 |
 |
 |
|
 |
For the value of most of the letters in this chapter we have no definitely positive proof, but our main clue is their current value. Our moderate certainty about the validity of the present values also in ancient times comes from the fact that there is virtually no negative indication, i.e., that the scanty evidence we have is not at odds with this correspondence.
I consider familiarity with the present phonology of Greek to be a prerequisite for the understanding of this, as well as the following chapters of this section.
The consonants of present-day Greek may be divided in two major families: basic consonants (which comprise the labial, dental and palatal/velar series or rather 2x2 matrices) and auxiliaries (which include the nasals, the liquids and the sibilants).The terms "basic consonants" and "auxiliaries" are my own denominations and not to be found in any book of Greek grammar, either modern or earlier. The reason why I consider the labials, dentals and palatals/velars (palatals being essentially allophones of the velars) as related and, hence, classifiable under the term "basic consonants" is that there is a one-to-one relationship among them, including their associated nasals, the only difference being the place of articulation (as illustrated in the accompanying graph); although modern sibilants may also be considered to have the same (2x2) structure, their different nature is, i.a., suggested by the lack of an associated nasal. A similar classification is apparently considered by the ancient Grammarians. Thrax considers (BEKK16, p. 631) that the term σύμφωνα (←συν+φωνή, con+sonus) is due to the fact that "αὐτὰ μὲν καθ᾽ ἑαυτὰ φωνὴν οὐκ ἔχει, συντασσόμενα δὲ μετὰ τῶν φωνηέντων φωνὴν ἀποτελεῖ|they have no voice by themselves, but when arranged together with the vowels they constitute [=produce?] voice" (in contrast to the vowels, the Greek name of which is "φωνήεντα|vowels (lit.: those that have voice)"); nevertheless, Thrax acknowledges the existence of some amount of "voice" in some of the consonants (ζ, ξ, ψ, λ, μ, ν, ρ, σ), which he calls "ἡμίφωνα|semi-sonants or semivowels (lit.: semi-voiced)", whereas all others (β γ δ κ π τ θ φ χ) are called "ἄφωνα|mutes (lit.: voiceless)".It should be noted here that the statements of the Greek Grammarians may bear no evidential value for the pronunciation of classical Attic (which precedes them by 4 centuries or more); it has also been pointed out that we cannot even be sure which pronunciation (if any) the Grammarians actually described. The statements quoted here, when they agree with the present pronunciation, are indications that the respective "modern-Greek" (i.e., orthodox) values were also valid in certain contexts and during a certain era, most likely the Hellenistic or Greco-Roman times of the Grammarians. The former correspond to modern "auxiliaries" and the latter to modern "basic consonants".
It is often argued that the term ἡμίφωνα refers to the continuants, whereas ἄφωνα refers to the stops (a.k.a. plosives). However, there is no evidence whatsoever that this was the intension of Thrax. The etymology of these terms is clearly linked to "φωνήν|voice" and does not at all allude to duration or time. The explanations provided by Thrax also lack the slightest hint that ἡμίφωνα are the protractible sounds: "ἡμίφωνα δὲ λέγεται, ὅτι παρ᾽ ὅσον ἧττον τῶν φωνηέντων εὔφωνα καθέστηκεν ἔν τε τοῖς μυγμοῖς καὶ σιγμοῖς|they are called semi-voiced because although they are less well-voiced than the vowels they come forward as mumblings and hisses"; in other words, their "special status" is attributed by Thrax to the fact that they sometimes relate to some unintelligible sounds that might be produced by humans (I suspect something like mmmmmm, sssssss, etc). The explanation of the ἄφωνα is even less suggestive of explosiveness: "ἄφωνα δὲ λέγεται, ὅτι μᾶλλον τῶν ἄλλων ἐστὶν κακόφωνα, ὥσπερ ἄφωνον λέγομεν τὸν τραγῳδὸν τὸν κακόφωνον|they are called voiceless [=mutes] because they are more cacophonous than the others, just as we say that a cacophonous tragedian is voiceless [=mute]"; the explanation relies merely on a (subjective) sense of euphony, rather than on occlusion of the vocal tract. The definitions provided by Halicarnasseus (ROBE10, pp. 138, 140) are equally hazy and probably contrary to the continuant/stop theory: "ἃ μὲν καθ' ἑαυτὰ ψόφους ὁποίους δή τινας ἀποτελεῖν πέφυκε, ῥοῖζον ἢ σιγμὸν ἢ μυγμὸν ἢ τοιούτων τινῶν ἄλλων ἤχων δηλωτικούς· ἃ δ' ἐστὶν ἁπάσης ἄμοιρα φωνῆς καὶ ψόφου καὶ οὐχ οἷά τε ἠχεῖσθαι καθ' ἑαυτά· διὸ δὴ ταῦτα μὲν ἄφωνα τινὲς ἐκάλεσαν, θάτερα δὲ ἡμίφωνα|those that by their nature give rise to some noise or other, a whizzing, a hissing, a murmur, or suggestions of some such sounds; those that are devoid of all voice or noise and cannot be sounded by themselves; for this reason, some have called these 'voiceless' and those 'semi-voiced' [Roberts' translation]", "ἡμίφωνα δ' ὅσα μετὰ μὲν φωνηέντων αὐτὰ ἐαυτῶν κρεῖττον ἐκφέρεται, καθ' ἑαυτὰ δὲ χεῖρον καὶ οὐκ αὐτοτελῶς· ἄφωνα δ' ὅσα οὔτε τὰς τελείας οὔτε τὰς ἡμιτελεῖς φωνὰς ἔχει καθ' ἑαυτά, μεθ' ἑτέρων δ' ἐκφωνεῖται|semi-vowels to all which are pronounced better in combination with vowels, worse and imperfectly when taken singly; mutes to all which by themselves admit of neither perfect nor half-perfect utterance, but are pronounced only in combination with others [Roberts' translation]"; it is difficult to perceive this description as specifying that the continuants (such as [m] and [s]) can be uttered with great difficulty without an accompanying vowel and that the stops cannot be uttered at all. In view of that, the assertion that these terms describe the distinction between continuants and stops follows only from an a priori acceptance that ζ, ξ, ψ, λ, μ, ν, ρ, σ are continuants and β γ δ κ π τ θ φ χ are stops.
There is 87.5% agreement between Orthodox and Catholics on the values of the auxiliaries; that is, they agree on the values of 7 out of the 8 respective letters, ζ being the only point of discord. Although the terms were not used by the early Grammarians, the auxiliaries may be classified into "ῥινικά|nasals" (μ, ν), "γλωσσικά|linguals" or "ὑγρά|liquids" (ρ, λ) and "συρριστικά|sibilants" (σ, ζ, ξ, ψ). The Grammarians do identify ζ, ξ, ψ as "διπλᾶ|doubles" and μ, ν, ρ, λ, σ as "ἁπλά|singles".
The agreement rate on the basic consonants is lower: 33.3%, or 3 out of the 9 respective letters, β, γ, δ, κ, π, τ, θ, φ, χ. There are two ways to divide those: one based on the place of articulation and one on the manner of articulation. The classification of the modern sounds is illustrated in the following table:
| modern | labials | dentals | gutturals (velars) | Grammarians |
| voiceless stops | π | τ | κ | ψιλά|tenues (lit. thin or plain) |
| voiced continuants | β | δ | γ | μέσα|mediae (i.e., middle) |
| voiceless continuants | φ | θ | χ | δασέα|aspiratae (lit. thick or dense) |
Each column from left to right corresponds to a different place of articulation from outer to inner mouth. The terms "labials" etc are not used by the early Grammarians, but the descriptions provided by Halicarnasseus (ROBE10, pp. 148-150) clearly reflect the above (and, hence, the modern) classification.cf. Roberts' observation (ROBE10, p. 149, n. 15): "Dionysius does not actually use Greek equivalents for the adjectives labial, dental, and guttural ; but he clearly knows the physiological facts in which those terms have their origin". Each row is related to a different manner of articulation. The present properties of each row are listed on the left, with the distinctive feature in bold, whereas the name assigned to each triplet by the ancient Grammarians is listed on the right. There is coincidence of Orthodox and Catholic values only for the first row, the consonants called by the Greek Grammarians "ψιλά" and by the Roman ones "tenues". The term "ψιλά" has the general meaning of "simple", "bare" or "unadorned", obviously referring to the absence of one or more features present in the consonants of the other two rows (in the Orthodox phonological model, the missing features would be voice and duration). Thus, κ, π, τ appear to be "plain vanilla" consonants, which is consistent with their present values, namely voiceless stops.
After the above analysis of the Grammarians' testimony and the classification of the basic consonants, there is not much to add about the tenues, Κ, Π, Τ. One can hardly find any evidence suggesting different values from those they have today.We have seen seen that Sturtevant establishes (STUR20, pp. 98-101) a full correspondence of latin B, D, G with greek Π, Τ, Κ, respectively, in terms of "energy"; however, this conclusion is based on examples from some Romance languages and from some transliterations of proper names between (unofficial) Latin and Greek. His conclusion does not seem to be endorsed by subsequent researchers, such as Allen, and does not explain what the reason might be to use P, T, C in the vast majority of other borrowings, if B, D, G better corresponded to Π, Τ, Κ, not to mention that the feature referred to by the terms "fortis" and "lenis" are not sufficiently defined, to enable a better phonetic description of the corresponding sounds (perhaps intentionally so, so that everyone interpret it as everyone sees fit). Fluctuations between consonants of the same class (labial with labial, dental with dental, velar with velar) are normal in the case of foreign borrowings (as we have seen), without having to assume any metaphysical property ("energy") shared between the evidently different sounds used at both sides of the transliteration ([p]→[b], [t]→[d], etc). Their orthodox (i.e., present) values are, therefore, not disputed and are assumed to always have been valid since Greek was Greek (and even before that, in the times of PIE).
This is to say, Π was a voiceless (bi)labial stop:
ΠOC: Π = [p]
Τ was a voiceless dental stop:
ΤOC: Τ = [t]
and Κ was a voiceless velar stop:
ΚOC: Κ = [k]
There is virtually no evidence for a voiced value of π, τ, κ after the nasals μ, ν, γ.In modern Greek, progressive assimilation of voice has resulted in the cluster μπ standing for any of [mb], [b] or rarely [mp], the cluster ντ similarly representing [nd], [d] or rarely [nt] and γκ being pronounced [ŋg], [g] and almost never [ŋk], which led to the establishment of these digraphs as renderings of the sounds [b], [d], [g] mainly in words of foreign origin. The origins of this voicing are not known. A few variants of the type Ἀμβρακιῶται↔Ἀμπρακιῶται (cf. IG II² 236 and IG II² 403) are too isolated to draw conclusions and may be explained by different assumptions (e.g, that π=[p] and β=[b], as per catholic model, π being particularly prone to voicing between the two voiced "liquids" μ and ρ; or that π=[p] and β=[v], as in modern Greek, both being imperfect approximations of a "foreign" [b]). Similarly, for ἐντελέχεια and ἐνδελεχής Blass appears to be right: "the word must have been ἐνδελέχεια, but being of infrequent usage it was remodelled on the analogy of τέλος" (BLAS90, p. 97). However, Blass is too dogmatic in considering the voicing of π, τ, κ in ancient Greek an impossibility: to his question "how could the ancients have kept ἐντός and ἔνδον, ἀναφανδόν and -φαντο- so strictly distinct, as they certainly did?" the simple answer is "as they do today, by pronouncing one as a voiced stop and the other as a voiced continuant". Blass' question is based on the common identification ancient-Greek δ = modern-German/English d, but if we overcome such prejudice and allow for the possibility of δ in its present value ([ð]), we will realise that there is no secure way of establishing the possibility of post-nasal voicing either to the positive or to the negative.
Alveolars
We have seen that Sturtevant comments on the description of the basic consonants by Halicarnasseus that "this passage does not help us to decide whether δ, θ, and τ were alveolars or dentals". So, an "alveolar" value for τ is not to be a priori excluded. Both Sturtevant's and Allen's verdict is in favour of a "dental" value, based on its modern value, as well as transcriptions in Indian/Prakrit.
There is, however, one characteristic of East Greek (including Attic), which may suggest a more alveolar value for τ: "the characteristic innovatory 'assibilation' of original [t] before [i] (i.e. [ti] > [tsi] > [si])", e.g., -κάτιοι→-κόσιοι, Ἀφροδίτιος→Ἀφροδίσιος, Ἀρταμίτιος→Ἀρτεμίσιος (HORR10, pp. 19-20). Of course, the palatal vowel [i] can "pull" a dental τ towards an alveolar realisation, as it has "pulled" the velar (or at least palatal) c towards an alveolar sound ([s]) in Romance. On the other hand, the fact that West Greek did not exhibit this kind of assibilation and East Greek did would be more easily understandable, if the realisation of τ was closer to the palate in East Greek than in West Greek, at least before [i]. Certainly, the evidence is not conclusive, probably not even suggestive of an alveolar value, and the issue may further be an insignificant detail (I do not known many Greeks who can tell the difference between "true dentals, as e.g. in French, and [...] alveolars as in English"; ALLE87, p. 17).
All in all, the distinction between dentals and alveolars is probably more resolution than one might expect to derive for the ancient phonology.
Palatals
In present-day Greek, Κ has a palatal allophone [c] in front of [i] and [e] (the so-called "front vowels"), a value that is certainly used by modern Greeks when reciting ancient texts:
ΚO/: Κ/_Vfront|front vowel, i.e., ε or ι, also η, υ, ει, οι, αι = [c]
To the question whether the same kind of allophony was a feature of ancient Greek, Allen asserts that "There is no evidence in ancient times for the 'palatalized' pronunciation of κ as [ky] before front vowels which is normal in modern Greek" (ALLE87, p. 17). In addition to not providing any substantiation together with this statement, the wording is indicative of the westerners' perplexity with regard to the greek palatals.No statement is more representative of how unsuitable the Popes of Catholicism are for speaking about the Greek language than Blass' description of the Greek [c]: "in the καί of the present day a sound is heard somewhat like kye, in which the k is produced so far forward on the palate, that it approximates to t" (BLAS90, p. 98). In the history of "velar palatalisation", there is no resultant sound that "approximates to t" (not considering the Spanish [θ] and the Attic ττ, which are results of further "fronting", after the initial palatalisation); even Italian [t͡ʃ] does not qualify as an approximation of t. To describe [c] (the first step of velar "fronting") as a sound that "approximates to t" is a disgrace for a phonologist, even of the 19th century. The sound he describes as [ky] is actually the voiceless palatal stop (as already pointed out, his reluctance to use the IPA is regrettable for a professor of phonology) and his notation is indicative of some professionals' inability to grasp the difference between palatals and sequences of consonants (most often velars) with the front (or "palatal") semivowel [j].In fact, many Greeks consider [c] and the other modern-Greek palatals ([ɟ], [ç], [ʝ], [ɲ], [ʎ]) not to be single phonemes, but to stand for κι, γκι, χι, γι, νι, λι. This misconception is, no doubt, caused by the orthography, which has no way to represent the palatals other than as combinations of a velar (or ν, λ) with [i], as well as by the etymology, since the palatals in many cases are results of monophthongisation (or coarticulation) of κι̯, etc before a vowel. His placement of the term "palatalized" (curiously with the american spelling) between quotation marks implies that he does not consider the transformation [k]→[c] before front vowels as proper palatalisation; this is understandable, since western linguists fail to recognise a palatalised velar until it has become assibilated; this is the reason why the term "palatal" among them almost exclusively means post-alveolar sibilant, usually affricate ([t͡ʃ], [d͡ʒ]).
Apart from being phonologically embarrassing, Allen's assertion is furthermore not true. The first hint towards a diverging conditional value (before front vowels) is provided by Allen himself, a few lines before the blatant assertion: "As in many languages, the precise point of articulation of the velar series may have varied to some extend according to the following vowel, i.e. further forward before a front vowel and further back before a back vowel"; in other words, not only is a more forward value a possibility, it is also the normal state of phonology in many (if not all) languages (something that has already been pointed out again and again).
The second hint is also provided by Allen: "in the oldest Attic inscriptions one finds before the vowel ο the symbol Ϙ (κόππα), which had represented the Semitic uvular plosive [q] ('qof')". He considers this orthographic variant to be indicative of velar fronting before front vowels, i.e., an indication of a "back value" of κ before ο and a "front value" before other vowels (including the "front vowels").Blass offers us yet another... entertaining "scientific" moment when he declares that "The syllables κο κρο κτο were written with Ϙ, because the letter was called koppa, κα κρα etc. were on the other hand written with kappa for the same reason" (BLAS90, p. 98). He does not, then, try to explain why, e.g., σα στα were not written with san and σι στι with sigma. One may rightfully wonder what kind of back and front allophones Allen had in mind, if not [k] and [c]. The only other possible pair would be velar [k] vs uvular [q]; but a (normal) back value [q] would be incompatible with the previous (the PIE value was allegedly [k]) and later (i.e., modern) stages of the language. The only reason for considering an uvular value would be because Ϙ corresponds to "the Semitic uvular plosive [q] ('qof')"; however, this assessment of Allen's is premature, as it has not been established that qoph had an uvular value in Phoenician (in fact, qoph has a multitude of different realisations in the various modern semitic languages) and as Ϙ would not necessarily have the same value as qoph had in original Phoenician (cf. the values of Θ and Σ vs their phoenician counterparts). Admittedly, if the use of the two letters Ϙ and Κ was indeed indicative of an allophonic pronunciation, one would expect that a "back value" would apply not only before ο (the only back vowel of the old-Attic alphabet), but also before α (a central vowel), as it does today. For this reason, I consider it possible that the use of Ϙ before ο may not reflect a different place of articulation, but may be due to tradition, prejudice or even a different articulatory feature, e.g., lip rounding, which accompanies the back vowels.It appears that I am not the first to associate Ϙ with lip rounding. Miller believes that "the evidence is rather that qoppa was associated with lip-rounding". However, his argumentation is highly speculative, as it is largely based on similarity in shape between Ϙ and Φ, as well as some Linear-B symbols (!) representing two labiovelars and a velar (+vowel), it assumes a value of kw(h) (!) for Ϙ when the Greeks borrowed the alphabet, sound shifts from labiovelar to labial, re-assignment of letters (Π and Φ), re-borrowing of Ϙ from Phoenician qof... It is certainly an imaginative account, but quite an unnecessary one. It suffices to observe that Ϙ is to be found only before Ο, as well as Υ in Doric, the only common property of (the most likely values of) which is lip rounding (backness being shared also with Α, which does not receive Ϙ); in Miller's words, "the Greeks were quite fond of the use of qoppa in the context of lip-rounding". At least, it is positive that scientists are no longer thinking along the modern-Semitic lines (where often ק=[q]), as Allen appears to be doing. On the other hand, this practice is very reminiscent of the corresponding Latin practice of using C, K, or Q before I/E, A/consonant, or O/V. Again, the three-way distinction in Latin affords different interpretations, one being that K stood for the normal velar plosive, Q a labio-velar (due to the lip rounding of the vowels of the back axis) and C a "fronter" version (palatal or post-alveolar affricate, as reflected in modern Romance). All in all, the use of Ϙ before ο in Attic may not be an indication of a different place of articulation, but if Allen believes it to be one, then a velar-palatal distinction is the most likely differentiation.
A third hint comes from internal evidence that caused the emergence of the cluster σσ (attic ττ), which the philologists tell us was derived i.a. from "original" *κι̯, e.g., *φυλάκ-jω [phulákj-jo:] → φυλάττω, wherein [kj] > [tʃ] = [ttj] in Boeotian and [tt] in Attic (HORR10, p. 19). It is strange that κ before j (essentially a short, non-syllabic /i/, ι̯) is readily accepted to have been fronted all the way to the teeth or at least the alveolar ridge, but before the full vowel ι it would not even budge forward to a palatal value.
Finally, tradition is uniformly favouring a palatal value, as there is no modern dialect that features a velar value before front vowels, most having the standard-Greek palatal value [c] and some (most likely under foreign influence, as in Crete) an affricate [t͡ɕ], [t͡ʃ] or [t͡s], i.e., an even "fronter" value.
To summarise the above, a front value of a velar before front vowels is the normal realisation for all languages, an orthographic variant possibly due to difference of articulation is attested, extreme fronting of velars before the front semivowel is etymologically assumed in the case of σσ/ττ and there are no traces of a non-palatal value in modern dialects (including Tsakonian, which has evolved down a different path). Yet Allen sees no evidence at all for (what he thinks is) "the 'palatalized' pronunciation of κ as [ky] before front vowels". One would like to ask him what (other) kind of evidence he would consider satisfactory, but it is clear from his wording and notation that he does not understand the feature he is looking for. Unfortunately, the same goes for many catholic researchers, whose limited anglo-germanic phonological experience prevents them from grasping the peculiarities of the very object of their research (we have already seen another case of Allen's unfamiliarity with greek phonology, where he confuses γ's modern value [ɣ] with that of its palatal allophone [ʝ], which he further misrepresents as the front semivowel [j]; [y] in his notation).
All evidence considered, the orthodox conditional realisation of Κ as a palatal [c] is very likely also for the ancient language, albeit in no way proven beyond doubt.
The modern value of Μ is [m], a (bi)labial nasal. There is no evidence suggestive of a different value in any ancient Greek dialect. On the contrary, its labial and nasal properties are clearly reflected in the description provided by Halicarnasseus (ROBE10, pp. 144): "τὸ δὲ μ τοῦ μὲν στόματος τοῖς χείλεσι πιεσθέντος, τοῦ δὲ πνεύματος διὰ τῶν ῥωθώνων μεριζομένου|μ [is pronounced] by the mouth being closed tight by means of the lips, while the breath is divided and passes through the nostrils [Roberts' translation]". The letter that it was derived from (i.e., Semitic מ) and those that were derived from it (such as Roman M and Cyrillic М) stand (today) for the same value ([m]). Furthermore, μ is the only nasal found in front of the "labial" consonants (see above) inside words and inter-word assimilation (=sandhi) of final ν before labial consonants (see below) results in the conversion of ν to μ.It is true though that, in ancient inscriptions, "the etymological spelling [i.e., original ν, instead of μ, as the result of assimilation] is often retained in both positions", i.e., "in the interior and at the end of a word" (STUR20, p. 167). This may merely mean that the stone cutter was aware of etymology. There is, thus, consensus among both Orthodox and Catholics:
ΜOC: Μ = [m]
The same reasons that lead us to adopt the Orthodox value as the only candidate for μ lead us to a similar adoption for ν: the dental (actually alveolar) nasal [n]. The description of Halicarnasseus reflects its nasality, but not expressly the place of articulation (ROBE10, pp. 144): "τὸ δὲ ν τῆς γλώττης τὴν φορὰν τοῦ πνεύματος ἀποκλειούσης καὶ μεταφερούσης ἐπὶ τοὺς ῥώθωνας τὸν ἦχον|ν [is pronounced] by the tongue intercepting the current of the breath, and diverting the sound towards the nostrils [Roberts' translation]"; the obstruction of the mouth cavity, a necessary condition for the diversion of the airstream through the nostrils and the production of a nasal consonant, is described as achieved by the tongue, but the exact point of contact is not specified; the tongue might block the airstream at several points in the mouth, specifically at the teeth or gums ([n]), at the hard palate ([ɲ]) or at the soft palate ([ŋ]).Allen sees (ALLE87, p. 33) a "dental [n]" value "clearly described" in the following passage of Halicarnasseus: "τοῦ μὲν γὰρ ν περὶ τὸν οὐρανὸν γίνεται ὁ ἦχος καὶ τῆς γλώττης ἄκροις τοῖς ὀδοῦσι προσανισταμένης καὶ τοῦ πνεύματος διὰ τῶν ῥωθώνων μεριζομένου|ν is sounded on the arch of the palate, with the tongue rising towards the edge of the teeth and with the breath passing in separate currents through the nostrils [Roberts' translation]" (ROBE10, p. 220). While it does mention the tongue and the teeth cooperating for the production of ν, the conclusion that this passage clearly describes [n] might be premature. First of all, it specifies that the sound is produced around the palate, a very vague description, which would also apply to a palatal or velar nasal and possibly also to any kind of nasal (including labial), as it appears to specify nothing more than the fact that the mouth cavity reverberates the sound. Then, the interaction of tongue and teeth takes place at a location least expected for a dental nasal: "ἄκροις τοῖς ὀδοῦσι" evidently refers to the lower ends of the teeth, not their roots near the alveolar ridge; placing the tongue at the edge of the teeth, as in [θ] and [ð], produces no complete obstruction of the airstream and, thus, no clear nasal. Finally, the passage referred to relates to the specific ν in the expression "κλυτὰν πέμπετε" (p. 218, l. 22 ff.), namely the description may only relate to a word-final ν that is at odds with the first consonant (π) of the following word. Furthermore, in the previous paragraph (p. 218), Halicarnasseus states that χ and ν (in the expresion "ἐν χορόν") are also incompatible, which seems to exclude a velar value for ν, but at the end of the same paragraph (p. 222), he fails to make a positive connection of ν with the dental θ (claiming that it is unnatural for a semivowel to stand before any mute!). As a consequence, this chapter cannot be considered to constitute a robust description of any sound, much less of a dental nasal. The fact that in normative spelling ν escapes assimilation only before dental mutes, together with its modern value and those of the related letters in other alphabets (נ, N, Н, etc) tip the scales in favour of the dental/alveolar value:
ΝOC: Ν = [n]
Agma
Although Halicarnasseus' failure to specify the exact place of articulation of ν may be due to mere negligence, it is also possible that it was deliberately left unspecified, in order to also include a possible velar value.There is no evidence for a palatal value of ν. Presently, a palatal ν ([ɲ]) is only produced in combination with a non-syllabic /i/ (i.e., as νιV, νειV, νοιV, etc, V being any vowel sound) in official Greek, and together with a syllabic /i/ (i.e., in syllables νι, νει, etc) only dialectally; hence, today the sound [ɲ] is considered to correspond to the digraph νι. There appears to be no way to verify whether this was also the case in ancient Greek. In compound words, wherein the first word ends in ν (which is the only nasal allowed as word final) and the second begins with one of the mutes, the ν undergoes assimilation according to the mute's place of articulation: if the mute is a labial, ν becomes a labial nasal (cf. syn+pathos→sympathy, συμπάθεια) and is rendered as μ; if the mute is a dental, ν does not change (cf. syn+taxis→syntax, σύνταξις), as it is already a dental; if the mute is a velar, an analogous assimilation is expected to take place producing a velar nasal ([ŋ]), as is actually the case in many languages, including modern Greek and English. The latter assimilation is not always reflected in scripts that lack a symbol for the velar nasal (as the velar nasal is not an independent phoneme in the respective language, i.e., it is only found before a velar and never in other environments, such as between vowels); thus, English retains the "n" spelling even though the pronunciation is not that of a dental nasal, as did ancient Greek in early inscriptions (STUR20, p. 168), but from V BC until today the ν is (graphically) converted to a γ to indicate the difference in pronunciation (cf. syn+kope→syncope vs the Greek spelling συγκοπή).
This change of orthographic convention (from ν to γ for nasals before velar mutes) makes us suspect that since V BC and, most likely, much earlier (if not always) the only nasal before velars was a velar nasal. This almost certain value [ŋ] was, then, an allophone of the dental nasal /n/ and as long as ν was used to denote it (mainly before V BC), it was its second, conditional value:
ΝOC/: Ν/_(#|end of word)Cvelar|velar consonant, i.e., γ, κ or χ, also ξ = [ŋ]
The (#) denotes that the assimilation of the nasal into a velar may (but not always) take place even if the ν and the velar are in different words, mostly when one of the words is an enclitic or a proclitic.
The contemporary sound of λ is an "alveolar lateral approximant":I have never understood why this sound is an "approximant". It is probably a technicality, since the definition of "approximant" relies on the absence of "turbulence", but I do not intend to present my objections here (for one, the fact that the associated "fricatives", either voiced or voiceless, do not sound very "lateral", but seem to have a central airway, as sibilants do) and I will abide by the established terms. This is also the (present) value of its predecessor ל in the various Semitic languages. We have no reason to suspect a different value for Attic. Halicarnasseus only tells us that tongue and palate cooperate in its production, but not how: "τὸ μὲν λ τῆς γλώττης πρὸς τὸν οὐρανὸν ἱσταμένης καὶ τῆς ἀρτηρίας συνηχούσης|λ by the tongue rising to the palate, and by the windpipe helping the sound [Roberts' translation]" (ROBE10, p. 144). No "lateral" quality is alluded at, but, in the face of its present value, we cannot imagine λ standing for anything other than a lateral consonant:Theoretically, it could be something like the Japanese flap [ɽ], but that would hardly justify a distinction with ρ, which rather clearly corresponds to a... rhotic consonant.
ΛOC: Λ = [l]
There is no evidence in Attic or common Greek for a so-called "dark l" (actually velarised l) sound ([ɫ]), which appears to have existed in Latin (STUR20, pp. 79-80) and certainly existed in some Romance languages (cf. autre[FRA], otro[SPA] vs altro|(an)other[ITA], also steaua[ROM|Romanian] vs stella|star[ITA]); only dialectally is marginal evidence to be found (STUR20, pp. 166-167) and in modern Greek a "dark l" is only found among refugee populations originating in Anatolia (hence, under heavy Turkish influenceThis renders moot Allen's assertion (ALLE87, pp. 40-41, n. 69) that "an Asiatic Greek peculiarity" of velarising λ near back vowels may be reflected in the fact that Armenian dark l "tends to be used to transcribe λ in Greek words, more particularly in the vicinity of non-front vowels" and in the fact that "modern Cappadocian Greek shows developments of a labial or velar nature in such contexts"; Cappadocian is the Greek dialect with the heaviest Turkish influence, a language that is known to have "[l] [...] in complementary distribution with [...] [ɫ]; the former [occurring] adjacent to front vowels and the latter adjacent to back vowels"; the Armenian transliterations are (as Allen reports) inconsistent, raise the question "how do we know the (original) Armenian values for the two kinds of [l]?" and may well reflect the (not-uncommon and dialectally surviving) alveolar/palatal distinction of λ before back/front vowels, respectively.).
Similarly to ν, a palatal λ ([ʎ]) is presently produced when λ is combined with a non-syllabic /i/ (i.e., as λιV, λειV, λοιV, etc, V being any vowel sound) in official Greek, or with a syllabic /i/ (i.e., in syllables λι, λει, etc) only dialectally (as a consequence, the sound [ʎ] is considered to correspond to the digraph λι). It is not clear how it could be determined whether such a palatalisation was also a feature of Attic or common Greek.A /ʎ/-/ʝ/ merger, similar to Castilian ll/y merger, could provide such an indication, but such a merger never took place in Greek. Neither am I aware of any instances of confusion of λι and γι (the modern rendering of [ʝ]) or of omission of λ in λι (which would be an acceptable approximation of [ʝ]). Even today, [ʎ] and [ʝ] are sufficiently distinct sounds that (unlike Castilian [ʎ] and [j]) are never confused, just like in Italian, Catalan, etc. In absence of positive evidence, we will not consider the possibility of a palatalised λ.
Presently, ρ is an alveolar trill (like a "heavy" Scottish r). The Semitic letter ר that spawned it has various rhotic realisations today; of these, only the alveolar ones ([r] and [ɾ]) seem relative, as the uvular values ([ʁ] and [ʀ]) of Hebrew are obviously due to a germanic (i.e., yiddish) influence. An alveolar value is also clearly described by Halicarnasseus: "τὸ δὲ ρ τῆς γλώττης ἄκρας ἀπορριπιζούσης τὸ πνεῦμα καὶ πρὸς τὸν οὐρανὸν ἐγγὺς τῶν ὀδόντων ἀνισταμἐνης|ρ by the tip of the tongue sending forth the breath in puffs and rising to the palate near the teeth [Roberts' translation]" (ROBE10, p. 144). Some manuscripts, instead of "ἀπορριπιζούσης τὸ πνεῦμα", i.e., sending out the airstream in gusts, have "ἀπο(ρ)ραπιζούσης τὸ πνεῦμα", i.e., beating back the airstream (cf. ROBE10, p. 144, n. 24 and ALLE87, p. 41); while the former expression rather corresponds to a(n alveolar) trill, i.e. [r], the latter would be more appropriate for a tap, i.e. [ɾ]. Thus, of all the possible rhotic values, only the alveolar trill and tap appear to be consistent with the above evidence. We have no evidence suggestive of uvular or approximant values, except for cases of defective pronunciation, the most famous being Demosthenes (cf. ROBE10, pp. 145-146, n. 25). The following reference of Socrates to this letter rather tips the scales in favour of a trill: "ἑώρα γὰρ οἶμαι τὴν γλῶτταν ἐν τούτῳ ἥκιστα μένουσαν, μάλιστα δὲ σειομένην|for he observed, I suppose, that the tongue is least at rest and most agitated in pronouncing this letter [Fowler's translation]" (PLAT94, pp. 190, §426e). This is the value on which both Orthodox and Catholics agree:
ΡOC: Ρ = [r]
ῥ-, -ῤῥ-
There is an established orthographic tradition of writing a "δασεῖα" (a term that is usually translated as "aspirate", "rough breathing" or "spiritus asper" and the nature of which will be investigated later) over an initial ρ or over the second of a pair of ρ (the first receiving the "ψιλή"), i.e., ῥ and ῤῥ. The first explicit testimony of this practice comes from Herodian: "Τὸ ρ ἀρχόμενον λέξεως δασύνεσθαι θέλει, ῥά, ῥανίς, ῥάξ|ρ that starts a word must receive the 'δασεῖα', [as in] ῥά, ῥανίς, ῥάξ" and "Τὸ ρ, ἐὰν δισσὸν γένηται ἐν μέσῃ λέξει, τὸ μὲν πρῶτον ψιλοῦται, τὸ δὲ δεύτερον δασύνεται, οἷον συῤῥάπτω|if ρ becomes double in the middle of a word, the first receives the 'ψιλή' and the second receives the 'δασεῖα', such as συῤῥάπτω" (e.g., STUR20, p. 164). It should be noted that the testimony is rather late, from the time of the Antonines, and does not necessarily reflect a feature of classical Attic; nevertheless, everybody is eager to ascribe it to "Classical Greek" or to "Attic phonology" (cf. also ALLE87, p. 41, "classical Attic"). There are next to no hints for the same phenomenon in Attic: according to Threatte (THRE82, p. 25), a single word, Φρεά(ρ)ρhιοςIn the transcriptions, an h denotes an eta (Η) assumed to be used in its "consonantal value", i.e., representing the "δασεῖα", as it did in the old-Attic alphabet before the adoption of the Ionic alphabet (in which it has a vocalic value). The actual text of the inscription must have been something like ΦΡΕΑΡΗΙΟΣ, which should have been quoted, but I will occasionally use Threatte's and Allen's convention of using lower-case letters. (apparently a proper name), is the only example in Attic, with a single partial (hence, doubtful) instance in an inscription and the rest on ostraca from the Agora; there is no reported instance of word-initial ῥ in Attica. A couple of further examples from other Greek dialects (invariably Corcyrean ρhοϜαισι and Boeotian hραφσα[Ϝοιδοι; cf. STUR20, p. 165, THRE82, p. 25, ALLE87, p. 41) are cited as "evidence" for the theory's classical-time basis. To reinforce the argument, analogous instances of μh and λh are cited (e.g., THRE82, pp. 25-26), which allegedly indicate a similar variation in the pronunciation of word-initial μ and λ. The most serious evidence, however, is the Roman convention of transliterating word-initial ρ and the second of a pair of ρ's (i.e., those that Herodian suggests to spell ῥ) as rh, e.g., rhetor, Pyrrhus, etc; according to Threatte (THRE82, p. 25), the reverse transcription is also rarely attested, as "hr occasionally occurs instead of rh". The similar Armenian transliteration hr for ῥ (e.g., hṙetorThe actual spelling must have been ՀՌԵՏՈՐ. Note the different r's (Ռ and Ր) used to transcribe the first and last letters of ῥήτωρ (a difference not reflected in Sturtevant's Latin transcription hretor), yet the first is further preceded by Հ ([h]). Thus, the first r (ՀՌ) is not just an "aspirated" variant of the second (Ր), but a different letter, or possibly digraph.) is immaterial, as it postdates Herodian by more than two centuries and, as Sturtevant concludes (STUR20, p. 165), "it is likely that the h records a mere school tradition" (possibly based on the Grammarians' theories and/or the Latin spelling).
What does all this mean for the pronunciation of ῥ? Possibly nothing. It is clear that Herodian's statement refers to the script, namely that a δασεῖα must be used in writing, not necessarily to the pronunciation. However, one can hardly imagine that the Grammarians would prescribe spelling rules without reason, which would most likely be a difference in pronunciation.Only Threatte recognises (THRE82, p. 26) that this odd spelling "is likely to be a clue to actual pronunciation" (emphasis mine); everybody else takes it for granted that a difference in script, even if occasional (and, hence, liable to be a misspelling, etc), is indicative of a difference in pronunciation. In order to determine what kind of a difference this could be, we need to evaluate the other "evidence".
As already explained, the inscriptional evidence is far from conclusive. The only Attic example, ΦΡΕΑΡ(ΡΗ)ΙΟΣ, concerns one of the deme of Attica named Φρεάρριοι and is (almost?) exclusively encountered in connection with the name of Themistocles on several ostraca. I have only found two images of such ostraca, neither of which bears any sign of Η (the "aspiration"): the first one reads ΦΡΕΑΡΡΙΟ[, while the respective spelling on the second appears to be ΦΡΕΡΙΟΣ (or ΘΡΕΡΙΟΣ)! From the latter, it is evident that the spelling ΦΡΕΑΡ(Ρ)ΗΙΟΣ cited by Threatte, even if it can be clearly discerned on its ostracon, is not to be taken as indicative of the actual pronunciation (otherwise ΦΡΕΡΙΟΣ could also be taken as indicative of a... silent Α). Furthermore, since all Attic instances of (Ρ)ΡΗ cited by Threatte concern the same word, the Attic evidence can at most establish a special pronunciation for the particular word ΦΡΕΑΡΡΙΟΣ, but not necessarily for any instance of ρρ, much less for word-initial ρ.This peculiarity of the word Φρεάρριος may have to do with its etymology or paretymology from the highly irregular word φρέαρ (nom.), φρέατος (gen.), the derivatives of which should have the stem φρεατ- (e.g., φρεατιαῖος). The other two (non-Attic) instances are hapax legomena, namely too few (the Boeotian one is actually partial) to safely rule out mistakes, misspellings or misunderstandings and to declare ρh a pronunciation-based spelling.
Regarding the cases of word-initial μh and λh mentioned by Threatte (THRE82, p. 25: "Examples of μh are fairly frequent; λh is rare"), who attempts to link all cases of ρh, μh and λh together as indicative of the same kind of variation in the pronunciation of the sonorants, they may be less relevant than they appear at first sight. All instances of μh cited by Threatte are actually followed by ε, i.e., the spelling is ΜΗΕ; this could be merely due to a confusion caused by the vocalic function of Ionic Η as a (long) Ε (cf. the use of Η for ἑ in the Nikandre inscription; for other examples, see THRE82, pp. 45-47). Similarly, there is only a single instance of λh not followed by ε, the name Λhάβετος, which cannot be identified as Attic or as indicative of Attic pronunciation with sufficient certainty. At any rate, since the spellings μh and λh are not normative (i.e., they never found their way into the official orthography and are, in the words of Threatte, "entirely foreign to the more regularized orthography of official documents") but rather isolated cases, one does not need to explain them further. There is also no etymological reason for a different pronunciation in any of the cited instances, except possibly for λhαβ-, which Threatte associates with *hl (THRE82, p. 26) and Sturtevant with *sl (STUR20, p. 167).
The most important evidence for ῥ are, thus, the Latin transcriptions. It might be considered that Herodian's statement, coming as it does from the late Roman period, merely reflects the Roman practice, i.e., it relates to a transfer of Latin spelling into the Greek one. There are two possible reasons why the Romans would adopt this kind of spelling: either A) Greek ρ had a different value in ρ- and -ρρ- and Latin r had the same value in all environments or B) Latin r- and -rr- sounded different from -r- and the value of Greek ρ was uniform irrespective of environment. In other words, either Greek ρ or Latin r, but not both, represented a phoneme with two allophones and the orthographic convention aimed at making up for this discrepancy. Under assumption A, the Latin spellings rh- and -rrh- indicated the different (from normal Latin r) Greek sound ῥ; in scenario B, the insertion of h would mean that general allophonic rule (of Latin r) should not be observed. The latter case is not as unlikely as it may seem: the use of an h to "override" the usual allophonic value of a letter finds a perfect parallel in present-day Italian chi, che, which indicate that the c should not be pronounced with the (assibilated) value it has in the syllables ci, ce, but with the value it has in ca, co, cu. Furthermore, tradition rather favours scenario B: Greek has no surviving traces of a diverging pronunciation for ῥ; on the other hand, one can hardly overlook the Spanish distinction between r- or -rr- ([r]) and -r- ([ɾ]), which "is not limited to the Iberian peninsula alone, but is found in much of southern Italy, in Apulia, Calabria, Sicily and southern Corsica" (LLOY87, p. 246), facts that point to the possibility of allophonic r in Latin.
A plausible course of events that would explain the aforementioned facts under scenario B is the following:
There are a few weak points in the above theory, particularly that modern tradition is against the above phonological model (the values of modern-Greek ρ and Spanish r coincide, as [r], in the environments in which they are supposed to differ and differ wherever they should have coincided, according to the theory), that the Spanish -r-/-rr- opposition is allegedly "not a carryover from Latin quantitative distinctions"Lloyd postulates (LLOY87, pp. 244-246) that the equivalence of r- and -rr- is a later development in Spanish, similar to the equivalence of t- and -tt-, all word-internal single t's having been voiced to d. The problem with this theory is that a parallel development would be justified, if r- and -rr- were voiceless and -r- voiced (just as t- and -tt- remained voiceless, while -t- was voiced to -d-), which they are not! and that there is further orthographic evidence that Greek ρ- was equivalent to geminated -ρρ-, but not to (single) word-internal -ρ-. The latter is the spelling -ρρ- in compounds with the second constituent starting with ρ- or in inflected forms of verbs beginning with ρ-,SMYT20, p. 25, §80: "An initial ρ is doubled when a simple vowel is placed before it in inflection and composition. Thus, after the syllabic augment (429), ἔ-ρρει was flowing from ῥέω; and in καλί-ρροος fair flowing." e.g.,
ἀπό + ρέω → ἀπορρέω, κατά + ρίπτω → καταρρίπτω, ε-|aorist/imperfect augment + ραπίζω → ἐρράπιζον
Contrary to, e.g., συρράπτω←συν+ράπτω, where the first ρ of -ρρ- may be considered the result of assimilation of ν by the following ρ (cf. correlate←con+relate), there is no such phonetic reason for the reduplication of the word-initial ρ in the above examples. Instead, it appears that the reduplication aims to preserve the sound of (initial) ρ- and distinguish it from (medial) -ρ-, which the original ρ- would become, had the two constituents been merely put together (†ἀπορέω, †καταρίπτω, †ἐράπιζον). Thus, this orthographic tradition is rather suggestive of the correspondence of ρ- and -ρρ- and their contrast with -ρ-.
We will, therefore, consider the generally accepted scenario A above, i.e., that Latin had a uniform value for r (probably geminated in the case of -rr-) and that Greek word-initial ρ- was equivalent to double -ρρ-, but not to single -ρ-. The main question is: "What could have been the pronunciation of ῥ (in ῥ- and -ῤῥ-)?" I have not found any Orthodox who has attempted to answer this question; they appear to accept a uniform value for ρ, irrespective of environment. The (unanimous) answer provided by the Catholics is interesting: ῥ was a "voiceless r" ([r̥]), a sound that among the European languages is attested in Welsh and Icelandic.
ΡC/: Ρ/#|word boundary_ = Ρ/Ρ_ = [r̥]
It appears that the instigator of this view was Sturtevant: having identified the aforementioned instances of ρh (Corcyrean), hρ (Boeotian), rh (Latin) and hr|actually hṙ (Armenian), he concludes (STUR20, p. 165) that "the aspiration did not either precede or follow the ρ, but accompanied it throughout; ρ initial, after aspirates, or double was pronounced with aspiration instead of with voice; that is it was whispered" (emphases mine)! Although the wording "it was whispered" is clumsy and this alleged value of ῥ is not identified with the "voiceless r" of any other language, the expression "pronounced with aspiration instead of with voice" suggests that it lacked voice (if the term "voice" is intended by Sturtevant in the modern linguistic sense of modal voice) and further down in the same paragraph it is referred to as "a voiceless sound". Subsequent works on greek phonology cite Sturtevant's conclusion that the "aspiration" was simultaneous with the ρ and conclude that "ρh was actually some kind of voiceless [r]" (THRE82, p. 25) or "that the sound was a 'breathed' or voiceless [r]" (ALLE87, p. 42).
Sturtevant and his successors resort to the (questionable) secondary evidence (Corcyrean and Boeotian inscriptions, Latin and Armenian transcriptions) to justify the conclusion that the "aspiration" (whatever this term stands for) "did not either precede or follow the ρ, but accompanied it throughout". But this fact is patent from the primary evidence, i.e., the established greek orthography (since we have assumed that it was Greek that had the allophony and not Latin): the "aspiration" (="δασεῖα") is written over ρ, not before or after it.One may argue that "aspiration" was also written over the vowels, while it is generally accepted that in pronunciation it preceded the vowel. However, it would be premature to make this assumption without having investigated (in the course of the present research) the relevant evidence. Nevertheless, to provide a short answer, we can anticipate the relevant discussion and make the following observation: after the adoption of the Ionic alphabet, there was no sign for the "aspiration" (Η having been used for a vowel); a relevant sign was reintroduced in Hellenistic times in the form of a halved Η (𐅂) and, eventually, a semicircle (῾), which was consistently placed over the associated grapheme (as were also the accent marks ´,῀,`); had the "aspiration" corresponded to a feature that preceded or followed its associated letter, it would have simply been written in its respective place and there would be no reason to write it over the associated letter, a practice that was unprecedented at the time of its reintroduction. Having established that, the conclusion that rh (the Latin version of ῥ) was a "voiceless r" is an obvious attempt to reconcile [r] (the value of "normal" r) and [h] (the assumed value of h, based on its value in present-day, non-Romance languages) in the same sound. The orthographic conventions of some modern languages, which are not related to Latin orthography, like rhaff[WEL|Welsh] and hraka[ICE|Icelandic], has undoubtedly contributed to "confirming" the assumed value. In fact, Allen identifies ῥ- with Icelandic hr-, but he also points out that, unlike Icelandic, Greek ρ can only have one value in a certain environment (i.e., the difference is allophonic).
So far, all seems well with the Catholic theory of "voiceless r", which enjoys universal acceptance, particularly in academic cycles: it proposes a seemingly reasonable value for ῥ that appears to somehow correspond to the Latin transcription, which is also used by modern (i.e., Welsh) renderings of the same sound. However, it may create more problems than it solves, if adopted. The first paradox is the significance of the h, i.e., the δασεῖα ῾ or "aspirate". In this particular case (as well as in the cases of μh and λh), it is assumed to indicate voicelessness, which is odd in view of its commonly accepted value in the other cases (the δασυνόμενα/δασέα or "aspirated" vowels and consonants), where it is considered to represent a purely consonantal sound ([h]). Reasons of consistency would necessitate a uniform treatment of the "aspirate"; if it was employed as an indication of voicelessness in the case of the sonorants (ρ, μ, λ), as posited by the Catholics, there is no apparent reason why it would not indicate the same property in the other cases; however, this would mean that "aspirated" (δασυνόμενα) vowels, like ἁ, ἡ, ὁ, etc, would also have to be voiceless, something that Allen rejects outright (ALLE87, p. 54); it would also mean that the distinction between the "tenues" (π, τ, κ, which were rendered in Latin as p, t, c) and the "aspiratae" (φ, θ, χ, which were rendered in Latin as ph, th, ch) was the removal of voice in the latter, something that would be inconsistent with the universally accepted fact that both groups comprised voiceless consonants. Since the assumption of a uniform function of the "aspiration" is incompatible with the other commonly accepted assumptions (voiced character of the vowels, voicelessness of κ, π, τ), the "voiceless r" theory has as a logical consequence the awkward admission that the same grapheme ("aspiration") was used (by the ancient Greeks) for indicating two distinct features of different natures, namely a property (voicelessness) and a consonantal sound ([h]). I am not aware of any scholar who has ventured to provide even a remotely plausible explanation for such a blatant inconsistency in the use of the "aspiration".
But the most puzzling issue is the value of ῤῥ. In an obvious attempt to interpret the script, Allen asserts (ALLE87, p. 42) that "it may be, as the grammatical tradition has it, that only the second element was aspirated, i.e. that the geminate began voiced and ended voiceless". Such a weird construct of a Frankenstein geminate (as concerns voice) is not, to my knowledge, attested in any language, at least among those related to Greek; under this assumption, the emergence of a voiced ρ prefixed to the retained original (assumed voiceless) ῥ in the aforementioned cases of composition or inflection is unexplainable;In that respect, cf. the conversion in Welsh of the word-initial voiceless r (spelt rh) to a normal voiced r (spelt r), when it essentially becomes word-internal, as a result of composition with another word that appears to act as a proclitic, e.g., rhaff→hen raf, rhif→ei rif, etc. moreover, it contravenes the principle that consonant clusters are generally harmonised in Greek through a process of assimilation of voice(lessness), so that voiced consonants "do not combine in groups with voiceless sounds (thus e.g. λέγω but λέλεκται)" (ALLE87, p. 29).Admittedly, Allen's observation concerns the voiced mutes β, γ, δ. Sonorants μ, ν, ρ, λ (which are generally voiced) do combine with voiceless consonants, as in ἄλσος, ἁπλός, Ὄλυνθος, ἄρτος, κάπρος, νύμφη. Although there is no indication that the sonorants were devoiced in these cases, it should be noted that these particular clusters concern consonants of different classes (liquids or nasals with mutes); it is conceivable that there would be no reason for the geminate ρρ, which comprises two consonants of the same class, to not follow the general tendency of voice harmonisation in Greek. A value of mixed voiced-voiceless nature is, therefore, extremely unlikely for ρρ. In fact, Allen subsequently recognises that "this rule [of writing ῤῥ] could be artificial", which appears to be a much more reasonable explanation.Another, albeit less plausible, explanation would be that the rule is inspired by the Latin orthography, as per the scenario of case B above. We will, therefore, accept a uniform value for both ρ's of the geminate ρρ and will seek an explanation of the above-derived conclusion that ῥ- and -ῤῥ- were equivalent. Such an equivalence cannot be explained by the "voiceless r" theory, as it would necessitate that a repeated voiced sound (the ρρ, which may be the result of assimilation, e.g., θάρσος→θάρρος, i.e., a genuine geminate ρ and not a mere orthographic convention, or digraph, of a different sound) would be equivalent to a voiceless version of itself (the ρ-). There is no satisfactory explanation of this paradox.The modern voiceless realisation of the Portuguese rr in some dialects is not at all relevant. The voiceless sound in question is a velar, uvular or glottal fricative ([x], [χ] or [h], evidently the result of a series of sound changes, e.g., [r]→[ʀ]→[ʁ]→[χ]), which is significantly different from the assumed value [r̥] of ῥ and would not explain the convergence back into a voiced alveolar trill in modern Greek (which would have to be assumed, given the present status quo).
So, if the "voiceless r" theory does not provide satisfactory answers to these issues, which other theory would? I believe that the best clue is provided by Spanish, which exhibits the same kind of allophony of r (when word-initial and doubled, it is a trill [r], when word-internal, it is a flap [ɾ]) and reduplication of word-initial r in composition, e.g.,
Puerto Rico → puertorriqueño
The r-values of Spanish provide a better model for ancient-Greek ρ and a more credible answer to the geminate-ρ issue discussed above: if -ρ- is a flap, its duplication -ρρ- would be a series of (at least two) flaps, which is essentially the difference between the alveolar trill [r] and the alveolar flap [ɾ] (cf. LAGE96, p. 245: "From an acoustic point of view, a trill is not unlike a series of taps"). No unnatural hybrid (voiced-voiceless) gemination has to be assumed, no loss of voice upon gemination has to be explained. After this long discussion, I can now present my first point of divergence from the established (catholic and orthodox) models:
Ρalt: Ρ = [ɾ]
Ρalt/: Ρ/#|end of word_ = ΡΡ = [r]
One can only wonder how this simple explanation has escaped the attention of the Catholics. Well, it hasn't. The Spanish analogy has been cited by Blass in his second (BLAS82, p. 75) and third (BLAS82, p. 90) editions: "Analogies for the different values of p are furnished by Spanish, where also r when initial and when doubled in the middle of a word has a quite different and much more emphatic sound than medial r alone". Although his treatment of sonorant "aspiration" is inconsistent and probably erratic,As regards "aspirated" μ, he identifies (BLAS90, p. 88) μh with the Welsh voiceless m (spelt mh), obviously on account of the spelling, but he neglects to at least mention the same language's voiceless r when discussing ρh. Furthermore, in an outright misrepresentation of the phonology of his own language, he claims (p. 90) that "Aspiration of initial liquids is, not to speak of other languages, not unknown even in German, especially where we speak with much emphasis"; according to the translator (n. 2), Blass submits that the Germans "are accustomed to pronounce (in emphasis) t-hage (Tage), n-hein (nein), s-hage (sage), and even 'haber (aber), that is to say [they] pronounce the spiritus asper after the lenis" (to my understanding, this interpretation is purely artificial). how could the subsequent Catholics overlook the inspired suggestion of their venerable Pope and not even mention it as a possibility? I can think of no other explanation, but their conviction (or obsession) that h (and the "aspiration" in general) stood for its present (germanic) value [h] and that the script reflected the actual pronunciation (to the extend possible).It is worth noting that the claim that rh was a single voiceless sound is the first admission by the Catholics of a use of digraphs in Latin. But the value of Latin h and Greek δασεῖα is something that has to be investigated before it can be used as an axiom, as it creates more problems than it solves, at least in the case of ῥ- and -ῤῥ-. Once this prejudice is overcome and the alternative value of ῥ is adopted, everything else falls into place:
Of course, there is no definite proof for the correctness of the alternative values of ρ, but this is also the case for all (orthodox, catholic or alternative) conjectures about ancient-greek phonology (including the "voiceless r" theory) and the values proposed here fare much better with all the evidence, as explained above. In other words, if the allophony of ρ was a feature of Greek (instead of Latin), then the flap/trill theory (i.e., the values Ρalt and Ρalt/) is our best guess.
Today, σ is a voiceless coronal sibilant. I am using the term "coronal" instead of the established "alveolar", because I am not sure about the most appropriate place of articulation; to me it appears that the tongue "is placed behind the opening between the upper and lower teeth at a relatively small distance", which would describe it as "dental"; however, the sound associated with the (thought as) corresponding Latin letter s is described as "alveolar". Personally, I do not see any difference between Greek σ and English s ([s] in, e.g., "discuss"), but often I encounter claimse.g., ALLE87, p. 45, n. 80: "The modern Greek sound is rather more retracted [than English alveolar s]"; what he means by "retracted" is unclear, but it probably refers to the tongue approaching the palate at a point more to the interior of the mouth than the gums, thus more like the post-alveolar sibilant [ʃ]. that Greek σ may sound more like English sh ([ʃ]) or that it may be realised as "apical". Even though dialectally (e.g., in the area of my hometown) it may be pronounced as (something that sounds like a) [ʃ], in general, Greek lacks palato-alveolar sibilantsThese are only attested dialectally, as assibilated versions of the palatals ([c], [ç], [ɟ], [ʝ])→([t͡ʃ], [ʃ], [d͡ʒ], [ʒ]), particularly in areas (that were) under Italian influence. and the general impression is that Greek σ sounds (exactly) like English s and unlike sh. It can, therefore, be argued with certainty that the Orthodox value, which is also accepted by the Catholics, is:
ΣOC: Σ = [s]
Although it is generally accepted that the value of σ in antiquity was that of a voiceless dental (i.e., as it is today), it should be noted that a slightly different value cannot be ruled out. Irrespective of the values of the so-called "doubles" (ζ, ξ, ψ), which are believed to either represent sibilants or at least comprise a sibilant component, there is enough room for variation, which would go unnoticed, since σ was (under any scenarios for the "doubles") the only voiceless coronal non-composite sibilant; a similar argument has been put forward by Lloyd (LLOY87, p. 81) for the Latin s: "we still have no way of being certain whether the phonetic realization of Latin [s] was uniform in all parts of the empire. Being the only sibilant, it could have varied considerably from one region to another and over time without there being any difference in its phonemic status or in writing". The value considered by Lloyd for Latin s is interesting, as he posits the probability "that the sound of the sibilant in Latin was more that of an apico-alveolar sibilant, much like that of modern Castilian [ś], a sound found also in other conservative sections of Romania, and very probably inherited from Indo-European" (p. 80). The same apico-alveolar value (as far as I can tell, it sounds like [ʃ] without lip rounding) can be considered as a likelihood also for Greek, for the same reasons as for Latin: Greek could have also inherited it from Indo-European, some dialects do display rhotacism (e.g., Laconian word-finally and Eretrian intervocalically; cf. CHAT02, p. 339), which is also a feature of early Latin explained by Lloyd as a transition to "/r/, an alveolar flap",The same phenomenon is explained by Sturtevant (STUR20, p. 163) as an indication of "a voiced sibilant in these positions". However, it is not necessary that σ be voiced in order to become ρ: although Eretrian displays the same kind of (intervocalic) rhotacism as Latin (where it is suspected that the value of intervocalic s was voiced, as in Italian, French, etc), the existence of a voiced sibilant in Laconian only in word-final position, but not intervocalically (where it disappeared or "became h", according to Sturtevant), would be certainly curious. and it is reported that in present-day Greek "[s] and [z] can be apical ([s̻], [z̻])".
Another point that adds to the mystery of the value of Greek σ is its origin as a grapheme. It is said to derive from Phoenician shin, at least as far as its shape and alphabetic position are concerned. However, the value of its semitic counterpart, at least in Phoenician, is considered to have been [ʃ]. Had Greek σ the value [s] at the time (and place?) of adoption (or creation), one would expect it to rather correspond to samekh, which consistently has that value in the modern semitic languages.It is contended (cf. ALLE87, p. 171, n. 4; details in TAYL83, p. 98, n. 1) that samekh has lent its name and value to sigma. Other than the fact that there is an "s" and an "m" in samekh's name, in order for the argument to be plausible, one has to argue that the Greeks mixed up not only the vowels of the name "samekh", but also "kh" (a כ, i.e., /k/, that in word-final position is spelt ך and pronounced [x]) was "metathesised" (a popular claim among the linguists) with "m" and was voiced (and, if you are a Catholic, also fortitioned) to a γ, not to mention its nasalisation, if we are to believe Allen's value for γμ (cf. ALLE87, pp. 35-37). In addition to the random anagrammatisation of samekh's name in order to derive sigma, the proponents of this theory postulate that the Greeks took the name and sound from one phoenician sibilant and used it for another and vice versa, particularly between samekh and shin and between zayin and tsade. Any objective reader can decide for himself how (im)plausible such an unnecessary shuffling of the values, names, shapes and positions of the phoenician sibilants is. The second part of the thesis (zayin for tsade and vice versa) is particularly problematic: it assumes that ζ had the name and value of צ and that Ϻ (greek "san") had the name and value of ז. However, צ is believed to have been an "emphatic s", probably "pharyngealised", which did not exist in Greek and could never have been the value of ζ. An affricate value of צ, as per modern Hebrew, is too doubtful an assumption for ζ, as Greek appears (from internal evidence) to have entirely lacked dental-sibilant affricates. Conversely, the claim that Ϻ had the value [z] (the modern value of hebrew ז) and was adopted because "those Doric dialects that kept San instead of Sigma may have had such a pronunciation of /s/", implies that the Doric dialects that exclusively used San instead of Sigma had a voiced sibilant, but no voiceless one, which is not very likely, as it is reported that "The presence of [z] in a given language always implies the presence of a voiceless [s]". In fact, it appears that the names σάν and σίγμα were used interchangeably by the ancients to refer to the same letter (and, presumably, to the same sound), possibly using the former for its Doric shape Ϻ and the latter for its Ionic form Σ, as can be inferred from Halicarnasseus' reference to Pindar's verse "καὶ τὸ σὰν κίβδηλον ἀνθρώποις" in connection with the discussion of the sound of σίγμα (ROBE10, p. 148). The traditional view that "The name σίγμα for Σ may be a derivation from the onomatopoeic verb σίζω" (ALLE87, p. 171) is certainly more cogent and does not involve juggling the letters of the name of a different semitic letter.
Admittedly, Reverend Taylor, the instigator of the fanciful thesis regarding the pairwise confusion of names and values of the semitic sibilants, makes a strong case and provides reasonable justification for it (TAYL83, pp. 97-102). However, the loose ends in his theories can only be tied up in an ambience of genuine 19-century confusion. For example, in order to justify the association of ζ with Tsade, he calls the latter a "lingual sibilant" and assumes its value to be [t͡s], apparently ignoring that it was an "emphatic" consonant, and refers to the former as a "dental sibilant" that "had the power of σδ or δς" (p. 97, n. 1). Similarly, in order to make a connection with Shin, which he calls "the palatal sibilant ש, sh", he considers ξ to denote "the Greek sound σσ, afterwards becoming the guttural sibilant", not bothering to explain how ξ's "primitive power of s or ss" would be associated with (or, actually, lead to) a "guttural" value ξ=[ks]. How little phonological coherence there is in his theories is manifest in his treatment of the Greek (so-called) "aspirates": he appears to accept that θ, φ, χ "represented the complete tenuis followed by a distinct aspiration, like the aspirated mutes in Sanskrit" (p. 89, n. 1), but, in order to derive both θ and φ from Teth, he refers to "the well-known tendency to substitute for th the easier sound of ph", using as examples that "Children often use the easier sound instead of the more difficult, saying bof, erf, fink, for both, earth, and think" and that "The Russians change
Theodore into Feodor." (p. 89, n. 3), which however relate to (and only make sense for) the "fricative" values of θ and φ. Assuming an apico-alveolar value for σ answers this question satisfactorily, as the correspondence of greek σ and phoenician shin is paralleled (in the inverse direction, i.e., Indo-European→Semitic) by the "Transcriptions of Spanish words in Arabic letters[, which] most often represented s with the letter ش šīn, which stands for the palatal sibilant [š]", while "Arabic has another sibilant, س sīn which [has taken the place of samekh and] represents a dental sibilant [s] which could have been used to represent a Romance sibilant, if it had resembled the Arabic sibilant" (LLOY87, pp. 338-339); the correspondence of [s̺] and [ʃ] is also attested in several examples of confusion in various medieval borrowings, such as pousser[FRA]→push and sepia[LAT]→xibia[SPA].
The evidence, thus, points to a further possibility for σ:
Σalt: Σ = [s̺]
An apico-alveolar value for s perhaps also explains Halicarnasseus' aversion for its sound: "ἄχαρι δὲ καὶ ἀηδὲς τὸ σ καὶ πλεονάσαν σφόδρα λυπεῖ· θηριώδους γὰρ καὶ ἀλόγου μᾶλλον ἤ λογικῆς ἐφάπτεσθαι δοκεῖ φωνῆς ὁ συρυγμός|σ is an unattractive, disagreeable letter, positively offensive when used to excess. A hiss seems a sound more suited to a brute beast than to a rational being [Roberts' translation]" (ROBE10, p. 146); the apico-alveolar sibilant, which "to speakers of languages or dialects that do not have the sound, it is said to have a 'whistling' quality", would justify its portrayal as "unattractive" and "disagreeable". On the other hand, the description provided by Halicarnasseus about the sound corresponding to σ can barely be considered to describe an apical pronunciation: "τὸ δὲ σ τῆς μὲν γλώττης προσαγομένης ἄνω πρὸς τὸν οὐρανὸν ὅλης, τοῦ δὲ πνεύματος διὰ μέσων αὐτῶν φερομένου καὶ περὶ τοὺς ὀδόντας λεπτὸν καὶ στενὸν ἐξωθοῦντος τὸ σύρυγμα|and σ by the entire tongue being carried up to the palate and by the breath passing between tongue and palate, and emitting, round about the teeth, a light, thin hissing [Roberts' translation]" (ROBE10, p. 146; emphasis mine). For this reason, the alternative value proposed above is considered as a possibility that fares at least as well as the traditional value [s], which would however also be acceptable.
Affricates
Ancient Greek was intolerant of combinations of dental mutes with sibilants. As far as I can tell, a consonant cluster comprising one of τ, θ, δ as the first component and one of σ, ζ (or ξ, ψ) as the second is not attested in any Greek inscription or text. Even when morphology would require that a dental mute stand before σ, the mute is dropped, e.g., *ποδ|stem of the noun ποῦς, ποδός: foot-ς|ending of nominative masc. sing. 3rd declension→ποῦς (with compensatory lengthening, ο→ου, of the preceding vowel).According to Smyth (SMYT20, p. 28, §98), "A dental stop before σ is assimilated (σσ) and one σ is dropped". I am not aware of any evidence supporting the alleged assimilation as an intermediate stage. Perhaps the fact that in some instances, e.g., the dative plural *ποδ-σι→ποσί, there is no compensatory lengthening (i.e., *πουσί), can be explained as the result of such assimilation. In these cases, one would expect that, even in absence of the corresponding affricate per se from the basic word forms (e.g., stems) of the language, the affricate would be retained, even in slightly modified form, e.g., †ποτς; this is what essentially happens in English, e.g., cat-s|plural suffix→cats, even though [t͡s] is not part of the english native inventory. Only in late (=post-classical) antiquity are such clusters to be found, usually as attempts to represent foreign names, e.g., "Γεθσημανή" (Gethsemane), also "ΤΣΑΔΗ" for the name of the letter Tsade in the numbering of the book of Lamentations of the Septuagint (but that may be a later addition, not necessarily featuring in the original version of II BC). In the Middle Ages, the affricates appear to have been accepted as part of the Greek phonology, at least as far as the spelling of proper names (usually of foreign origin) was concerned, with the spelling τζ (e.g., Τζέτζης, Τζιμισκής in X AD) being used to indicate either [t͡s] or [d͡z] (or, indifferently, any of the two). Today, these two affricates are part of the Greek phonology and are represented by τσ and τζ, respectively, not only in foreign borrowings, but also as results of phonological processes in native greek words (e.g., elision of unstressed vowels: κάθισε|sit! (imperative 2nd person singular)→*κάθσε→κάτσε, presumably since XII AD; metathesis, dialectally: της|of the (fem. article genitive singular)→τση). A similar creation of dental/sibilant clusters as affricates is not observed in ancient Greek.
An easy solution to the problem of explaining the lack of dental/sibilant clusters in ancient Greek would be to assume that σ was an affricate itself. If, for instance, σ were [t͡s], then the simplification (a kind of haplography) of τσ=[tt͡s] (also of θσ and δς) to σ would make more sense; incidentally, it might also explain the lack of gemination of σ (SMYT20, p. 29, §107: "Two sigmas brought together by inflection become σ"; for σσ corresponding to Attic ττ, see below), even though affricate gemination is not uncommon in, e.g., Italian (cf. pizza=[ˈpit.t͡sa]) and disguised in German (cf. BLAS90, p. 104: "the Germans write quite correctly tz for double z[=[t͡s]]"). However, such a value for σ would not be supported by any of the other evidence: there is no realisation of σ as an affricate today (however, Psichari reports the forms ἀτσήμι for ἀσήμι and τέτσαρα for τέσσερα "among the Greek islands"; BLAS90, p. 116); the description of Halicarnasseus does not mention any contact of the tongue and the teeth or alveolar ridge, hence no stop component in σ; its predecessor shin does not have an affricate value in any semitic language. Instead, the most reasonable explanation appears to be that ancient Greek did not possess affricates and did not tolerate them, more or less like modern French.
To be honest, there are limited phenomena of assibilation that might be explained by an affricate value of σ: "Ionic [...] changes τ before ι to σ" (SMYT20, p. 2); also for Attic and common Greek: "τι̯, θι̯ after long vowels, diphthongs, and consonants become σ; after short vowels τι̯, θι̯ become σσ (not = ττ 78), which is simplified to σ" and "τ before final ι often becomes σ" (p. 30, §113, §115). It might be expected that, in these cases, σι would stand for [t͡si] (cf. politia[LAT]→polizia[ITA], tertia[LAT]→terza[ITA]). However, it is not necessary that the expected affricate be introduced into the phonological inventory; if the latter entirely lacked affricates, it is also possible that the closest available (sibilant) be used, in that case σ=[s]; cf. the orthographic merger of Latin -ti- (expected [t͡s]) and -ci- (expected [t͡ʃ]) and the development of these digraphs in western-Romance languages other than Italian (both are, e.g., [s] in French and [θ] in Spanish). Incidentally, if σ were [t͡s], it would be difficult to explain the loss of ν before σ in etymologies like *παντ-ι̯α→(*πανσα→)πᾶσα, γιγαντ-ς→γίγας, μέλαν-ς→μέλας etc (p. 28, §100; p. 27, §96).
A similar development that led to the claim that some kind of affricate was present in ancient Greek is that of Attic -ττ- (e.g., γλῶττα|tongue, θάλαττα|sea), which corresponds to common-Greek -σσ- (γλῶσσα, θάλασσα). Etymologically, this comes either from suspected borrowed words (like Ἁλικαρνασσός|Halicarnassus) or from combinations of voiceless velar (κ, χ) or dental (τ, θ) mutes with the front semivowel (ι̯=[j]), e.g., φυλακ-ι̯ω→φυλάττω|I guard/φυλάσσω, μελιτ-ι̯α→μέλιττα|bee/μέλισσα, sometimes also with the back semivowel [w], as in the case of (*kʷetwóres→)τέτταρες (SMYT20, p. 29, §112, p. 30, §114; ALLE87, pp. 60-61). In fact, a special character (Ͳ, later Ϡ) is found in a few pre-classical inscriptions from Ionia in lieu of the common-Greek digraph (?) -σσ- (attic -ττ-) and a reasonable conjecture is that it was employed for representing a sound (evidently a foreign one, e.g., Carian), that did not correspond to one of the letters of the Greek alphabet.This sound does not have to be an "original" or "intermediate" sound for dialects other than asian Ionic, i.e., it may have nothing to do with the history of the corresponding sound in other greek dialects. It could have been a sound evolved out of -σσ- (if gemination was phonemic and not orthographic; cf. ll→[ʎ], nn→[ɲ] in Spanish); or it could represent a dialectal diversion in Ionic (perhaps under foreign influence?) of the "original" *κι̯, *χι̯, *τι̯, *θι̯, *τϝ. In either case, the new sound might have happened to coincide with a foreign one and was, therefore, used in borrowed words as well (cf. the development of voiced stops out of μπ, ντ, γκ in Greek and their use to describe the foreign sounds [b], [d], [g], e.g., Μπους for Bush). It is also postulated that the corresponding symbol Ͳ may be a foreign (Carian?) invention adopted by the Ionians. Incidentally, the theory that Ͳ was derived from semitic tsade does not seem to hold water: first of all, it is not certain that tsade had the value that modern-hebrew צ has ([t͡s]); furthermore, in (one of) its earliest attestations, an abecedarium from Samos, it has the last alphabetic position, not that of tsade (after π), which must mean that its was a later addition (like υ, φ, χ, ψ, ω, which were also added at the end of the alphabet) and not a borrowing from Phoenician. However, this reasoning cannot be extrapolated to the value of -σσ-, i.e., to assert that -σσ- stood for the same sound. It could very well be that the (assumed) original *κι̯, *χι̯, *τι̯, *θι̯, *τϝ developed differently in different dialects, similarly to the different evolution of the labiovelars before front vowels in Lesbian and Attic and the diverging development of Latin ce, ci in Italian, French and Spanish. Whatever the case for asian Ionic might have been and regardless of the etymological origins of -σσ- (and attic -ττ-), Allen's observation that "it is scarcely credible that the existence of an affricate sound would not have been revealed in any inscriptional spelling outside those mentioned above (e.g. as τσ), nor the tradition of it survive in the account of any grammarian" is reasonable and we can but concur with his assessment of the sound value of the digraph: "there seems therefore no need whatever to assume that the ττ of Attic or the σσ of other dialects mean anything more than they appear to" (ALLE87, p. 61).
To sum it up, Greek's intolerance to dental-mute/sibilant clusters suggests that ancient Greek most likely possessed no (dental) affricates; if it had, these clusters (created, e.g., by inflection) would either remain as they stood, at most converted to, e.g., τσ (or θσ; cf. old Attic χσ and φσ for κσ, γσ and πσ, βσ), or they would be converted to some other symbol indicative of a dental affricate (cf. common Greek ξ and ψ vs the aforementioned old-Attic χσ and φσ). Such a symbol is only to be found in eastern Ionic (Ͳ), however not in lieu of dental-mute/sibilant clusters, but for (suspected) palatalised dental or velar mutes. Otherwise, the result of these clusters is in all dialects a mere sibilant (σ), which suggests that the stop component of the expected sibilant was dropped, unless we are to assign an affricate value to σ itself, which would be too hard to explain in view of the present phonological state of Greek (σ being a pure sibilant and τσ, τζ being introduced as affricates).
Voiced σ
Generally, σ was voiceless. This is understood from the use of the terms "συρυγμός" (why not συριγμός, as spelt in STUR20, p. 163? could it be a spelling mistake by Roberts?) and "σύρυγμα", i.e., whistling or hissing sound, by the ancients (e.g., Halicarnasseus above) which only makes sense for voiceless sibilants;cf. also Plato's description (in the words of his character Theaetetus): "τό τε σῖγμα τῶν ἀφώνων ἐστί, ψόφος τις μόνον, οἷον συριττούσης τῆς γλώττης|σ is a voiceless letter, a mere noise, as of the tongue hissing [Fowler's translation]", particularly as contrasted to β: "τοῦ δ᾽ αὖ βῆτα οὔτε φωνὴ οὔτε ψόφος|β again has neither voice nor noise [Fowler's translation]"; "ψόφος|noise" here appears to relate to the feature of (modal) voice in its modern sense, actually to lack thereof. had it been voiced, e.g., [z], then its sound would be more appropriately described as a buzz. This is also true in present-day Greek; with one exception: before voiced consonants (μ, β, γ, δ), it becomes voiced ("assimilation of voice"), i.e., identical to ζ=[z]. There are traces of the same kind of assimilation in ancient times, including testimonies in the literature, orthographic variations in inscriptions and transliterations in foreign languages, mostly dating from post-classical or even post-Christian times. Based on such evidence, the Catholics are eager to accept the orthodox conditional value of σ:
ΣOC/: Σ/_Cvoiced|voiced consonant, i.e., μ, β, γ (theoretically also δ, ν, λ in compounds) = [z]
Here, it must be noted that the catholic claim is that σ always had this diverging conditional value ("The case, where a medial or liquid follows the σ is and was an exception", BLAS90, p. 91) or at least since the classical period (as appears to be suggested by Allen and Threatte, who present "evidence" from 445 BC and <479 BC, respectively). It is interesting to note that the Catholics, in general, need to see sufficiently numerous evidence before they concede a "modern-Greek" feature deviating from the graphical representation (such as a voiced allophone of an otherwise voiceless sound, in the present case) and even in the face of sporadic evidence, they are only willing to accept this as an isolated case and not a general tendency or practice; however, in the case of "voiced σ" they are willing to extend the validity of the orthodox conditional value not only to time period covered by the scanty evidence, but even before that, despite the lack of evidence. If one is to apply the same strict principles for scrutinising the evidence attesting to the modern pronunciation of σ before voiced consonants, these may not be as convincing as we are expected to believe.
Starting from the most recent evidence, there are provided a number of transliterations in Gothic, Armenian, Hebrew, etc using native characters that (are generally thought to) stand for [z] whenever σ is found before voiced consonants in the Greek word, but these are relatively late (from III AD on) and could reflect the phonotactics of these languages, not necessarily those of ancient Greek (cf., e.g., the general modern-Greek tendency to pronounce "Smith" as [zmiθ], resulting in a possible greek spelling Ζμιθ).
Regarding the literary evidence, there is no direct description of a voiced σ in the works of the Grammarians. The clearest reference is probably provided by Lucian (II AD), in his comedy lawsuit of σ vs τ, wherein σ declares "ὅτι δὲ ἀνεξίκακόν εἰμι γράμμα, μαρτυρεῖτέ μοι καὶ αὐτοὶ μηδέποτε ἐγκαλέσαντι τῷ ζῆτα σμάραγδον αποσπάσαντι καὶ πᾶσαν ἀφελομένῳ [σ]μύρναν|That I am a longsuffering letter, you yourselves are my witnesses, since I have never brought suit against Zeta for taking an emerald from me and robbing me of all [myrrh] [Stu's translation, my correction]".The published text has "Σμύρναν" alluding to the city of Ionia Smyrna, as does Sturtevant's translation, but this is rather implausible, as I have already argued. In isolation, this statement may be taken as an explicit acknowledgement of σ's voicing before μ and possibly before all voiced consonants; however, in the context of the entire work's spirit, it merely indicates that σ in these words was replaced by ζIt is clear from the other examples of the same play (e.g., τέτταρα for τέσσαρα, τήμερον for σήμερον, ξυγγραφέα for συγγραφέα, μυρρίνας for μυρσίνας, ἐντελέχειαν for ἐνδελέχειαν) that Lucian (or rather "sigma") refers to words that afforded or had established alternative spellings. Presumably, the spelling variations reflected alternative pronunciation. and was apparently pronounced accordingly (tentatively [z], ζ's modern value, which must however be confirmed in the corresponding chapter). A similar allusion to alternative spellings is provided by Lucian's contemporary, Sextus Empiricus, in his portrayal of the grammarians: "ὅταν σκεπτώμεθα πότερον διὰ τοῦ ζ γραπτέον ἐστὶ τὸ σμίλιον καὶ τὴν [σ]μύρναν ἤ διὰ τοῦ σ|when we are pondering with which one of ζ and σ σμίλιον and [σ]μύρνα should be written".Bekker's edition again has Σμύρναν, obviously implying the city (with which modern editors appear to be obsessed), but again σμύρνα (the substance) is a more reasonable assumption, particularly since together with σμίλιον they are related to Sextus' profession. These testimonies suggest that there was orthographic confusion in post-Christian times, at least in the cited examples and probably in all cases of σ before μ.
As a matter of fact, the inscriptional evidence verify that σ interchanges with ζ almost exclusively before μ: of the 22 examples provided by Threatte (THRE80, pp. 548-549; 20 "ζ for σ before μ, β" and 2 "σζ for σ before μ", all post-classical), 21 concern ζμ for σμ, 15 corresponding to the two words mentioned by Lucian and their variants;Ζμύρνα appears as early as III BC, while ζμάραγδος only from I AD on, as well as in its Latin form zmaragdus (BLAS90, p. 91, n. 4). The disproportionately frequent ζμάραγδ-, ζμύρν- may be the reason why they were brought up by Lucian. There is no example of the ζμίλιον mentioned by Sextus other than the (late) Armenian transliteration zmelin cited by Allen (ALLE87, p. 46). there is only one doubtful case of ζβ for σβ and no variants of σγ (at least in the Attic inscriptions investigated by Threatte, Πελαζγικόν mentioned by Sturtevant probably being non Attic).
For the sake of completeness, it should be mentioned that Allen's claim (ALLE87, pp. 45-46) that voicing of σ before δ "is suggested by the fact that Ἀθήνας+δε is written αθεναζε (=Ἀθήναζε, 445 B.C. etc.)", as well as "Boeot. διοζοτοσ = Διόσδοτος", is based on inscriptional evidence of a different kind: instead of ζδ (cf. ζμ, ζβ, ζγ in the case of the other voiced consonants), plain ζ appears for "expected" σδ. Similarly, Threatte submits τοισζ(ε) from early V BC (THRE80, p. 510), which he assumes to stand for τοῖσδε or rather τοῖσσδε (in analogy to Ἀρίσστων for Ἀρίστων);Teodorsson submits for this particular example (TEOD74, p. 140) that "The writer had written Τ, then changed it to ɪ" (Η rotated 90°, one of ζ's earlier forms), which may mean that the reason for writing ζ was not phonetic, but a clumsy effort to reconcile what the writer had in mind when carving the stone (perhaps τὲ→τ'?) with what should have actually been carved. again, σδ is not replaced by ζδ as in the examples of the other voiced consonants, but by σζ. Both claims are rather based on the a priori conviction about "the special symbol ζ = [zd] for σδ",This is the epitome of cyclic reasoning, since the proof for the voiced value of σ in ζ=σδ is that "the [z] element would be a normal voiced variant of the /s/ phoneme as in, for example, Λέσβος (cf. p. 46)" (ALLE87, p. 57), where page 46, to which the reader is referred, is the part of the book that "proves" a voiced σ in σβ (of Λέσβος) by stating that "The case of σδ [based on ζ=[zd]=σδ] makes it virtually certain that the same applied before other voiced plosives"!!! which could (allegedly) be freely used by the ancients whenever they did not feel like writing σδ! Contrary to σμ, σβ and σγ (cf. ᾆσμα, Λέσβος, φάσγανον), σδ never appears in proper greek morphemes; disregarding the self-proving identity ζ=σδ, the only instances of σδ are found across morpheme boundaries in composition (e.g., προσ-δέχομαι) or enclisis (e.g., τῆσ-δε). There are only a couple of cases (exactly those mentioned by Allen) where this trans-morphemic cluster appears to have been replaced by ζ, but this may be just an illusion. In the case of Ἀθήναζε (for which I have not seen any inscriptional instance of the assumed *Ἀθήνασδε), -ζε appears to have developed into an "inseparable Suffix", used in lieu of the most common -δε, as is evident from Ὀλυμπίαζε, Μουνυχίαζε, etc (wherein there is no etymological reason for -ζε to stand for -σδε); how this -ζε came about can only be a matter of speculation, but the surmise that it represents exactly [zde] and was used because in this particular case the Athenians were too lazy to write -σδε (but not in the similar cases of ὅσδε, τούσδε, for which there are no instances of *οζε, *τουζε) is unsubstantiated; it could be a mere case of "fusion" of σ and δ to produce a new sound, similar to the aforementioned process that produced σσ (Attic ττ), for which no-one claims that σσ/ττ stood for any of the "original" *κι̯, *χι̯, *τι̯, *θι̯, *τϝ. The second example, the non-attic διοζοτοσ, Allen equates with Διόσδοτος, but no such name is to be found in the Athenian Onomasticon, which only lists the variants Διόζοτος and Διόδοτος (also Διοδότη); it appears that the assumed Διόσδοτος exists only in the catholic imagination and that the ζ of Διόζοτος corresponds to the δ of Διόδοτος, not to any σδ; in fact, the similar variants Θεόζοτος, Θεοζοτίδης (again Boeotian, according to Threatte) for Θεόδοτος, *Θεοδοτίδης are discussed by Threatte (THRE80, p. 550) as suspected cases of ζ for δ.Curiously, the catholic zealots use exactly this variant, Θεόζοτος, as proof of ζ=[zd]: "Some inscriptions have -ζ- written for a combination -ς + δ- resulting from separate words, e.g. θεοζοτος for θεος δοτος 'god-given'". The assumed underlying form "θεος δοτος", comprising two constituents in the nominative, is peculiar for Greek, as Greek usually employs either the genitive of the first constituent (cf. Allen's grammatically correct, albeit apparently unattested, Διόσ-δοτος, as well as Διόσ-κουρος, Κωνσταντινού-πολις) or its stem, often augmented with an epenthetic -ο- (cf. Δι-ό-δοτος, Ἑρμ-ό-δοτος, Ἑρμ-αφρόδιτος). Thus, the normal form of the name is Θεόδοτος, Θεόσδοτος not appearing in the name list of attested proper names. An adjective θεόσδοτος seems to appear in literature (if the surviving texts do not comprise errors by the copyist/editor), but merely as "poet. and later Prose for θεόδοτος". This example of assuming a proper name (Θεόσδοτος), which is not attested and is based on a rare literary variant, is characteristic of how evidence favouring the orthodox thesis (Θεόζοτος for Θεόδοτος makes more sense if ζ=[z] and δ=[ð], as today) are clumsily forged in support of the catholic one. A case of ζ for δ is also Threatte's τοισζ(ε) for τοῖσδ(ε), which is the standard form; the underlying τοῖσσδε postulated by Threatte in analogy to Ἀρίσστων is yet another case of unprovable speculation. All in all, voicing of σ before δ is proved by the Allenian arguments only if one has already accepted that ζ=σδ=[zd], in which case the voicing is already part of the assumption and and does not need to be proved!
It is not clear how we can interpret the inscriptional evidence. If, for example, we accept that the earliest Attic example of ζ for σ, [ε]ἰργαζμένο[ν (THRE80, p. 548), suggests a voiced pronunciation of σ before μ (or at least in this particular word), should we not also conclude that its contemporary Σεύς (STUR20, p. 190; this reading is doubted by Threatte on THRE80, p. 548) suggests a voiced pronunciation of σ at the beginning of the words or a voiceless pronunciation of ζ in the same environment or in this particular word? Even if we accept the first kind of evidence (ζ for voiced σ) and reject the second (σ for ζ) in the face of the subsequent (i.e., modern) state of greek phonology, the evidence may merely suggest that voicing of σ was sporadic in the attested words, but not general: to provide a parallel from English, the spelling kozmic would be an indication that s is voiced before m in the particular word cosmos and its derivatives, but not in all cases, to which smear, smile, etc, can bear witness. Indeed, the existence in English of word-initial clusters of s=[s] with voiced consonants, like "smother", "snail", etc, and the lack of voicing in Spanish (which, like the "reconstructed" Attic, has no stand-alone voiced sibilant) when s is followed by a voiced consonant, such as "Esmeralda", demonstrate that there is no physical necessity for /s/ to be always assimilated by voice. Nevertheless, the Catholics are willing to concede that the orthodox conditional value was valid at least from classical times, based on 22 inscriptional instances (literary a handful, particularly if one disregards the recurring σμυρν-, σμαραγδ-) after 350 BC. The main idea here in that, since the orthodox conditional value is compatible with the evidence (a handful of post-classical instances of confusion of σμ, σβ, also non-attic σγ and ζμ, ζβ, ζγ) and since we have no counter-evidence (what kind of counter-evidence could it be?), the orthodox value is acceptable. This is strange, because, in other cases where an orthodox value is suggested by more numerous and earlier examples (cf., e.g., TEOD74, p. 49, n. 47: "There are 48 instances of 'confusion between η and ι' before 200 B.C."), the Catholics are not so easily convinced. It is evident that their eagerness to accept a voiced allophone of σ before voiced consonants stems from their desire to establish a classical value of [zd] for ζ (the argument goes like this: "if ζ were [z], there would be instances of ζ for voiced σ in inscriptions from classical times").
The reasonable conclusion is, therefore, that the provided pieces of evidence are not at all suggestive of a voiced allophone of σ in classical Attic and very little indicative thereof in post-classical times. In the case of σβ and σγ (σδ is not at all an issue, as we have seen), we can find virtually no supporting evidence until very late (the foreign transliterations of III AD or later). I would, thus, venture to disagree with both creeds and declare that the conditional value of σ has not been and cannot be proven for the classical language, if it were not for a peculiar variant that I ran into: the familiar word ὀσμή|smell, odour appears to be an "Att. form of the older ὀδμή", which was used by Homer, Herodotus and Pindar; irrespective of whether they may represent developments of two related stems along different paths (e.g., ὄδωδα→ὀδμή, ὄζω→ὀσμή)In fact, Smyth states that "ὀσμή odour stands for ὀδ-σμη" (SMYT20, p. 26, §87). If so, then the two variants, ὀδμή and ὀσμή, may be the results of two "euphonic" (in Smyth's terminology) laws, namely that σ is dropped between consonants (p. 28, §103) and that dental mutes disappear before σ (p. 28, §98), respectively, depending on which law applied first. or different variants of the same word (ὀδμή↔ὀσμή), a voiced value of σ would be the most reasonable explanation (not to mention that this example is even more understandable, essentially as a case of reverse lisping, under the other orthodox value δ=[ð]cf. the evidence from Elean submitted by Chatzidakis on CHAT02, p. 441, with ζ used for δ in "ζίκαια", "ϝειζώς", "ζίφιον", "Ὀλυμπιάζων", "Ζί", "ζέ", "οὐζέ", "ζᾶμον", etc, from which he (and Blass with him; BLAS90, p. 113) concludes that δ was a fricative in Elean.). I have not had the time and the means to run an exhaustive search on further variants σμ↔δμ, but even this single example renders the value Σ/_μ=[z] more possible for the classical era than the evidence submitted by Allen et al.In fact, Liddell-Scott alludes that σμ↔δμ was a regular transformation in "aeolic and ionic" and submits also ἴδμεν for ἴσμεν.
In brief, while the catholic evidence is not convincing, a voiced allophone of σ, at least before μ, is likely to have existed in classical times.
Today, ξ and ψ do justice to their appellation "doubles" ("διπλᾶ"), as they are not single sounds, but (voiceless) sequences of a stop and a sibilant (in both cases [s]). ξ comprises a velar stop:
ΞOC: Ξ = [ks]
while ψ's stop component is labial:
ΨOC: Ψ = [ps]
Both Orthodox and Catholics are happy with the present values of ξ and ψ. For the Orthodox this is understandable, because they do not have to change their habits. The Catholics have found in the modern values of ξ and ψ all the excuses they needed for the exotic values they propose for the other "double", ζ. Their confidence seems to be based on the following facts about the "doubles":
At first sight, the logical conclusion of (D1) and (D2) is that the common values ΞOC and ΨOC must have been valid in antiquity: if the "original" Greek values were [ks], [ps] and the present Greek values are also [ks], [ps], then all intermediate Greek values (including "ancient Greek") must have been [ks], [ps]. There is one problem in this reasoning though: it treats "Greek" as if it were a single uniform language without local phoneme variations. However, the "Greek" spoken by the various ancient tribes differed not only in vocabulary and morphology, but occasionally in phonology. Phonological differences among the various Greek dialects are suggested by evidence in various cases; for example, Allen postulates a different pronunciation between Attic and Cretan θ already at the beginning of his treatise (ALLE87, p. xiv), even though they both have the same origin (assumed PIE /bʰ/) and the same fate (there is but one realisation of θ today: [θ]). In the particular case of Ξ and Ψ, potential evidence of dialectal divergence can be found in the regional orthography:
The various renderings of ξ and ψ in the various local-script groups are summarised in the following table (each group receiving the colouring assigned to it by Kirchhoff):
| Southern | Other Eastern | Eastern | Western | |
|---|---|---|---|---|
| ξ | ΚϺ | ΧΣ | Ξ | Χ |
| ψ | ΠϺ | ΦΣ | Ψ | ΦΣ |
It is odd that such graphemic diversity has never been considered as a possible indication of an underlying phonemic diversity. After all, the Catholics do not hesitate to follow orthography for inferring pronunciation. For example, the Boeotian spelling η for the Attic diphthong αι (e.g., STUR20, p. 125) is interpreted as evidence that the corresponding sound was a monophthong in Boeotian (and still a diphthong in Attic). We have also seen above that the introduction of Ͳ in positions where a common-Greek σσ (Attic ττ) would be expected is taken to mean that, instead of σσ, Asian Ionic had developed (or retained) an affricate in this case. It would, thus, be reasonable to consider the possibility of a similar development for "Eastern" Ξ and Ψ and for "Western" Χ.
The distinct nature of Ξ and Ψ is also suggested by their origin, as it is almost certain that the two letters were Greek additions to the alphabet after the transmission. This is clear for Ψ, which was placed after the borrowed letters, in all alphabets that included it. The same can also be inferred for ξ from the alphabet of Euboea, which is believed to be the place of transmission (or at least seems to be the oldest alphabet comprising symbols for any of Ξ and Ψ) and where the symbol X corresponding to (Ionic) Ξ is placed at the end of the alphabet, after the Semitic letters and together with the other innovations Φ, Ψ. "Eastern" Ξ is an odd case, as it has the shape and alphabetic position of semitic Samekh. However, a phonetic connection between the two letters cannot be established (putting aside Taylor's phonologically nonsensical claim that ξ stood for "the Greek sound σσ, afterwards becoming the guttural sibilant ξ[=[ks]]"; TAYL83, p. 97, n. 1), as semitic tradition appears to be unanimous: ס=[s]. It is rather more likely that the Ionians "recycled" semitic Samekh or assigned a new value to it at the time of adoptionAnalogous developments can be seen in the case of ionic Η (a vowel), as well as the principal greek vowels Α, Ε, Ο (and partly Υ), all of which (presumably) correspond to semitic consonants: Η=ח, Α=א, Ε=ה, Ο=ע (and Υ=ו). (the reuse being possibly due to a desire to retain the order of the letters that were also used as cardinal numbers). Thus, the Greeks or rather the Ionians (or the Euboeans) introduced Ξ and Ψ (or Χ) into the alphabet borrowed from Semitic as new letters, which served the obvious purpose of representing sounds that were not adequately rendered by the existing semitic letters. The consensus view that these sounds were ΞOC and ΨOC, i.e., consonant clusters, does not answer satisfactorily one fundamental question related to their raison d'être:
"Why would anyone need special symbols for [ks] and [ps], when symbols for [k], [p] and [s] were available?"
One possible answer would be to assume that these were inherited. Occasionally, during alphabet transmission superfluous letters are received by the target language together with the essential ones: this was the case with the early Etruscan alphabet, which comprised all Greek letters, the superfluous ones being later dropped; similarly, early Coptic and early Cyrillic alphabets comprised the unnecessary symbols for [ks], [ps] (carryovers of Greek Ξ and Ψ) and the digraph ΟΥ for [u(ː)]. This could not, however, be the case for "original" Greek Ξ and Ψ, because no symbol of the source Semitic is identified as standing for a consonant cluster and it is furthermore suggested by the alphabetic evidence presented above that Ξ and Ψ were greek innovations.
Allen (ALLE87, p. 59) acknowledges the paradox ("it is in any case surprising that special symbols should have been adopted for these combinations when they could very well have been written as κσ, πσ, and are in fact so written in some alphabets") and attempts to provide an answer to this fundamental question. One of his suggestions is that "They may have been introduced after the analogy of the other combinations of plosive+fricative, viz. ζ for [dz]". However, the value ζ=[dz] is postulated by some Catholics as the original value (before ζ was... metathesised to [zd]!) that originated from PIE *dj, *gj, *j- and was (allegedly) comparable to italian giorno|day[ITA]←diurnus[LAT]; the result of such a "palatalisation" cannot be a sequence of consonants, namely [dz] as believed by Allen & Co., but at most an affricate, i.e., [d͡z].It is unfortunate that IPA does not have single-character symbols for the affricates, but uses combinations of two consonants (even if linked with a connecting bow above or below), thus misleading many scholars to consider them "stop-fricative sequences" (which they are not). That they are actually single (albeit non-uniform) sounds is suggested by the use of single characters to represent them in many languages, e.g., ç=[t͡ʃ] in Turkish, z=[t͡s] in German, đ=[d͡ʑ] in Serbo-Croatian (not to mention j=[d͡ʒ] in English), and even related digraphs, such as dž and dz are "considered to be single letters" (as is ch in Spanish, not to mention English). We have seen that Greek had an aversion to affricates and an affricate value for ζ (or any other letter) would not be consistent with greek phonotactics, but in the improbable case that this was indeed the "original" value of ζ, it is highly unlikely that it would ever be perceived as a consonant cluster (of which consonants really?) by the Greeks and it would never be considered "analogous", as Allen claims, to the sequences [ks] and [ps], which are not affricates. The confusion of affricates with consonant clusters, even by professional linguists, has already been... lambasted as a notable case of linguinistics.Note also that the true sequence [dz], as in "endzone" endzonesource Wordsmyth and "deadzone" deadzonesource Cambridge is notably different from the true affricate [d͡z], as in "Godzone"(="God's own") godzonesource Collins, "Godzilla" godzillasource Wikipedia and (most perceptibly) ozono|ozone[ITA] ozonosource LEO (note the syllable division [o'dzɔno]). Allen (perhaps repeating a view of Kuryłowicz') also attempted to explain the adoption of single characters for consonant clusters by alleging a special nature for ξ=[ks] and ψ=[ps]: "it may also be noted that these groups do have a structural peculiarity in that they can occur in both initial and final position, and to this extent are comparable in Greek with single consonants rather than with other groups". While this observation is true (if we first accept that ξ=[ks], ψ=[ps]), it can barely be seen how it relates to the perception of the alleged clusters as single consonants (after all, not all single consonants are allowed as word final, but only ν, ς, ρ) and why the renderings κσ, πσ would be considered inadequate. Furthermore, why would Greeks, who made the quantum leap towards true alphabetic writing (the introduction of vowels being a monumental innovation) and strived to produce truly phonetic transcriptions, suddenly abandon the concept "one symbol for one sound" and burden their alphabet with unnecessary symbols (something they did not even do with, e.g., the diphthongs, which also start and end words and moreover belong to the same syllable, unlike [ks] and [ps] that are usually split between syllables by the Catholics to explain metric patterns)? And, if this (i.e., [ks], [ps] being the only clusters that "can occur in both initial and final position") was the reason for the Ionians introducing the special symbols ξ and ψ, why did the Euboeans opt to introduce only Χ(=ξ) in the "western" alphabets, but no letter equivalent to ψ?
Another attempt to provide a credible answer was made by Woodard. In order to back his thesis about "a Cypriot origin of the [Greek] alphabet" he suggests that "the presence of a single symbol with the sequential value of [k]+[s]", which is "a remarkably strange choice for inclusion in the new alphabetic writing system - a system also equipped with symbols having the values /k/ and /s/ individually", maybe due to the fact that "Comparable syllabic symbols (i.e., graphemes representing [k]+[s]+vowel) occur in the Cypriot Syllabary: unlike the alphabetic script, where xi is otiose, the corresponding ksV-syllabic symbols are essential, being uniquely required by the spelling mechanism of the Cypriot script". While interesting, this thesis raises more questions. Why was there a need only for "ksV-syllabic" symbols and not for psV (corresponding to ψ+vowel) in Cypriot? What "spelling mechanism of the Cypriot script" prevented the spelling of the consonant cluster [k]+[s] in the same way used for other clusters (cf. the spelling "a-to-ro-po-se" for "ἄνθρωπος")? But most importantly, how do we know that the symbols of the Syllabary identified as "ksa", "kse", "kso" (or rather "xa", "xe", "xo") phonetically represented [ksa], [kse], [kso] (and not, e.g., [ʃa], [ʃe], [ʃo])? Since the decipherment of the Cypriot Syllabary was based primarily on "a Phoenician-Cypriot bilingual inscription" and secondarily on limited Greek-Cypriot "biscripts", were those symbols identified as corresponding to Phoenician clusters (e.g., כש or כס) or were they identified as corresponding to Greek ξ? The former case would only be possible in case of transliteration of proper names comprising the cluster [ks] in their phoenician script (in order to strongly suggest a [ksV] value); however, no such name appears to have been the basis of decipherment. In the latter case, the value of the symbols would be determined by the value of ξ and they would correspond to [ksa], [kse], [kso] only if we a priori accept that... ξ=[ks] (q.e.d)! In fact, it appears that these values (at least [kse] for "xe") were assumed, in order to make the decipherment compliant with an expected Greek form: in one case (POPE99, p. 134, f. 83) the symbol identified as "xe" corresponds to the ending of the expected (from the phoenician text) word "king" ("ϝάναξ") and in another (p. 125, f. 79) it corresponds to the ending of the proper name (known from the accompanying greek text) "Κάρυξ". With such evidence, the value of the symbol identified as "xe" depends on the very value that we want to determine, namely that of ξ, and, had that been a monophthong in cypriot Greek, there would consequently be no "Comparable syllabic symbols (i.e., graphemes representing [k]+[s]+vowel)" in the Cypriot Syllabary. Actually, the hypothesis that ξ was a monophthong fits better with the evidence from the Cypriot Syllabary, as it implies that "xa", "xe", "xo" were normal CV syllables and obviates the need to assume that the Cypriots devised symbols only for one consonant cluster, but not for others.
And this brings us to the only viable explanation that both the Orthodox and the Catholics hate to admit, but have not been able to refute either: the simple answer is that there is no need for such symbols, unless they stood for sounds hitherto unexpressed in the semitic alphabet and not renderable as combinations of other existing letters. An extreme solution would be to suspect that the last part of the statement was not true, i.e., that there were no symbols for [k] and [p] and/or [s], essentially doubting the assumed values for Κ, Π and/or Σ; however, this would turn the assumed phonological system upside down and, more importantly, would lead to severe difficulties, e.g., the values of Ξ and Ψ are related to those of Κ, Π and Σ, as per (D1) above. It, therefore, follows that Ξ and Ψ could not stand for [ks] and [ps], as commonly assumed, but for some other sound, most likely a monophthong.
It is strange that, as yet, nobody has followed this simple reasoning to propose a single (monophthongal) value for ξ and ψ. Occasionally, some analysts appear to arrive at the same conclusion, but after a closer look it becomes clear that they simply misinterpret the modern pronunciation. For example, Pernot sees (PERN21, p. 70) "un son simple|a simple sound" in ionian Ξ and Ψ based on the use of a single symbol for the sounds rendered elsewhere as ΚΣ, ΠΣ or ΧΣ, ΦΣ, but what he has in mind is certainly not a monophthong, but a... monophthongal pronunciation of /ks/ and /ps/! Indeed, he claims that "Il en est encore ainsi en grec moderne|It is still like this in modern Greek" (which it is not) and compares it to the (affricate) ts of dialectal "métsin"(="médecin|doctor"), as well as the clusters of "accent", "absent", which, he contends, are pronounceable "avec un ks (x) et un ps simples|with a simple ks (x) and a simple ps"! It is evident that he could use a course in (proper) linguistics. An interesting theory is put forward by the Scholiast of Thrax, who appears to be positing a single (albeit long) value for the "doubles", "ἀπεδείχθη γὰρ ταῦτα μὴ συγκεῖσθαι ἐκ δύο συμφώνων, ἀλλ’ εἶναι μακρὰ σύμφωνα|for it has been proven that these are not composed of two consonants, but they are long consonants" (BEKK16, p. 823); however, his "proof" (p. 814) appears to rely on the script alone, as it is based on the observation that ζ, ξ, ψ are "στοιχεῖα|elements" (i.e., letters), which are by definition (or rather axiom) indivisible and cannot be partitioned in two distinct consonants. Consequently, this is rather yet another case of Grammarians' mirage.
Undoubtedly, the present pronunciation is the reason why such a revolutionary proposal has not been made yet and why there are only some clumsy efforts to interpret ΞOC and ΨOC as something that they are not (i.e., "single sounds" instead of consonant clusters). If ξ and ψ had a different value in antiquity, how can we explain their uniform pronunciation as [ks] and [ps] today? In other words, what theory would be consistent with the evidence (D1)-(D6) above? Here is an attempt to reconcile all the contradictory facts:
Under this scenario, everything falls into place; it explains both the introduction of new letters for what are now consonant clusters and the divergent orthography of the epichoric alphabets. One point that I hesitate to include as part of the "official" thesis is my guess about the renderings ΧΣ and ΦΣ. Oddly, unlike Ξ and Ψ, in this case the Catholics interpret the graphic evidence literally, squeezing an "aspiration" between the two elements of the cluster in their reconstructed pronunciations "[pʰs] and [kʰs]" (for, in their "reconstruction", the aspiratae Χ and Φ are [pʰ] and [kʰ] respectively), without bothering to explain how and why it intruded into the original [ks] and [ps] (particularly when they derive from κ+σ, γ+σ and π+σ, β+σ, respectively). Well, such linguistic abracadabra, where "aspiration" suddenly appears without apparent reason, as if by parthenogenesis, is not very credible (unless one is already convinced that the Catholics are in possession of the absolute truth). If we disregard the catholic doctrine and consider the current values Χ=[x], Φ=[f], then we are faced with a possible case of regressive assimilation, the assimilated feature being duration (or, more officially, "stricture"): the stops Κ, Π might have changed to the respective (under the assumed orthodox model) continuants Χ, Φ to match the (undoubtedly) continuant Σ. Such an assimilation may seem unlikely from the modern perspective, but this is only due to the fact that the modern language is fond of dissimilation with respect to duration rather than assimilation (thus, φθάνω→φτάνω). However, Attic exhibits the exact opposite behaviour, as "a labial or velar plosive is regularly aspirated by assimilation when it comes to stand before the -θη- suffix of the aorist passive, e.g. in ἐλήφθειν (from λείπω)" (ALLE87, p. 26; translation: -πθη- and -κθη- expected in conjugation become -φθη- and -χθη- respectively). Thus, a stem like πλεκ- produces ἐπλέχθην in 1st per. sing. aorist passive (cf. μπλέχτηκα in modern Greek).That the assimilated property is not "aspiration", but most likely duration, is suggested by the fact that "Before a /tʰ/ (aorist passive stem), velars become [kʰ], labials become [pʰ], and dentals become [s]" (translation: before θ of the aorist passive stem, κ, γ become χ and π, β become φ, while τ, δ become σ), a "detail" that Allen conveniently omits; in other words, if the result of the assimilation of dental tenues was an undoubtedly continuant and by no means "aspirated" consonant σ, it would appear odd to claim that the result of the same kind of assimilation for the other tenues was a non-continuant, "aspirated" consonant χ or φ. More on the possible values of the aspiratae θ, φ, χ in the respective chapter. Hence, the renderings ΧΣ, ΦΣ might as well be correct representations of the mutated clusters /ks/, /ps/ in Attic, perhaps an intermediate stage before their monophthongisation (to Ionic Ξ, Ψ).
As for the exact value of the ionic monophthongs Ξ, Ψ, we do not have enough data to make an educated guess. The reasonable conjecture would be to assume that the former was a "back sibilant" and the latter a "front sibilant" (κ and π "pulling" σ towards their respective place of articulation), but the exact values would be a matter of imagination or taste rather than factual inference. For the sake of making the discord official, I venture to submit the following (provisional) alternative values, which are proposed only for the original use of the specific letters and digraphs in the respective alphabets (i.e., for Ξ, Ψ in Ionic, for ΧΣ, ΦΣ in Attic), not for the later common use of the two letters:
Ξalt: Ξ ≠ [ks] = [ʃ](?)
ΧΣalt: ΧΣ = [xs]
Ψalt: Ψ ≠ [ps] = [s͎](?)
ΦΣalt: ΦΣ = [fs]
This bold proposal (baseline scenario of divergent evolution of /ks/, /ps/ together with the [xs], [fs] hypothesis for Attic etc) is not as far-fetched as it might seem at first sight. A glimpse at the development of /ks/'s reverse cluster /sk/ in West Germanic is illuminating: in High German it developed into a monophthong [ʃ] (no, I am not making that up!); in Dutch it mutated, but remained a consonant cluster [sx] (cf. attic ΧΣalt!); in Frisian it appears to have retained its original value [sk]. To mention a few cognates:
| Frisian | Dutch | German | |
|---|---|---|---|
| "sheep" | skiep ([skiːp] ?) | schaap ([sxaːp]) | Schaf ([ʃaːf]) |
| "school" | skoalle ([skʊ̯alə]) | school ([sxoːl]) | Schule ([ˈʃuːlə]) |
| "man" | minske ([mẽːskə]) | mensch ([mɛnsx] ?) | Mensch ([mɛnʃ]) |
In the above table the three Germanic languages are colour-coded according to the colouring of their corresponding ancient-Greek dialect groups. If the two developments (of greek /ks/ and germanic /sk/) are indeed comparable, then the Germans would be the counterpart of the Ionians and the Dutch could claim to be the modern Athenians, while the predominance of [ks] in common Greek would be equivalent to the Frisian pronunciation overtaking all other west-germanic ones (not very likely given the present demographics). Furthermore, even though the above proposal rests on the assumption that the attic pronunciation (ΧΣalt and ΦΣalt) was sidelined in Koine (which was based on Attic), this is not as implausible as it might sound, since the prevalence of a non-attic form is paralleled by the dominance of -σσ- (γλῶσσα, θάλασσα, etc) over attic -ττ- (γλῶττα, θάλαττα, etc).
While the above prove that the alternative theory is plausible, I believe that (at least based on the currently available evidence) it is not provable. Its biggest advantage is that it has none of the shortcomings of the consensus view that all of Ξ, ΧΣ, ΚΣ and Ψ, ΦΣ, ΠΣ stood for [ks] and [ps] or the even more extravagant view that ΧΣ, ΦΣ comprised an "aspiration",Sturtevant believes that the attic orthography suggests that, in the clusters /ks/ and /ps/, "κ and π were pronounced with more energy than elsewhere" and proclaims "That the same pronunciation was current in Hellenistic times appears from Armenian transcriptions such as k'sest=ξέστης, k'sip'i=ξιφίας, p'senas=ψῆνας, p'siat'=ψίαθος" (STUR20, pp. 188-189). However, the "Hellenistic times" were over half a millennium before the Armenians even had an alphabet and to claim that this spelling (whatever the value of the armenian consonants represented as p' and k' might have been) reflected actual pronunciation in V AD (or later) rather than mere tradition (cf. Sturtevant's verdict on the similar armenian rendering of greek ῥ on p. 165: "it is likely that the h records a mere school tradition") or convention is liable to be a self-defeating argument, particularly considering that the greek orthography (and not pronunciation of the individual letters) was the model for the armenian spelling at least in the case of ow (ու) = [u]. something that even the Catholics do not really believe (cf. ALLE87, p. 60: "It seems unlikely, however, that full aspiration was involved", "Certainly there is no contrast between aspirate and non-aspirate in this position, and any degree of aspiration that may have existed here can be ignored by the modern reader without any danger of confusion"; also the very advocates of the bizzare theory: "the aspiration of the first element was phonologically irrelevant").
Even though no positive verification of the theory is in sight, one can (following the familiar catholic fashion, where a value is first declared as unequivocally true and then "proved" by means of unilaterally interpreting ambiguous evidence) cite a number of secondary evidence that are easier explained by Ξalt and Ψalt than by ΞOC and ΨOC.
It may be objected that the argumentation starts to be awfully reminiscent of that presented earlier, which I admitted to be bogus. However, there are substantial differences between the two, even though the conclusion is partly the same. First of all, the thesis here is not that the universal values of ξ and ψ in ancient Greek were different from the ones today, but that different values emerged in certain dialects as a result of diverging evolution of the clusters /ks/ and /ps/, leading to the introduction of special symbols (Ξ, Ψ in eastern alphabets or Χ in western alphabets) for the new sounds, which were later used by the other Greeks in the course of the panhellenic adoption of the Ionic alphabet, despite the difference in pronunciation (in other words, Greeks that maintained the "original" pronunciation of /ks/ and /ps/ pronounced in their own fashion, but wrote in the Ionic fashion); a double tradition might also have survived for some time. Secondly, contrary to the catholic practice, I do not consider the above to constitute "proof" of the proposed alternative values; I contend that the presented evidence are more compatible with a (partly) different pronunciation of ξ and ψ than with their traditional values. Finally, I consider the above thesis to be exactly that: a theory that can be neither proved nor disproved based on the currently available evidence; yet, it can be rather safely claimed that (under the Incompatibility Principle) it has better chances than the universally accepted present values, as the latter are not compatible with some of the evidence (particularly, the descriptions of the Grammarians and the reasons behind the introduction of these special symbols).
When it comes to describing vowel sounds, nothing beats the triangle. This refers to the two-pronged shape that results when one aligns the principal vowels of the latin alphabet: a-e-i and a-o-u. Both sequences relate to the distance of the tongue from the walls of the mouth, the first along the "palatal" or "front" axis and the second along the "velar" or "back" one. Their common point is the sound [a] ("open" or "low" vowel, depending on whether one looks at it from the perspective of the air-duct diameter or of the tongue's vertical position, respectively), which is produced when the mouth is wide open and the tongue furthest apart from any mouth wall. Their endpoints are the vowels [i] and [u] ("closed" or "high" vowels), which correspond to the shortest distance from the hard and the soft palate, respectively, before they are considered almost consonants, i.e., the palatal and (labio-)velar semivowels [j] and [w]. Somewhere along the two segments, conventionally at their middle, are the remaining two vowels [e] and [o] ("mid" vowels). It appears that all possible vowel sounds, when represented by a point in the diagram to the right (the horizontal coordinate representing the place of articulation and the vertical the tongue height; *g*=[g], *γ*=[ɣ], *g*=[ɟ], *γ*=[ʝ]), have to be within the triangle. There is no escape from the triangle!In order to accommodate the diversity of their languages in terms of open vowels, the western scholars have stretched or pruned the triangle to a trapezium; however, there is no need to adopt such a concept for the study of Greek, but we will rely on the good ole triangle, which is commonly used in all studies, from Blass (BLAS90, pp. 19-20) to Allen (ALLE87, p. 5).
Today, Greek exhibits exactly the above-cited arrangement of vowel sounds: symmetric and evenly spaced (cf. /i/, /e/, /a/, /o/, /u/ in the diagram to the right; the vowels of the back axis are accompanied by complimentary and proportional "lip rounding"). Interestingly, the same arrangement seems to have been valid for Greek at the time of adoption of the alphabet, as regards the principal vowels Α, Ε, Ι, Ο, Υ (note that Η and Ω are later additions, used by all Greeks only after the panhellenic adoption of the ionic alphabet). Indeed, there is a one-to-one correspondence between these five greek vowels and the latin ones (since the latin alphabet was based on the greek one, the only difference being the new shape of Υ in Latin, as V), for which we have no reason to believe that they differed from their present values in, e.g., Italian and Spanish.I am talking about the pronunciation of the letters and not of the phonemes originally represented by them, as we know that some phonemes have been transformed to others, e.g., "short i" to e in Spanish. Thus, it seems reasonable to consider, at least as working hypothesis, that the three corners of the triangle were represented by the letters Ι, Α, Υ and that Ε and Ο stood for phonemes generally placed near the middle of each of the two axes. The only difference of this model (blue letters) with the present situation in Greek (red letters) is the value of Υ, which appears to have "slipped" to the front of the triangle to coincide with Ι. For the other four, the Catholics and the Orthodox generally agree on their values. Disregarding "quantity" for the time being, there is no discord about the open vowel:
ΑOC: Α = [a]
neither about the front closed vowel:
ΙOC: Ι = [i]
The Catholics have attempted to accurately pinpoint the values of the mid vowels, variably as:
ΕOC: Ε = [e], [ɛ], [e̞]
for the front mid vowel, and as:
ΟOC: Ο = [o], [ɔ], [o̞]
for the back one; however, in the absence of any "competition" from other vowels, this distinction is moot and all three alternatives may be considered good approximations of their actual values (which might have differed dialectically and/or for each person).
At this juncture, there is no point in further discussing the structure of the ancient-Greek vowel system and we can use these values, together with the conjecture about the intended use of the vowels in the original greek alphabet, as a (tentative) guide for the discussions of the next chapters.
In this chapter, we have briefly reviewed the letters that are pronounced more or less the same by both Orthodox and Catholics. The values of the consonants Κ, Π, Τ, Μ, Ν, Λ, Ρ and the vowels Α, Ε, Ι, Ο have been "confirmed", in the sense that we do not have reasons to doubt that they have been retained in present Greek. For Σ, an alternative value [s̺], was proposed, which may be faintly suggested by some evidence (e.g., origin, rhotacism), but the traditional value [s] is equally acceptable. The mystery of the introduction of (present) consonant-cluster shorthands Ξ and Ψ in the first real alphabet (one sound, one letter) is only resolved if we assume monophthongal values in the dialect and time of their invention; the traditional values are acceptable only for the (ultimately prevailing) common language and only in certain dialects before that.
We have also considered some conditional values thereof, which often are points of divergence. Of these, we have strong evidence for the existence of a velar allophone of ν ("velar nasal" spelt γ from V BC) before velars, as unanimously accepted. Having ruled out the possibility of σσ (attic ττ) having an affricate value (and the existence of affricates altogether), we realised that we do not have sufficient evidence supporting the present voiced allophone of σ before voiced mutes in classical Attic and only presumptive evidence for σ before μ. An imaginative account is the catholic theory that ῥ was a "voiceless r", which has no basis on the ancient facts, but rather on modern orthographic practice in unrelated languages (such as Icelandic and Welsh). The catholic answer to the question of whether κ had a palatal allophone before front vowels, as it does today, is indicative of the linguistic confusion of the great Masters of "reconstruction": while accepting a "further forward" value, Allen fails to realise that modern-Greek palatal [c] (which he describes as "palatalized" and inappropriately denotes as "[ky]") is the first candidate for a "fronter" velar.
The conclusions of the present chapter are summarised in the following table:
| Letter | Proposed Values | Verdict |
|---|---|---|
| Π | Π = [p] | accepted |
| T | T = [t] | accepted |
| Κ | Κ = [k] | accepted |
| Κ/_Vfront|front vowel, i.e., [i] or [e] = [c] | very likely | |
| Κ/_Vfront|front vowel, i.e., [i] or [e] = "further forward" [k], but not "[ky]" | epic fail | |
| Μ | Μ = [m] | accepted |
| Ν | Ν = [n] | accepted |
| Ν/_(#|end of word)Cvelar|velar consonant, i.e., γ, κ or χ, also ξ = [ŋ] | accepted | |
| Λ | Λ = [l] | accepted |
| Ρ | Ρ = [r] | accepted |
| Ρ/#|end of word_ = Ρ/Ρ_ = [r̥] | unlikely | |
| Ρ/#|end of word_ = ΡΡ = [r], Ρ = [ɾ] | reasonable | |
| Σ | Σ = [s] | acceptable |
| Σ = [s̺] | probable | |
| Σ/_Cvoiced|voiced consonant, i.e., μ, β, γ (theoretically also δ, ν, λ in compounds)|250 = [z] | likely only before μ | |
| Ξ | Ξ = [ks] | only for common Greek |
| Ξ = [ʃ](?) ≠ [ks] | it's a theory | |
| Ψ | Ψ = [ps] | only for common Greek |
| Ψ = [s͎](?) ≠ [ps] | it's a theory | |
| Α | Α = [a] | accepted |
| Ι | Ι = [i] | accepted |
| Ε | Ε = [e], [ɛ], [e̞] | accepted |
| Ο | Ο = [o], [ɔ], [o̞] | accepted |
By examining and agreeing on the values of the (more or less) commonly pronounced letters, I have set the foundations upon which the further investigation can be based. By pointing out the shortcomings of the catholic theories on some conditional values of otherwise agreed-upon letters, I hope to have sown the first seeds of doubt about the accuracy of the "scientific" arguments. In the next chapters, we will scrutinise (and challenge the interpretation of) the evidence put forward in connection with the letters whose pronunciation does not coincide in the two creeds. Prepare for a journey into the Twilight Zone of linguistics!
|
|
|
|
|