Thus, to state that φ and χ "represented Egyptian ph and kh, and not the fricatives f and ḫ" is imprecise, since the "Demotic Egyptian text" certainly did not use these graphemes (ph, kh, f, ḫ), but the proprietary Egyptian script. Obviously Allen refers not to Latin letters f etc, but rather to sounds [pʰ], [kʰ], [f], [x], but this is like comparing apples and oranges, since φ and χ are graphemes and [pʰ] etc are phonemes (or rather "phones"). Hence, the hidden premise is that the Egyptian graphemes corresponding to Greek φ and χ had exactly those values ([pʰ], [kʰ]).

Allen does nothing more than continue the linguistic custom of using Latin letters instead of the inconvenient native symbols. However, this practice is misleading and relies on an a priori identification of the native symbol with the (approximate) sound corresponding to a Latin letter. Allen, like most of us, was neither an expert Egyptologist nor familiar with the original Egyptian script; instead, he and most (if not all) scholars using transliterations between Greek and a third language rely on that language's Latin transcription established by the pioneers of the study of that language and assume that the transcription indicates pronunciation. If one resorts to the actual experts, specifically on Egyptian phonology (PEUS99), one sees that this (i.e., correspondence of Latin transcription and sound) is not always the case. To begin with, a transcribed text may conceal important details of the original text.PEUS99", p. 46: "The transcription of a specific word does not indicate whether it is written phonographically, semographically, or by a combination of both methods. Neither does it usually show whether all consonants are explicitly written in the actual case or whether some of them are merely supplied by presumably more 'complete' phonographical renderings of the same word form in other texts. The huge graphical variability of Egyptian is therefore concealed as soon as words are put into transcription." Then, a certain transcription may have been taken up for reasons other than phonetic, yet non-experts take it literally and use it at face value for developing further dubious theories.PEUS99, p. 80: "When BRUGSCH & ERMAN (1889) established a convention of transcribing Egyptian, they decided to render class 1 stops with the letters for tenues, class 2 stops with the letters for mediæ of the Latin alphabet. It was clearly expressed then that this practice served nothing but typographical convenience and did not anticipate a decision on the phonetic interpretation of these signs. [...] This transcription system, being propagated by the influential Berlin school, soon became firmly established. Subsequently, a number of scholars began to believe in the voiced character of the stops of class 2, and the suspicion can hardly be avoided that the mere transcription convention encouraged them to do so. Based on this assumption, dozens of Egyptian-Afroasiatic etymologies were proposed which seemed to corroborate the voice hypothesis. Today this hypothesis is prevalent primarily in non-specialized literature on Egyptian from other fields than Egyptology." (emphasis mine)

The question of what conclusions can be drawn from a transliteration between Greek and the Latin transcription of a third language is much more complicated than simply declaring the value of the Greek letter as equal to that of the corresponding Latin letter of the transcription. In the specific example of Allen's, there are the following questions that must be answered before we reach a conclusion about the sound represented by the Greek letter: What were the original letters represented by "Egyptian ph and kh, [...] f and ḫ"? How do we know that their pronunciation was exactly that? I am afraid the answer to the last question is not very encouraging for the use of Egyptian as a source of evidence for Greek phonology. It appearsPEUS99", p. 26: "Beginning with Jean François Champollion who deciphered the hieroglyphs in 1822, the first Egyptologists identified numerous hieroglyphic groups with words of Coptic which was already well known at that time, but they were not yet fully aware of the differences between Egyptian and Coptic as languages. When the linguistic distance between both languages became more and more evident, extensive research on Egyptian phonetics began. Much of this early work is no longer acceptable today and will not be discussed in this book." that the "decipherment" of the Egyptian script and consequently its "reconstructed" phonology was based on the wrong identification of Egyptian with Coptic. Although the mistake was later identified and our conclusions on Egyptian were revised, there still remains the problem that the original decipherment was based on a misunderstanding and may not be valid anymore. And there is an even bigger problem: by the time it was used for the decipherment, Coptic was already a dead language.PEUS99", p. 31: "O'LEARY (1934: 243-249) assumes that Coptic had already ceased to be in spoken use in the 13/14th centuries AD. Certain sources which, however, are of questionable reliability suggest that Coptic was still spoken in the 17th century." How do we know its phonology? I am afraid the answer might be "through our knowledge of Greek" (e.g., by assuming that the Coptic letters that have the same shape as the Greek ones also represented the same sounds), resulting in a useless recursive argument.

Too Many Questions
Before transliterations in another language can be used, there is a series of questions that need to be answered about its "reconstructed" phonology.

Testimonies

In some ancient works, there are direct or indirect references to the sound of particular letters. Sometimes they are occasional and non-comprehensive, as divergences from the main topic of the work. More often, they are more systematic and constitute the very subject of the work, which thus can be identified as a "Grammar".

Occasional

The references to phonology in ancient literary works are scant and often superficial. The work with the most such references has to be Plato's Cratylus (PLAT94), where Socrates describes some qualities of the various sounds represented by the Greek letters. Other works have isolated allusions to phonology or are the source of the aforementioned transcriptions of natural sounds, wordplays, alliterations, transliterations of foreign names, etc. These are apparently synchronic evidence, i.e., contemporary with the period of interest (Plato and Aristotle lived in Athens of V-IV BC). But are they really?

Consider, for example, Aristotle's following passage cited by Sturtevant (STUR20, p. 157): "εἴπερ μὴ καὶ τὸ ὄρος καὶ ὅρος τῇ προσῳδίᾳ λεχθὲν σημαίνει ἕτερον. άλλ' ἐν μὲν τοῖς γεγραμμένοις ταὐτὸν ὄνομα, [...] τὰ δὲ φθεγγόμενα οὐ ταὐτά|unless both ὄρος and ὅρος pronounced with the breathing ["prosody" in the original] have one of the two meanings. But in writing they are the same word, [...] while in pronunciation they are not the same [Stu's translation]". This passage is celebrated by both Sturtevant and Allen as unequivocal evidence of "the phonetic distinction between this [i.e., ὅρος] and ὄρος". Nevertheless, Allen notes that "There is, however, some doubt about this example. It has been suggested that Aristotle wrote not ὅρος but ὀρός ('whey'), which would be distinguished from ὄρος by accent and not by breathing" (ALLE87, p. 52, note 101). How can we have doubts about what Aristotle wrote? Did he smudge the respective word in his book, so that we cannot discern whether he wrote ὅρος or ὀρός? Did he have a strange handwriting?

The truth is that we have never seen Aristotle's handwriting. We have neither seen any of his original works. Neither his nor any of his contemporaries'. The earliest manuscripts of ancient literary works are not dated before medieval times. They are copies of earlier manuscripts, which were seldom original works and almost always copies of other manuscripts. It is, therefore, normal that these copies of copies (of copies) are not necessarily faithful replicas of the original works, but contain a whole lot of misspellings and inaccuracies, which intruded thereinto during the copying process, due to the copyist's carelessness, misunderstanding, pronunciation or ignorance. In fact, different copies of the same work have different versions of the same text. The works we are consulting today are the results of editing, which involves collecting as many manuscripts of the same text as possible and deciding on a common text (either by selecting the most plausible variant or by emending patently erroneous wording). This common text is the one published, either as such (e.g., BEKK16) or with footnotes indicating the non-preferred manuscript variants (e.g., ROBE10).

In the particular case of "ὅρος" vs "ὀρός", there is a further factor that adds to our uncertainty. Aristotle could have written neither "ὅρος" nor "ὀρός", as Greek minuscules were not "invented" until VII-IX AD; furthermore, the "inventor" of the diacritics for the "breathings" (῾ and ᾽) and the "accents" is considered to be Aristophanes of Byzantium, who lived one century after Aristotle.The cited passage of Aristotle's mentions the use of some diacritics by some of his contemporaries ("κἀκεῖ δ'ἤδη παράσημα ποιοῦνται|but nowadays they put distinguishing marks beside them [Stu's translation]"). However, if it is indeed part of the original text (it may as well be an observation of the editor/copyist about the use of diacritics by the contemporaries of the editor/copyist), then it most likely confirms that no such diacritics were used with the term "ΟΡΟΣ" in the original text; otherwise, the reader would have already seen them and would not need to be informed of their usage (instead, a statement like "markings as used herein" would make more sense). This is also corroborated by the statement "ἐν μὲν τοῖς γεγραμμένοις ταὐτὸν ὄνομα|in writing they are the same word", which would not be true if "ΟΡΟΣ ΚΑΙ ΟΡΟΣ" were accompanied by diacritics. In a later article (STUR37, p. 116), Sturtevant abandons the interpretation of "παράσημα" as "markings", aligns himself with Laum in interpreting the term as "a mark in the margin to indicate that the passage involves a grammatical problem" and accepts Aristophanes of Byzantium as the person that introduced diacritics for at least the "rough breathing". On the contrary, all surviving ancient manuscripts (such as the oldest surviving document, the Derveni papyrus) include but capital letters. All diacritics and punctuation (as well as minuscules) that we encounter in the reference publications (such as BEKK16, DION83, PLAT94, POST66, QUIN56, ROBE10) are aesthetic additions by later copyists, scholiasts and editors. It is also worth noting that modern editors, faced with diverging variants, erroneous syntax, and unclear meaning in the manuscripts, proceed to so-called "emendations", which are sometimes conjectural; thus, even the published text is not flawless and certainly not to be trusted blindly (Jannaris JANN97, for one, has proposed many further, seemingly reasonable, emendations of the "official" texts).

It is, thus, highly unlikely (if not impossible) that Aristotle wrote "ὄρος καὶ ὅρος" (but rather "ΟΡΟΣ ΚΑΙ ΟΡΟΣ") and very misleading to treat the published text of a work of Aristotle's as if it were a faithful copy of the original.cf. Pighi's observations (as quoted by Lloyd) that "rusus n'est pas attesté «chez Plaute», mais qu'il est «attesté dans les manuscripts de Plaute»: différence que les philologues et les linguists oublient très souvent|rusus is not attested «in Plautus», but that it is «attested in the manuscripts of Plautus»: difference that the philologists and the linguists very often forget"; "Nous connaissons la plupart des anciens écrits par le moyen de copies tardives, du IVe-Xe siècle; une petite partie seulement est gardée par des copies du IVe-Ve siècle|We know the majority of the ancient works by means of late copies, of the 4th-10th century; only a small part is held by copies of the 4th-5th century"; "l'ecriture du palimpseste ne prouve point par elle-même que celle-là fut la graphie du Cicéron, mais elle prouve seulement que dans le Ve-VIe siècle un éditeur, pour tradition ou conjecture ou préciosité archaisante, employait cette graphie en faisant la copie d’une oeuvre de Cicéron|the writing of the palimpsest by itself does not prove at all that this was Cicero's spelling, but it only proves that in the 5th-6th century an editor, out of tradition or conjecture or archaic preciousness, employed this spelling while copying a work of Cicero's" (LLOY87, p. 75, note 14).

General Relativity
Any conclusions drawn from the works of ancient authors are subject to the (unverifiable) assumption that their published text is the original.

Systematic

A small part of ancient literature relates to the study of (the Greek) language and comprises extensive (as compared to the other literary works) references to phonology. The respective authors are known as the "Grammarians", although "'Grammar', as now generally understood - the art concerned with the principles and rules of language in speaking and writing - was viewed differently among the ancients" (JANN97, p. 34). The earliest surviving Greek grammar, "Τέχνη Γραμματική|Art of Letters" by Dionysius Thrax (DION83), a rather short treatise (15 pages in Bekker's edition, BEKK16, pp.629-643) dating from late II BC or early I BC, provides the following definition: "Γραμματική ἐστιν ἐμπειρία τῶν παρὰ ποιηταῖς καὶ συγγραφεῦσιν ὡς ἐπὶ τὸ πολὺ λεγομένων|grammar is the knowledge of the usual subject-matter and diction found in the (classical) poets and prosaists [Jannaris' (free) translation]". Thus, for the ancients, the term "Grammar" appears to refer exclusively to the study of literature, rather than language structure, as it does today. It is for this reason that their remarks on phonology are, in the words of Jannaris, "incidental"; for instance, in the treatise of Dionysius Thrax, phonology occupies barely more than one page (out of 15). Nevertheless, those limited remarks are often quoted as "proof" for the correctness of one or the other doctrine (i.e., for the Catholic as well as for the Orthodox pronunciation). The use of the teaching of the ancient Grammarians is not free from pitfalls, though.

The first thing to note about the Grammarians (other than the aforementioned fact that, being ancient authors, their published texts are not necessarily the originalcf. CHAT02, p. 347: "ὡς ἐπὶ τὸ πλεῖστον δὲν περιεσώθησαν ἡμῖν αὐτὰ τὰ ἔργα τῶν παλαιῶν Γραμματικῶν, ἀλλὰ μόνον διδάγματά τινα αὐτῶν μνημονευόμενα ὑπὸ μεταγενεστέρων ἢ καὶ Βυζαντινῶν Γραμματικῶν, Σχολιαστῶν, Λεξικογράφων κλπ|for the most part, we are not in possession of the works of the old Grammarians, but only of some of their teachings mentioned by later, even Byzantine Grammarians, Scholiasts, Lexicographers etc") is that they flourished after the main period of interest, namely the Classical times.Jannaris observes (JANN97, p. 33) that "they are post-Christian scribes (excerptors, commentators, copiers) and cannot speak with authority of the pronunciation of 'classical' Greek"; although they are not all "post Christian" (a few flourished in the 100 years before Christ), they are certainly far from the classical times. It is, therefore, not certain how their teaching relates to classical Attic. Some modern scholars assume that the ancient grammars are extensions of earlier classical works, if not exact copies thereof; therefore, they consider that their teaching also applies to the classical language.Papadimitrakopoulos, for one, proclaims (PAPA98, p. 47) that "παρὰ τοῖς πεπαιδευμένοις καὶ τοῖς Γραμματικοῖς [...] μέχρι τέλους διεσῴζετο καὶ ἐν τῇ θεωρίᾳ καὶ ἐν τῇ πράξει ἡ γνῶσις τῆς παλαιᾶς προφορᾶς|the knowledge of the old pronunciation survived among the educated and the Grammarians, both in theory and in practice, until the end" and that (p. 73) "ἡ μόνη ὁδὸς ἡ ἀσφαλῶς ἄγουσα εἰς εὕρεσιν τῆς γνησίας και ἀληθοῦς προφορᾶς τῶν Ἑλλήνων εἶνε ἡ διδασκαλία τῶν Ἑλλήνων Γραμματικῶν, τών ἀμέσων διαδόχων τῆς ἐγκρίτου ἀρχαιότητος καὶ τῶν ἀρίστων ἑρμηνευτῶν τῆς πατρῴας φωνῆς|the only path that safely leads to the discovery of the genuine and true pronunciation of the Greeks is the teaching of the Greek Grammarians, the direct successors of the prestigious antiquity and the excellent interpreters of our ancestral voice". Others believe that their teaching applies only to the classical languagecf. CHAT02, p. 345: "οἱ μεταγενέστεροι συχνότατα ἐπαναλαμβάνουσι θεωρίας παλαιοτέρων καὶ τὰ ἐκείνων ὡς ἴδια προβάλλουσιν οὐδαμῶς ἀντιδιαστέλλοντες τὰ ἴδια τῶν ἀλλοτρίων, πολλάκις προφέρουσι διδάγματα ἀληθῆ μὲν προκειμένου περὶ παλαιοτέρων χρόνων, ψευδῆ δὲ περὶ τῶν χρόνων αὐτῶν|the later Grammarians very often repeat theories of the older ones and they project the older facts as [if they were] their own without discriminating between their own [pronunciation] from that of others; often they adopt teachings that are true as far as older times are concerned, but not valid at their times" and "δῆλον ἄρα ὅτι ὑπῆρχε κατὰ παράδοσιν τοιαύτη ἀρχαία διδασκαλία, ἣν οὗτοι ἐπαναλαμβάνουσιν, εἰ καὶ ἐπ’ ἀληθείας ἄλλως τοὺς φθόγγους ἐσχημάτιζον|it is, therefore evident that this ancient teaching existed due to tradition, which they [i.e., the later Grammarians] repeat, even though they were actually formulating the sounds differently". or only to the later stages of ancient Greek.cf. CHAT02, p. 346: "οἱ περὶ τὴν γραμματικὴν τῆς Ἑλληνικῆς ἀσχοληθέντες ἤκμασαν οὐχὶ ἐν τοῖς δοκίμοις χρόνοις, ἀλλ’ ἐν τοῖς μετέπειτα [...] Ὅσα ἄρα περὶ τῶν φθόγγων τῆς γλώσσης διδάσκουσιν, ἀνάγκην ἀπαραίτητος εἶναι νὰ μὴ ἀποδιδῶνται ἀβασανίστως τοῖς δοκίμοις χρόνοις, ἀλλὰ μόνοις τούτοις, καθ’ οὓς οἱ ἄνδρες ἐκεῖνοι ἐδίδαξαν|those who busied themselves with the Greek grammar flourished not in the classical times, but in the later ones [...] Hence, it is imperative not to hastily ascribe everything that they teach on the sounds of the language to the classical times, but only to those during which those men taught" At times it appears that not even the Grammarians themselves knew which phonology they were describing.

Out of Time
The phonetic descriptions provided by the Greek Grammarians may relate to the phonology of their times, earlier times or no time at all!

But the most important thing is that the ancient Greek Grammarians are overrated. They were not good phoneticians; far from it.cf. CHAT02, p. 342: "οἱ ἀρχαῖοι πρῶτον μὲν ἠγνόουν ὅλως τὸν φυσιολογικὸν σχηματισμὸν τῶν φθόγγων· διὰ τοῦτο δὲ οὐδέποτε προσηκόντως περιγράφουσι τὰς κινήσεις, ἃς τὰ φωνητικὰ ἡμῶν ὄργανα πρὸς παραγωγὴν ἑκάστου φθόγγου ἐκτελοῦσιν. Αἱ περιγραφαὶ αὐτῶν εἶναι ὅλως ἐξωτερικαὶ καὶ πλημμελεῖς, ἐντεῦθεν ἐπιδεκτικαὶ πολλαπῆς ἐρμηνείας|first of all, the ancients were not acquainted with the physiological formation of the sounds; for that reason, they never describe appropriately the movements, which our vocal organs perform for the production of each sound. Their descriptions are superficial and deficient, hence they afford many interpretations"; p.346: "οἱ ἀρχαῖοι Γραμματικοὶ ὑπῆρξαν ἀδυνατώτατοι παρατηρηταὶ καὶ ἐρευνηταὶ τῶν φθόγγων τῆς τε ἀρχαίας καὶ τῆς συγχρόνου αὑτῶν γλώσσης|the ancient Grammarians were very weak observers and researchers of the sounds of both the ancient and their contemporary language". Describing the correct way to pronounce a certain sound is a daunting task for (even modern) phoneticians and not easy for everyone to apprehend.Cf. for example my own attempts at describing the Greek vowels and consonants and decide for yourself how lucid a description which is not based on a reader's experience with other languages or the IPA can be. Not surprisingly, the ancient pioneers (if they can be considered as such) did not fare better at describing the sounds of their time (or of any time they might refer to). Unable to grasp concepts such as phonemic splits (at least σ and ν had already two allophones each, [s] and [z] the former and [n] and [ŋ][ŋ] was rendered as γ and is, therefore, sometimes mistakenly considered as an allophone of γ (allophony refers to phoneme variations and the corresponding phoneme for the velar nasal [ŋ] is the dental nasal, as seen, e.g., in the transformations ἐν|in+γάμος|matrimonyἕγγαμος|matrimonial, σὺν|together+χέω|I pourσυγχέω|I mix). the latter) and mergers (for one, ει came to coincide with and was ultimately used for "long" ι by the time of the Grammarians), they relied more on the script and less (if at all) on the sound.cf. CHAT02, p. 344: "Διὰ τὴν ἄγνοιαν δὲ ταύτην τοῦ φυσιολογικοῦ σχηματισμοῦ τῶν φθόγγων οἱ ἀρχαῖοι συγχέουσι τοὺς φθόγγους πρὸς τὰ παριστῶντα τούτους γραπτὰ σημεῖα ἢ γράμματα|Because of that ignorance about the physiological formation of sounds, the ancients confound the sounds with the written elements or letters representing them". For example, Thrax, in describing the "στοιχεῖα|elements", speaks of seven "vowels" α ε η ι ο υ ω, excluding from the list the digraph ου, which was used to render [u(ː)] (a vowel sound in phonological terms) and classes together the digraphs ει and ευ as "δίφθογγοι|diphthongs", even though we have strong evidence that the former was monophthongised at least four centuries before his time, whereas the latter still stands for two distinct sounds ([ev] or [ef] today and apparently [ew] originally). One should, therefore, agree with Blass (BLAS90, pp. 19-20) that most likely "in this entire theory writing rather than sound has evidently been the guide". Even when they did attempt descriptions of the sounds corresponding to the various letters, they had trouble identifying their actual phonological features.Chatzidakis, commenting on the descriptions provided by Dionysius of Hallicarnassus on the consonants, remarks that he "λαλεῖ οὐχὶ περὶ τοῦ φυσιολογικοῦ σχηματισμοῦ αὐτῶν ἀλλὰ περὶ τοῦ αἰσθήματος τοῦ κατὰ τὴν ἐκφώνησιν τούτων|speaks not of their physiological formulation, but of the feeling [they cause] during their pronunciation" (CHAT02, p. 343).

Optical Illusion
A great part of the teaching of the Greek Grammarians refers to the script (graphemes) rather than to phonology (phonemes).

The above are nicely summarised by Chatzidakis, who observes that the Grammarians were "ἐν μεταγενεστέροις χρόνοις ἀκμάσαντες, ὅτε ἡ προφορὰ ἐν πολλοῖς εἶχεν ἀλλοιωθῆ, καὶ ὄντες παντάπασιν ἄπειροι τῆς φυσιολογίας τῶν φθόγγων καὶ δὴ κάκιστοι ἑρμηνευταὶ τῆς πατρῴας φωνῆς|thriving in later times, when the pronunciation had long ago changed in many respects, being totally ignorant of the physiology of sounds and actually lousy interpreters of the ancestral voice" (CHAT98, pp. 393-394). Thus, to take their words literally would probably be "to attribute too great a sophistication to Greek phonological theory" (ALLE87, p. 30).

Direct

A considerable number of written documents have survived until the present day. These include many inscriptions on durable material, such as stone, ceramic, metal, etc. Less durable materials were certainly employed in antiquity for everyday use, but the vast majority of these (being vulnerable to the harsh environment of Greece) did not survive.For example, the oldest (and one of the few) surviving papyrus found in Greece is the Derveni papyrus, dating from some time before mid-IV BC.

One point that needs to be emphasised is that inscriptions are the ONLY source of synchronic evidence. They are the only direct relics of the era under investigation, all other types of evidence being the result of transmission, during which any kind of modification might have occurred, as seen above. Despite their importance, they were ignored in the early (i.e., pre-XIX AD) argumentation, basically due to their unavailability. It is clear, e.g., from Pickering's article (PICK18, p. 266 ff.), that the existence of inscriptions and their significance as evidence of the ancient pronunciation only began to be realised in XIX AD.

Directly related with their importance is the question of dating the various found inscriptions. Inscriptions are direct relics of the era during which they were carved, but how do we know WHEN they were carved? Most of the published inscriptions (for example, Inscriptiones Graecae) are accompanied by an exact or a less precise (i.e., identifying the century, decade or time range) date which is very rarely earlier than VI BC.cf. HORR10, p. 18: "continuing dearth of alphabetic material from before the 6th century BC". Various dating techniques have been used, mostly based on the context (e.g., archaeological stratum) in which an inscription was found or its content. I am not convinced that the archaeologists and epigraphists can be safely accurate in their estimates.For instance, the official chronology may be in need of revision. Furthermore, the argumentation used is not always convincing.For example, the oldest surviving Greek-alphabet inscription, the Dipylon inscription, is (ARVA37, p. 37) "κεχαραγμένη ἐπὶ πηλίνου ἀγγείου [...] δι' ἀκιδωτοῦ ἐργαλείου, ἐπὶ ὠπτημένου ἤδη τοῦ πηλοῦ. Χρονολογεῖται, ἕνεκα τοῦ ῥυθμοῦ τοῦ ἀγγείου, ὡς ἀνήκουσα εἰς τοὺς περὶ τὸ 850 π. Χρ. χρόνους|carved on a clay jug [...] by means of a pointed tool on the already baked clay. It is dated, due to the jug's style, as belonging to the years around 850 BC" (the date provided by Arvanitopoulos is actually about a century off). It is clear that the dating concerns the bearing object and not the inscription per se. It is also clear that the text was carved after the jug was prepared. How much later we do not know; it might have been carved a few seconds, days, years, decades or centuries later, even after its unearthing in 1871 AD! However, it is not my intension here to doubt their findings.

Not Carved In Stone
Dates accompanying published inscriptions may not be definite, but will be taken at face value for the purposes of the present investigation.

The main question regarding inscriptions is how can they prove useful in the quest for ancient pronunciation. After all, no descriptions of phonology were ever found carved in stone. The answer is to be found in the inscriptions' orthography (lit. "correct writing"). This refers to the use of the alphabet for representing the spoken language. When there is a one-to-one correspondence between the available graphemes and phonemes, there is no question of correct and incorrect writing, since there is only one way of writing. But when the correspondence is not (or no longer) exact, there may be more than one way to render a spoken pattern. Such a situation usually occurs when the process of sound change causes two formerly distinct sounds to coincide; one of the most notable examples of converged sounds, or phonemic mergers, is that of V (Latin consonantal "u") and B in Late Latin (LLOY87, p. 132). In that case, the graphemes (B and V in the Latin example) that used to represent them end up having the same phonetic value and one might be used in lieu of the other without any consequence on the pronunciation. However, for various reasons the old spelling, called historical spelling, is preferred by the learned, even though it does not describe the actual pronunciation better than the alternative one. Speakers of many modern languages (such as English, French, Greek, but also Spanish and German) take great pains to learn the correct spelling of every word, but they cannot always avoid occasional misspellings. A future reader who encounters these misspellings (deviations from the "norm") can infer that the substitute grapheme had the same phonetic value. This is how inscription texts are used for drawing conclusions about phonology: when two graphemes are interchanged, this is taken as a possible indication that "writing does not keep pace with the changes of sound" (BLAS90, p. 7) and that the sounds formerly represented by these graphemes have come to coincide.Phonemic mergers are not the only reason for coincidence of sound. It may simply be that the design of the alphabet was "faulty" from the beginning, i.e., it may comprise redundant letters. For example, there seems to be no reason for the introduction of Y in the German alphabet, as the various sounds it represents could readily be rendered with Ü or J/I (Mythos, Bayer, Meyer could have been written as Müt(h)os, Bajer, Maier).

Although is it tempting to take isolated cases of grapheme interchange as proofs of sound coincidence (as was my own reaction upon encountering the spelling ΠΟΛΗΑΣ instead of ΠΟΛΙΑΣ in the book of Arvanitopoulos) or at least similarity, there are various possible explanations for such "orthographic variations" (as Teodorsson calls them):

  1. Pure Chance: Sometimes, scribes or stonecutters committed an honest mistake in that they carelessly wrote or carved the wrong letter. Even though mistakes of this kind were not as common as they are now (that, e.g., a neighbouring key may be accidentally pressed instead of the intended one), they were still possible, e.g., due to absentmindedness.CHAT02, p. 337: "καὶ οἱ χαράκται ἦσαν ἄνθρωποι καὶ δὴ ὅτι ὅπως ἡμεῖς ἀντιγράφοντες καὶ μάλιστα πάντων ὅπως οἱ στοιχειοθέται στοιχειοθετοῦντες, οὕτω καὶ ἐκεῖνοι μεταφέροντες καὶ χαράττοντες τὰ πρὸ αὑτῶν γράμματα ἐπὶ τοῦ μαρμάρου ἤ μετάλλου πλεῖστα ἠδύναντο νὰ σφάλλωνται ἐκ μετεωρίας, οὐχὶ δ' ἐκ τοιαύτης ἤ τοιαύτης προφορᾶς|the stonecutters were also human and as we ourselves while copying or above all the typesetters while typesetting, they too, while transferring and carving on the marble or the metal the letters that lied in front of them could very well err out of forgetfulness, not due to this or that pronunciation" Teodorsson (TEOD74) excludes this kind of errors from the "orthographic variants" he cites, but Chatzidakis (CHAT02, p. 337) does mention a few instances that are evidently results of pure chance (e.g., ΓΡΑΦΣΑΝΤΙ→ΗΡΑΦΣΑΝΤΙ). Had all "alternative" spellings been results of random errors, there should be a uniform distribution of reasonable (i.e., according to later sound developments, such as Η↔Ι=[i]) and unreasonable (e.g., Η↔Γ) transpositions. This, however, does not seem to be the case, as most inscriptional mistakes appear to follow the sound identities of modern Greek (e.g., Η=ΕΙ=Υ=ΟΙ=Ι).
  2. Alternative Pronunciation: In some particular cases, it might be that the graphical representation represents the actual pronunciation (used by the author), even though it diverges from the established norm. This may occur, if a particular word (but not a particular sound in general) has undergone some linguistic change, such as assimilation, dissimilation, metathesis, etc. For example, there are many instances of ἥμυσυ instead of the more common ἥμισυ (TEOD74, p. 63) and assimilation is suspected to be the most likely cause, i.e., ι and υ are considered to have represented distinct sounds and, in this particular word ἥμισυ, the ι assumed the sound of υ under the influence of the following υ. Another cause may be that the author was not a native speaker of the language/dialect he was writing in and failed to grasp the difference between two phonemes that are not distinguishable in his mother tongue.Ruijgh (RUIJ78, p. 84) makes a convincing case for the influence of one's native phonology on one's perception of another language's sounds, e.g., "il est naturel qu'un homme dont la langue maternelle comportait le système vocalique /a i u ā ī ū / ait fait dans sa forme de l'attique des confusions telles que ε/ι/υ, η/ει/ι/υ, ω/ου|it is natural that someone whose mother tongue comprised the vowel system /a i u ā ī ū / confused in his version of Attic the sounds ε/ι/υ, η/ει/ι/υ, ω/ου" (the "Attic" referred to by Ruijgh here is one devoid of phonemic mergers and redundant graphemes). Even though Sanskrit provides a different example (PIE [e(:)] and [o(:)] merged not with [i(:)] and [u(:)], but with [a(:)]; cf. CHAT02, p. 87), it is a rather credible theory. Personally, I used to spell bijou|jewel[FRA] as bisou|kiss[FRA], due to the absence of post-alveolar sibilants ([ʒ], [ʃ]) from Greek; another example is the "acoustic" spelling Sadiago for Santiago by a fellow countrywoman, as there is no phonemic distinction among the sounds [nt], [nd] and [d] in Greek.
  3. Genuine Confusion: This presupposes the existence of two or more graphemes that represent the same sound. Only then is it conceivable that the author would be prone to use one of the graphemes in lieu of the other. The most usual cause for such a confusion should be unfamiliarity with the established norm ("orthography") of a certain word, e.g., as a result of semi-illiteracy: a (native) speaker who has been taught the letters and the sound each represents, but not orthography (i.e., the customary way of writing particular words) may not use the correct one of the two or more graphemes in some instances.A typical example is the handwriting of the Greek-revolution protagonist Makriyannis, who learned how to write at an advanced age, too late to master the complex Greek orthography and, as a consequence, uses the letters based merely on their acoustic values. Even learned individuals may occasionally err in keeping with orthography, e.g., as a result of fatigue, momentary lapse of concentration, etc.

Unfortunately, we cannot know for sure which of the three kinds of causes is responsible for each individual case of deviation from the orthographic norm. Some orthodox scholars (e.g., Caragounis) are too eager to conclude that the main cause is the last one, forgetting that graphical redundancy is merely a sufficient, but not necessary condition for an orthographic variant to occur. Thus, they consider each instance of, e.g., use of Η for Ι and vice versa as unequivocal proof of their phonetic identity without bothering to mention the other factors that may have caused it, much less to explain why they are not likely causes. On the other hand, catholic zealots overemphasise the other causes in the cases where these are even remotely likely and consider the remaining cases as insignificant or, even worse, barbaric! This last conclusion is the favourite aphorism of the devout Catholics.Chatzidakis is too quick to explain away all instances that cannot be otherwise justified, as either "ψελλίσματα ἀλλογλώσσων βαρβάρων|splutters of foreign-language barbarians" (CHAT02, p. 333), i.e. products of foreign ignorance, or solecisms, i.e., corrupted language that "ὁ ὄχλος τῆς κατωτάτης τάξεως|the mob of the lowest classes" spoke (p. 334), who are not worth considering as bearers "τῆς γνησίας προφορᾶς τῆς Ἑλληνικῆς γλώσσης|of the genuine pronunciation of the Greek language" (p. 336). He does not (feel obliged to) provide any evidence for his claims. How foreigner workers or slaves were entrusted with such an important task as, e.g., carving decisions of the city on stone is something they do not feel obliged to explain.

As we have seen, Teodorsson was probably the first one to have attempted a systematic treatment of the "rare orthographic deviations" found in Attic inscriptions. He characterised each of them with factors (other than universal coincidence of sound) that might have caused the misspelling, classified them in groups according to the exchanged graphemes and the number of factors and drew conclusions based on the number of instances in each group for each pair of graphemes. His work has been disregarded by the Catholics, mainly because it largely confirmed the orthodox model, as far as the vowel system is concerned. However, it appears that no-one else has attempted a different systematic approach, at least one with a superior theoretical basis. Instead, the catholic "response" to Teodorsson, primarily represented by Threatte, relies on the traditional (blind) counting of the instances of "confusions", explaining away most as possible cases of "assimilation" etc. and rejecting the rest as "'clerical mistake[s]' or 'careless omissions'" (MILL11, p. 143); only when the number and nature of misspellings is overwhelming do they concede coincidence of sound. Such is the case of Egyptian papyri, which were private documents written in Greek by (evidently) non-native speakers and comprise a large number of "iotacisms" (i.e., misspellings caused by an underlying phonological model that resembled that of present-day Greek). In that case, the Catholics are willing to accept that the barbarians of Egypt had already moved towards the orthodox model, as long as they do not have to acknowledge the same for the cradle of "correct" pronunciation, Attica, where they assert that Orthodoxy did not prevail until II AD (THRE82, p. 148).

I find both approaches unsatisfactory, albeit to a different degree and for different reasons. Teodorsson, for one, seems to rely too much on orthographic variations found in public inscriptions. "All data from documents of strictly private character will be marked with factor (1)", which denotes that "The writer may be supposed to have had a defective ability to read and write implying partial ignorance as to what grapheme(s) normally corresponded to a given phoneme and consequently chose the grapheme(s) in question more or less at random" (TEOD74, p. 56)! This is too strict and too arbitrary a consideration, if it is not supported by further evidence in the same inscription, such as confusion of other, not phonetically related letters. Misspellings in (official) public inscriptions are more likely to be caused by non-phonetic factors, as they must have been written with more care than those of "strictly private character". Furthermore, it seems that he does not take into account the nature of the inscription, the correlation of misspellings in the same inscription, the context of the inscription, etc., but he treats all misspellings as equal (bar the few factors that he marks each instance with, which are also "regarded as equivalent"; p. 62).He claims (p. 62) that the reason for equal treatment is "to avoid any ad hoc treatment of instances" and "to keep the individual subjective judgement of the investigator away from the instances themselves". While the merits of this approach are evident, it seems that even he himself did not resist the temptation of often resorting to specific instances (albeit in light of the statistical results) for reinforcing his conclusions. Moreover, it appears that, in the case of "rare orthographic variants", this methodology merely shifts the objectivity from the "ad hoc treatment of instances" to an ad hoc treatment of the statistics. In the end, one may at least have the feeling that the new method does not differ substantially from the traditional one of counting the instances and deciding based on their number, the only differences being the classification of the instances according to the number of markings and the (very) objective threshold for considering a number high enough for drawing conclusions about the "phonemic system of the dialect spoken by the majority of the population living in the province of Attica" (p. 13). Finally, he does not provide any indication about the number of verifiable "false alarms", i.e., the instances that could not have been caused by phonetic factors (such as Ι for Χ or Τ), since "Writings where non-phonetic causes can be proved with certainty are of course irrelevant and have not been included in the data" (p. 56). If, along with the some 3000 instances he lists, we had at least some indication of the number of non-phonetic instances, we would have a better idea of the evidential value of the listed data; e.g., if there are only a few tens (or even hundreds) of the latter, then one would expect that the major cause for the listed instances must have been phonetic similarity, but if their number is comparable or higher, then there is a considerable probability that pure chance is responsible for most of them and only those were selected, which complied with the orthodox model. The same can be said for instances within the same inscription: if it contains other, non-phonetically caused misspellings, it is more likely that it was written carelessly and irrespectively of the actual phonemic system.

Bearing that in mind, it appears that the catholic treatment of the misspellings found in the direct evidence (inscriptions, etc) is merely the other side of the same coin: Teodorsson's "factors" are excuses for the Catholics to exclude the associated misspellings from the phonetically relevant evidence, while the required threshold for considering the number of instances enough to substantiate coincidence of sound is much higher in the catholic view. The most conservative among them have placed the threshold so high that they are not convinced until the instances number in the range of hundreds.In that matter, Miller criticises Threatte's obstinate rejection of the evidence: "One wonders how many 'clerical mistake[s]' or 'careless omissions' (Threatte 1980: 358f.) of the final Ι of Ἀθηναίᾱ for Ἀθηναίᾱι 'to Athena' [c5b+] must be found before Threatte is forced to acknowledge a trend toward monophthongization of /āi/ " (MILL11, p. 143). This approach has as a consequence the other major catholic principle: lack or scarcity of misspellings (as regards letters that now correspond to the same sound in Greek) is taken as evidence confirming the catholic phonological model. This view is flawed in at least the following respects:

  1. The main assumption is that the text faithfully reflects the actual sound, i.e., if one writes, e.g., Η, it is because he pronounces something that is not, e.g., Ι, otherwise one would have written Ι. This is invariably substantiated by denying the existence of (historical) orthography in classical Greek.BLAS90, pp. 9-10: "a crystallization of orthography can only occur where the word forms have stamped themselves firmly by much reading and writing ; where there is but little reading and writing, as in Greece in the classical period and in western Europe in the Middle Ages, unless the sound is very stable and well defined, the orthography is extremely shifting".
    Also CHAT02, p. 324: "μέχρι τοῦ τέλους τῆς Ε´ ἑκατονταετηρίδος δὲν ὑπῆρχεν ἱστορικὴ ὀρθογραφία τῶν λέξεων οἵα παρ᾽ ἡμῖν|until the end of the 5th century [BC] there was no historical spelling of the words as there is for us". The only reason provided is the transition from the old-Attic to the Ionic alphabet in 403 BC: "ὅπου ἔμπεδος ὀρθογραφία, ἐκεῖ οὔτε ἐπιχειρεῖται οὔτε κυροῦταί τις αὐτῆς μεταρρύθμισις. Ἔμπεδος ἱστορικὴ ἀμετακίνητος ὀρθογραφία καὶ ἅμα μεταρρύθμισης αὐτῆς καὶ μάλιστα μεταρρύθμισις τοσοῦτον σπουδαία καὶ κατισχύσασα εἶναι ἔννοιαι καθ᾽ὁλοκληρίαν ἐναντίαι καὶ ἀσυμβίβαστοι|where there is an established orthography, no kind of reform is attempted or validated. Established historical fixed orthography and reform thereof, especially a reform so important and prevailing are completely opposite and incompatible concepts" (p. 327). The assertion that the decree of Eucleides constituted an alphabetical "reform" instead of a mere adoption of the already established Ionic alphabet is a blind repetition of Blass' unfounded claims. Furthermore, his dogmatically expressed views about lack of reform in an established orthographical system were disproved a few decades after his death, when the Greek "established historical fixed orthography" was disembarrassed of the useless "breathings" and the variety of tones (leaving only a single symbol of stress), as well as the (de facto) abandonment of the iota subscript.
    But to treat the heyday of ancient times, the classical period, as the starting point of literacy is tantamount to believing that there was no professional swimming before Michael Phelps. Instead, the long period since the adoption and adaptation of the phoenician alphabet (period A above) provides ample time for "crystallization of orthography" to occur.
  2. It's Never Too Soon
    A pre-classical crystallization of orthography is not to be a priori ruled out.

  3. Inscriptions cannot tell us which sound a letter corresponded to; at most they can tell us that two graphemes corresponded to the same or very similar sounds, when one of the graphemes is often found in lieu of the other (but still not what that common sound was). As the catholic model does not foresee redundancy of graphemes (i.e., corresponding to the same sound), no inscription can provide a positive verification of the catholic model; on the other hand, only the orthodox model, which does comprise redundant graphemes, can be (at least partially) verified by inscriptional evidence.
  4. Inscriptional Doctrine
    Inscriptions are biased towards Orthodoxy: they can only (partially) reject the catholic model or confirm the orthodox one.

  5. Speakers of a language using a writing system as imperfect as the orthodox one may still produce error-free texts, provided they master its orthography; the fact that most people can write English (which has the most wicked orthography of all) virtually without misspellings is a confirmation that proper education may constitute a powerful tool in the most adverse conditions. If written evidence is produced almost exclusively by those that have received high education, then we should expect to have a negligible number of misspellings.This appears to be the case for modern Greek until, probably, my generation: despite the highly redundant writing system, it is hard to find any texts that comprise a considerable number of mistakes; this is undoubtedly due to the fact that the production of books and articles was the task of highly educated individuals and that the relatively smaller portion of the population that visited schools received an education of high quality; once the quality of education deteriorated, the participation rate reached almost 100% and nearly every individual contributed to the production of texts (proliferation of printed material due to private/small-scale typesetting, the Internet, etc), the spelling monstrosities multiplied and are now ubiquitous. With respect to the ancient documents we have so far found, the ones with the highest density of misspellings are obviously the papyri. Among them are found private letters meant to only be seen by another correspondent and this non-official character of theirs encourages semi-literate authorship and favours less attention in their production; not surprisingly, they contain a high number of errors, comparable to those committed by present-day Greek speakers in informal texts.cf. VESS10, p. 2: "papyri were not meant to last as were inscriptions, and this allowed for a higher degree of inaccuracy". Nothing comparable to Egyptian papyri has been found in Greece proper, as they "have all perished in the humid soil of Greece" (JANN97, p. 5); instead, the vast majority of ancient-Greek (thus, also Attic) inscriptions are of permanent and official character; hence, they were produced by educated authors with considerable care.This contrast with the Egyptian evidence has often been pointed out by several scholars who contest the main catholic narrative that change towards the "modern-Greek pronunciation" initiated in Egypt (as if linguistic change from the assumed catholic model is in a one-way road towards an inevitable "modern-Greek pronunciation" or as if the Egyptian tail wagged the Greek dog by imposing the Egyptian pronunciation in mainland Greece, comparable to American pronunciation overtaking England). Threatte has attempted to counter the argument of "The Alleged Conservatism of Attic Epigraphical Documents" (THRE82) as compared to the Egyptian papyri; he observes (p. 150) that "in some cases the types of examples found in the much later Attic texts are quite similar to those seen in papyri much earlier" and poses the question "whether the [...] sound changes actually occurred considerably later in Attica than in Egypt". It is evident that his intention is to show that Attica acted as a bulwark of the correct pronunciation and resisted longer to barbaric change. But if the question is answered to the positive, i.e., if the same changes occurred first in Egypt and then in Attica, this can only mean one of two things: a) either Athens was subdued to the authority (and phonology) of Alexandria (which we know is not true, as it was still considered the cradle of philological studies at least until the time of the Cappadocian Fathers) or b) was invaded by Egyptians (which is certainly ridiculous); if neither of the two is true (as actually appears), then we are to believe that Greek phonology is the only one of its kind that evolved regionally, but did not diverge! At any rate, it appears that the only valuable conclusion that can be drawn from this article of Threatte's is that the ancient inscriptions that are commonly thought as being of "private character" are actually more carefully written than one would expect: "The dipinti were placed on the vases before firing, and were intended to be permanent", "the ostraca with the same name can be grouped according to hands, which shows that they were pre-made for distribution to citizens at the time of the vote"(p. 152), similarly for the "defixiones" and "graffiti" (p. 153).
  6. All Inscriptions Were Not Created Equal
    The percentage of carelessly and/or ignorantly written texts among ancient-Greek inscriptions is fairly low.

A proliferation of misspellings may be, thus, due to factors other than sound change: "looser" education, character of the evidence found, etc.I have run into a comment that summarises the above pretty nicely: "any increase in the consistency of the confusion of η and ι in later (Hellenistic) times might have something to do with an increase (a) in population and therefore in the written records, (b) in the greater numbers of semi-literate writers who "learned" common spelling practices from their equals, and (c) in the numbers of records salvaged/discovered and available to us today. Had the climatic and geographic conditions in Athens been the same as in Egypt, chances are that heaps of papyrical writings would have likewise been discovered in Athens in the manner they were in Egypt. Not only climate, but also human vice (wars, vandalism, theft, carelessness), and other factors must have all contributed to the loss of ancient records. Without the classical stone inscriptions we would have had even fewer records available today." It would, therefore, be too precarious to draw any conclusions on phonology from the lack of misspellings in ancient inscriptions. I can only think of one possibility where this might suggest (partial) lack of coincidence of sound: if an inscription contains a considerable amount of misspellings of one kind (thus revealing a lack of familiarity with orthography), but none of a second kind, then it would be pretty safe to conclude that the pronunciation of this particular individual comprised phonetic mergers corresponding to the first kind, but not to the second kind. I have not as yet heard of any Attic inscriptions that fit this description.

Case Study: The Inscriptions From The Academy

One of the relatively recent developments was the discovery, during excavations in the archaeological area of the Academy of Athens in 1958, of several slates made of schist, one of them bearing an inscription with the words "ΑΘΙΝΑ ΑΡΙΣ ΑΡΤΕΜΙΣ ΔΙΜΟΣΘΕΝΙΣ" (for "Ἀθηνᾶ Ἄρης Ἄρτεμις Δημοσθένης"). It was reported that "Elles dateraient au plus tard du début du IVe siècle, mais pourraient remonter au milieu du Ve|They would date to the beginning of IV BC at the latest, but they could go back to the middle of V BC" (DAUX59, p. 579). The orthography, particularly the use of Ι where Η is expected, is surprising and the most reasonable explanation is that the person who carved the inscription pronounced Η like Ι (presumably [i(ː)]).

Such a strong indication of "iotacism" in classical Athens was not received with enthusiasm (to say the least) by the Catholics. Allen played the statistical card and reported it as "Startling by quite aberrant" (ALLE87, p. 74, n. 32), namely he downplays it as an unimportant nuisance. Teodorsson objected to this "method of minimising existing evidence practised by nineteenth century scholars [which] has not yet been abandoned" and observed that "Allen does not even mention that this inscription is the only one published out of about a hundred, which - according to the editor - contain a large number of iotacisms of the same sort" (TEOD74, p. 49, n. 47); although this is an overstatement, since there are "about a hundred tablets" (BALA91, p. 145) "of the same sort", but "Only on sixteen there were scratchings which could be identified as letters" (p. 145, n. 5), it is true that the existence of further inscriptions "of the same sort" reinforces its validity, rules out accidental variation and adds more weight to its evidential value. Nevertheless, Teodorsson also treats it as yet another statistic, since the three misspelled words appear (p. 90) as separate (uncorrelated) entries in his list of data (marked, according to his rules, with factor (1), i.e., for a suspected "defective ability to read and write").

Next in line was Threatte (Allen's "New Meisterhans"), who essentially rejected it stating that "The best explanation for these spellings is the boys' incomplete mastery of the alphabet" (reflecting the reported view that it was a schoolboy's tablet). Miller (MILL11, p. 145) aptly observes that "It is curious that the only major problem remaining in their 'mastery' involves the proper representation of the sounds /ɛ̄/ [presumably the value of Η] and /ī/ [presumably the value of long Ι], which on independent grounds can be expected to have overlapped in various environments by this time". Indeed, the tablet (viewable in DAUX59, p. 581, fig. 14) has no other misspelling but Ι for Η. While it is remotely possible that the schoolboys had learned to write all other letters but Η (which was, nevertheless, only the seventh letter in a 24-letter alphabet), it is more plausible to attribute a storm in New York to a butterfly in Beijing than to consider this a satisfactory explanation!This is, however, not plausible, since Η is found on another tablet of the same bunch (no. 3, as reported in BALA91, p. 146).

After wasting more than two decades on rudimentary argumentation, the Catholics launched a better planned offensive. Reportedly (BALA91, p. 150), Lynch was the first to doubt the only potentially questionable aspect of the tablets, its dating. He used a number of arguments, but these only relate to peripheral and perhaps extravagant claims of the excavators (the identification of the adjoining wall as the "wall of Hipparchos", of the location as "γραμματοδιδασκαλεῖον|(elementary) school", of the plates as "schoolboy tablets" and of the names on the second line of some tablets as "signatures of the writers"), which are typical in this field.As regards the alleged "wall of Hipparchos", a similar practice can be seen in the entire history of excavations, from Schliemann and Evans down to Andronikos, all of whom linked their discoveries to mythical/historical figures (Priam/Agamemnon, Minos and Philip respectively); it is certainly more prestigious to associate your work with a known personage than to an unknown individual that merely lived in the past. Balatsos, in addition to publishing the entire contents of the 16 inscribed plates, addresses Lynch's objections and argues rather convincingly that "there is little or no evidence for a post-classical dating" (BALA91, p. 153).

Of course, the Catholics would not give in without a fight. Threatte, who appears to have taken the matter personally, has come back to this topic in 2007 with an entire book, its apparent intension being to cast doubt on the validity of the classical dating. I have not been able to get a hold of it yet but, judging from its table of contents, it appears that the discussion of the dating occupies less than 10% of the book (!) and the moderate titles of the corresponding chapters give the impression that no conclusive evidence for a later dating are presented.In fact, Balatsos points out (BALA91, p. 151) that "The objects found together with them are of different date and of various kind but, to judge from the information given, no objects belonging to post-classical periods were found among these at the depth of 1.80-2 m., nor in the strata above this level", so that stratigraphy does not favour any post-classical dating.

By deciding to scrutinise this particular set of inscriptions, Threatte contravenes the very first rule defined above: "we know dating is not indisputable, but we are not excessively fastidious about it". As long as the dating has been done in accordance with the (imperfect, but established) normal practice, there does not seem to be any reason to vehemently contest it. If one decides to scrutinise the "inconvenient" inscriptions and cast doubt on their assigned date based on general skepticism (if that is indeed the case in Threatte's book), it is only decent that this be extended to all other inscriptions. Thus, Threatte seems to be opening Pandora's box and essentially to be shooting himself in the foot, since his "opus magnum" (THRE80) is concerned with the analysis of Attic inscriptions as these were originally dated by the excavator/publisher; without secure dating, all his conclusions drawn from epigraphical data are precarious. At any rate, unless Threatte provides very convincing arguments for a much later dating (which seems unlikely), the inscriptions from the Academy remain an important indication of early iotacism.

Metre

The term "Metre" refers to the rhythmic pattern of poetry, which traditionally (i.e., excluding some sorts of modern "poetry", which rather relate to versified prose) is repetitive. It can be said that metre is to poetry as percussion is to music. In present-day (rhythmic) poetry, the "drumbeat" is provided by stress: the basic pattern, or foot, is a certain combination of stressed and unstressed syllables, the most common being one stressed syllable followed by one unstressed syllable ('_ '_ '_ '_, like Pink Floyd's "We don't need no education") or vice versa (_' _' _' _', like "Amazing Grace! how sweet the sound"). This is a comprehensible mechanism, as the increase of volume (=stress) corresponds to the familiar beat of the drum.

In ancient Greek, the accented syllable (which was neither marked before III BC nor considered by the Catholics to correspond to stress) played no role in a verse's metre; instead, the drum beat was determined by the "length" of the syllable. The ancient foot consisted of a combination of "long" and/or "short" syllables. In principle, "μακρὰ συλλαβή|long syllable" is considered one that comprises a "long" vowel (a concept which also includes "diphthongs") and "βραχεῖα συλλαβή|short syllable" one that comprises a "short" vowel. Indeed, Dionysius Thrax defines (BEKK16, pp. 632-633, §9-10) these two kinds of syllables as "φύσει μακρά/βραχεῖα|long/short by nature"; he defines three cases for the former:

  1. "ὅταν διὰ τῶν μακρῶν στοιχείων ἐκφέρηται, οἷον ἥρως|when it is expressed by [means of] long letters [i.e., η and ω], such as ἥρως"; or
  2. "ὅταν ἔχῃ ἕν τι τῶν διχρόνων κατ᾽ ἔκτασιν παραλαμβανόμενον, οἷον Ἄρης|when it has one of the variable ones [i.e., α, ι and υ] in its expanded form, such as Ἄρης"; or
  3. "ὅταν ἔχῃ μίαν τῶν διφθόγγων, οἷον Αἴας|when it has one of the diphthongs [i.e., ει, αι, οι, ευ, αυ, ου and possibly also υι], such as Αἴας".

and two for the latter:

  1. "ὅταν ἔχῃ ἕν τι τῶν φύσει βραχέων, οἷον βρέφος|when it has one of the short ones by nature [i.e., ε and ο], such as βρέφος"; or
  2. "ὅταν ἔχῃ ἕν τι τῶν διχρόνων κατὰ συστολὴν παραλαμβανόμενον, οἷον Ἄρης|when is has one of the variable ones [i.e., α, ι and υ] in its contracted form, such as Ἄρης".One note regarding the last case of both kinds of syllable: The Old-Attic alphabet did not have separate graphemes for "long" and "short" vowels, but used Α, Ε, Ι, Ο, Υ to denote all vowels, irrespective of "length"; after the adoption of the Ionic alphabet, the (new) graphemes Η and Ω stood for "long" vowels, while the (old) graphemes Ε and Ο stood for "short"; the other vowel graphemes, Α, Ι and Υ, stood for both "long" and "short" vowels and were called δίχρονα|variable [lit. bitemporal or bimodal]. However, a "δίχρονο" found in a certain place in a certain word is always considered only "long" or only "short"; e.g., α in δῶρα|gifts is always "short" (hence the circumflex), whereas in γλώσσας|languages/tongues [acc.] it is always "long" (hence the acute). In general, it is considered that "Δὲν ἐπιτρέπεται νὰ θεωροῦμε ὅτι ὁ ποιητὴς μπορεὶ, ἀνάλογα κάθε φορὰ μὲ τὶς ἀνάγκες τοῦ στίχου του, νὰ χρησιμοποιῆ τὰ δίχρονα φωνήεντα τῶν ἐπιμέρους λέξεων ἄλλοτε ὡς μακρὰ καὶ ἄλλοτε ὡς βραχέα|We are not allowed to consider that the poet can, depending each time on the needs of his verse, use the variable vowels of the specific words sometimes as long and sometimes as short" (LYPO07, p. 17). However, this is exactly what the most ancient of the Grammarians seems to consider in the above cited passage, since the same syllable of the word (ρης|Mars) is used as example of both a "long syllable" and a "short syllable", according to whether it is "expanded" or "contracted"! This is also suggested by Dionysius' Scholiast: "Βραχεῖα δέ ἐστι σημείωσις συστελλομένου φωνήεντος καὶ βραχυνομένου, οἷον Ἄρες Ἄρες βροτολοιγέ. τὸ γὰρ δεύτερον Ἄρες συνεσταλμένον ἔσχε καὶ τὸν τύπον τῆς βραχείας|Short, on the other hand, is notation of a contracted and shortened vowel, like Ἄρες Ἄρες βροτολοιγέ; for the second Ἄρες has a contracted one and is of the short type" (BEKK16, p. 692) and "καὶ τὸ ι ὡς μακρὸν καὶ βραχύ, μακρὸν μὲν ὡς τὸ φίλε κασίγνητε, βραχὺ δὲ ὡς τὸ ἀλλὰ φίλον περ ἐόντα|and ι as long and short, long as in φίλε κασίγνητε, short as in ἀλλὰ φίλον περ ἐόντα" (p. 822). So, sometimes the poet did feel free to treat a "variable" vowel as he saw fit.

Thus, it appears that all a poet has to do for composing a dactylic foot (composed of three syllables: "long", "short", "short") is to place two "short" vowels after a "long" one. If only things were that simple! There is more in ancient metre than a mere arrangement of "long" and "short" vowels. The first and most important exception to "short by nature" is a "θέσει μακρά|long by position" syllable, five cases for which are provided by Thrax:

  1. ὅταν εἰς δύο σύμφωνα λήγῃ, οἷον ἅλς|when it ends in two consonants, such as ἅλς; or
  2. ὅταν βραχεῖ ἢ βραχυνομένῳ φωνήεντι ἐπιφέρηται δύο σύμφωνα, οἷον ἀγρός|when two consonants follow a short or shortened vowel, such as ἀγρός; or
  3. ὅταν εἰς ἁπλοῦν σύμφωνον λήγῃ καὶ τὴν ἑξῆς ἔχῃ ἀπὸ συμφώνου ἀρχομένην, οἷον ἔργον|when it ends in a simple consonant and has a following one [i.e., syllable] starting with a consonant, such as ἔργον; or
  4. ὅταν διπλοῦν σύμφωνον ἐπιφέρηται, οἷον ἔξω|when a double consonant follows, such as ἔξω; or
  5. ὅταν εἰς διπλοῦν σύμφωνον λήγῃ, οἷον Ἄραψ|when it ends in a double consonant, such as Ἄραψ.

In other words, a syllable may be considered "long", even if it comprises a "short" (i.e., ε, ο) or "shortened" (i.e., "short" α, ι, υ) vowel, as long as the vowel is followed by two consonants (either graphically distinct or disguised in the form of a "double" consonant ξ, ψ, ζ), irrespective of whether they are in the same syllable as the vowel or in the following one (also considering the following word). Further exceptions are defined by Thrax under the term "κοινὴ συλλαβή|common syllable", which means that "ἐναπόκειται τῷ ποιητῇ νὰ μεταχειρισθῇ αὐτὴν κατὰ βούλησιν, ἤ ὡς μακρὰν ἤ ὡς βραχεῖαν|it is up to the poet to treat it at will, either as long or as short" (RHAN92, p. 9). These include:

  1. ὅταν εἰς φωνῆεν μακρὸν λήγῃ καὶ τὴν ἑξῆς ἔχῃ ἀπὸ φωνήεντος ἀρχομένην|when it ends in a long vowel and the following one begins with a vowelRhankabes calls this kind of syllable "θέσει βραχεῖα|short by position" (RHAN92, p. 8), but I could not verify whether this is an established term.; or
  2. ὅταν βραχεῖ ἢ βραχυνομένῳ φωνήεντι ἐπιφέρηται δύο σύμφωνα, ὧν τὸ μὲν δεύτερον ἀμετάβολον, τὸ δὲ ἡγούμενον καθ᾽ ἓν ἄφωνόν ἐστιν|when two consonants follow a short or shortened vowel, whereof the second one is unchangeable [a.k.a. liquid, μ, ν, λ, ρ] whereas the preceding is itself voiceless [a.k.a. mute, β, γ, δ, κ, π, τ, θ, φ, χ]The "shortening" of a "θέσει μακρά" syllable in this case is known as "Attic correption".; or
  3. ὅταν βραχεῖα οὖσα καταπεραιοῖ εἰς μέρος λόγου καὶ τὴν ἑξῆς ἔχῃ ἀπὸ φωνήεντος ἀρχομένην|when, being short, it stands at the end of a part of the speech [=word?] and the following [syllable] begins with a vowel.

This is more or less what we know from the ancients about metre in ancient-Greek poetry. There are a few more cases of occasional exceptions (nicely summarised in RHAN92, p. 8-11), but these are more rare and we will refer to them only if necessary. Thus, the three rules for deciding whether a syllable is "long" or "short" together with their corresponding exceptions (wherein the syllable is "common", but most usually the opposite of what the rule stipulates) can be simply defined based on the vowel it comprises and its consonantal context as

RuleTypeNucleus|vowel part of the syllableContext|environment of the nucleusException
#1"long by nature""long|η, ω, ῃ, ῳ, ᾳ, αι, ει, οι, υι, αυ, ευ, (α), (ι), (υ)"-/_V|before a vowel
#2"long by position""short|ε, ο, (α), (ι), (υ)"/_{CC,ζ,ξ,ψ}|before two consonants or ζ, ξ, ψ/_CmuteCliquid|before a consonant cluster of a mute (β, γ, δ, κ, π, τ, θ, φ, χ) and a liquid (μ, ν, λ, ρ)|250
#3"short by nature""short|ε, ο, (α), (ι), (υ)"/_{CV,V}|before a single consonant or a vowel/_#V|at the end of the word, the next one begining with a vowel

Based on this information (and the texts of ancient-Greek poetry that have survived, as transmitted to us), what (kind of) conclusions can we draw on the pronunciation? I believe that, in order to be able to draw any conclusions, we need to address the following questions:

We should not forget that poetry is (and has to be) different from normal speech; otherwise, it would not qualify as art, i.e., some special sort of speech. Sometimes, the pronunciation used in normal speech is modified, in order to accommodate the needs of an artistic style. The most typical example must be the pronunciation of word-final e(s) in French singing (as exemplified i.a. by the well-known kid song "Frère Jacques"), which is at odds with the actual everyday pronunciation. That something similar could have happened with ancient-Greek poetry is alluded at by Thrax: "ἀνάγνωσις ἐντριβὴς κατὰ προσῳδίαν|reading in accordance with prosody" and "ἀναγνωστέον δὲ καθ᾽ ὑπόκρισιν, κατὰ προσῳδίαν, κατὰ διαστολήν. ἐκ μὲν γὰρ τῆς ὑποκρίσεως τὴν ἀρετήν, ἐκ δὲ τῆς προσῳδίας τὴν τέχνην, ἐκ δὲ τῆς διαστολῆς τὸν περιεχόμενον νοῦν ὁρῶμεν· [...] τὰ γὰρ μὴ παρὰ τὴν τούτων γινόμενα παρατήρησιν καὶ τὰς τῶν ποιητῶν ἀρετὰς καταρριπτεῖ καὶ τὰς ἕξεις τῶν ἀναγιγνωσκόντων καταγελάστους παρίστησιν|it must be read in accordance with expression, prosody, and pauses; from expression we discern the virtue, from prosody the art and from the pauses the intended content [meaning?]; [...] if it is done without due observance of these rules, the merits of the poets are degraded and the habits of readers become ridiculous" (BEKK16, p. 629, §1-2). Prosody (which was translated by the Romans as "accent") relates to special diacritical marks introduced in writing (but not necessarily invented) during the Hellenistic (i.e., post-Alexander) period; there were seven kinds thereof (SIAM88, pp. 537-541, §2,679-692), namely the so-called τόνοι|tones (οξεῖα|acute ´, βαρεῖα|grave ` and περισπωμένη|circumflex ῀), πνεύματα|spirits (ψιλή|lenis ᾽ and δασεῖα|asper ῾) and χρόνοι|times (μακρά|long ¯ and βραχεῖα|short ˘). While the marks of the first kind had no impact on metre and those of the second kind very rarely did soFor example, Rhankabes submits (RHAN92, p. 11) that the "common" syllable of exception #3 is most often "long", if the starting vowel of the next word receives a δασεῖα., the marks of the last kind are certainly related to metre, since they were used to indicate "long"/"short" vowels (later only the variable/indeterminate ones, α, ι, υ) or syllables. In principle, these marks might have related to pronunciation (at least during recitation) or they could have merely been technical notations serving to facilitate scansion (i.e., the identification of metre).It is often suggested that the "prosodic marks" were introduced to facilitate non-native speakers (mainly the Egyptian students of the Alexandrian school, where most of the "Grammarians" belonged to) in learning the proper pronunciation of Greek. I have not seen any substantiation of this claim, e.g., by reference to any hellenistic text that at least hints that this was the case. On the contrary, Teodorsson's observation (TEOD74, p. 279, n. 278) that the use of these marks was not limited to the schools where Greek was "taught", but they were adopted by the entire Greek-speaking world, illustrates rather convincingly that this theory does not hold water. But the choice of words by Thrax is rather suggestive of their role as "technical marks" rather than indicators of actual pronunciation: if the objective is to read "according to prosody", then this must mean that the "natural" or "normal" way of reading would be without due observance of prosody; furthermore, Thrax warns against this ("unprosodic") way of pronouncing only in connection with the reading of literary works, particularly poetry (he mentions as examples tragedy, comedy, elegy, epic and lyric poetry and dirges); finally, he specifically describes prosody as the main instrument of "τέχνη|art" (a term obviously referring to metre).

Phantom Prosody
It is possible that prosody merely relates to an ancient pronunciation, an artificial pronunciation or no actual pronunciation at all.

The simplest explanation is usually the correct one. The most straightforward conclusion that can be drawn from the terminology used in the description of Thrax (e.g., "τῶν μακρῶν στοιχείων|of the long letters (lit. elements)", "κατ᾽ ἔκτασιν|by expansion" and "τῶν φύσει βραχέων|of the short by nature", "κατὰ συστολὴν|by contraction", but above all "δίχρονα|(lit.) bitemporal"), is that the terms "short" and "long" correspond to duration, i.e., that a "long" vowel takes more time than a "short" one. Since the same terms are used for describing the various kinds of syllables, the simplest explanation is that syllabic length relates to temporal duration of the syllable, i.e., that a "long" syllable takes more time than a "short" one.This is also the impression that the Scholiast got: "τὸ γὰρ μακρὸν καὶ βραχὺ κατὰ χρόνον λέγει καὶ διάστημα φωνῆς|because long and short is said with respect to time and span of voice" (BEKK16, p. 797); although the expression is perversely convoluted (as is often the case in ancient-Greek authors), it rather undoubtedly associates length to temporal duration. Let us express this provisional conclusion with the mathematical form:

d(L) > d(S) (1)

wherein d(·) is a function representing the temporal duration of its argument (i.e., whatever is within the parentheses) during pronunciation, while L and S stand for "long syllable" and "short syllable", respectively. A similar relationship can be expressed for "long vowels" Vlong and "short vowels" Vshort:

d(Vlong) > d(Vshort) (2)

Since we have entered the realm of math, let us be a bit more mathematical: the duration function d(·) has the distributive property. Since the phonemes are pronounced not simultaneously, but consecutively, the duration of the syllable that they form must equal the sum of the durations of each phoneme:

d(P1...PN) = d(P1)+...+d(PN) (3)

wherein P1, ..., PN are the phonemes constituting the syllable P1...PN (e.g., for the syllables/words ἅλς and βοῦς mentioned by Thrax P1=ἅ, P2=λ, P3=ς and P1=β, P2=οῦ, P3=ς, respectively). A consequence of this is that the duration function d(·) should also have the commutative property:

d(P1P2) = d(P2P1) (4)

e.g., the syllables/words σέ and ἐς should have the same duration. This extends to larger syllables comprising N>2 phonemes, for any permutation of the same N phonemes.

There is no absolute definition of the duration of the "long" and "short" vowels and syllables. Length is only defined in relative terms, i.e., a "long" vowel or syllable is longer than a "short" vowel or syllable, as expressed by (2) or (1), respectively. Some have attempted to define an exact relationship between the two. For example, the Scholiast of Thrax, based, e.g., on the observation (BEKK16, p. 797) that Homer uses δέελον for the customary δῆλον, identifies a 2:1 ratio between long and short vowels (p. 798: "'Βραχέα δύο, ε καὶ ο.' Δηλονότι οἱ παλαιοὶ ταῦτα παρατρέχοντες ἐφθέγγοντο, καὶ οὐ τοσαύτην ὥραν, ὅσην ἐπὶ τοῦ η καὶ τοῦ ω, ἐποίουν ἐν τῷ ἐκφωνεῖν, ἀλλὰ τὸ ἥμισυ μέρος|'Short [were] two, ε and ο.' It is evident that the ancients pronounced those rapidly and took for their pronunciation not as much time as for η and ω, but [only] have of it"), i.e., he refines (2) as:

d(Vlong) = 2·d(Vshort) (5)

A similar relationship is argued for the "long" and "short" syllables, based i.a. on the observation (e.g., RHAN92, p. 20) that the (two "short" syllables of a) dactylic foot ("long", "short", "short") can (almost invariably) be replaced by a (single "long" syllable, thus creating a) spondee ("long", "long"):However, the place of a normally long syllable (i.e. the first of the dactylic foot) is never occupied by two short ones (except for cases of synizesis, which corresponds though to a long syllable; RHAN92, p. 21), as we would have expected if a long syllable were truly equivalent to two short ones.

d(L) = 2·d(S) (6)

At first glance, yes. If (6) is true, then the pure dactylic hexameter ("long", "short", "short" x6) would look like this in musical notation (ignoring any tone embedded in the syllables):

each half note denoting a syllable with a "long" vowel or "diphthong" and each quarter note standing for a syllable with a "short" vowel. This illustrates the underlying pattern, which must have dominated ancient (epic) poetry. However, in order for (6) to be universally true (i.e., for all "long" and "short" syllables), all "long" syllables should be of the same duration and so should all "short" syllables (equal to half the duration of the "long" syllables). Resorting only to the examples provided by Thrax, this would mean that the syllables "ἥ" of "ἥρως" and "ρης" of "Ἄρης" should have the same duration, even though the second comprises the first plus two consonants; this is not possible under (3), unless one assumes zero duration for the consonants; but this cannot be the case, since every sound has some duration, even if infinitesimal.

The major fallacy of (6) is that it assumes a fixed duration for every "long" syllable and a corresponding fixed duration for every short syllable. The fact that some "short" syllables are actually longer in duration than others is pointed out by Dionysius Halicarnasseus (ROBE10, p. 152), who observes that the first syllables of δός, Ῥόδος, τρόπος, στρόφος are progressively longer, but they are all considered "short", and that ἡ (or ἤ) and σπλήν are two monosyllabic words, both considered "long", even though the latter is certainly longer than the former ("μήκους δὲ καῖ βραχύτητος συλλαβῶν οὐ μία φύσις, ἀλλὰ καὶ μακρότεραί τινές εἰσι τῶν μακρῶν καὶ βραχύτεραι τῶν βραχειῶν|There is more than one kind of length and shortness of syllables : some are longer than the long and some shorter than the short [Roberts' translation]").

The Syllable Farm
All "short" (resp. "long") syllables are equal, but some are "shorter" (resp. "longer") than the others.

Not quite yet. The mathematical theory we have developed (based on common assumptions about metre) seems like a dead end. But we are not ready to give up yet; just to give up on (6). In other words, we conclude that any "long" syllable cannot be twice as long in duration as any "short" syllable. We will cling onto (1) though, namely that the duration of every "long" syllable is longer than the duration of any "short" syllable. This will merely mean that the half notes of the above musical notation are not all the same, but they stand for something more than the quarter notes (which are also not the same); thus, the pattern may not be perfectly even, but it will still be some pattern.

Enter "long by position". Since the syllables of that kind can be readily used in lieu of the normally "long" ones ("long by nature"), this means that technically there is no distinction between "long by nature" and "long by position", i.e., a syllable that is "long by position" is as "long" as one that is "long by nature". This is puzzling, because a syllable that is "long by nature" still comprises a "short" vowel. For example, in the word νέκταρ|nectar, mentioned in the verse "νέκταρ ἐοινοχόει: τοὶ δὲ χρυσέοις δεπάεσσι" of the Iliad (4.3), the first syllable ("νε") is scanned (i.e., identified metrically as) "long" by position, since its vowel is followed by two consonants, while the second ("κταρ") is scanned "short", its vowel being followed by a single consonant. This is puzzling, because both syllables comprise a "short" vowel ("ε" and "α", respectively), but the first has two consonants less than the second. In view of (3), there is no way to justify (1) in this case.

One objection that mainstream scholars would raise against my previous example is the partition into syllables. I have used the traditional rules of Greek syllabification, according to which two consecutive consonants belong to the following syllable if they may be found at the beginning of a(ny) Greek word; otherwise the first consonant is allocated to the previous syllable;This is an ancient tradition dating back to the Grammarians and a reasonable one: if one is to recite a verse for the purpose of making the rhythm (="metre") apparent, one needs to divide the verse into distinct syllables, which would be equivalent to redrafting the verse as a series of monosyllabic words. Each of these mini-words should naturally obey the rules of Greek phonotactics, which means that it may only begin with a consonant cluster acceptable as word-initial; at the same time, it is quite unnatural for Greek words to end in consonants (only ς, ν, ρ, ξ, ψ are allowable as word final), which would be way too frequent, if consonant clusters were always divided between syllables. this principle is evidently accepted by Thrax, who speaks of a syllable ending in a consonant, the next one beginning with another consonant, in the case of ἔργον (ἔρ-γον) and of a short vowel (of the syllable in question) followed by two consonants (obviously of the next syllable), in the case of ἀγρός (ἀ-γρός); it is also expressly stated by Herodian (as quoted in ALLE87, p. 166) and summarised by Smyth (SMYT20, p. 35, §140). However, the official narrative is that "Greek speakers in the classical era only pronounced a consonant as part of a preceding syllable when it was followed by other consonants, due to the rules of Greek and Latin syllabification. In a consonant cluster, one consonant ends the preceding syllable and the rest start the following syllable".This statement has no apparent basis in ancient literature; yet Allen (ALLE87, pp. 105-106) feels compelled to correct the Grammarians, because the rule of keeping two consonants together if they may introduce a Greek word, "is quite contrary to phonetic division in Greek", without specifying on what kind of experience with Greek or other evidence his assertion is based. One can see how unnatural this division is by considering the examples of the so-called "double consonants", e.g., ἔξω mentioned by Thrax, which (assuming that ξ had the modern value of [ks]) would be allegedly divided in pronunciation as ἔκ-σω, thus making the use of a single letter (ξ) meaningless! According to this theory, the syllabification of νέκταρ would be νέκ-ταρ. Since the two syllables still have the same number of consonants and short vowels, in order for the first syllable to be considered longer than the second, the argument has to be taken one step further by considering the verse as a single word divided in syllables as νέκ-τα-ρἐ-οι-νο-χό-ει etc. Only when the verse is chopped up in this brutal way, can we consider it to conform with the "duration theory".The Scholiast of Thrax has surpassed the Catholics by further mutilating the syllable following a "long-by-position" one (BEKK16, p. 823): "τότε γὰρ γίνεται ἀνάγκῃ μακρά, ὅτι ἡ βραχεῖα ἕνα χρόνον ἔχει καὶ ἕκαστον τῶν συμφώνων ἡμιχρόνιον, ὥςτε ταῦτα συνερχόμενα ἀποτελεῖν δύο χρόνους. ἡ δὲ μακρὰ ἐκ δύο χρόνων συνίσταται, ὥστε γίνεσθαι θέσει μακράν|because then it becomes long out of necessity [forcedly?], since the short [syllable] has one time-length and each of the consonants half a time-length, so that these together constitute two time-lengths. the long [syllable] consists of two time-lengths, so that this [the syllable comprising the short vowel and the two consonants] becomes long by position". Interesting as this theory may sound (consonant=1/2, "short"=1, "long"=2), it does not fare better than the Catholic syllabification, since it chops up the words in an even more unnatural way and allows for doubly-long syllables (such as σπλήν, which would span 1/2+1/2+1/2+2+1/2=4 time-lengths), something that would certainly distort the rhythm.

This ugly syllabification does have some advantages: it restricts "short" syllables to at most two letters, a consonantal onset and a "short"-vowel nucleus, whereas a "long by position" has additionally a consonantal coda and is, therefore, "longer". However, there are syllables without an onset and when they are "long-by-position" they would have exactly the same composition as a "short-by nature" syllable. For example, in the Homeric verse "εἴ που ἔτι ζώει γε Νεοπτόλεμος θεοειδής" (Iliad 19.327), the first two syllables of "Νεοπτόλεμος" (if we divide it according to Catholic syllabification, Νε-οπ-τό-λε-μος) comprise a single consonant and a short vowel and should, under (4), have the same duration; yet the first is scanned as short and the second as long (by position). There is no way to reconcile this case with (1)-(4).Nor would it help to argue that the difference in duration may be owing to the particular vowels and consonants, which are different in each of the two syllables. All letters/phonemes of each of the three categories (long/lengthened vowels/diphthongs, short/shortened vowels, consonants) defined by Thrax and the other Grammarians are deemed equivalent (with the irrelevant exception of Attic correption) and should have the same effect on duration. Furthermore, if anything can be argued on the duration of the consonants in this particular case, it is that the "short" syllable's consonant ν, being a continuant, cannot have a duration smaller than the "long" syllable's consonant π, which is a stop! This fact alone suffices to show that the second syllable cannot be forced to have a longer duration (unless one meddles with the short vowel and pronounces it as something that it is not, e.g., ω or ου). I believe that it has been sufficiently demonstrated that the "duration theory" (as most kinds of Lingo-Math) cannot provide a reasonable explanation of the rules of metre, particularly the case of "long by position".

The "duration theory" can now be declared bust.

As explained above, the "duration theory" appears to be trapped in the speculative interpretation of metre provided by the ancient Grammarians. Based on the observation that "long" vowels (η, ω) and diphthongs (often) coincide with the prominent part of the foot ("ictus"), they have concluded that the associated syllable must also be long (i.e., "μακρά", a term that alludes to duration) and that any syllable not comprising a "long" vowel or diphthong was made "long" by virtue of the following consonants. The dead end reached, if duration is assumed as the intrinsic property providing the rhythm ("metre"), has been apparently recognised by modern scholars (e.g., ALLE87, pp. 110-112), who have attempted to develop different theories that would be based on other principles. As a first step, modern scholarship has abandoned the terms "long" and "short" syllables and uses instead the terms "heavy" and "light" syllables, respectively, terms that are evidently based on "the terminology of the ancient Indian Grammarians" (ALLE87, pp. 104-105), while they do not speak of "length" or "duration", but of "quantity" (cf. ALLE87, p. 110: "Quantity, like vowel length [...], should not be considered as a simple matter of duration"). Here, "heavy" stands for what was defined as "long by nature" and "long by position" and "light" stands for "short by nature". But, so far, there does not seem to be any major divergence from the "duration theory", except for a change of names. What was the (phonetic) significance of "heavy" and "light"? How did they make metre stand out?

An attempt to answer this question is Allen's theory of "syllable arrest" (ALLE87, pp. 2, 6, 112), perhaps the most prominent theory of this kind. It builds on the aforementioned Catholic syllabification and, if I understand it correctly, is based on consideration of the syllable's termination, rather than the syllable as a whole. Allen confirms that the duration of a "heavy" (formerly "long") syllable may not always be greater than that of a short oneHe actually cites Halicarnasseus' example στρο-, which is considered "light" (formerly "short"), as compared to ὠ-, which is "heavy" (formerly "long"), and, using the arbitrary allocation in terms of time-lengths: consonant=1/2, "short"=1, "long"=2, computes the duration of the former to be 2.5 time-lengths, i.e. greater than that of the latter (which is, by definition, 2 time-lengths). However, the assignment of half a time-length to consonants (a common illusion of ancient "rhythmicians") is unfounded and evidently wrong (i.a., it assumes that both stops, such as κ, π, τ, and continuants, such as σ, μ, ν, ρ, λ, have the same duration) on one hand and στρο- is not a syllable that would be scanned in any epic as "light" (in the middle of the verse, its first consonant would have to be allocated to the previous syllable according to Catholic syllabification, while at the beginning of the verse it would have to be considered "heavy", as first part of the dactyl; cf. RHAN92, p. 10, item ι´) on the other. and concludes that "quantity is not concerned so much with the duration of the syllable as a whole (though, in general, heavy syllables will have been of greater duration than light), but rather with the nature of the syllabic ending". He, thus, considers that what matters for the characterisation of the syllable as "heavy" or "light" is the part from the vowel on, i.e., the vowel and the coda, disregarding the inconvenient onset. Thus far, Allen merely rephrases the rules provided by the ancient Grammarians, but does not explain the effect induced by "heavy" syllables. Then, he attempts to make the connection between theory and practice by asserting that "the movement of a light syllable is 'unarrested', whereas that of a heavy syllable is 'arrested' (by the chest-muscles in the case of a long-vowel ending, by the oral constriction in the case of a consonant ending, or by a combination of the two in the case of a diphthongal ending, according to Stetson p. 7, n.)". This appears to suggest that the main difference is that the expiratory syllabic process ends abruptly in the "light" syllables and more smoothly in the "heavy" ones. This is as much as one can make out of Allen's description; yet it does not provide a satisfactory explanation of metre for at least the following reasons:

  1. The insistence on the Catholic syllabification as the only truth without providing the slightest piece of evidence about its application in ancient Greek is nothing but "Proof by Blatant Assertion" (ZEMA94). It also has some inexplicable implications, such as the need to divide the so-called "double consonants" ξ and ψ, which represented (as accepted by both Orthodox and Catholics) the consonant clusters [ks] and [ps], respectively, which the ancients had joined into single symbols; had the two consonants of the cluster belonged to different syllables (according to the Catholic syllabification), it would be extremely unnatural to even consider them as a single entity.
  2. Allen does not convince that he understands (or has a clear idea of) the very theory he is promoting. In the above statement (p. 112), he clearly suggests that "light" syllables (which end in a "short" vowel) lack "arrest" ("the movement of a light syllable is 'unarrested'"); but in the introduction to the theory (p. 6), he describes the "oral arrest", one of the three kinds of "arrest" in the above-cited passage ("the movement [...] of a heavy syllable is 'arrested' [...] by the oral constriction in the case of a consonant ending"), as a feature of the "short" vowels ("An oral arrest, on the other hand, is a relatively rapid movement and so is associated with short vowels"); which one is it? He also claims (p. 6) that "Short vowels may be also be associated with a type of movement in which the release of the following syllable overtakes the arrest of the preceding, rendering it effectively unarrested"; why is this cancelation of the preceding syllable's "arrest" not taking place when that one ends in a consonant ("long by position")? should this not apply also to the case of the preceding syllable ending in a "long" vowel (in which case, a "long" vowel followed by a consonant cluster would automatically lose its "arrest", i.e. it would be "shortened")?
  3. The whole theory of "arrest" is based on the work of Stetson, which Allen himself portrays as "suspect" ("much of the detail of Stetson's experimentation has been considered suspect", p. 2)! Without in-the-field verification, it amounts to yet another "thought experiment" (in Lloyd's terms) that produces a fancy theory, which may nevertheless have little to do with reality. Furthermore, Allen states (p. 2) that Stetson's theory "approaches the problem from the standpoint of the physiology of the syllabic process rather than its acoustic results", which means that any effects of "quantity" are likely to be felt by the speaker, but not necessarily heard by the listener. Are we, then, to believe that metre was an... esoteric process, meant as a magical experience for reciters, but not for their audience?

Not really. Since its exportation to Venice and the West by fleeing Eastern-Roman (="Byzantine") scholars in XV AD (cf. MENA08, pp. 8-9), metre had been a puzzle. By then, poetry was dominated by stress-based patterns and people could only think of rhythm in terms of stress. In order to feel the rhythm, they stressed the "long" syllables of the metre (e.g., the first syllable of the dactylic foot), irrespectively of whether those received an accent mark (i.e., οξεῖα|acute ´, βαρεῖα|grave ` and περισπωμένη|circumflex ῀, which in current speech correspond to a stressed syllable) or not. This is the reason why Rhankabes proposed (e.g., RHAN81, pp. 42-45) to stress the "long" (now "heavy") syllables and disregard the actual accent, in order for the metre to be more prominent (and in accordance with modern practice).

He is not alone in this consideration. Even Allen, while addressing the question of "ictus", concludes that there may have been some amount of stress associated with the "heavy" syllables (ALLE87, pp. 114-115: "it seems possible that heavy syllables were liable [!] to bear stress in the language - but that not all [!] such syllables were stressed; that light syllables tended [!] to be unstressed; and that verse rhythm was based on the alternation of stressed and unstressed syllables"). The same consideration has led the Catholics to deny any association of the ancient-Greek accent with stress, otherwise the two kinds of stress (ictus and accent) would cancel each other (wherever they do not coincide). However, the assumption that the "heavy" syllables were associated with any amount of stress (not only in poetry, but also) in normal speech leads to further problems; for one, it allows a word to comprise two "heavy" syllables and, thus, two stressed syllables, which is contrary to the present rules of Greek (except for the special case of proparoxyrones followed by enclitics); how could the language tolerate several stresses in the same word (note that unlike Germanic, where long words are essentially concatenated short ones, each retaining its stress, many Greek words are and always have been inherently polysyllabic), even side-by-side (when they contain at least one spondee), and then suddenly develop an intolerance to multiple occurrences of stress in the same word? The question of quantity in Greek will be addressed in a separate chapter, but for the time being suffice it to say that the association of quantity and metre with stress is only indicative of lack of imagination, undoubtedly due to limited experience with human-language phonetics: there are several syllabic properties beside stress, which could be considered as candidate qualities of "heavy" (formerly "long") syllables (perhaps some are hitherto still undiscovered), but if one's only experience with metre is based on stress patterns, one is liable (to borrow Allen's term) to seek stress in any metric technique.

Actually, Rhankabes did not claim that "long" (now "heavy") syllables received stress in actual speech, but only in poetry, as dictated by the metre (RHAN81, p. 42: "Im anapästischen Versmass sollte es [d.h. πολλάς] ποολ-λὰς, im trochäischen πόλ-λας, im jambischen πολ-λὰς gelesen (gesungen) werden, um den Rhythmus nicht zu überwerfen|In anapaestic metre it [i.e. πολλάς] should be read (sung) ποολ-λὰς, in trochaic πόλ-λας, in iambic πολ-λὰς, in order not to destroy the rhythm"). He argues that all kinds of verse were meant to somehow be sung, not spoken and that the stress pattern was (artificially) imposed by the accompanying music ("Diese Geltung des Nachdrucks wurde dadurch gesteigert, dass die Verse eigentlich nur zum Gesang bestimmt waren. Die epischen wurden von den Rhapsoden, die lyrischen von Chören gesungen, und der dramatische Dialog auch als ein leichtes Recitativ mit Begleitung der Flöte deklamiert. Da ist es begreiflich, dass die Musik den Nachdruck der den Rhythmus bezeichnenden langen Silben aufzwang, und so auf diese als die hervorragendste wies|This application of stress [emphasis?] was reinforced by the fact that the verses were actually intended for being sung. The epics were sung by the rhapsodists, the lyrics by the choruses and the drama dialog was also declaimed as a lightweight recitative accompanied by the flute. It is, thus, understandable that the music imposed the stress [emphasis?] on the long syllables that indicated the rhythm and so made them stand out as the most prominent ones").

A similar thesis has been put forward by Caragounis (CARA95, pp. 175-176) in his attempt to deny the existence of "long" vowels (and quantity, in general) in Greek. He considers that the rhythm in poetry was observed by the setting and raising of the foot (much like timekeeping while playing a musical instrument), whence the term "foot" for the metrical unit (something that seems plausible). He then argues that "length" is merely a feature owed to the position of a syllable within the foot: "if a syllable was placed in an 'accented' position within the foot, it was considered long by position (θέσει μακρά); if in an unaccented position, it was considered short (ἄρσει βραχεῖα)"; thus all "long" syllables were "long by position" (of the foot)! He also claims that the category "long by nature" is a misunderstanding caused by the adoption of Η and Ω (which, he reckons on p. 159, "appear to have been adopted originally (Vth century B.C.) as mere technical, compensatory marks for Ε and Ο respectively in accented (and therefore lengthened) position"): "syllables containing these letters came to be considered as naturally long (φύσει μακραί)". Essentially, the thesis is that no vowel was short in actual speech, but only when sung (or recited) were those that happened to be in emphasised position (=ictus) perceived as "long".

There is but little evidence for this theory. A disregard of the actual "long" or "short" nature of the syllables is probably more evident in lyric poetry, wherein short and long syllables may be treated as variable, with metre imposing the "length" (cf. ROBE10, p. 129, n. 17: "in Greek lyric metres, so far as they come under what we have seen called μέλη and ῥυθμοί or 'rhythmi,' long and short syllables alike were more or less variable" and "To reproduce the rhythmical pattern which the poet had in mind, the singer, if not also the reader, made some long syllables longer and others shorter than two χρόνοι πρῶτοι, and made some short syllables longer than one χρόνος πρῶτος"). In other words, there was little effort to arrange the syllables so that their usual "quantity" be in conformity with the intended metre. A similar practice may also be seen (albeit less often) in epic poetry, the poets often devising strange tricks to use syllables against their (metrical) nature, such as arbitrary doubling of consonants to render a "short" syllable "long" (cf. RHAN92, p. 10, item ια´), using the same syllable of the same word here as "short" and there as "long" (cf. RHAN92, p. 9-10, items ς´ and ζ´, as well as Ἄρης mentioned by Thrax as example of both "shortened" and "lengthened" syllables) or not using a syllable as "long by position" despite its "short" vowel being followed by a double consonant or a pair of consonants (cf. RHAN92, p. 9, item δ´); in general, the epic poets often showed no respect for the "normal" quantity of syllables (cf. JANN97, p. 522, n. 1 quoting Kühner-Blass: "The Homeric language shows an almost unlimited freedom in the use of long and short as occasion arises, and that so, that the quantity is or is not represented by the script"). Nevertheless, these cases were exceptions rather than the norm and we should not consider them as generalised. As a rule, the poets strived to arrange the "long" syllables (whether "by nature" or "by position") in the same position within the metrical foot. Had quantity been a mere matter of position and, hence, of chance, the same syllable of the same word should be expected to feature in various locations within the metrical foot. However, this is not the case; on the contrary, its distribution is overwhelmingly in favour of its "normal" quantity. This consistency in the placement of syllables is difficult to explain as pure chance.

Caragounis' views are greatly influenced by those of Jannaris, whose theory on quantity (JANN97, pp. 519-540) he has probably misunderstood or misrepresented. In that very interesting theory, Jannaris starts (p. 523) with the observation that quantity-based poetry was being produced throughout the entire Antiquity, until several centuries in the post-Christian era. Since we know "that post-classical speech, at least since Greco-Roman times, had lost all feeling of quantity" (he actually means vowel length), it follows that "quantitative versification ever since has been a purely artificial fabric" and the post-classical (or at least post-Christian) poets "followed a purely artificial system alien to the nature of their native speech". Based on some further observations from the classical times, he reaches (p. 524-525) the similar conclusion that "versification with the [classical] Attic poets [...] was an artificial method founded upon precedent practice, a system developed in the course of previous antiquity and duly handed down to them".Although some of his arguments are dubious (like the confusion of accent and quantity), this conclusion could be reached (as a possibility) by applying the Invariability Principle: if quantity-based metre was observed in post-Christian times, without any basis in the spoken language, but merely as continuation of an established tradition, the same might have happened at earlier times, as long as there was enough time for a tradition to settle in. He then claims (p. 528) that "long" vowels (including the spurious diphthongs ει and ου) were introduced "for restoring or rather indicating the disturbed rhythm", due to linguistic phenomena known as "compensatory lengthening (antectasis), contraction, crasis, etc", which result in the creation of "long" vowels and "are virtually nothing else than various manifestations of one and the same principle: metrical compensation". Then comes the most extravagant part of his theory: he asserts (p. 529) that, in Attic, "the lost sound" (consonant in the case of antectasis, vowel in the case of crasis) was indicated by "a vertical (straight) stroke", which was ultimately confused for a normal letter either on its own or in combination with the previous vowel; he suggests that after Ε it led to the "diphthong" ΕΙ, but after Ο the resulting ΟΙ was (somehow?) converted to ΟΥ; even more extreme is his claim that in some cases the two graphemes of EΙ and ΟΙ were joined as glyph for Η, the precursor of Η, and Ω, after being rotated by 90°! He finally postulates (p. 531) that, due to "the political convulsions of those times" (V BC), these "hitherto technical or compensatory symbols" were "mistaken for real phonetic symbols". In other words, in order to prove that "before the period or stage of 'contraction' and 'antectasis', Greek, like other languages, showed only two classes of syllables: syllables naturally short and syllables long by position" (pp. 535-536), Jannaris would like to convince us that the concepts of "long" vowels and duration-based metre are basically products of a misunderstanding!

Although taken to the extremes to reach ludicrous conclusions, Jannaris' imaginative theory is not completely worthless. It points to an important fact that may be the link between "long by position" and "long by nature": the latter is, in some cases, produced when one of two consonants in an impermissible consonant cluster (e.g., νσ or δσ) is dropped, causing "compensatory lengthening" of the previous vowel. The following thesis might, therefore, be a more appropriate basis for explaining ancient metre:

  1. The differentiating feature among the various syllables was the number of consonants (possibly including glides [j] and [w], as second elements of "diphthongs") following a vowel. This would produce an audible distinction (now called "ictus"), as consonant clusters would appear periodically, where the metre required an emphasised (now "heavy") syllable.
  2. The simplification of some consonant clusters (loss of one of their consonants) produced, in the oldest works of poetry that were orally transmitted from one generation to the next, a lengthened vowel in some syllables that occupied the place of a "heavy" syllable (i.e., originally followed by two consonants).
  3. A metrical tradition may have been established (probably even before Homer or at least the time that his works were recorded and "normalised") based on the observation that the emphasised positions (receiving the "ictus") in the foot were either "long by position" or "long by nature" (in the sense defined by Thrax).
  4. From then on, every poet followed the established tradition, in the same fashion that did the poets of late antiquity, much like the unnecessary (but established) syntax rules of modern German, which are nevertheless obeyed by every speaker of German.German, being an inflecting language, ought to have a free syntax, like Greek does. A rigid syntax is only necessary for languages deprived of inflexional variety, like English, wherein the words of the sentence "The cat ate the mouse" cannot be freely rearranged (e.g. "ate the mouse the cat" or "the cat the mouse ate") without introducing ambiguity about who ate whom. In Greek on the other hand, any permutation of the words ("Ο γάτος|the (male) cat έφαγε|ate το ποντίκι|the mouse", "Ἐφαγε ο γάτος το ποντίκι", "Το ποντίκι ο γάτος έφαγε", "Το ποντίκι έφαγε ο γάτος", etc) is permissible and does not leave any doubts about the meaning (the mouse is always the victim). However, German syntax rules unnecessarily require that the verb always be placed in the second position in main clauses ("Der Kater|the (male) cat aß|ate die Maus|the mouse" or the rarer "Die Maus aß der Kater") and at the end in subordinate clauses ("dass der Kater die Maus aß"), while all other words may be freely moved around as long as they do not interfere with the position of the verb. I have not been able to identify the cause of this rule, but, whatever the reason may be, it is an established rule that has (at present) no apparent raison d'être in the language and is, nevertheless, obeyed by all native speakers.

This theory is more solid that the one put forward by Jannaris: it does not rely on a denial of existence of "long" vowels in Greek and it is not based on controversial assumptions like the never-attested "vertical (straight) stroke" of Jannaris; however, it is (and will always be) mere speculation, since it appears impossible to find evidence that would support it.

Absolutely none. The problem with all the aforementioned theories is that they are mere speculations. Even though their instigators pretend they know exactly how ancient phonetics and ancient poetry worked, the truth is that we are in the dark, since we cannot observe ancient-Greek metre as recited by the ancients. The proposed theories fail to illuminate all aspects of ancient metre:

  1. The "duration theory" fails to explain why a syllable ("long by position") with the same number and kinds of letters as another one ("short by nature") is considered "longer"; it would only make sense, if all consonants are assigned to the syllable of the previous vowel (i.e., if all syllables are onset-less), but this kind of syllabification would have no basis whatsoever in the actual language.
  2. Allen's theory of "syllable arrest" is also based on unnatural (at least for Greek) syllabification, as well as on "experimentation [that] has been considered suspect", it fails to explain why a stop (such as κ, π, τ) provides the same kind of "arrest" as the "long" vowels, whereas a "short" vowel does not and it rather appears to relate to a sensation of the reciter, but not of the listener.
  3. Caragounis' theory of "isochronous syllables" and "stressed positions" within the foot ("ictus") fails to explain the consistent correspondence of normally "long" syllables (in both senses) with the "stress positions".
  4. Jannaris' theory of "vertical strokes" is too far-fetched and hard to defend.
  5. Finally, my own theory of "misunderstood lengthening" is speculative and practically unfalsifiable.

In fact, since none of us has direct experience with ancient Greek, the only honest answer is to admit that "δὲν ξέρουμε τὶ ἐσήμαινε ἀκριβῶς - ἀπὸ τὴν ἄποψη τῆς προφορᾶς - μακρὰ ἤ βραχεῖα συλλαβἠ|we do not know what it meant - in terms of pronunciation - long or short syllable" (LYPO07, p. 16) and to treat metre the same way that it was dealt with by the post-Christian scholars: as a set of constraint rules that must be obeyed, as an ancient Sudoku!

Ancient Sudoku
The only conclusions that can be drawn from ancient metre are the very rules that define it, all theories based on them being unproven speculations.

Conclusion

I hope it is clear from the above that for almost every kind of evidence, there is inherent uncertainty regarding its significance for ancient phonology. One must take all these shortcomings into account, before proclaiming the problem of the determination of ancient-Greek pronunciation solved. Sturtevant, like most Catholics, also recognises (STUR20, p. 11) that "most of the available evidence falls short of definite proof", but claims that the uncertainty may be compensated by statistical inference ("the force of our evidence is cumulative") and that "the inference from all the items combined is in many cases practically certain". I believe that, by means of the "accumulated" evidence suggestive of a value [ʃ] for Ξ, I have illustrated how uncertain evidence can be manipulated to back an extravagant conclusion that is anything but "practically certain".

One may object that I have been too harsh on the various kinds of evidence and that, within this strict framework, making anything useful out of the evidence looks like a mission impossible. This would be similar to the (self-)criticism considered by Chatzidakis: "Δυνατὸν αἱ μεθοδολογικαὶ αὗται ἀρχαί, περὶ ὧν ἐγὼ κρίνω ὅτι εἶναι ἀπαραίτητον νὰ τηρῶνται κατὰ τὴν ἐξέτασιν τοῦ περὶ τῆς ἀρχαίας προφορᾶς ζητήματος, νὰ φανῶσιν εἴς τινας ὅτι ὄζουσιν ὑπερκριτικῆς τινος, καὶ ὅτι καθιστῶσι τὴν ἔρευναν πάνυ δύσκολον|It may appear to some that the methodological principles, by which I consider necessary to abide during the examination of the question of ancient pronunciation, reek of excessive criticism and that they render the research extremely difficult" and the answer cannot be other than that the framework outlined here: "δὲν πρόκειται περὶ εὐκολοτέρας ἐρεύνης, ἀλλὰ περὶ μεθοδικωτέρας καὶ δὴ ἀσφαλέστερον εἰς τὴν ἀλήθειαν ἀγούσης|is not a case of easier research, but of a more methodical one and actually one that more safely leads to the truth" (CHAT02, p. 370).

This "truth" will be the subject of the chapters of the following section. Unlike Chatzidakis though, I will show that almost no claim, neither Catholic nor Orthodox, can be (positively or negatively) proven with any degree of certainty; my position about drawing conclusions is rather... Socratic, namely no conclusion can be considered "safe". The honest answer to the question "How did ancient Greeks/Athenians pronounce the letters?" is simply "We do not know". But even if we had to assign probabilities to the candidate phonemes corresponding to each letter, in most cases the suggested "reconstructed" (i.e., Catholic) phonology does not fare better than the received (i.e., traditional or "modern Greek" or Orthodox) pronunciation.

Notes

Go to previous page Go to main page Go to next page