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


After having dealt with linguinistics in the previous chapter, it is time to examine another supernatural feature of the catholic narrative: Quantum Linguistics. There is no better example of this extravagant theory than the so-called "aspirate(s)".

The "aspirate" consonants θ, φ, χ are presently voiceless continuants ("fricatives"):

ΘO: Θ = [θ]

ΦO: Φ = [f]

ΧO: Χ = [x]

The set of continuants [f], [θ], [x] is a difficult class of consonants, as it does not exist in its entirety in many western languages (e.g., [θ] is not part of the german inventory and [x] is not part of the english one, not to mention French, Italian, etc). It is, therefore, understandable that the western scholars have sought (evidence for) alternative values. The opportunity was provided by a feature associated (as it is described with the same adjective and it causes the emergence of aspiratae from the respective tenues) with the "aspirates", the so-called "spiritus asper" or "rough breathing", for which no symbol existed in the (prevailing) ionic alphabet, but notations were later invented, which ended up having the form of a reverse apostrophe () placed over some initial vowels (and ultimately also ρ). The "rough breathing" has been removed from the modern-greek writing system (in the 80s) and before its extinction it bore no phonetic significance (other than that of a reminder to turn a preceding π, τ, κ into φ, θ, χ, respectively). It was, therefore, ignored by the Orthodox when reciting ancient texts:

O: ῾ = Ø

This is the primary point of discord, for the Catholics are convinced that the "spiritus asper" stood for the voiceless glottal fricative:With this term I refer to the realisation of [h] in English and German, which is the model for the catholic pronunciation. It appears that some linguists use "h" as an umbrella term encompassing a variety of sounds and/or features. For example, the expression "There are other languages [such as Hebrew and Arabic] which show a more definite displacement of the formant frequencies for h, suggesting it has a [glottal] constriction associated with its production" implies that what is considered as an "h" in those "other languages" is actually a (possibly related, yet) different feature or sound. This is possibly another example of confusion of script and sound, since it appears that it applies to the various sounds represented by or uses of the grapheme <h> in different languages, which may not coincide. As IPA sounds ought to be language independent (e.g., an [a] ought to be the same sound in all languages that are considered to comprise it; otherwise, it should be represented as [ɐ], [ä], [ɑ], [æ] or [ɒ], as appropriate), I will not consider these alternative definitions, but I will focus on the actual pronunciation of [h] employed by the catholic scholars when reciting ancient Greek.

C: ῾ = [h]

Starting from this identification and for reasons we will shortly mention, the Catholics consider that there is some h in the aspiratae (the latin name of the "δασέα" φ, θ, χ) and specifically, that these were aspirated consonants. In correct IPA notation, the values of the aspiratae according to the Catholic model are:

ΘC: Θ = [tʰ]

ΦC: Φ = [pʰ]

ΧC: Χ = [kʰ]

In other words, according to the official catholic model, the aspiratae were also voiceless (i.e., as also per the orthodox model), but of the plosive type followed by exhalation, or better with an exhalatory release.

We shall examine first the evidence that hinted towards a possible alternative (to the traditional) realisation of the aspiratae and point out the inconvenient "details" that were overlooked or silenced by the Catholics in their eagerness to accept these alternative values. Subsequently, we shall investigate the nature of the "spiritus asper" and seek evidence that may confirm or rule out its catholic value. Finally, we shall explain why the traditional values not only survive the compatibility test, but are additionally the most likely values, at least for the historical period of Greek.

A Quantum of Lingo-Math

The main clue that incited the Catholics to adopt these particular values must have been the transliterations of the greek aspiratae in Latin. The use of Latin as the arbiter for the greek pronunciation is a recurring phenomenon in the study of ancient Greek by the Westerners. In this particular case, the spellings involved make use of the letter h and specifically:

(ΦΘΧ1) The normative transliterations of ῾, φ, θ, χ in Latin are h, ph, th, ch

Since latin h, p, t, c were "reconstructed" as (or assumed to represent) [h], [p], [t], [k], the Catholics concluded that the values of ancient-Greek ῾, φ, θ, χ were exactly what the latin transliterations stood for, namely [h], [ph], [th], [kh]. By applying the principles of their Lingo-Math, this conclusion may be expressed (for the case of φ↔ph) as:

(p = [p])|p was pronounced [p] ∧|and (h = [h])|h was pronounced [h] ∴|therefore ph = [ph]|ph was pronounced [ph]

A similar conclusion is drawn from some Greek evidence, specifically that the aspiratae were produced when the corresponding tenues were followed by the spiritus asper (e.g., οὐκ + ὅπως → οὐχ ὅπως):

(ΦΘΧ2) π῾, τ῾, κ῾ → φ, θ, χ

as well as the spelling conventions of some non-attic/ionic (primarily Melean and Theran; cf. CHAT02, p. 450) alphabets lacking the symbols <Φ> and <Χ>:

(ΦΘΧ3) Some insular inscriptions render φ, χ as ΠΗ, ΚΗ

Here, the grapheme <Η> was used in its old (e.g., pre-Euclidean) function of representing the (later common-greek) "spiritus asper" ῾, which corresponded to latin h (see ΦΘΧ1 above) and is assumed to have had the phonetic value [h].Chatzidakis is the only authority making the claim (e.g., CHAT02, p. 450) that ΘΗ was also written for Θ, but there appears to be only one example ("Θℎαρ(ρ)ύμαϟℎος" in IG XII, 3 763) of ΘΗ for normative (?) Θ, which may easily have been a stonecutter's spelling mistake. Otherwise, the inscriptions of Melos and Thera make the expected use of Θ (see also DAWE95, p. 52), i.e., without the Η/h. Contrary to what is claimed by Jannaris (JANN97, p. 22, §3 and p. 26, §12), there is no instance of ΤΗ for Θ, the latter being readily available in the received phoenician alphabet and not necessitating a surrogate digraph. The "math" governing the conclusions from Greek can be expressed (in the case of <ΠΗ>↔<Φ>) as:

(π = [p])|π was pronounced [p] ∧|and (῾ = <Η> = [h])|῾, formerly represented by Η, was pronounced [h] ∴|therefore π῾ = <ΠΗ> = [ph]|the grapheme <ΠΗ>, representing π῾, was pronounced [ph]

A common premise among the reasonings above is that the value of latin h and of the greek "spiritus asper" ῾ was [h]. This claim will be investigated in a separate section, but let us for argument´s sake concede the catholic value h=῾=[h]. It turns out that, even with this concession, the above simple conclusions are not as sound as they might seem at first glance.

<Xh> ≠ [Xh]

The main concept behind the above conclusions (in addition to the identification ῾=h=[h]) is the conviction that the script faithfully represents the sounds, more specifically the dogma "one symbol, one sound". However, it is not true that in a script every symbol has to represent a separate sound. Consider the following:

Why, then, is it not conceivable that h (<H>) was used in the same function in Latin to indicate that the sounds of the preceding p(=[p]), t(=[t]), c=([k]) should be converted to the respective continuant ([θ], [f], [x]), or equivalently pronounced without contact of the tongue? Why can ph, th, ch and ΠΗ, ΚΗ not be digraphs devised to rendered sounds for which no letter was available?This is actually the interpretation of ΦΘΧ3 by Allen, who asserts (ALLE87, p. 29, n. 36) that "it is simply a matter of a digraph being used for a single sound, where no special single symbol had been inherited". Of course, his view is based on the identification of φ, θ, χ as "aspirated plosives" and he draws the parallel of the same sounds in modern Indian, where they are rendered with single characters in "a Sanskrit script" and with two (plosive+[h]) in a Perso-Arabic script". Nevertheless, the argument works with any kind of single-sound values for the aspiratae φ, θ, for which "no special single symbol had been inherited", examples of uni- and digraph renderings of non-aspirated sounds being abundant (e.g., spanish ñ vs. portuguese nh for the single sound [ɲ], for which no latin grapheme was available).


While the previous observation means that ΦΘΧ1 does not preclude the orthodox values, a clarification made by the most recent Catholic Pope should be enough to destroy all illusions that ΦΘΧ1-ΦΘΧ3 are decisively favouring the catholic values.

In 19th-century notation, the "reconstructed" values of φ, θ, χ were rendered as ph, th, kh, a practise which misled some Catholics to link the latin spellings with the corresponding consonant clusters of their native (germanic) languages (which are almost exclusively found in compounds). In a manner very reminiscent of their inability to distinguish between the series [dz] and the affricate [d͡z], some Catholics still identify their assumed ancient-greek values of the aspiratae with the sequences [ph], [th], [kh]. For example, in "A Compendium of Greek Phonology", it is claimed (p. 69, §14.1) that "they were pronounced as in English uphold, hothouse, and inkhorn" (emphasis in the original).

However, this common misconception, which is at the root of the catholic "reconstruction", is expressly (and correctly) denounced by Allen. Based on the fact that φ, θ, χ behave in (almost) all cases as single consonants rather than consonant clusters (as [ph], [th], [kh] actually are), he concludes that "the aspirated plosive is one sound and not two [...] and so is quite different from the pronunciation of English words like saphead, fathead, blockhead, where the plosive and the [h] are divided between separate syllables" (ALLE87, pp. 28-29).

The main reason why he considers the aspiratae as single sounds is their metrical behaviour, e.g., that "the preceding syllable in a word like σοφός is regularly light and not heavy". In general,

(ΦΘΧ4) φ, θ, χ almost never make position

This fact is, of course, not conclusive, but particularly suggestive of a single sound. Not conclusive, because it may simply be yet another rule governing the metrical Sudoku. However, had φ, θ, χ been (clear) sequences of two consonants (plosive+[h]), they would have been expected to behave like all other consonant clusters and, if they didn't, the Grammarians would have said a word or two about this oddity.

But I believe that the main hint for single-sound values of φ, θ, χ are the old spellings of greek words in Latin, namely before the establishment of the digraphs ph, th, ch (e.g., STUR20, p. 179; ALLE87, p. 22):

(ΦΘΧ5) In old-Latin inscriptions, φ, θ, χ of borrowed greek words correspond to p, t, c

If φ, θ, χ were discernible consonant sequences [ph], [th], [kh] (and if latin h was [h]), why would the Romans refrain from representing them clearly as ph, th, ch (and, thus, distinguish them from p, t, c) right from the start and choose to ignore the second consonant ([h]), which allegedly was part of their phonology? Quite the contrary, the use of the single letters p, t, c to render both π, τ, κ and φ, θ, χ is understandable only if each of the latin letters was only an approximation for at least one of the corresponding two (tenuis and aspirata); as we admit identity between greek and latin tenues, the reasonable conclusion is that φ, θ, χ were single sounds approximated by latin p, t, c.

As a minor clue against a double-consonant value of φ, θ, χ, it can be observed that words like έχθρός would comprise an unprecedented sequence of five consonants (i.e., [ekhthros]), which would be very atypical for Greek.

There is only one safe conclusion from these considerations: φ, θ, χ were NOT consonant clusters, certainly NOT the clusters [ph], [th] and [kh]. In view of this conclusion, ΦΘΧ2 suggests that the laws of mathematics do not apply in linguistics: when a consonant (π, τ, κ) meets another consonant (῾=[h], as axiomatically accepted), the result is a single consonant; in other words 1(consonant) + 1(consonant) = 1(consonant)! This already renders moot the Lingo-Math equations above. The naive interpretation of ΦΘΧ1, ΦΘΧ3 and especially ΦΘΧ2 as indicating that the aspiratae were equal to the succession of the corresponding tenues and [h] does not hold water.

How, then, are these facts to be interpreted? We have seen that even the Catholics admit that ΦΘΧ1 and ΦΘΧ3 are cases of using digraphs for rendering unrepresentable (by the limited symbol sets, i.e., the latin and theran/melean alphabets) sounds. So, the question is what kind of linguistic phenomenon would justify the emergence of a single sound out of the coming together of two consonants, as specified in ΦΘΧ2 (under the assumption ῾=[h]).

One mechanism that produces single sounds out of two different ones is their fusion (merging, coalescence, you-name-it) into a sound that usually shares some characteristics with one and some with the other. We have seen this kind of behaviour in the case of Germanic [sk]→[ʃ], where the resulting sound has the sibilant nature of the first consonant and the (approximate) place of articulation of the second consonant. A similar behaviour is to be found in modern Greek, where a non-syllabic ι=/j/ is incorporated into a previous "palatalisable" consonant [k], [g], [x], [ɣ], [n], [l] yielding the corresponding (single) palatal consonant [c], [ɟ], [ç], [ʝ], [ɲ], [ʎ] (e.g., χιόνι|snow=/ˈxjoni/→[ˈçoni], ήλιος|sun=/ˈiljos/→[ˈiʎos]).This is not always recognisable by native speakers, who think that the greek palatals are combinations of a consonant and ι (i.e., they refer to [c] as κι, to [ʝ] as γι, το [ɲ] as νι, etc), most certainly because of the spelling (there are no special symbols for the palatals in the Greek alphabet). Many (old-school) linguists also have trouble understanding the nature of the greek palatals; for example, Allen refers to the palatalised voiceless velar [c], which is the value of κ before the front vowels [i] and [e], as "the 'palatalized' pronunciation of κ as [ky]" (ALLE87, p. 17). The reason is either that they are not capable of understanding these sounds because their native phonology lacks them or that they had not received proper linguistic training (or both). If the incorporation of the palatal approximant /j/ induces a palatal articulation on the preceding consonant, then it should not come as a surprise if the glottal fricative [h] (assuming this was the value of the "spiritus asper") could lend one of its qualities (i.e., imperfect closure of the vocal tract, a.k.a. sustainability) to a preceding stop [p], [t], [k] to produce the corresponding continuants [f], [θ], [x] as single sounds. The orthodox values are, thus, not ruled out by ΦΘΧ2, even under the assumption ῾=[h].

Pis & Tetas

While the orthodox values [f], [θ], [x] provide a satisfactory solution to the single-sound riddle, whereas the originally ascribed catholic sequences [ph], [th], [kh] do not, the Catholics did not feel like discarding their proposal. They had to adjust it though and the amendment involved the "promotion" of the h to superscript status, as already mentioned. In the words of their most revered Pope, the model used for the catholic sounds is "the voiceless plosives of English, when these begin a stressed initial syllable (as in pot, table, etc.)" (ALLE87, p. 28).Despite this observation, the notation he uses for the "reconstructed" aspiratae is exactly the one he condemns (e.g., p. 18: "the ph, th, kh of Sanskrit and the modern Indian languages"; p. 19: "φ, θ, χ here stand for aspirated [ph], [th], [kh]"). IPA eloquence was not one of his strong points. That is to say that, according to the catholic model, the Φs & Θs ("Phis & Thetas" in traditional transcription) of classical Greek were actually pronounced like "Pis & Tetas" do in English. Despite the... obscenity of this expression in Spanish, it is a more cogent theory than the old one: being single sounds, the phonemes [pʰ], [tʰ], [kʰ] are compliant with ΦΘΧ4To explain a few exceptions to ΦΘΧ4, where φ does make position, Allen essentially argues (ALLE87, p. 28, n. 35) that the h is dethroned from its superscript position, turned into a normal consonant [h] and separated from [p] into a different syllable. His notation "ph" for the aspirate certainly helps him to make this weird claim (or prevents him from seeing the paradox of it). The ancient authorities he mentions have a better explanation, which appears to be rather compliant with φ=[f]. For example, a scholiast considers that this happens "διὰ τὴν σφοδρότητα τοῦ πνεύματος|due to the intensity of breath", which makes sense if it refers to the airstream produced by a prolonged [f], but hardly any for [pʰ], no matter how intensely one pronounces the [ʰ]. And Marius Victorinus ascribes it to a "lengthening of the φ", which is only possible for a continuant (like [f]), but not for a plosive (like [pʰ]); note also that he is speaking of lengthening, not gemination (which would yield [ppʰ] for an aspirated φ) or separation of the plosive and the glottal fricative into the sequence [ph]. and ΦΘΧ5; the spellings with h (ΦΘΧ1) or Η (ΦΘΧ3) would be paralleled by Urdu spellings of the Indian aspirates, which "employs the unaspirated consonant-symbols combined with h" (ALLE87, p. 29, n. 36) and possibly some old spellings th for word-initial (aspirated) t- in German (e.g., BLAS82, "Thatsächlich" on p. 6 and "Thierlauten" on p. 14).Probably, also, the pronunciation [ˈt(ʰ)ɒm.əs] for Thomas (from greek Θωμάς=[θoˈmas]) is due to a spelling pronunciation caused by such as convention.; the transformation of the "spiritus asper" (always assumed to represent [h]) to an h-appendage (ΦΘΧ2) would be paralleled by the nasalisation of a preceding vowel by a following n or m in French (however, this happens only when the vowel and the nasal are in the same syllable, whereas the original π/τ/κ and the ῾ are in different words, e.g., οὐκ + ὅπως → οὐχ ὅπως; moreover, in Greek the merger takes place in the syllable onset, wheres in French the nasal is part of the coda, which is a more vulnerable position).

This is a noble attempt to reconcile the naive interpretation φ=ph=[ph], θ=th=[th], χ=ch=[kh] with the incompatible conclusion that φ/ph, θ/th, χ/ch represented single sounds. Nevertheless, it is apparent that the connection between the "reconstructed" values of the "spiritus asper" ([h]) and the aspiratae ([pʰ], [tʰ], [kʰ]) is not as direct as the Catholics initially believed. The assumed [h] of the spiritus asper is not merely appended to the preceding tenues to produce the corresponding aspirata, but it fuses therewith to produce the single sound that is required to explain ΦΘΧ4 and ΦΘΧ5. According to the Catholics, this fusion involves the marginalisation of the assumed [h] to the superscript. However, if one needs to discard the straightforward solution (φ=ph=[ph], θ=th=[th], χ=ch=[kh]) to ΦΘΧ1-ΦΘΧ3 and accept some degree of fusion, then the "fully fused" values [f], [θ], [x] are as good solutions as the "partially fused" ones [pʰ], [tʰ], [kʰ].

Mowing Grassmann Down

Despite the acknowledgement that the sequences [ph], [th], [kh] are not appropriate solutions, they are often used to justify further weird Lingo-Math put forward as evidence of the "aspirated" nature of φ, θ, χ. For example, Allen attempts (ALLE87, pp. 15-16) to use Grassmann's Law:

(ΦΘΧ6) Two consecutive syllables of the same word cannot bear "aspiration"; if something's got to give, it is always the first oneThere are a few exceptions to this rule:
- in the first aorist passive imperative, it is the second one that gives, when it follows -θη-: *λύθηθι→λύθητι (SMYT20, p. 31, §125b);
- in aorist passive and the perfect infinitive, -θην, -θαι and -θι do not affect a previous "aspirate": ἐθρέφθην, πεφάνθαι, γράφηθι (SMYT20, p. 31, §125N);
- word-initial ὑ- (which is always "aspirated") is not affected by a following "aspirate": ὕφος, ὑφίστημι (cf. ἔχω);
- isolated cases, such as ἁφή.

to prove the relationship between the tenues and the aspiratae; after comparing the pair ἔχω/ἕξω (which he identifies as "[ekhō]"/"[heksō]", thus anticipating the very point that he wants to prove!) with τριχός/θριξί, he concludes that "τ is to θ as zero is to [h]-in other words τ stands for [t] as θ stands for [th]". There are several levels of arbitrariness, intentional ambiguity and circular argumentation here. As is typically the case, the value of the "spiritus asper" ῾ is axiomatically taken to be that of the germanic consonant h; since it is also our working hypothesis (until the time comes to place it under scrutiny), we can live with this harmless assumption. The similar adoption of the value "kh" for χ (which is essentially the point to be proved) is also not critical, as it is not used in the conclusion. However, the main concern with this argumentation is the interpretation of Grassmann's Law. All the "law" tells us is that two consecutive "aspiration"-bearing syllables were not tolerated (whatever "aspiration" may stand for); thus, the pairs ἔχω/ἕξω and τριχός/θριξί do not tell us more than we already know: the second ἕ- comprises "aspiration" (δασεῖα) and the θ- is an "aspiration"-bearing consonant (σύμφωνον δασύ), while on the other hand the first ἔ- bears "non-aspiration" (ψιλῆ) and the τ- is a "non-aspiration"-bearing consonant (σύμφωνον ψιλόν). But Allen appears to believe that a linking relation is implied. At first sight, the wording ("x is to c, as a is to b") seems to suggest analogy, of the kind applied in the simple rule of three:

᾽|spiritus lenis/῾|spiritus asper = τ/θ → θ = τ ῾/᾽ = [t] [h] / Ø,

but this kind of math is meaningless, as multiplication and division cannot be defined for phonetic values (particularly division by Ø). Instead, Allen considers that Grassmann's Law suggests that the difference between ᾽ and ῾ is the same as between τ and θ, the term "difference" having its strict mathematical meaning:

῾|spiritus asper-᾽|spiritus lenis = θ-τ → θ = τ + ῾ - ᾽ = [t] + [h] - Ø = [t] + [h] = [th]

Here, too, we have the axiom that addition results in concatenation, i.e., into a consonant sequence ([t] + [h] = [th]), which, paradoxically, has to be a single sound (let us not forget that, in "reconstructing" θ, Allen has rejected the [th] of "fathead" and has proposed "the initial [...] t [...] of English", which is [th]). The messed-up notation ([th] instead of [th]) allows Allen to sell the "aspirated-plosive" tale under a scientific (mathematical) guise, but a careful reader would realise that he ends us contradicting himself. In other words, the Grass(mann) is not greener on the catholic side of the fence.

Looking For Greener Pastures

We have seen that any attempt to apply mathematical logic does not verify, at the same time, the catholic value [h] of the "aspiration" and the (adjusted) values [pʰ], [tʰ], [kʰ] of the aspiratae. It is evident that we have been applying the wrong principles or made the wrong assumptions about the nature of "aspiration" (or both). As far as the mathematical operations are concerned, it can be safely concluded that what appears as addition (+) in all of the above equations is not concatenation ([t] + [h] ≠ [th]), but involves some kind of alteration of the nature of the sound(s) combined with the "aspiration". One possible candidate for this alteration has been identified as the "fusion" of the sound with the assumed value [h] of the "aspiration" (to the catholic [pʰ] etc or to the orthodox [f] etc or to whatever other single sound). However, this conjecture still rests on the assumption that "aspiration" is the consonantal sound [h]. It is time to investigate this working hypothesis.

From Grassmann To Grasshopper

Let us start the investigation by revisiting the catholic interpretation of Grassmann's law. It is true that Allen does not use the above equation (῾-᾽ = θ-τ) verbatim. He may imply it ("τ is to θ as zero is to [h]"), but he is careful enough not to engage in clear argumentations that may disprove his point. His actual words when he makes use of Grassmann's law to establish the value of θ are "the relationship between the values of e.g. θ and τ was the same as that between [h] and zero, i.e. presence and absence of aspiration" (p. 20). This ambiguous expression is a more reasonable conclusion, but it hardly proves that θ was τ appended with another sound, namely [h]. Grassmann's law indeed tells us that two consecutive syllables may not bear "aspiration", but not what "aspiration" actually is. The law would make sense for the catholic definition of "aspiration"=[h], i.e., an individual sound, if that sound introduced each "aspirated" syllable, as a case of intolerance to alliteration. However, if ῾, φ, θ, χ were [h], [pʰ], [tʰ], [kʰ] (as per Catholicism), then the [h] would be internal (and "marginalised" to an exponent) in the syllables of, e.g., *θριχός(=[tʰri.kʰos])→τριχός and initial in the first syllable of *ἕχω(=[he.kʰo])→ἔχω. To be intolerant to the consecutive presence a floating sound (jumping from exponent to baseline and vice versa) is certainly curious and cannot be associated with any known linguistic phenomenon. Grassmann's law is usually advertised as a typical case of (systematic!) dissimilation; yet, the whole argument relies on the representation of the aspirates as [ph], [th], [kh] and the treatment of the aspirated release [ʰ] as an individual sound of the consonant sequence. In other words, back to square one!

Another mystery associated with the "aspiration" as the individual sound [h] is the theory that, when a tenuis (κ, π, τ) meets an "aspiration"-bearing (δασυνόμενον) vowel, the "aspiration" leaves the vowel and is "transferred to the consonant" (ALLE87, p. 20), yet this same "aspiration" [h] is able to jump over other sounds, specifically ρ (e.g., τετρα|quad+ἵππος|horseτέθριππον|four-horse chariot, also προ+ὁδός→φροῦδος, προ+ὁρά→φρουρά; ALLE87, pp. 20, 43), or split in two while jumping (νύκτα+ἡμέρα→νυχθήμερος)Smyth (SMYT20) provides two more examples of this kind: νύκτα ὅλην→νυχθ᾽ ὅλην (p. 31, §124) and ἑπτά+ἡμέρα→ἑφθήμερος (p. 26, §82). Note the violation of Grassmann's law in the second example! or to spontaneously reproduce (ε-λειπ-θην→ἐλήφθην, ε-δερκ-θην→ἐδέρχθην; ALLE87, p. 26). No math (of the kind we have seen so far) can justify this behaviour, if "aspiration" is a sound [h].

Here is Allen's explanation of the compositions of the first kind (τέθριππον, φροῦδος, φρουρά): "the ρ in these words will first have become aspirated (de-voiced) before an aspirated vowel (which then loses its aspiration in the compound: i.e. ρἱ, ρὁ → ῥι, ῥο), and this in turn will have required that the preceding plosive be aspirated" (ALLE87, p. 43). Note how, according to this theory, the spiritus asper ῾ leaves the vowel (ι, ο) and lands on the ρ, as if it were a flee or migratory locust. Certainly, Allen has in mind a transformation of the kind ρ-hι, ρ-hο → ρh-ι, ρh-ο, where the "aspiration" is detached from the front of the vowel and attached at the end of the ρ, making it an "aspirated" ρ; this way of thinking is identical to the reasoning examined at the beginning of this chapter, according to which the spiritus asper [h] of an aspirated vowel is combined with a preceding plosive [p], [t], [k] to yield φ=[ph], θ=[th], χ=[kh]; again, back to square one (or rather... two: ΦΘΧ2). However, the analogy ends there: in the case of π/τ/κ+῾, the assumed consonantal value [h] of ῾ is allegedly reduced to an aspirated release (superscript ʰ); in the case of ρ+῾, the catholic (allenian) thesis is that the effect of the aspiration's ([h]) incorporation into ρ was "that the sound was a 'breathed' or voiceless [r]" (ALLE87, p. 42), a theory that we have practically ruled out in the respective section; then, this allegedly causes a mutation of the sound before it, specifically necessitating "that the preceding plosive be aspirated". However, if a "voiceless [r]" is to induce a modification on a preceding sound that would not be induced by the normal (voiced) ρ, the modification would have to relate to the only difference between these two ρs, which is (according to Allen) voice. In other words, if ρ's "voice" played any role in the mutation, it follows that the mutae preceding the normal voiced ρ (i.e., the τ, π in the original τετρ-, προ-) would be voiced and that those preceding the "voiceless" ρ (i.e., the θ, φ in the resulting τέθριππον, φροῦδος, φρουρά) would be voiceless. Yet, according to both orthodox and catholics, voice is the distinctive characteristic of the mediae (β, δ, γ), while both the tenues (π, τ, κ) and the aspiratae (φ, θ, χ) are voiceless; instead, the catholic point of view is that θ, φ differ from τ, π only in the presence of "aspiration". Therefore, Allen's explanation essentially suggests that voicelessness (of ρ) does not induce voicelessness, as would be expected, but aspiration! Where is the linguistic basis for that? What kind of phonetic mechanism would account for this behaviour? Where is a similar behaviour attested?

In the compositions of the third kind (ἐλήφθην, ἐδέρχθην), the transformations π→φ, κ→χ are obviously due to assimilation: the plosive preceding θ assimilates to the consonant type of θ, namely aspirata. In the second kind (νυχθήμερος), the transformation κ→χ may also be attributed to assimilation, after the following τ had been turned to θ due to the spiritus asper of η (of the second constituent ἡμέρα); in that case, if vowel "aspiration" indeed corresponds to [h], the reason for becoming "aspirated" would be different for each plosive: τ→θ would be caused by the incorporation of [h] into τ (fusion), but κ→χ would be induced by the resulting θ (assimilation). Although strange, there is at least one example of a similar behaviour in modern Greek: a non-syllabic ι̯ (/j/) may merge with a preceding velar ([k], [g], [x], [ɣ]) to produce the corresponding palatal ([c], [ɟ], [ç], [ʝ]), which in turn palatalises a preceding nasal ([ŋ]→[ɲ]), e.g., μάγκας|wiseguy=[ˈmaŋgas], but μαγκιά|wiseguy behaviour=[maɲˈɟa]. A similar explanation (merging of [h] with the closest consonant, induction of aspirated articulation on the other) could be claimed for the compositions of the first kind (τέθριππον, φροῦδος, φρουρά), if we could come up with a reasonable definition of an "aspirated" ρ; however, the catholic identification of ῥ with a "voiceless r" and the resulting claim that voicelessness induces aspiration are not well-substantiated theories.Allen cites "the voiceless pronunciation of r (and l) after the aspirated allophones of English voiceless plosives [...] e.g. in pray, please" (ALLE87, p. 43, n. 74) as English parallels to the statement of Choeroboscus that ρ's "aspiration" conforms to that of a preceding muta (CHOEPR, p. 25, §44: "τὸ ρ μετὰ τῶν δασέων δασύ ἐστι|ρ is thick [='aspirated'] after the thick [='aspirated'] [consonants]"). However, besides the facts that r in English is not an alveolar trill (as postulated for ancient-greek ρ) and that the (alleged?) de-voicing also concerns l (and reportedly vowels and other sonorants, as well), which has no parallel in ancient Greek, the cited phenomenon, if not an illusion, is a case of an aspirated consonant causing voicelessness on r and not vice versa. Thus, this english "parallel" would relate to τέθριππον, φροῦδος, φρουρά at all, only if the primary step would be the conversion of the tenues (τ, π) to aspiratae (θ, φ), which would then cause ρ's voicelessness. But a conversion of the tenuis as a first step would require that the (alleged) [h] of the second constituent (ἵππος, ὁδός, ὁρά) jump over the ρ, in order to have an effect on the former. In other words, to justify this conversion, we have to assume either a paranormal assimilation ([h]→voiceless ρ→aspirated τ/π) or a grasshopper [h] (transferred from after the ρ to before it).

Coup de... Grass

We have seen that the assumption ῾=[h] leads to a paradox or two, particularly parthenogenesis and/or mobility of [h], which cannot be explained even under the "adjusted" values of the aspiratae [pʰ], [tʰ], [kʰ]. However, there is a more important property of "aspiration" which practically precludes it being a consonant and is conveniently overlooked or brushed aside by the Catholics.

We have seen under ΦΘΧ2 that a tenuis, when followed by a vowel bearing the "spiritus asper", is converted to the corresponding aspirata. This is, however, possible only in the case of οὐκ, which is the only greek word ending in a tenuis.Another word ending in κ is the preposition ἐκ; yet, before words starting with a vowel, it has the form ἐξ, so it never results in κ+῾. In all other cases of conversion, there is an intervening vowel between the tenuis and the spiritus asper, which should normally prevent any interaction between the two consonants (if, indeed, ῾=[h]). Most of the times, the intermediate vowel is dropped, i.e., elided (as in κατὰ ἡμέραν→*κατ' ἡμέραν→καθ' ἡμέραν), which is not a typical behaviour before consonants, but nearly universal before vowels. Often, the intervening vowel is merged with the following "aspirated" vowel, in a process called crasis (e.g., καὶ ὅπως→χᾥπως, also the already seen προ+ὁδός→φροῦδος, προ+ὁρά→φρουρά), not being bothered by the consonant [h] that allegedly stands between the two vowels, but is miraculously "transferred" to the other side of first vowel (and even a preceding ρ). This can be summarised (in Allen's words) as:

(ΦΘΧ7) "Aspiration" does not prevent elision and crasis

In other words, the phonotactics of ancient Greek behaves as if the (consonant) [h] were not there. Well, if there is no smoke, there is probably no fire either.cf. the similar behaviour of English, where the expression "an honour", as opposed to, e.g., "a horror", indicates that the h "is not sounded", i.e., there is no [h]. That the "spiritus asper" in these cases could not have the value of English h is evident if one considers modern coined words, like antihero and antihistamines: in accordance with ancient-greek phonotactics (ΦΘΧ7), the final vowel of anti should have been elided and the τ turned into a θ (ἀντί+ἥρως→*ἀνθήρως, ἀντί+ἱσταμίνη→*ἀνθισταμίνη; cf. ἀντί+ἥλιος→ἀνθήλιον=anthelion, ἀπό+ἥλιος→ἀφήλιον=aphelion); however, this elision did not come natural to the coiners of these (modern) words, obviously due to their not being philologists (thus, ignoring the ancient-Greek rule) and to their pronouncing h as [h].

There appears to be no way to reconcile ΦΘΧ7 with the catholic value C. For this reason, most catholic scholars choose to ignore ΦΘΧ7 or propose still more exotic characteristics for the "aspiration", e.g., that its application on a vowel results in a "voiceless vowel" (rather than [h]+vowel)! Yet Allen has ventured a sorry excuse for a linguistic theory: "there is nothing to prevent the same type of sound functioning as a consonant in one language and as a 'prosody' in another", (ALLE87, p. 54). Here, Allen is trying to hide behind the ambiguous term "prosody" (προσῳδία), which is used by ancient Grammarians to describe the "spiritus asper", as well as the accent marks (´ ῀ `) and the feature of vowel "length" (μακρά and βραχεῖα); the latter two are clearly "simultaneous feature[s] of the vowel[s]" and, hence, do not interfere with the interaction between a vowel bearing these features and a preceding vowel, i.e. with elision and crasis. Having that in mind, Allen's sibylline statement affords a number of (self-contradictory) interpretations:

In all interpretations of the aforementioned statement, the common point is that "aspiration" is something (a consonant), but at the same time it is not (in order to justify ΦΘΧ7)! Such behaviour is typical in the subatomic level, but is not justified macroscopically, so that it would be surreal to establish a branch of... Quantum Linguistics. Without any substantiation with scientific evidence and with no backing by real-world examples, Allen's statement is equivalent to saying "there is nothing to prevent the same humans existing as material beings on Earth and as spiritual ones on Jupiter". The theory that a sound can be here a "consonant" (i.e., a segment) and there a "prosody" (i.e., a suprasegmental feature) is a typical case of Phoneytics.

A typical treatment of [h] that (in the catholic minds) would justify its disregard for the purposes of elision is to regard it as something elusive. Allen subscribes to Gimson's characterisation of [h] as "a strong, voiceless onset" (ALLE87, p. 54, n. 108), whereas the guru of phonetics describes it as a segment with only laryngeal specifications. However, any phonetic particle (be it an "onset" or a segment) that intervenes between two vowels would prevent their interaction (and, thus, elision and crasis of the two vocalic nuclei).The need to eliminate any intervening sound, in order for crasis to take place, is implicitly recognised in the etymological derivation of the "regular active infinitive of verbs in -ω", i.e., -ειν or -ην "both arising by contraction < -ε-εν [-e(h)en] < *-e-sen" (HORR10, p. 10). Nobody dares propose that the "contraction" of the two ε’s took place in the presence of an intervening segment (/s/ or its alleged mutation [h]), but their direct adjacency is assumed to be a prerequisite. Note that, even under the curious theory of aspiration throwback, if there were an [h] (as a result of /s/ mutation), it would have jumped to the previous onset, which is not suported by the surviving forms, e.g., πλέκειν, not †πλέχειν.

It has also been suggested to me by a linguist that elision of vowels before consonants is not that uncommon, "is not"→"isn't" in English and "από το σπίτι"→"απ᾽το σπίτι" in (modern) Greek bearing witness to that. Nevertheless, the mentioned examples are not cases of apocope, i.e., elision of a word-final vowel, but of syncope, i.e., elision of an unstressed word-internal interconsonantal vowel (the greek example being essentially a single word augmented by two unstressed proclitics, the accent mark on από being an unfortunate convention of modern-Greek orthography). Moreover, if syncope were the reason behind the disappearance of vowels before the consonant [h], this could not be selective (i.e. only before [h]), but it would have to occur before other consonants as well, e.g., ἀπό τούτου→†ἀπ᾽ τούτου beside ἀπό ὅτου→ἀφ᾽ ὅτου. The answer can, thus, not be syncope.If I were to confuse apocope with syncope in some linguistic forum, the professionals would eat me alive and throw my bones to the cyberdogs! Instead, the underlying mechanism that produces ἐφ᾽ὅσον (ἐφόσον) from ἐπὶ ὅσον must be the same as the mechanism that produces ἐπ᾽αὐτοῦ from ἐπὶ αὐτοῦ. And this mechanism is universally identified as "expulsion of a short vowel at the end of a word before a word beginning with a vowel" (SMYT20, p. 23, §70). It is evident that, if ῾ were "pronounced as h, which is sounded before the vowel" (p. 9, §9), then the second word would not begin with a vowel, i.e., the precondition for the elision would not be satisfied and the elision would not be justified.

Hit the Hay, [h]

In two full centuries of intensely theorising about the "aspirated" nature of φ, θ, χ, the Catholics have not been able to come up with a credible explanation of ΦΘΧ7, i.e., why ancient Greeks allegedly had a sense of the "function" of the sound [h] different from that of the English and the Germans, after whose pronunciation the sound of greek ῾ was modelled. In a parallel universe, it may be possible that elision occurs also before [h], but in the present universe no such case is reported. A single example of a living language that ignores an intervening [h] for the interaction between vowels, yet maintains it after the vowels have "interacted" (so that it can be attached to a consonant preceding the first vowel), still remains elusive.

Since the early days of Greek studies, ΦΘΧ7 has been considered a consequence of Attic prejudice against hiatus, which is defined as "the immediate succession of two vowel sounds in adjoining syllables" (SMYT20, p. 18, §46). It is evident that, under the Catholic assumption that ῾ is a [h] preceding an "aspirated" vowel, there is no "immediate succession" of the two involved vowels.It is true that, in Latin, elision is not prevented by an intervening final m, e.g., "multum ille et"→"multillet". However, this is not comparable to ΦΘΧ7. First of all, it is not clear what the status of word-final m was in Latin (cf. Italian molto and Portuguese muito vs. Latin multum, no Romance language preserving the -m), at least in the works that treat it as if it were not there (e.g., it might have been purely orthographic, the actual pronunciation following some "Vulgar Latin" model). But most importantly, m is also elided together with the vowel, whereas ῾ in ΦΘΧ7 survives elision. It is, therefore, imperative that C be regarded as extremely suspect, at least until elision before [h] is verified in observed data.

In short, there seems to be NO WAY to justify elision (apocope) in front of [h], unless the argumentation enters the realm of metaphysics or... quantum physics!!!

It is also worth pointing out the inconsistency of treating the two πνεύματα of ancient Greek, the ψιλή (᾽) and the δασεῖα (῾), by the Catholics. We have seen how Allen (as well as Thumb; THUM88, p. 91) discards the theory that the former was the glottal stop (promoted i.a. by Chatzidakis) by aptly observing that this "is almost certainly ruled out by the fact that unaspirated initial vowels in Greek permit elision and crasis, which would be highly improbable if they were preceded by a stop articulation" (ALLE87, p. 56). It is not clear why [h], the assumed value of the δασεῖα, would be more "transparent" than [ʔ], the rejected value of the ψιλή, so that it can be ignored in the interaction of the vowels, but also be able to itself interact with the preceding consonant. For reasons of scientific honesty and consistency, particularly without providing any evidence to the contrary in any living language, the assumed value of [h] for the δασεῖα should have also been declared "highly improbable".

The Constant Gardener

Instead of building on the only reasonable conclusion from the above (namely that a consonantal value for the δασεῖα is extremely problematic), the Catholics chose to adapt their theories to the preconceived value ῾=[h], thus following the Recipe towards linguistic fraud. The most deceitful is the second step, namely that of "Evidence Selection", where ΦΘΧ7 is usually ignored, as it is not compatible with the assumption ῾=[h], or declared as not decisive, like Allen’s unfortunate statement. Thus, the various catholic treatises invariably point out that an aspitata is produced when a tenuis meets the δασεῖα (ΦΘΧ2), without mentioning that a prerequisite for such a confluence of the two sounds is that an intervening vowel get out of the way (ΦΘΧ7).

There is, however, a single (to my knowledge) exception to this rule. Kehrein provides an elegant explanation of ΦΘΧ7 (always under the assumption that ῾=[h]) in the course of his "Prosodic Account" theory (KEHR02; pp. 66-212). The main points of the latter are essentially:

On this basis, the examples of elision and crasis, as well as others that involve a traveling "aspiration" (assimilation, metathesis), are explained (pp. 199-201) as attempts of "aspiration" to escape elimination (or "neutralisation"). In particular, the explanation stipulates that whenever the feature [spread] is found (due to composition, proclitics or mutation *s→[h]) in a position that it cannot be sustained, it attempts to move to a more favourable position under the constraints that [spread] in ancient Greek:

C1 is essentially identical with the observation of Choeroboscus that a syllable does not end in an aspirata (CHOEPR, p. 25, §44: "οὐδέποτε οὔτε συλλαβὴ οὔτε λέξις εἰς δασὺ καταλήγει|never does a syllable nor a word end in an aspirata"). C2 explains the appearance of δασεῖα in the word ἱερός from *iseros→*iheros (the word-medial [h] allegedly moves to the onset of the previous syllable ι, where it is sustainable). C3 explains the derivation of ἅρμα from *arsma→*arhma (as h cannot survive in the lone company of sonorants). C4 explains why πολυ+ἵστωρ yields πολυΐστωρObviously, Kehrein’s rendering of this word as "poluistóòr" (with a "circumflex" on the last vowel) is due to a misunderstanding (obviously misinterpreting Buckley’s symbol ǭ, i.e., "long ǫ"=ω, for ῶ) and betrays his unfamiliarity with the actual language, which nevertheless does not prevent him from developing fancy theories based on the prevailing doctrine, i.e., the propagation of the catholic dogma. and not †φολυΐστωρ ([spread] does not survive in the previous onset either, but does not move further forward).

The phenomena of elision and crasis when a [h] (the presumed value of δασεῖα) is involved would, thus, be explained under the assumption that the [h] parasite first gets out of the way, in an attempt to find a more favourable host, thus leaving the way open for the normal operation of elision and crasis: προ+ὁρός→*῾(πρ)οορός→φρουρός (where ῾(πρ)=[pr]+[spread]=φρ and the crasis of οο yields [oː]=ου), τετρα+ἵππος→*τε῾(τρ)α-ιππος→τέθριππος (α being elided before ι).

In essence, Kehrein's theory stipulates that the order of events is the reverse of that assumed by Smyth: "A smooth stop (π, τ, κ) brought before the rough breathing by elision, crasis, or in forming compounds, is made rough, becoming an aspirate (φ, θ, χ)" (SMYT20, p. 31, §124). It is, therefore, an attempt to reconcile ΦΘΧ7 with the incompatible assumption C by giving up as little as possible, i.e., abandoning the conventional view of elision before δασεῖα in order to maintain the assumed value of δασεῖα.

Although commendable and more scientific than the other catholic theories, Kehrein’s explanation is guilty of the same abuse: cherry picking ("Evidence Selection"). In particular, the survival instinct of [spread], as stipulated by C4, cannot be selective, i.e., only in the cases examined by Kehrein, but should apply universally in all cases of composition with a (word-initial) δασεῖα. Consider the following examples: πρόωρος (προ+ὥρα), προσάπτω (προς+ἅπτω), ἔνορκος (εν+ὅρκος) and Πανέλληνας (παν+Ἕλληνας) in Hom. Il. 2.530. In all these cases, the (always word-initial) "aspirated" vowel (presumably [h]+V) is found in word-medial position after composition, where it cannot "survive" (in the first case it would be intervocalic, in violation of C2, and in the others it would have to combine with a sonorant, in violation of C3). If C4 held true, the feature [spread] would have migrated to the previous onset, which provides a favourable environment in all cases (as †φρο-, †φροσ-, †ἕν-, †Φαν-, respectively), which it does NOT! It is, therefore, fairly safe to say that assumption C4 is not justified by the evidence.The other two cases (*iseros→*iheros→ἱερός, *arsma→*arhma→ἅρμα) of δασεῖα emerging from an etymological *s (allegedly mutated into a [h]) are not important for the present investigation, as they have nothing to do with ΦΘΧ7. Nevertheless, in addition to being inherently speculative (as they involve rather doubtful etymologies), they fail to explain why, e.g., *esmi has developed into εἰμί (deletion of *s and compensatory lengthening of *e) and not to †ἑμί (*esmi→*ehmi→*hemi), as would have been expected, in analogy to the alleged *arsma→*arhma→*harma.

I am not aware of a single instance of composition involving the transfer of "aspiration" to a preceding onset without an accompanying interaction of the intervening (between the onset and the "aspiration") sounds, specifically crasis or elision. The only reasonable explanation would, therefore, be the traditional view that the crasis/elision is responsible for any interaction between the tenuis (essentially the entire onset) and the "aspiration" and not any metaphysical desire of the "aspiration" to migrate when endangered.

In addition to its non-universality, I have doubts about the plausibility of the sequence of events that would be required according to Kehrein’s theory, which necessitates three individual steps in the case of composition, such as επι+ἅπτω→ἐφάπτω and συν+ἅπτω→συνάπτω: bringing the two constituents together (*επιἅπτω, *συνἅπτω); realising that [h] is not sustainable in word-medial position and moving it to the previous onset (*επ῾ιάπτω, *σ῾υνάπτω); determining whether it is sustainable in the new position and merging it with a tenuis, if it is (ἐφάπτω) or deleting it, if not (συνάπτω). This would make sense, if it could be seen as a series of linearly applied sound laws, wherein the assumed intermediate forms (*επιἅπτω, *συνἅπτω; *επ῾ιάπτω, *σ῾υνάπτω) would have survived for some time before the new law kicked in. However, composition such as κατα+ἡμέραν→καθ’ἡμέραν must have occurred in the course of an individual’s daily life, and would not merely be mutated relics of long-established terms. Since composition is a productive process, the second and third steps would have to be performed as soon as the first takes place, i.e., as soon as the two constituents are brought together. It appears a rather heavy cognitive burden for a speaker to have to realise the non-sustainability of [h] in the original position, to make sure that it not be moved further than one onset ahead, to decide whether the new position is favourable and to take care not to iteratively apply the second and third steps, if the new position is not favourable; and all that in real time.

As a further remark on this prima facie elegant, yet problematic attempt to justify the catholic preconception ῾=[h], I would like to point out that Kehrein also relies on C4, namely the rush forward of "aspiration", to explain Grassmann’s law (particularly the variants ἔχω↔ἕξω and θριξ↔τριχός) as a case of "laryngeal movement" (p. 200). But it cannot explain Grassmann in reduplication of aspiratae, where there is loss of (the first) "aspiration" instead of migration. Thus, C4 is essentially the comeback of the "aspiration-throwback" (ATB) theory, which was put forward by some indian Grammarians, but apparently is rather problematic and, consequently, not very popular among modern scholars.

After having done away with the last of the catholic excuses, the evidence is incontrovertible: to pronounce the δασεῖα as the English h is horrible, hideous, abhorrent!


Since we have invalidated the working hypothesis ῾=[h], upon which the discussion relied so far, we need to take a step back and examine whether there is other evidence that would lessen the impact of ΦΘΧ7 and would make C less of an improbability.

Declaration of Independence

At this juncture, it is imperative to discuss nomenclature, as there is the danger of being prejudiced by the current conventional meaning of the word "aspiration" and its derivatives. The feature under investigation in this subsection is called "δασεῖα" in Greek and the (word-initial) vowels it accompanies are called "δασυνόμενα" (φωνήεντα). As we have seen, the letters φ, θ, χ, the value of which is to be determined in this chapter, are called "δασέα" (σύμφωνα). As is evident, these terms share the same etymology and could be considered to relate to the same property, conventionally called "δασύτης" in Greek. A correct rendering of these terms would probably be "aspirate" for δασεῖα, "aspiration" for δασύτης, "aspirated" for δασυνόμενα and "aspiratae" for δασέα. However, the current catholic practice is to refer to both vowels and consonants as "aspirated", to use "aspiration" not for the property, but for the feature, for which the term "aspirate" is also occasionally used. To make matters worse, the "aspirate" is axiomatically identified with the segment [h] (cf., e.g., ALLE87, p. 52, "The aspirate* [h]"), which leads them to analyse "aspirated" vowels as sequences of segments (i.e., ἁ as [ha]) and "aspirated" consonants variably as sequences of segments or imperfectly released stops (i.e., φ both as [ph] and [pʰ]), as they see fit. Since we have seen that the identification of the δασεῖα with [h] or any other (consonantal) segment is not a tenable theory and since the notion of "aspiration" is so firmly established as an indicator of h-like quality, I will henceforth cease to employ the english/latin terms and I will abide by the traditional greek terminology.

Starting in the Hellenistic age, diacritical marks known as "πνεύματα" ("Spirits" or "Breathings" in English), either the δασεῖα ῾ or the "ψιλή" ᾽ (referred to as "Spiritus Lenis" in Latin and "Smooth Breathing" in English) were placed over all word-initial vowels. These were evidently complementary features, as was the case with the other prosodies of ancient Greek ("βαρεῖα", the grave accent, being the complement of "ὀξεῖα", the acute accent; "μακρά", the long timing, being the complement of "βραχεῖα", the short timing), so that a word-initial vowel could be either "δασυνόμενον" or "ψιλούμενον" (φωνῆεν). Similarly, we have seen that the letters π, τ, κ are called "ψιλά" (σύμφωνα), considered to be antipodal to φ, θ, χ (however, in the case of the ancient-Greek basic consonants, there was a third, intermediate possibility, the "μέσα" or "mediae" β, δ, γ). This complementary property of the ψιλά will be referred to as "ψιλότης". All these greek terms will also be preferred over the english/latin ones, when the relevant features are referred to.

Free from the prejudice of the conventional terms, we can now proceed to examining the facts that led to the present misconception.

The Root of All Evil

As long as there was Catholicism, δασεῖα was associated with the segment [h]. No alternative value was ever considered. The Catholics’ conviction was so firm that they even ignored strong evidence to the contrary (ΦΘΧ7) that was in front of their eyes. And even in cases they did not ignore it, the alternative solutions they sought were mere variations of the [h] theme, i.e., such as "voiceless" or "aspirated" vowels. Orthodoxy never questioned this identification, at least not for "sufficiently old" times, and confine their argumentation in proving that the [h] pronunciation was not observed in classical times or shortly afterward. Where does this conviction come from? There is no doubt that the primary reason for this assumption is the correspondence of greek δασεῖα and latin h.

Indeed, in all non-Ionic alphabets (with the exception of Crete), the character Η (or its variant glyph for Η) was employed for the function of (later) δασεῖα. This is the origin of the latin h, either directly or via Etruscan. This fact alone is not enough to establish identity between δασεῖα and h; for example, the greek letter Η (the precursor of δασεῖα and h) stems from the phoenician Heth, but nobody claims that the two were identical. Nevertheless, there is an orthographic tradition in classical times to use h before all δασυνόμενα (essentially the first part of ΦΘΧ1), which speaks for the equivalence of h and δασεῖα.I am not sure about the extent of the reverse transliteration, namely the rendering of latin h with greek δασεῖα. For example, Hannibal and Hamilcar are rendered as "Ἀννίβας" and "Ἀμίλκας" respectively (both with ψιλή), whereas I could not find a spelling "Ὁρατιος" for Horatius. Furthermore, there is no report of any passage in the ancient literature (the Greek or Roman "Grammarians") mentioning any substantial difference between the two. It is, therefore, not unreasonable to conclude that h was at least a good approximation of δασεῖα.

Q.E.D., right? No! We are again defying the rules of logic. ῾=h does not mean that ῾=[h]. What is missing is to prove that h=[h], the hidden premise in our faulty reasoning. But this is evident, right? Again no! Do not let the script mislead you. IPA symbols using latin and greek letters are not necessarily based on their pronunciation in Latin and Greek; for example, [v] is not the sound of latin V (the consensus view is that it was [u(ː)] in its vocalic function and [w] in its consonantal function), while [χ] is an uvular consonant, but greek χ was never uvular.

But have the scholars not established this value of h? Hardly. Their main concern is to show that h was still pronounced in the classical era and beyond, but this would at most mean that h≠Ø, not that h=[h]. Allen (ALLE78, pp. 43-45), after preaching about the nature of h (wherein he utterly confuses <h>, /h/ and [h]), points out the substantial differences (!) between latin h and the english consonant h (that the former does not make position, permits elision and does not prevent contraction, essentially like the greek δασεῖα), acknowledges that it was "on the way to being lost" even before the classical period (based on morphology, inscriptions and literary accounts), mentions that there is no trace of it either in the Romance languages or in early german loans, but surprisingly suggests to pronounce it as the english letter "with perhaps even greater consistency than the native speaker" (what do the Romans know?)!

The last (and only?) scholar that bothered to provide something like a proof must have been Sturtevant, who submits that "The approximate character of Latin h is fixed [!] by its frequent description as aspiration [...] and by its correspondence with the Greek rough breathing in loan-words (Homerus, hydropicus)" (STUR20, p. 69). The latter part (h↔῾) is certainly a circular argument (we claim that δασεῖα is [h], because it is equivalent to latin h, which is [h], because it... corresponds to δασεῖα). The former is equivalent to kicking the can down the road, similarly to determining the value of δασεῖα from its correspondence with the latin h: we still need to prove that "aspiratio" was a term referring to an exhalation like [h]. As we will shortly see, this is in no way established; in fact, it could probably not be further from the truth.

There is, in fact, no positive indication that latin h corresponded to the sound [h]. Among the Romance languages, i.e., the scions of Latin, h has a phonetic value (officially [h], but I suspect that it also stands for, at least original, [x]) only in Romanian and a functional value (h aspiré) in French, in both cases for words of non-Latin (basically germanic, slavic or hungarian) origin, i.e., none of the two can serve to establish the value of latin h. Otherwise, "The Romance languages contain no trace of it" (STUR20, p. 70) and wherever it is written it has no phonetic value (other than occasionally modifying the pronunciation of the previous letter in digraphs like ch, nh, lh, which again is no latin practice). Therefore, tradition does not favour the catholic thesis h=[h]. The very facts mentioned by Allen, above all the fact that latin h is also characterised by ΦΘΧ7, should have led the scholars to reject any identification of latin h with a segment, an approach also taken by latin Grammarians, who do not consider it a letter (ALLE78, p. 43).

The only favourable clue for the catholic cause is the germanic tradition, where h is uniformly pronounced as [h]. The vast majority of the proponents of catholicism being native speakers of a germanic language, it is no wonder that they took h’s value for granted. But whence does it follow that the germanic letter had the exact value of the latin one? Certainly not from oral tradition, as the two language families share no common linguistic history in post-PIE times and, as pointed out by no less a Catholic than Allen, "nor is there any evidence of it in early loans to Germanic" (ALLE78, p. 44). Most seem to assume that there was an orthographic continuation, where the germanic peoples made use of the same letter for the same sound, which latter disappeared among the Latin speakers, but such an assumption is not justified.

Whatever latin h stood for, it did not survive long in the imperial times: "In vulgar Latin h seems to have been lost completely in Pompeii in the first century A.D., and not much later everywhere in the empire" ("STUR20", p. 70). Even allowing for the unlikely development that the loss of "aspiration" (like the assibilation of /k/ before front vowels) post-dated the spread of the language from Latium to the entire known world and propagated like a virus that affected all speakers of Latin (rather than assuming that the uniformity is the result of a change that had taken place in the language while confined in its place of origin), if we also take into account that the evidence only supports "the retention of initial aspiration [in Greek] until about the 2 c. A.D." (ALLE87, p. 53), it is evident that by all accounts the entire empire (i.e., both the Latin-speaking West and the Greek-speaking East) lost its "aspiration" long before any German, Goth or Viking had their own writing system (cf. GRAN07, p. 142, §343: "Germanic h appeared when Latin h had long been silent in popular speech"). Even allowing for the often-asserted "schoolroom tradition" (which must have been orthographic rather than phonetic), to claim that "aspiration" was revived in the mouths of the invading barbarians in its exact original form would be tantamount to "discovering" the value of latin H in the present value of greek Η(=[i]) or cyrillic Н(=[n]).

I have not found any evidence why Germans adopted H for their [h] sound, so it seems that we can only speculate about it. One possibility is that it was just a case of recycling a letter that had no actual value, just like the Ionians recycled Η (that denoted the δασεῖα) as one of the two new symbols for the long mid vowels. A more likely course of events that might have spawned the establishment of the letter H for rendering the "glottal fricative" [h] is based on its phonological provenance. In Germanic, [h] is essentially a degenerate /x/, possibly an allophonic realisation of the velar fricative. Germanic /x/ was regularly rendered by means of the digraph <ch> (which even the Catholics admit to have had the "late-Greek" value /x/), as is evident from the renderings of germanic names, such as Cherusci and Chlodovechus; if there was the need to distinguish between the full sound and its debased version, then the suppression of <c>, which essentially represented the supra-laryngeal features of /x/, and the retention of only <h> is a reasonable choice, particularly at the time that the phonetic significance of latin h was not even a distant memory. At any rate, such speculation is beyond the scope of this investigation. The crucial point is that the claimed "continuation" of the sound of latin h in german mouths is not warranted.Irrespective of how and why the use of h for Germanic [h] was established, since any phonetic value it might have had in Latin had long been forgotten, it is no wonder that some may have made the same mistake modern scholars do, namely think that when it was something more than an empty letter in Latin it had the only value they had heard in connection with this letter, [h]. This would, then, explain the (isolated?) "michi" for mihi|to me (which by then was pronounced [mi(ː)] or [mi.i]; cf. Romanian mie) as a case of (excessive) hypercorrection, a fact essentially acknowledged by Allen (ALLE78, p. 45).

Spirits, Sprites & the Holy Exhalation

We have just seen that the Catholics read "exhalation" (the sound [h]) in the latin term "aspiratio". A similar assertion is made about the greek term "πνεῦμα". Nothing illustrates better this illusion than Allen’s statement on the terminology used by the greek Grammarians: "in discussing the pure aspirate [h] the grammarians adopted the same terminology [i.e., δασύ and ψιλόν], calling it not merely πνεῦμα but more specifically and pleonastically πνεῦμα δασύ ('spiritus asper', 'rough breathing'), and then referring to its absence by the self-contradictory πνεῦμα ψιλόν ('spiritus Lenis', 'smooth breathing')", (ALLE87, p. 55). Right! So, "πνεῦμα", "aspiratio" and "δασύ" (since it "pleonastically" characterises πνεῦμα) are synonyms of "exhalation" and "ψιλόν" is an antonym (in order for πνεῦμα ψιλόν to be "self-contradictory"). I hate to play the philologist, particularly for a language that I have never actively used, but in this case the fallacy is patent by a mere inspection of relevant passages.

The word πνεῦμα has various meanings. Its most popular use nowadays is in religious contexts, particularly for the Holy Ghost (Ἅγιον Πνεῦμα) and other immaterial beings. In general, all its meanings are somehow related to a current of air, including breath. The term features in the works of the ancient Grammarians with various frequencies. Thrax mentions it only once: "μέση δὲ σημεῖον πνεύματος ἕνεκεν παραλαμβανόμενον|the semicolon is a sign of where to take breath [Davidson’s translation]" (DION83, §δ´); it is clear that it is used for describing the act of breathing and not exhaling. Halicarnasseus though makes more extensive use of it; let us review the passages we have already seen:

There is more (ROBE10, pp. 140, 148, 150):

It is evident that "πνεῦμα" is described as a feature of (nearly) every sound of the Greek language. Which reasonable person would conclude that it refers to the exhalation [h]? Anyone with elementary knowledge of phonetics would understand that in all these instances "πνεῦμα" refers to the airstream, the air pushed out of the lungs (πνεύμονες), which is a mandatory feature of all vowels and pulmonic consonants. This simple truth has escaped (?) the Catholics, even though it was pointed out more than a century ago: "Ἡ μεγίστη πλάνη τοῦ Blass περὶ τὴν ἐξέτασιν τῆς περὶ τῶν δασέων διδασκαλίας τοῦ Ἀλικαρνασέως Διονυσίου καὶ τοῦ μουσικοῦ Ἀριστείδου τοῦ Κοϊντιλιανοῦ κεῖται ἐν τοῦτῳ, ὅτι, ὅπου οὗτοι ἐν τῇ διδασκαλίᾳ αὑτῶν περὶ συμφώνων χρῶνται τῇ λέξει πνεῦμα, νομίζει ὅτι ἑννοοῦσι τὸ «δασὺ πνεῦμα», μὴ παρατηρήσας ὅτι τῇ λέξει πνεῦμα χρῶνται ἐπὶ τῆς ἐκφωνήσεως παντὸς φθόγγου, σημαίνοντες διὰ ταῦτης τὸν ἐκ τῆς ἀρτηρίας ἤ τῆς φάρυγγος προαγόμενον εἰς τὸ στόμα ἀέρα, οὗ ἄνευ οὐδεμία φωνὴ εἶνε δυνατή. Τούτου δὲ ἕνεκα οὐδέν ἄπορον ἄν κακῶς ἡρμἠνευσε τὴν διδασκαλίαν τῶν εἰρημένων Τεχνογράφων, καὶ ἐν τῇ φράσει τοῦ Διονυσίου «τὰ δασέα ἐκφωνεῖται πολλῷ πνεύματι», ἐνεῖδε τὴν ὕπαρξιν δασέος πνεύματος, πράγματος ὅλως ἀγνώστου τῷ Διονυσίῳ ἐν πάσῃ τῇ διδασκαλίᾳ αὐτοῦ περὶ τῆς ἐκφωνήσεως τῶν συμφώνων καὶ φωνηέντων|The greatest deception of Blass regarding the examination of the teaching by Dionysius Halicarnasseus and the musician Aristides Quintilian on the δασέα is that wherever those use the word πνεῦμα he thinks that they mean «δασὺ πνεῦμα»[='spititus asper'], not having observed that they use the word πνεῦμα in connection with the pronunciation of every sound, therewith meaning the air that is pushed from the windpipe or the pharynx to the mouth, without which no speech sound is possible. Because of this, it is not strange that he misinterpreted the teaching of said writers of rhetoric and in the expression «τὰ δασέα ἐκφωνεῖται πολλῷ πνεύματι»[=the δασέα are pronounced with a lot of πνεῦμα] of Dionysius he saw the existence of 'spiritus asper', which is completely unknown to Dionysius in all his teaching about the pronunciation of the consonants and the vowels" (PAPA98, p. 26, n. 1). For the misinterpretation of the expression «τὰ δασέα ἐκφωνεῖται πολλῷ πνεύματι», as well as the meaning of the terms δασύ and ψιλόν, we will come back later, when examining the pronunciation of the δασέα. For the time being suffice it to point out that there is not the faintest implication of the sound [h] in the use of the term πνεῦμα, which Roberts sufficiently correctly translates as "breath" and more precisely denotes the airstream.

(ΦΘΧ8) πνεῦμα is a feature of every sound and apparently denotes the airstream

My limited knowledge of Latin does not allow me to carry out a similar investigation on the latin term "aspiratio". Nevertheless, there is at least one testimony that is not quite compatible with the [h] interpretation: "Velius Longus, in the middle of the second century, says that the u in ualente is pronounced 'cum aliqua aspiratione'" (GRAN07, p. 135). The meaning of "aspiratio(ne)" here cannot be that u (<V>) is "aspirated" ([wʰ], which is not reported for any language). Instead, Lindsay regards it as a clear description of a pronunciation "as a consonantal spirant, not as a half-vowel" (emphasis mine). It is interesting that the term could be, at least in this instance, an indicator of "friction", the very property that characterises the δασέα in modern Greek! But the most important consequence for now is that it cannot be safely considered a synonym of [h].

He Is The One

As already mentioned, before the establishment of the Ionic alphabet as the de facto "Greek alphabet", the letter Η was used in almost all other epichoric alphabets wherever we later find the δασεῖα (the former before a δασυνόμενον vowel, the latter over it). When Greeks adapted the semitic alphabet, it appears that they generally strived to employ the same symbols for the same sounds (thus, Κ↔כ=/k/, Π↔פ=/p/ Τ↔ת=/t/, etc). There is a particular symbol in the semitic alphabets that appears to universally represent [h] (or /h/) in all modern semitic languages: He (ה). Curiously (if you are a Catholic), the Greeks bypassed this letter and used the next one: Ḥet (ח). Specifically, Ε has the (original) shape and alphabetic position of ה, while Η (in its original form glyph for Η) has the (original) shape and position of ח. We can, therefore, state as a fact that

(ΦΘΧ9) Greek Η derives from semitic Ḥet (ח), not from He (ה)

The official catholic explanation (e.g., ALLE87, pp. 53-54) is that ה was already taken for the vowel Ε and, therefore, the Greeks used the (phonetically) closest available of the remaining letters. This theory is problematic in more ways than one. All semitic alphabets are and (almost certainly) were consonantal, which means that the first step taken by the Greeks in the adoption of the alphabet must have been to adopt the consonants that stood for (almost) the same sound in Greek; indeed, the three mid/open vowels Α, Ε, Ο correspond to symbols that are assumed to have represented laryngeal consonants, which were most likely not part of the greek phonological inventory (Ι and Υ correspond to their associated semivowels י and ו, the latter spawning both the vowel Υ and the semivowel Ϝ). It is, therefore, unlikely that the assignment of vowels preceded that of consonants, but it was most likely carried out with the available letters after the homologous consonants were already assigned. Even if the Greeks came up with the idea of using some of the semitic consonants for their vowels at the time of adaptation of the semitic alphabet, such a mixup of symbols and sounds is unlikely. Specifically, if δασεῖα were the same as the phoneme represented by ה, it would have been weird for the Greeks not to notice the identity of sound, but use the symbol for one of their vowels, then when δασεῖα’s turn comes to realise that the ideal symbol was "taken" and to have to make do with a very coarse approximation (ח). It follows that almost certainly ה and δασεῖα did not correspond to the same sound. Since ה universally stands for [h], the odds are against a value [h] for δασεῖα at the time of the transmission of the alphabet.

The Catholic reaction to this inevitable conclusion is (to me) incomprehensible. The most surreal statement is that of rev. Taylor: "In Greek this difficult sound [of Ḥet] readily weakened [!] into the rough breathing h, thus forming a substitute for he, the fifth Phoenician letter, which had lost its aspiration [!!], and had lapsed into the vowel E, originally called εἶ, and afterwards ἔψιλον" (TAYL83, p. 85). It sounds as if the Semites imposed the sounds together with the symbols of the alphabet to the Greeks, who in their endeavour to reproduce the imposed sounds mispronounced [x] (his assumed value of Ḥet), which was a "difficult sound" (if you are an Englishman), as [h], but at the same time not being able to preserve the "aspiration" of He, which... disappeared! I guess everything is possible in the world of religion.

Sturtevant states (STUR20, p. 156) that Ḥet "denoted a spirant" and relies on the hoax that Ξ=hs in "early Naxian inscriptions" (despite acknowledging that the graphemes used for Η and for the first part of Ξ are different) to speculate that "the orthography must have originated at a time when glyph for Η denoted a spirant (compare German Ochs, etc.)"! Twisting the facts to accommodate one’s theories is not exactly scientific. Note that the (probably irrelevant) German forms "Ochs, etc." are considered comparable to the unfounded conclusion Ξ=hs in Naxian for establishing the value of Η as [h], but not to the Attic digraph ΧΣ (for Ionic Ξ) as suggestive of a "spirantic" value of Χ.

A similar "palatal fricative [ç]" intermediate value of the δασεῖα is postulated by Allen (ALLE87, p. 54) for explaining away Η’s origin in ח instead of ה, but the chronology does not add up. Allen himself states that any intermediate value must have been replaced by "its purely aspirate value" (as if [ç] were an "impurely aspirate value") by "the time of the operation of Grassmann's Law". Furthermore, it is generally considered that adoption of the alphabet by the Italians (Etruscans, Romans) followed suit shortly after its adoption by the Greeks; since latin H appears to be equivalent to greek δασεῖα (formerly Η), did the same "palatal fricative" follow down the same path in Latin or was there enough time for greek Η to degenerate to [h] before being transmitted to the Italians? And it should also be noted that Allen postulated his [ç] value as an intermediate stage between PIE *j and [h]; what about δασεῖα’s other PIE origin, the most frequent *s? did it remain a /s/ until it and [ç] fell to [h]? did it develop to [h] first, remaining unexpressed until [ç] joined it and lent it its grapheme? was there enough time (between the adoption of alphabet and the alleged phonological changes) for any of these to happen?Allen submits another requirement, namely that "the Greek phoneme must have developed its purely aspirate value" by "the time of the operation of Grassmann's Law". How much more can we squeeze in the (relatively short) period after the adoption of the alphabet (and before, say, Homer)? All in all, the "spirant/fricative Η" theory seems to merely be extravagant speculation.

If there is something worth pointing out with respect to the provenance of <Η> (other than the implausibility that it stood for [h]), it is that (the sound corresponding to) Ḥet is the first real fricative in the semitic alphabet.

Spanish Lullaby

As just mentioned, δασεῖα is considered (e.g., STUR20, p. 156) to have derived from PIE *s and *j. The former appears to be a more frequent source than the latter.

(ΦΘΧ10) In cognates, δασεῖα corresponds mostly to /s/ and less often to /j/

It is in the first part of this phonological provenance that some Catholics see another confirmation of their [h] value, for they claim that the "debuccalisation" of [s] gives [h],Misled by Kehrein’s definition that "'debuccalization' means the deletion of all subsegmental material except for the LAR NODE" (KEHR02, p. 193), I had my doubts about the correctness of the term "debuccalisation" for the conversion of [s] to [h]: if [s] is stripped of supralaryngeal material, we are left with exactly... nothing; only [sʰ] would yield [h]. Nevertheless, either Kehrein’s definition is wrong or it refers to the specific case of Kashaya discussed in the relevant part of his book. citing as proof the development of /s/ in some dialects of Spanish. Needless to say that this is another case of analogy as proof: just because a certain phoneme has a certain development in one language does not mean that it has to develop alike in all languages. In this case though, the argument is flawed in more respects.

First of all, I doubt the existence of [h] as a degenerate stage of /s/ in Spanish. It is true that, in some areas (such as Andalusia and the Americas), "se tragan las eses|the s’s are swallowed". The very term "tragar|to swallow" is indicative of what happens: when something is "swallowed", nothing is left behind; in other words, the speakers of (mainstream) Spanish interpret the phenomenon not as mutation, but as disappearance of the [s]. I have had a lot of experience with various spanish accents and I have yet to hear a [h] in lieu of a [s]. Even if my ear is not trained enough to detect an almost imperceptible sound (that might be picked up by a spectrogram), I can certainly tell the difference between, say, English "heir" ([eə(ɹ)]) and "hair" ([hɛə], [hɛɹ]), and the deleted /s/ in Andalusian etc sounds nothing like the h of the latter, but rather like that of the former (i.e., las otras|the other ones (fem.)=[la.ˈo.tɾa]). I did hear the sound [h] in Cuban Spanish, but only as a replacement of mainstream [x] (thus [moˈhito] for mojito=[moˈxito]), i.e., the same origin as in Germanic. I, therefore, contend that the fate of "debuccalised" /s/ in Spanish is no different than in the similarly derived French côte(←costa), maître(←maistre←magister), épée(←espee←spatha) etc, i.e., a mere deletion of the consonant.

Irrespective of the validity (or not) of the transformation [s]→[h] in Spanish, there is a more important fact that is conveniently overlooked. Spanish /s/ degrades only in coda position; that is, in a word like sois|you are (pl.), the second s may be left out, but the first one is always pronounced. This is exactly the opposite of what (is said to have) happened in Greek; for instance, in the derivation of ἕξ from PIE *swéḱs, it is the first s that undergoes transformation, while the second remains unaffected. More specifically, it was a single *s (or a *sw, as per *swéḱs→ἕξ, *sweh2d-→ἡδύς, etc) before a vowel that was transformed into the δασεῖα word-initially and to nothing word-internally (e.g., *genesos→γένεος/γένους), but *s was preserved before another consonant (e.g., *h2stḗr→ἀστήρ) or word finally (e.g., the ending -(ο)ς of nominative singular).It is for this reason that I doubt Kehrein’s derivation of ἅρμα from *arsma (by transformation of *s to *h and migration of the latter to the previous onset). I have not seen any example of disappearance/mutation of a PIE *s before *m, but it is more reasonable to assume its origin to be *sarma (be it with metathesis of inter-consonantal s or whatever). The effect of the voicing of the (vocalic) nucleus need not be the same on the coda and the onset. A weakening (i.e., lax pronunciation) of a voiceless fricative after the voicing has ended is understandable, but to argue that the same also occurs in anticipation of the voicing (i.e., in onset position) is unfounded.

What happened to PIE *s, what (if anything) it mutated to and which path it took on its way to its final value I do not know (nor can anyone claim that they do). But to liken the fate of pre-greek syllable-initial *s to that of modern-spanish syllable-final /s/ is to compare apples and oranges.The same can be said for sanskrit visagra, which "is an allophone of /r/ and /s/ in pausa". Probably the only interesting point to keep from the assumed phonological origins of δασεῖα in *s- and *j- is that they were prevocalic fricatives.

Now You See It, Now You Don’t

We have seen above that the transliterations from and to Latin are not conclusive; there is evidence of correspondence between the greek δασεῖα and the latin h, but the correspondence of latin h and germanic [h] (which is the catholic practice) has not been established, at least not for the time when h might have had some phonetic significance in Latin. But there must be transliterations from/to other languages that confirm the value [h] for δασεῖα, right? Jein|German for Ja+Nein, Yes and No at the same time!

There are, indeed, gothic transliterations like Haibraius (for Ἑβραίους|Hebrews (acc. pl.)?), Herodes (for Ἡρώδης|Herod), etc, that may suggest correspondence between germanic h (which could only be a segment corresponding to /x/ or its degenerate form [h]) and δασεῖα. But at the same time, there are spellings like Helias (for Ἠλίας|Helias), Her (for Ἤρ), Haileisaius (for Ἐλισ(σ)αῖος|Elisha), etc, i.e., use of h in Gothic where there is no δασεῖα in the original greek form, which compel Sturtevant to admit that "Clearly the rough breathing was little more than schoolroom tradition in northern Greek of the fourth century A.D." (STUR20, p. 160); that is to say there is no phonetic reason behind the use of gothic h for the δασεῖα.There is no apparent reason for specifying "northern Greek" other than the region’s proximity (as compared to the rest of the Greek-speaking world) to the homeland of the Goths. Sturtevant appears to be assuming that the Goths learned these words from their most proximate Greeks, but there is no need to make such an assumption. It is very likely that these spellings were introduced by Ulfilas himself, who did not speak "northern Greek" (he was a Cappadocian) and also lived in the "South" (he was consecrated bishop of the Goths in Antioch). The actual reason for the "correct" use of h to transliterate dασεῖα is most likely the association of both with the latin h, the latter as a long-established orthographic tradition (already mentioned above) and the former through its graphic origin in latin h (for which I have provided plausible explanations).

Sturtevant draws the same conclusion from "the Armenian and Rabbinical texts": "by the fifth century A.D. the rough breathing was in the East no longer pronounced although still taught in the schools" (STUR20, pp. 160-161). As reason for the convention of using Armenian and "Rabbinical" (Hebrew? Aramaic?) symbols of [h] (if, indeed, this was the value of the used symbols) for representing the δασεῖα we may assume the already established gothic/germanic convention. The transliterations of IV-V AD and later are well into the era of no phonetic value for the δασεῖα (even with catholic standards; cf. BLAS90, p. 94) to claim that they were based on any living pronunciation.

Marginally more recent are the "Coptic transliterations of Greek words [...] of the second century A.D.", which Sturtevant considers as proving "the survival of the rough breathing in Egyptian Greek of second century A.D." (STUR20, p. 160). Too little information is provided as to the actual nature of these transliterations. Blass states that they "represent the spiritus in Greek loan-words almost without exception with their Ϩ (h): hoste, hina, hote etc." (BLAS90, p. 94). Nevertheless, the evidence from Egyptian Greek suggests that δασεῖα was of no phonological significance since at least a couple of centuries before the Coptic script was invented. Sturtevant himself admits that "the Egyptian Greek papyri show many variations from Attic usage in the employment of the rough and smooth mutes before initial vowels", but tries to explain it away by asserting that "In most cases, however, the difference from Attic represents an analogical change; καθʼἔτος is due to καθʼἡμέραν, καθʼὥραν, etc". How likely is it that the writer pronounced κατά ([kata]) and ἔτος ([etos]) correctly, but when he put them together he forgot this pronunciation and inserted a δασεῖα because he pronounced one in ἡμέραν and ὥραν?! The spellings were probably "analogical", but most likely because the writer was not sure which words take a δασεῖα and which do not. This is actually the conclusion drawn from the study of the papyri, e.g., by Gignac (BUTH08, p. 224).I do not know to what extend the papyri employ diacritics for the δασεῖα and the ψιλή. I suspect that they do not, but use exclusively plain capital letters (cf., e.g., here). In that case, it is not possible to draw conclusions about the correct pronunciation of the δασεῖα but indirectly, i.e., from the correctness of the conversion of the ψιλά to δασέα, as undertaken by Gignac. It is, therefore, evident that the use of coptic Ϩ for a symbol that was not pronounced is pure orthographic convention and was not based on any identity of sound. Here is some further information on coptic Ϩ (all quotes from PEUS99):

All in all, Ϩ is used often, but not consistently (Ϩ1), for a greek feature that most likely disappeared long before the introduction of the coptic script, its use was not even consistent in its own language (Ϩ2), has its phonetic origin in a sound alien to Greek (Ϩ3) and appears to have occasionally had an auxiliary, non-phonetic significance (Ϩ4). I would say that we are still too far from establishing that δασεῖα was the phonetic equivalent of Ϩ, much less that it had a value of [h] when the coptic alphabet was introduced or even earlier. As to why represent the δασεῖα (even merely orthographically) by a symbol originating in two ancient-egyptian phonemes which "In Egyptian-Semitic transliterations, [...] consistently correspond to Semitic /h/ and /ħ/ [I assume he means ה and ח] respectively" (p. 98), this does not necessarily mean that it was due to a pronunciation of δασεῖα as [h] or a memory thereof; first of all, these two values of the egyptian phonemes are probably undertaken exactly because they are assumed to have been identical with the (presumed) semitic values (we have seen that the phonology of Egyptian is not exactly carved in stone);One should also not forget that there is no oral coptic tradition, since Coptic as a living spoken language passed away before any phonologist could observe it (PEUS99, p. 31). Graphic evidence is all that Coptic has left behind. then, it could have been exactly that correspondence with ח that set the link between Ϩ (or even its predecessor graphemes that are considered to have represented [ħ]) and the greek δασεῖα (which, when it was a letter, Η, derived graphically from ח); this may seem like an empty theory at first glance, but how else can one explain the second of the renderings "Η(Ι)ΤΑ, ϨΗΤΑ" (p. 59) for the eighth letter of the Coptic alphabet, which comes from greek Η, namely ἦτα that (ever since it was used in its vocalic value) had no "aspiration" (cf. ALLE87, p. 172, n. 7)? Therefore, the evidence from Coptic are most likely educated transcriptions using established conventions.

The only evidence appearing to be synchronic with a time period during which δασεῖα might have (still) had some phonetic value comes from indian languages. Allen sees "in astronomical terms such as horā ( = ὥρα) in Sanskrit" the retention of [h] "up to at least the beginning of the Christian era" (ALLE87, p. 53). This statement is very puzzling, as we have seen that there was no direct contact between Sanskrit and Greek, much less any interaction between the two in hellenistic times. I have not been able to find evidence that would justify a derivation of sanskrit horā from greek ὥρα or vice versa. Instead, it may be mere coincidence, false etymologyThis may be the most likely cause. Note that the term horā appears to always have had an astrological meaning in Sanskrit, whereas the term ὥρα and its derivatives appear to have always been used in their temporal sense in Greek, their astrological meaning having been introduced by either Ptolemy (II AD) or at most Manetho. Both of them relied on the previous astrological advances in the Middle East, possibly reusing its terminology, which may have borrowed the term from Sanskrit through earlier contact. Either of them could have (wrongly) associated sanskrit horā with the similarly sounding ὥρα, a kind of misunderstanding known from terms like Zoroaster. or aberrant development of a PIE form. We can, thus, say that sanskrit horā is a transliteration of greek ὥρα, as much as latin ferō is a transliteration of greek φέρω.

Of more actual weight is Sturtevant’s submission that "The Greco-Indian coins of the last two centuries B.C. show such forms as Heliyakreyasa=Ἡλιοκλέους, Hipastratasa=Ἱπποστάτου [sic], Heramayasa=Ἑρμαίου" (STUR20, p. 160). The Greek spoken in Bactria at that time (I BC) had been isolated from the rest of the Greek-speaking world for more than a century, due to the rise of the Parthian Empire; however, the use of h for δασεῖα may be indicative of at least its approximate value or a "remembered" value, even if that had mutated in bactrian Greek. One problem with the data provided by Sturtevant has already been identified: the Indians did not use the latin alphabet and the quoted forms "Heliyakreyasa", etc are nothing but our "decoding" of their native script. I have managed to find one example of the relevant coins, which bears inscriptions in Greek and Kharoṣṭhī alphabets on its sides. I have tried to "decipher" the front (greek) side based on my familiarity with the greek alphabet and the rear (Kharoṣṭhī) side with the help of this decryption key. The first observation is that the letters are not exactly typed. There is considerable variation and uncertainty, which sometimes has to be resolved by guessing the most likely symbol based on other knowledge. For example, the greek text on the front side appears to read "ΒΑΣΙΛΕΩΣ ΔΙΚΑΙΟΥ ΗΛΙΟΚΛΕΟΥ|of just king Heliocles", but the last letters of the king’s name are not very legible. The text accompanying the picture is "ΗΛΙΟΚΛΟΥ", but there are certainly more than two letters following the clearly visible Λ. The correct form should be "ΗΛΙΟΚΛΕΟΥΣ", but I cannot discern a Σ, as the last letter of the name seems to be an Υ (perhaps the missing Σ is the one inexplicably appearing in the middle of the rear side). These problems are exacerbated on the rear side. Greco-Bactrian Coin I have managed to identify the three words of the provided translation "Maharajasa dhramikasa Hiliyakreasa" (see image to the right). Not all syllables are clearly identifiable. The ones provided as "dhra" and "kre" are a mystery; I assume they are rather "dhṛ" and "kṛ", but cannot associate them with the Kharoṣṭhī symbols. The symbol of interest is the first of the king’s name, which is allegedly a "hi". Curiously, the vowel is most likely not "i", as the provided transcription suggests, but "e"; as seen in the syllable "mi", the diacritic for "i" is a vertical stroke running through the main symbol, whereas in the first symbol of the king’s name the stroke stops when it meets the main symbol, which rather represents "e" (according to the key); therefore, the provided transcription is most likely not correct. The Catholics may rejoice at this observation, because it may mean that the first symbol stands for "he", the normative latin transliteration of Ἡ (the first syllable of Ἡλιοκλέους). However, this does not necessarily follow; the stroke is applied on a main symbol, but it is not easy to tell which sound this main symbol represents. Specifically, it is not clear whether this main symbol looks more like the symbol for "a" (bottom left side) or that for "ha" (right side). Only in the second case would the indian inscription stand for the desired "Heliyakreasa", but the first case, which would stand for "Eliyakreasa" (i.e., no [h] for δασεῖα), cannot be ruled out. It appears that the "decipherer" has chosen the former interpretation, only because he was expecting a transcription [h] for the δασεῖα. In other words, the transcriptions provided by Sturtevant’s sources were made under the presumption that the Indians would transliterate Ἡ as "He" and not as "E" and thus bear no significance for a "verification" of the catholic value of the δασεῖα, much less for its "retention".The other coins that I have seen online (e.g., this and this) have even less legible symbols and the provided indian transcriptions are obviously based on expectation rather than observation.

From the above it is evident that transliterations of δασεῖα in non-Latin languages are probably optical illusions. Most importantly, these are all secondary "discoveries" of the Catholics, after they had convinced themselves (based on its correspondence with the latin h) that δασεῖα stood for [h]. As such, they cannot bear more weight than the various "confirmations" I devised for my preconceived theory about the values of Ξ and Ψ.

Getting There

We have come a long way in the investigation of the nature of the δασεῖα. To summarise, we have no positive evidence that δασεῖα was a [h]. The primary clue for the adoption of an [h] value, its Latin connection, is an incomplete argument; it does not even have the marginal evidential value of the C↔Κ transliterations, because nobody remembered to first prove that h=[h] in Latin! It is remarkable that the pitiful catholic interpretation of the word πνεῦμα as descriptive of [h] has not been scorned by any philologist in the last hundred years. No satisfactory explanation has been provided for the graphic origin of δασεῖα’s predecessor Η, particularly why the semitic h-symbol ה was avoided. The invocation of the development of /s/ in some Spanish dialects as a modern reflection of the fate of pre-greek *s is just another case of little-thought linguinistics; spanish /s/ mutates only in codas, pre-greek *s lost its status only when alone in onsets. Finally, evidence from transliterations of δασεῖα in third languages is not as sound as we are expected to believe. There is, therefore, no strong reason to believe in the catholic doctrine ῾=[h] other than out of an a priori conviction about the value of latin h.

It is true that we do not need to seek alternative explanations for the aforementioned circumstantial evidence, if we simply accept an [h] value for the δασεῖα. However, we cannot simply turn a blind eye to the fact that elision and crasis are not affected by an intervening δασεῖα (ΦΘΧ7), because in this case the catholic value C (as well as any identification of δασεῖα with a segment) violates the Incompatibility Principle. In accordance with Sturtevant’s ingenious test, we can declare that "The only way to establish that thesis is to find a language in which elision actually occurs in that position. Until then we may assume that Greek vowels with [δασεῖα] were absolutely initial" (STUR37, p. 119, my adaptation within square brackets).The quote comes from Sturtevant’s article on the so-called "smooth breathing", where he discusses the theory (promoted, among others, by Chatzidakis) that ψιλή corresponded to a glottal stop ([ʔ]). He ends up dismissing it on account of ψιλή’s invisibility to elision and crasis. He acknowledges the lack of effect of both πνεύματα on elision and crasis, but states "that aspiration and glottal stop are very different sounds, and that the existence of elision before aspiration establishes no presumption that elision is also possible before the glottal stop". He then requires (in the above-quoted passage) the identification of a language that permits elision before [ʔ], in order to consider the glottal-stop theory as likely. One would like to point out to him that there is no language that permits elision before [h], either. It seems that the only language that "permits" this kind of elision is... (catholic) ancient Greek! In fact, the case of [h] is even less defensible than the case of [ʔ], since he sees a way out for the latter: "The only possible way to rescue the theory of an initial glottal stop in Greek is to assume ad hoc the loss of this sound after a vowel!"; for [h] not even this loss would be enough, as the sound would have to reappear after elision has taken effect, in order to affect a preceding tenuis (ΦΘΧ2), i.e., one would have to assume Houdini-style teleporting!

On the face of this conclusion, one can feel the desperation of the catholic hordes, who were brought up with the dogma ῾=[h]:

"If ῾ is not [h], what else could it be?"

The short answer to this question is "we do (and can)not know", as we can no longer observe spoken "ancient Greek". Our inability to be sure about something (i.e., what δασεῖα is) does not mean that we have to adopt a solution (i.e., [h]) that is patently incompatible with the evidence (particularly ΦΘΧ7). We, therefore, have to look for alternative solutions.


A revolutionary proposal is made by Geldart: "in ancient Greek the only difference between the rough and the smooth breathing may have been that it was the custom to turn κ, π, τ into χ, φ, θ before words which had the rough breathing, whereas before the smooth breathing they remained unaltered" (GELD70, p. 39). This is essentially the orthodox practise of denying δασεῖα any phonetic value (Ο) and only observing its functional significance (ΦΘΧ2). Even though some may wince at such a seemingly naive theory and point out that there is no phonetic justification for different behaviour before the two πνεύματα, if those did not differ phonetically, Geldart already provides a verifiable paradoxical parallel: in French, there is a distinction between two kinds of word-initial h, one that comes from latin and one in words of foreign (principally germanic) origin; both have absolutely no phonetic value (Ø) and are represented by the same letter; the former (known as h muet|mute h) is purely orthographic (it could have been left out without any impact on the language), but the latter (known as h aspiré) has an effect on the language, in that it represents hiatus, i.e., it prevents elision of a preceding vowel. I have always been wondering how the French know when to avoid elision in this case, since they do not pronounce anything and their linguistic instinct should drive them to drop a word-final vowel before another (word-initial) vowel. It appears that "prescriptively correct usage requires a considerable amount of memorization", much like it used to be until 30 years ago in Greek, when Greeks had to memorise the few hundreds of words that take the δασεῖα.

There is no doubt that for a long time in the history of the Greek language (at least since II AD, even by catholic accounts) δασεῖα had a notational function, but no phonetic value. Thus, Geldart’s theory is valid for the largest part of Greek’s literate history. The question is whether it also reflects the state of things since the introduction of the alphabet. It is hard to imagine that Greeks added the δασεῖα (in the form of Η) in every word that might in a certain environment convert a previous tenuis to an aspirata, even if the word is not in that environment (note that Η is written in the two oldest greek inscriptions, less clearly in the Dipylon inscription, but certainly on Nestor’s cup, in positions where it would not have any effect on its surrounding letters, there being thus no reason for it to be used, if it had no phonetic value). Geldart points out that δασεῖα "stood properly for some letter which had been left out at the beginning of a word, more especially for σ" (cf. ΦΘΧ10), implying that it may have been used as a replacement (or "placeholder") for a deleted letter. This, too, is not unheard of: it is reported that in medieval Latin "the letters 'n' and 's' were often omitted and replaced by a diacritical mark above the preceding or following letter". While it is possible that the medieval-Latin scribes were not the first ones to employ graphical marks for deleted letters, it should be emphasised that the presumably deleted *s and *j are never attested in any ancient-Greek dialect (i.e., no text or inscription has ΣΕΞ for the number 6, but the only forms found are ΕΞ and ΗΕΞ). Their presumed deletion must, therefore, predate the introduction of the alphabet and it is not easily conceivable how the memory of a deleted letter (or rather sound) could have survived without written records.

Most importantly, Geldart's theory fails to account for the very phenomenon it claims as its foundation, namely the conversion of a preceding tenuis to an aspirata. In French, the "h aspiré" prevents elision because once upon a time it was a segment ([h]) that separated the two vowels and modern Frenchmen honour this tradition, even though the reason that prevented elision (i.e., the sound of h) has disappeared. If δασεῖα (or rather Η) owed its function to its prior status as full segment, this would mean that this prior segment also allowed elision, which is even more unlikely for [s] than it is for [h]; if elision was allowed because the [s] had disappeared, then there would be no reason for a preceding tenuis to be converted to something else. In other words, elision and conversion are not simultaneously justified for either of the assumed prior state ([s]) and the final state (Ø) of δασεῖα. In order to explain how the tradition of treating δασεῖα as a mere reminder for tenuis conversion was established, it is necessary to assume an intermediate value that simultaneously justified elision and conversion (to whatever, as we have not yet established the value of the aspiratae). The Catholics contend that such an intermediate value was [h], but this value only justifies conversion (and that through a further stage of fusion), but not elision. We have seen that elision of a word-final vowel requires that it be followed by another vowel, so that no consonantal value of the δασεῖα would justify elision.

A (semi)vowel?

If δασεῖα was a segment, only a vocalic value would do the trick (of elision). Marginally likely is also a semi-vocalic value, since semivowels behave as vowels in some languages, e.g., "l'oiseau|the bird[FRA]" (not "le|the (masc.)" before /w/), "l'huître|the oyster[FRA]" (not "la|the (fem.)" before /ɥ/), as compared to, e.g., "a woman" and "a one-man show" (not "an" before /w/) of English. But what vowel or semivowel could it be? It appears that δασεῖα may be combined with all vowels without having any special behaviour for some of them (which would hint us towards a pronunciation related to that vowel). The only fact that may be relevant is that

(ΦΘΧ11) All words beginning with Υ receive the δασεῖαI do not consider the similar habit of placing the δασεῖα over every word-initial Ρ a "fact". It is rather a post-Christian orthographic convention, the origins of which cannot be securely associated with any classical pronunciation.

One problem is that there is no consensus (between Orthodox and Catholics) about the classical value of Υ; the catholic view is that it stood for [y], but (as we will see in the corresponding chapter) this proposal comes from scholars whose main linguistic experience was with German and French and their imagination was limited by the phonological models of those languages; what we may consider almost certain is that its original (vocalic) value was [u] and that its present value is [i], but we can only guess what happened during the transition from [u] to [i]. Could this mean that δασεῖα was some semivowel related to Υ, possibly [w]? This is rather unlikely, as there was a specific letter, the digamma Ϝ, which almost certainly represented that sound. What about [ɥ], the semi-vocalic equivalent of the catholic value of Υ? This would mean that all (word-initial) Ὑ would be pronounced [ɥy], an unlikely combination, or [ɥu] or [ɥi], two combinations that cannot be reasonably accounted for. Moreover, it cannot be explained how a rounded semivowel can have anything to do with /s/ and /j/ (ΦΘΧ10).

An unrounded front semivowel though is a much more reasonable alternative. After all, /j/ is supposed to be one of the sources of δασεῖα (ΦΘΧ10). As for the major assumed source of δασεῖα, /s/, its transformation into [j], although unusual, is not impossible; one needs to realise that the environment in which /s/ mutated is identical with that in which german /s/ is voiced; indeed, word-initial /s/ is [z] in Salz|salt=[zalt͡s], Süd|south=[zyːt], etc, as is intervocalic /s/ in gesund|sound, healthy=[ɡəˈzʊnt], Wisent|Bison=[ˈviːzɛnt], etc; a conversion of [z] to [j] would be the opposite of what usually happens (cf. the fate of latin /j/ in French and Portuguese), but [j] would be a reasonable approximation of a voiced sibilant, for example assuming a substratum whose phonology lacked the latter (cf. the spanish rendering of foreign [ʒ] and [d͡ʒ] as y). A two-stage transformation consisting of an internal (assimilation of voice) and an external (approximation of a difficult sound by substratum) phase could plausibly account for the conversion of /s/ to /j/. The resulting semivowels would not prevent elision of preceding vowels, which would bring them into direct contact with... pre-preceding tenues κ, π, τ and converting them either into "palatalised" stops [kj], [pj], [tj] or a "fused" version thereof. We know that eventually word-internal semivowels disappeared, whereas word-initial ones ended up marking the conversion of a potential preceding tenuis to aspirata, otherwise becoming phonetically extinct, too. The relic of this (always hypothetical) process would be a sign (Η originally, the δασεῖα ultimately) that did not have any significance other than the conversion of a previous tenuis.

I cannot resist the temptation of pointing out that, if δασεῖα were related to /j/, the δασεῖα and the ψιλή would correspond to the two yers of Russian, the soft sign (Ь) and the hard sign (Ъ) respectively, which have no (stand-alone) phonetic value, but indicate the presence or absence of palatalisation of the preceding consonant. Although intriguing, I do not want to pursue this theory any further for the following reasons:

The above together with the fact that an identification of the δασεῖα with the palatal semivowel would raise more questions (why does a δασεῖα or a δασύ not appear in cases of synizesis? how is a word like ἱερός to be pronounced? etc) make the only remotely viable identification of δασεῖα with a segment less appealing.

A property?

Our investigation so far has indicated that any identification of δασεῖα with a segment (even the independent laryngeal-only segment [h]) is incompatible with its behaviour in cases of composition, particularly with its ability to teleport, i.e. to get out of the way and reappear at another position; the marginal possibility that it stood for the front semivowel /j/ has other problems of its own, which prevent us from giving it serious consideration. Therefore, the only viable option that remains, if δασεῖα had any phonetic relevance, is what Allen rejects without second thought, namely a (simultaneous) property of the segment it accompanies. Prosody, a term misused by Allen, fulfils these requirements. If we look at the other two "prosodies" of ancient Greek (i.e., other than the πνεύματα), namely accent (be it pitch or stress) and length (be it duration, weight or whatever), neither would stand in the way of elision and crasis.

The accent mark (´) in modern Greek indicates the presence of stress and, like δασεῖα until a few decades ago, is written over the stressed vowel. It does not inhibit the interaction with a previous vowel, when the stressed vowel is word initial. In fact, stress is "transferred" to the previous vowel after the elision of the stress-bearing vowel, for example του έδωσα|I gave (to) him→τού’δωσα ([tu.ˈe.ðo.sa]→[ˈtu.ðo.sa] (note the IPA notation of stress as ˈ before the stressed syllable and how it migrates to the previous syllable). A similar preservation of accent is observed in ancient Greek at least in the case of contraction (cf. SMYT20, p. 40, §171I believe that the strange rules established by Smyth (in SMYT20, §174) about "throwback" of accent in elision is a literal interpretation of the spelling conventions of post-Christian Greek, which may have been graphical rather than phonetic. Note, for example, that all oxytones receive the grave in writing, which, if interpreted literally, would mean that these words were unaccented (the grave ` being the complement of the acute ´) and there would be no (acute) accent to "throw back".). Length may also be considered a binary distinction indicating the presence/absence of the "long" property (in fact, it is very common in modern renderings of Latin to indicate long vowels with a bar, ā, ē, ō and short vowels without one, a, e, o) and it does not get in the way of crasis and elision.However, it is not possible to claim a "transfer" of length in elision and crasis similar to the transfer of stress, as the elided vowel is always the short one (cf. SMYT20, p. 23, §70) and contraction always results in a long vowel (cf. SMYT20, pp. 20-21, §59). If δασύτης is assumed to be a property or feature similar to accent and length, its presence being indicated by the δασεῖα and its absence by the ψιλή, ΦΘΧ7 would be no more an unprecedented paradox. This would mean that we simply have to interpret the teaching of the Grammarians, who classify δασεῖα together with accent and length as prosodies, literally, instead of adopting the bizarre interpretation of Allen’s. A property and not a segment is, therefore, what we should consider as a likely value of the δασεῖα.

In addition to stress and any other prosodic feature, any vowel property, such as height, backness, roundedness, nasality, phonation and pharyngealisation, would fulfil the requirement of "transparency". In other words, there is a host of features that could be considered as candidates for δασύτης and δασεῖα, once we are freed from the [h] prejudice.

One might be tempted to identify δασεῖα and δασύτης with the feature [spread] (spread glottis) defined by Kehrein, in which case a δασυνόμενον (φωνῆεν) would essentially be a so-called "voiceless vowel" ([ḁ]) or "breathy-voiced vowel" ([a̤]) or something in between (KEHR02, p. 84). While such a possibility (contrary to what Allen professes) cannot be ruled out, such an identification would stem from a need to reconcile the a priori identification of the aspiratae with [pʰ], [tʰ], [kʰ], i.e. the preconceived idea that the δασέα comprise the feature [spread]. Furthermore, if δασεῖα referred to the feature [spread], this would imply a number of peculiarities for ancient Greek:

Other than the preconceived idea that φ, θ, χ were stops with an "aspirated release" involving spread glottis, we have no factual reason to favour the solution δασύτης=[spread] over, e.g., the solution δασύτης=[constricted] (creaky voice, implosives). Come to think of it, even [voice], the third "laryngeal feature" considered by Kehrein, makes a good candidate for δασύτης, as it is known to assimilate in both directions (post-nasal <π>, <τ>, <κ> → [b], [d], [g] and <σ>→[z] before <β>, <δ>, <γ>, <μ> in Greek; prevocalic <s>→[z] in German; intervocalic <p>, <t>, <c> → <b>, <d>, <g> in Romance); admittedly, this feature is reserved for the mediae β, δ, γ, but there is nothing (other than the present voiced character of β, δ, γ) to prevent us from claiming a different distribution of [voice] among the consonants of ancient Greek.

The laryngeal features ([spread], [constricted], [voice]) are not the only solutions that could be considered. Any other feature, such as those mentioned above, could be considered, as long as it relates to the following two important attributes:

Most features can be assimilated (A1), but not always in the right direction or order; for example, nasality is usually (e.g., in French) induced from a nasal consonant of the coda to a vocalic nucleus (not from the nucleus to an onset, as is the case with δασύτης).

The requirement that the property be the same for vowels and consonants alike (A2) is too strict. It suffices that δασύτης refer to (or be perceived as) similar or related, but not necessarily identical features. For example, consonantal pharyngealisation is said to trigger vocalic pharyngealisation in some languages (i.e., the same property), but in the case of Syriac it appears that it has an impact on vowel height and backness (i.e., different properties). Furthermore, vowel roundedness is reportedly affected by adjacent-consonant labialisation and vice versa.

Under these considerations and since we have virtually no hint for the phonetic realisation of δασυνόμενα (vowels), we can only hope to be able to infer the nature of δασύτης from the value of δασέα (consonants). Until we complete our investigation on the latter, we have to content ourselves with the only solution that is consistent with all data, particularly that δασέα were single sounds (ΦΘΧ4, ΦΘΧ5) and that δασεῖα does not prevent direct vowel interaction (ΦΘΧ7):

alt: ῾ = a(n unspecified) property

We will resume the investigation at the end of this section.

Syllable quality?

Before concluding the investigation on the δασεῖα, it might be worth examining an interesting observation submitted by a scholiast to Thrax, who describes both δασεῖα and ψιλή as "ποιότης συλλαβῆς|syllable quality" (BEKK16, p. 692). That is, the πνεύματα are modes of articulation that characterise not only the vowel over which they are written, but rather the entire syllable. Certainly, we cannot rely on the testimony of a scholiast, who probably lived at a time that δασύτης was long forgotten phonetically (albeit not graphically), neither are we allowed to assume that he was reciting some older theory (of which we have no other trace in the Grammarians),Some Catholics (e.g. STUR20, pp. 157-158) have seen a confirmation of their [h] theory in the preceding description of the scholiast: "τὸ σημεῖον τῆς δασείας [...] τίθεται ἐπάνω φωνήεντος δασυνομένου, ἤγουν ἐκ τοῦ θώρακος μετὰ πολλῆς τῆς ὁρμῆς ἐκπεμπομένου|the mark of the rough breathing [...] is placed above a vowel pronounced with aspiration, that is, expelled from the breast with much force [Stu’s translation]" (here Sturtevant over-translates, using the terms "rough breathing" for δασεῖα and "pronounced with aspiration" for δασυνόμενον, the circular definition of the scholiast that the mark of δασεῖα is placed over a δασυνόμενον, which by definition is a vowel that... receives the δασεῖα!). This is a case of seeing signs, as the description rather matches any pulmonic sound (which originates in the thorax) and only when one is pre-convinced that δασεῖα is [h] can a "verification" be read in the aforementioned statement. What’s more, this is yet another case of blatant cherry picking, where the context of the statement is plainly ignored; indeed, the scholiast provides a description not only of δασυνόμενον, but also of ψιλούμενον (vowel) as "ἐξ ἄκρων τῶν χειλέων προφερομένου|expelled from the tips of the lips [Stu’s translation]"; if one sees a preceding [h] in the description of the former, scientific decency (and consistency) would necessitate that a labial value (e.g., [w]) of the ψιλή should be concluded from the latter description. The subsequent statements of the scholiast are even more confusing, if taken at face value, as he states "Ἔστι γὰρ ἡ μὲν ψιλὴ ποιότης συλλαβῆς, καθʼ ἣν ἄκροις τοῖς χείλεσι τὸ πνεῦμα προφέρεται, οἶον Αἴας· ἡ δὲ δασεῖα ποιότης συλλαβῆς, καθʼ ἥν ἀθρόον ἐκ βάθους χειλέων τὸ πνεῦμα ἐκφέρεται, οἶον ἥλιος|For the smooth breathing is a quality of a syllable, according to which the breath is expelled with the tips of the lips, as Αἴας; whereas the rough breathing is a quality of a syllable, according to which the breath is expelled all at once *from the depths of the lips*, as ἥλιος [Stu’s translation bar **]" (again Sturtevant conveniently omits "ἐκ βάθους χειλέων"); does this mean that ψιλή is more... "lippy" than δασεῖα? does one originate from the inside of the lips and the other from their outside? but this theory is worth investigating, because it may explain i.a. the assimilation mysteries, the transformation of tenues to aspiratae (ΦΘΧ2), even Grassmann’s Law (ΦΘΧ6).

Indeed, if we assume that ψιλότης and δασύτης were two different manners of pronunciation (e.g., two different ways of modulating the airstream) characterising the entire syllable (in accordance with the rules of Greek syllabification), then many oddities related to δασύτης receive a simpler explanation:

Although the consideration of δασύτης as a suprasegmental syllabic feature is appealing due to all above-mentioned advantages, I hesitate to promote it until a minor (?) problem is resolved: if the theory is true, syllables with aspirata(e) in the onset and/or nucleus should also have aspirata(e) in the coda; however, there is at least one counterexample that I can think of, namely the word ἵππος (ἵπ-πος), as is also evident from the aforementioned syllabification τέ-θριπ-πον. This and any similar exampleI have not found any further examples, but the only possibility seems to be a geminated voiceless stop (ππ, ττ, κκ), the first element of which forms the coda of a δασύτης-bearing syllable. I do not consider θάρρος (i.e., θάῤ-ῥος in post-Christian orthography) a similar example, as I have doubts about the relation between orthography and pronunciation: do the variants ῤ and ῥ reflect a greek or a latin tradition? is ῥ a δασύ (consonant) in the same sense as φ, θ, χ or merely a denser/fuller variant of ῤ (whence its consideration as δασύ)? is the first ῤ truly ψιλόν or a mere orthographic convention? have to be explained away, either as singularities or as results of some other law, e.g., Choeroboscus’ law. Until then, the theory may only be considered a proposal for investigation.

Ceasars and Maharajas

As we cannot reach any definite conclusion about δασύτης (other than its most likely being a property), let us examine the main evidence against the orthodox values. These can be classified as evidence from Latin, from the Indian languages and from the testimonies of the Grammarians.

What the F?

In a recurrent theme, Latin was used as the benchmark of correctness of a candidate (invariably the orthodox) pronunciation. The main argument here relates to the transliterations of the aspiratae into Latin (ΦΘΧ1). We have already discussed this and concluded that the representation of the (single) sounds φ, θ, χ as ph, th, ch does not preclude the latin digraphs standing for the present ("fricative") values of the greek letters. Nevertheless, the argument has a second, more important leg, which essentially questions the reasons behind the selection of the particular digraphs: if φ, θ, χ had their present (fricative) values, why did the Romans not use their fricatives to represent them, specifically f (STUR20, p. 176), s (cf. ALLE87, p. 24), h (cf. CHAT02, p. 348)? As far as θ↔s and χ↔h are concerned, the argument is faulty in almost every possible way:

While the argument that fricative Θ and Χ would have been represented by S and H is absurd, the choice of PH instead of F for representing Φ is a more reasonable basis for (catholic) objection. After all, in all languages that it is used today, F has the same value ([f]) as the orthodox Φ. This is actually the main argument of the Catholics for rejecting ΦO: "Most significant is the failure of the Romans to represent φ by f, as they would certainly have done if the Greek sound had been a spirant" (STUR20, p. 176). As far as normative orthography is concerned, the choice of PH instead of F may be explained as due to reasons of symmetry (i.e., to be consistent with the representations of Θ and Χ by stop-H digraphs). Nevertheless, it would be expected that, if F and PH stood for the same sound, they would be confused by ordinary people (who might be unaware of etymology/orthography) in unofficial inscriptions. Well, the truth is that such confusion does exist: "From the 2 c. A.D. the representation of φ by Latin f becomes common" (ALLE87, p. 24). For this reason, the Catholics accept a coincidence of sound of F and Φ from II AD, but not before that. Before II AD, the frequency of misspellings is certainly lower, but not zero: "the first clear evidence for a fricative pronunciation comes from the I c. A.D. in Pompeian spellings such as Dafne" (ALLE87, p. 23; same in STUR20, p. 181), but also in the first half of I BC, "Trois inscriptions républicaines présentent un f en regard d’un φ|Three republican inscriptions have a f for a φ" (earliest before 88 BC).

From the above, it is fairly certain that there was no difference between F and Φ after I AD. The increased frequency of spelling confusions from II AD does not have to be due to a shift in pronunciation, as many unrelated factors may have contributed. Nevertheless, the Catholics submit further evidence suggestive of a difference between the (original) pronunciations of F and Φ. The most prominent one is Quintilian’s: "nam contra Graeci aspirare F ut φ solent, ut pro Fundanio Cicero testem, qui primam eius litteram dicere non possit, irridet|for the Greeks unlike us aspirate f like their own phi, as Cicero bears witness in the pro Fundanio, where he laughs at a witness who is unable to pronounce the first letter of that name [Butler’s translation]" (Institutio Oratoria, Book 1, 4.14). This is a clear indication that, in Quintilian’s time (I AD) and even more so in Cicero’s time (I BC), the sound represented by F was not part of the Greek phonology and Greeks could not pronounce it properly, but (apparently) pronounced a φ instead.

We do, therefore, have a process of convergence that may have spanned two or more centuries and was almost complete by II AD. Since both letters have today the value [f], it is reasonable to conclude that this was also the value towards which they had converged (otherwise, we would need to assume that F and Φ, after having converged to a different value, independently evolved to their present value). Based on this, the Catholics conclude that Φ did not have the same value as F and only became [f] around II AD.

By now, you should be able to spot the fallacy in the just-mentioned reasoning: Φ≠F means that Φ≠[f], only if F=[f] and the proof is not complete until we prove the latter identity. We have seen this kind of "neglect" in the case of H. Let us check whether the Catholics have done a better job in the case of F. For this letter, Sturtevant states (STUR20, pp. 90-91) that "The later grammarians describe f quite clearly as a labio-dental spirant, that is, as equivalent to English f" and quotes passages by Terentianus Maurus, Marius Victorinus and Martianus Capella, which rather clearly describe the cooperation of lower lip and upper teeth in the production of its sound. One point that is silenced though is the timeline involved: the earliest one (Terentianus) lived in II AD and the other two (Marius and Martianus) in IV-V AD! It is therefore no surprise that they describe the sound [f], which was the point of convergence of F and Φ from II AD. In other words, these "later grammarians" merely confirm the conclusion we have reached by examining the synchronic data (inscriptions and the rate of confusion between PH and F): Φ=PH=F=[f] from II AD.The description of Terentianus, at least as translated by Sturtevant, may be considered as suggesting that there was still a difference between the pronunciations of F and Φ: "Imum superis dentibus adprimens labellum, spiramine leni, velut hirta Graia vites, hanc ore sonabis|Pressing the lower lip against the upper teeth you will sound this letter with a smooth breath, as if avoiding the Greek rough mutes [Stu’s translation]" (STUR20, p. 90). I am not sure about the correctness of the translation of "hirta Graia" (singular) as "Greek rough mutes" (plural); it seems to me that the text speaks about some "rough Greek", but not necessarily a letter, much less a mute and even less an aspirata. But even if the intension was to emphasise some difference with Φ (which is not explicitly mentioned in the text), we know from the inscriptional evidence that such a difference eluded the speakers of Latin at that time (and it would not make sense for Terentianus to point to a difference not understood by at least some of his readers). It may well have been a relic of older descriptions (from a time when F and Φ stood for distinct sounds) or an artificial illusion caused by the convention of rendering Φ by PH, namely a grapheme different from F.

Regarding earlier times, the Catholics only quote the following description of Quintilian’s as hinting at the same sound: "nam et illa, quae est sexta nostrarum, paene non humana voce vel omnino non voce potius inter discrimina dentium efflanda est|for the sixth letter in our alphabet is represented by a sound which can scarcely be called human or even articulate, being produced by forcing the air through the interstices of the teeth [Butler’s translation]" (Institutio Oratoria, Book 12, 10.29). Sturtevant acknowledges that "Quintilian’s description is less clear than the others, but his statement that the sound is 'blown out through the interstices of the teeth' is true only of labio-dental, not of bilabial, f" (STUR20, p. 91). It is curious that the only alternative to a labiodental value considered by the Catholics is a bilabial one, namely [ɸ]. For all we know, if one is to take Quintilian’s description as suggesting that the air is only guided through the interstices of the teeth (which does not follow from the textIn fact, I do not know how we can be sure whether "discrimina dentium" refers to the openings between the upper teeth or to the space between the upper and lower teeth of a slightly open mouth (e.g., as per pronunciation of [s]).), then at least the interdental sounds [θ] and [ð] should have been considered as equally likely. The reason is evidently that Sturtevant advocates a bilabial value [ɸ] for F as its original value, based on older "assimilations as im fronte, CIL i. 1104, and comfluont, i. 199.13", i.e., based on the use of the bilabial nasal M instead of the expected N. Although Allen’s objection (ALLE78, p. 35) that the truly assimilated nasal before F=[f] would be a labiodental one (i.e., [ɱ]) and both M and N would be inaccurate representations thereof (the choice of either of them being "purely a matter of orthographical convention") is reasonable, there are other indications that may be suggestive of a bilabial value (cf. LLOY87, pp. 80, 213). Getting back to the Quintilian quote, to see in it a confirmation of a [f] value for F is wishful thinking; moreover, Quintilian’s aversion to its sound and the use of expressions like "non humana voce" is incomprehensible for a sound so stable and common as [f] is. It is, therefore, fair to say that there is practically no evidence of a labiodental pronunciation of F before II AD.

The lack of positive evidence does not necessarily mean that F≠[f] up to I AD. Nevertheless, consider the following statement of Allen’s: "From the 2 c. A.D. [...] Latin grammarians have to give rules when to spell with f and when with ph" (ALLE87, p. 24). In the earliest case mentioned by Allen, Caper (II AD) specifies that PH is to be used for greek names (nouns?) and F for latin ones. This is taken by the Catholics as an indication that Φ had "developed its fricative pronunciation" by II AD. However, this narrow-minded view disregards the fact that it is PH, not Φ, that rivals F. In other words, the spelling uncertainty is an internal issue of Latin (i.e., the recommendation is addressed to speakers of Latin), not a case of transliteration norm (as it would be if it were a recommendation to Greeks how to spell their words in the Latin alphabet). Consider the following two scenaria:

1. Latin borrows words from Greek and renders Φ by PH, e.g., ΦΑΛΑΓΞ↔PHALANX
2. Initially Φ=PH=[f], F≠[f]2. Initially Φ=PH=[pʰ], F=[f]
3. After some centuries of use, the borrowed words become part of the latin vocabulary
4. By II AD, F→[f] in Latin4. By II AD, Φ→[f] in Greek
5. PH continues to be pronounced [f], as it always has5. PH→[f] in Latin, either parallel to or in imitation (!) of the development of Φ in Greek
6. Genuine spelling confusion is caused by the converged values of F and PH, necessitating the establishment of orthographic rules

The highlighted step (#5) is a necessary assumption for justifying the misspellings F↔PH (#6); if the two graphemes maintained their distinct values ([f] and [pʰ] according to the Catholics), there would be audible discrimination and no reason for any confusion. Need I point out which scenario is reasonable and which defies the rules of logic? Which serious scholar would subscribe to the view that the Romans changed the pronunciation of their PH (particularly in words that were already in use in Latin for centuries) in accordance with the changed pronunciation of Φ in Greek? Unless one lives in the realm of the supernatural, one should concede that it was the sound of F that converged to the value of Φ/PH by II AD and not the other way around. And since we have concluded that the common value had to be [f], it follows that PH was a digraph that had the value [f] from day one of its existence (the same value that was represented by Φ in Greek) and that its establishment was an attempt to render graphically a sound that did not exist in Latin, as a (fricative) variant of the voiceless labial consonant P.

I know that it is difficult for a Catholic to digest this conclusion. The natural question is

"If F was not [f] before II AD, what was it?"

and my natural answer would be

"I do not care; let the Catholics figure it out"

However, since the Catholics are incapable of coming up with any alternative other than a bilabial fricative (which, as Sturtevant correctly points out, is "so similar [to the labiodental fricative] that only a trained phonetician would care to distinguish them in writing"), in this particular case I can make a plausible proposal. Let us begin by observing that if F were [f] it would be an outlier in latin phonology, as it would have no counterpart in the Matrix, neither in the X-axis (i.e., voiced [v]) nor in the Z-axis (i.e., fricatives [θ] and [x] at the other articulators, except possibly [s]); in other words, [f] does not fit in the symmetry of the phonological model (cf. also LLOY87, p. 213). In Romance languages of today, F=[f] is the voiceless counterpart of V=[v] (in Spanish, where V does not have exactly this value, a symmetry was established along the Z-axis, after the development of the sibilants C, G/J into dental and velar fricatives); however, (consonantal) V in classical Latin is universally accepted to have had the value [w] (cf., e.g., STUR20, pp. 39-41). What if classical F was to classical V as modern F is to modern V? Here is, therefore, a proposal for an "original" value of F:

Falt: F = [ʍ]

namely, the voiceless labiovelar semivowel (a sound known in English as the "original" sound of wh in words like "whine"). This value is not only a hunch, but also fits well with other observations:

The assumed development [ʍ]→[f] is also not unheard of. Labiodentalisation of wh(=[ʍ]) to [f] has occured in some dialects of English, such as Scotts. It also appears that the possible change (F=)[ʍ]→[f] in Latin is paralleled by a similar development in "xv [=[xw] or [xʷ] or [wʰ] or [hʷ] or simply [ʍ]?] > f in many Bulgarian and Serbian dialects". I believe that nothing stands in the way of discarding the catholic value F=[f] for the classical times.

To sum it up, the representation of Φ, Θ, Χ as PH, TH, CH in Latin (ΦΘΧ1) does not preclude them having their present values ([f], [θ], [x]) at the time of establishment of these digraphs. S and H, even if they had their present germanic values ([s] and [h]), were too distinct from [θ] and [x] to be considered as acceptable approximations and were, at any rate, already employed as transliterations of Σ and δασεῖα, respectively. The value of F, which is the basis for the main catholic argument, was [f] with sufficient certainty only from II AD, which is the time it started being used more consistently for rendering greek Φ; there is no positive evidence of the same value before II AD; the interpretation by the Catholics of the fact that

(ΦΘΧ12) From II AD on, Latin Grammarians provide orthographic rules about when to use F and when PH

is counterintuitive, as it requires that the value of not only the greek letter Φ, but also of its latin rendering PH changed from [pʰ] to [f]. The scientifically honest interpretation of ΦΘΧ12 is that the axiom F=[f] before II AD does not hold water, but F converged to [f] from another, sufficiently distinct value (certainly not [ɸ]), our best guess being [ʍ]. Under this consideration, the use of P, T, C for originally rendering the Φ, Θ, Χ of Greek (ΦΘΧ5) would be no surprise if these were the present fricatives [f], [θ], [x], as it would be a simple case of using the homorganic voiceless consonants (similar to English pronouncing foreign ch=[x] as [k]; Blass also reports the "Slavonic and Lithuanian representation of late Greek φ or German f as p", BLAS90, p. 102).Note that we also have Quintilian’s testimony that CH was an (old) alternative spelling of C in some latin words, such as choronae, chenturiones and praechones; similarly Gellius for lachrumas and sepulchrum (STUR20, pp. 72-73). If this is true, it may mean that C, but also possibly P, T, were allophonically or alternatively also realised as CH, PH, TH (whatever the digraph stood for) and their early use for what was later rendered as CH, PH, TH (i.e., the greek letters Χ, Φ, Θ) is a natural consequence of this practice. The spelling refinement by appending H to P, T, C (to produce the digraphs PH, TH, CH) could be seen as an attempt to indicate that Φ, Θ, Χ were fricative variants of the respective voiceless stops (P, T, C); this consideration of the nature of H, which is described by the Latin Grammarians as "a(d)spiratio(ne)", would also be consistent with the use of the term by Velius Longus to indicate the friction distinguishing [v] from [w]. The admission that PH, TH, CH had the fricative values [f], [θ], [x] since their introduction in II-I BC (STUR20, p. 70) would free us from making the unfounded assumption that their pronunciation changed to reflect the (alleged) new fricative pronunciation of Φ, Θ, Χ by the Greeks and from having to explain why these digraphs (assumed by the Catholics to stand for "aspirated stops") were used in transcriptions of foreign names from languages that lacked aspirated stops.For example, Ceasar’s "Cherusci" (Caes. Gal. 6.10.5) or reportedly "Latin-script renditions of late Punic includ[ing] many 'spirantized' transcriptions with th and kh in various positions". Chatzidakis also concedes that it is "πιθανὸν ὅτι πρὸ τῆς ἁλώσεως τῆς Κορίνθου [=146 π.Χ.] τὰ δασέα φ, χ, θ δὲν ἐξεφωνοῦντο ἐν τῇ ἐπισήμῳ Ἑλλ. γλώσσῃ ὡς κλειστὰ μετὰ πνεύματος ΠH, ΚH, ΘH|likely that before the fall of Corinth [=146 BC] the aspiratae φ, χ, θ were not pronounced in the official Greek language as stops with aspiration ΠH, ΚH, ΘH" (CHAT02, p. 348), which means that even the Catholics accept that the prerequisite (i.e., pronunciation [pʰ], [tʰ], [kʰ]) behind their interpretation of the digraphs PH, TH, CH (i.e., that they were specifically selected because they better expressed the nature of aspirated plosives) is likely not true!

In fact, it seems that the entire corpus of latin evidence (Grammarians, inscriptions) makes sense only if PH=[f], TH=[θ], CH=[x], h="aspiratio"=friction (a property), f≠[f] (before II AD) and little under the official catholic narrative.

Mahāprāṇa & Mahābhārata

The identification of the greek aspiratae with the so-called voiceless aspirated plosives (as opposed to the sequences of voiceless plosives+/h/) is not a recent development. When the Catholics (erroneously) concluded that Φ could not have been a fricative, because it differed from F, and the (artificial) need for an alternative pronunciation of φ, θ, χ arose, "the ph, th, kh of Sanskrit and the modern Indian languages" (ALLE87, p. 18) provided a convenient model. It may be safely argued that this identification was a consequence of the 19th-century scholars’ fascination with Sanskrit and its (axiomatic) consideration as a conservative language that preserved the majority of PIE features.

There is a tiny problem though: the phonemes (presently) realised as [pʰ], [tʰ], [kʰ] and (misleadingly) represented as <ph>, <th>, <kh> in latinised renderings of indian languages are not the phonological counterparts of Φ, Θ, Χ; as already outlined (in the case of Φ), the greek aspiratae correspond to <bh>, <dh>, <gh> of latinised Sanskrit/Indian, which are presently realised as breathy-voiced stops [bʱ], [dʱ], [ɡʱ] or [b̤], [d̤], [ɡ̈]. So, there is no identity between the greek and the indian phonemes that would suggest they both had value [pʰ].

(ΦΘΧ13) Φ, Θ and Χ's indian counterparts are <bh>=[b̤], <dh>=[d̤] and <gh>=[ɡ̈]

The argument, therefore, shifts to another level, namely that of PIE: Φ, Θ, Χ stem from PIE *bh, *dh, *gh (variably also rendered as *bʰ, *dʰ, *gʰ or *bʱ, *dʱ, *ɡʱ), which were at a first stage devoiced to [pʰ], [tʰ], [kʰ], the "original" values of Φ, Θ, Χ. Needless to say that, even accepting the reconstruction as correct, this is pure speculation and a typical case of Couch Linguistics; how a phoneme developed from a hypothetical value ([bʱ]) to the present value ([f]) is something that we cannot a priori know, as several equally (un)likely possibilities exist, e.g., [bʱ]→[b]→[v]→[f], [bʱ]→[β]→[ɸ]→[f], etc (theoretically speaking); during the immense time (>2000 years) between the common proto-language and the period of interest, anything could have happened.

To make matters worse, the very reconstruction is a product of imagination or preference rather than hardcore science. To illustrate this, consider the already mentioned case of corresponding graphemes <bh> (Sanskrit), <Φ> (Greek) and <F> (Latin); these three roughly contemporaneous languages are a good starting point for determining the value of the assumed common phoneme of PIE, let us call it /¥/ (to preclude any prejudice). In order to determine how /¥/ surfaced in PIE, we need to know the values of its scions in the three languages; as the value of <Φ> is the unknown variable to be (partly) determined by the value of /¥/, we have to rely on the other two graphemes; assuming that these had their present values ([bʱ] and [f]), how is the "reconstructed" value [bʱ] (instead of, e.g., a fricative value) justified? in fact, it appears that the reconstruction either relies on the assumption Φ=[pʰ] to reinforce the significance of the sanskrit sound (in which case, it makes no sense to seek further for the value of Φ, which has been adopted as an axiom) or on the axiomatic assumption that the sanskrit sound was more "ancient" (in which case, we hardly need comparative linguistics).

Even if we take into account the corresponding sounds of other IE language families, we note that except for Sanskrit (and excluding the object of our investigation, Greek) no language exhibits any kind of phonation that would justify the reconstructed *bh.Curtius claims that one of the corresponding sounds of Χ is that of Zend gh (similarly for the other aspiratae); however, this language is not reported to have possessed any aspirated consonants at all! How, then, is the reconstructed value justified, if not by either prejudice (about the superiority of Sanskrit) or recursive reference to the (sought) Greek value?One cannot speak of "scientific work" and at the same time regard PIE reconstruction as the unquestionable truth. An honest (hence, more "scientific") approach is that of Aitchison, who makes it clear (AITC04, p. 177) that a) "there are always enormous gaps in the data available", b) "it is not always possible to deduce the actual pronunciation from written texts", c) "no parent language is ever a single, homogenous whole", d) "daughter languages sometimes undergo independent, parallel developments", and e) "borrowing from neighbours can distort the picture" and concludes that "reconstructions merely represent the best guesses we can make about the parent language in the light of current knowledge". How different from the conviction of many modern scholars that PIE was exactly as the handbooks tell us! I can, therefore, but repeat my general conclusion that any reconstruction is of little significance for determining the classical phonology.

But even if the "reconstruction" /¥/=[bʱ] is correct, the link between that value and the catholic value Φ=[pʰ] is anything but straightforward. It might have been for the 19th-century scholars, who misleadingly referred to the two sounds as "aspirates" and represented them as "bh" and "ph" respectively, evidently based on the latinised rendering of Hindi (which they, like myself, had probably never heard). But linguistics has gone some way since then and, besides the realisation that the proposed value for the latter is actually a single sound ([pʰ]), has provided more accurate descriptions of the former: "the two types of sounds involve two distinct types of voicing and release mechanisms. The series of so-called voice aspirates should now properly be considered to involve the voicing mechanism of murmur" and has acknowledged that "From an articulatory perspective this terminology[=voiced aspirates] is incorrect, as breathy voice is a different type of phonation from aspiration". It is, therefore, evident that, in order to explain the derivation of the catholic Φ=[pʰ] from *bh=/¥/=[bʱ], a mere devoicing is not enough; we need to assume a (significant) change in phonation. It is reported that there are cases where "breathy-voiced stops have subsequently developed into voiceless aspirated stops"; the only trouble is that (at least in the IE family) the main example that proves this statement is... (yes, you guessed it right) Greek!In discussing the reliability of reconstruction, Aitchison states that "Similar [to sanskrit, greek and latin] written records of Albanian or Armenian might dramatically change the picture" (AITC04, p. 177). Sadly, it appears that official linguistics is not willing to adapt the theory to new data, but the data to pre-conceived theories. After discovering the true nature of indian "voiced aspirates", they merely adjusted the PIE phonation to the hindi one, but did not bother explain how it relates to the originally assumed "aspiration". Oh, I forgot; the transformation from breathy to aspirated is known from... ancient Greek! To illustrate the fallacy, consider the following dialogue:
- Bambi was red, because it was a deer and all deer are red.
- But we know of no red deer. In fact, all deer seem to be brown.
- OK, fine. But a brown deer can easily turn to red, as evidenced by Bambi.

If you have reached this far, you should expect the next objection: we have seen that the assumption F=[f] used above is probably wrong; how do we know that the modern indian pronunciation of "bh" was the same in Sanskrit? I have laid out my concerns about Sanskrit. First of all, there are submissions (by Rhankabes and Jannaris) that its phonology is assumed, but not proven. Then, unexpected terminology is reported (by Allen) that is interpreted against intuition and according to preconception. Whence is the certainty about sanskrit phonology derived? Other than the bias on the special status of indian phonology, the phonological model of Sanskrit may only be based on the theories of the indian Grammarians. The only relevant description seems to be the grouping of the consonants, essentially the indian version of The Matrix. Nevertheless, one can at most deduce therefrom that, e.g., the "aspirates" ("bh", "dh", etc) were a (natural?) class (as were and still are φ, θ, χ in Greek), but not what their exact phonetic values where. The "voiced aspirates" (actually breathy-voiced today) and the "voiceless aspirates" are described as "mahāprāṇa" ("big breath", according to Allen), a term also used in connection with the "fricatives"; if one is to give weight to this classification, the most reasonable conclusion is that the consonants of these three distinct (according to modern hindi phonology) classes shared the same characteristic. How, then, can we exclude that they all were of the same type, e.g., fricatives, that developed differently (i.e., the voiced to breathy voiced, the voiceless to aspirates, the sibilants unchanged) in the few millennia between Pāṇini and Sir William Jones? Not only are the modern values not verified, but there are actually indications to the contrary.It may be argued that the non-continuant nature of the "bh" series and the "ph" series is indicated by the term "spṛṣṭa", allegedly meaning "plosive", that accompanies them. I do not know whether this is the normal meaning of the term or whether it is a meaning ascribed to it by modern scholars, given the modern phonology (much less whether the term is actually used by the ancient Grammarians). It is, however, reported that "The ancient Sanskrit grammarians classified the vowel system as velars, retroflexes, palatals and plosives rather than as back, central and front vowels" (emphasis mine); if this is true (for I have not been able to verify it outside Wikipedia), it may mean that the term does not refer to a plosive quality, but possibly to consonantal nature. Note further that the term is also used for the nasals, which can hardly be considered as plosives and are certainly continuants.

Considering that a) the assumed aspirated values for the aspiratae of ancient Greek involve a phonation of a kind different from that of the breathy-voiced stops of modern Hindi, b) the only evidence (other than the assumed ancient-Greek "aspirates") supporting some kind of laryngealised quality of the original PIE phoneme comes from a single language family (i.e., the modern indian languages), and c) the phonetic values of sanskrit consonant are apparently proved "by blatant assertion" (cf. ZEMA94 for the term) and cannot be safely considered as known, the alleged indian justification of the catholic values for the greek aspiratae is probably the longest standing fairytale since... Mahābhārata!It is not too difficult to trace the origin of this fairytale to "the movable nature of the breathing which (a) is easily separated from the explosive element, and leaves the explosive element behind, e.g., πέφυκα for φέφυκα, but (b) just as easily unites with another explosive, and though its position varies, does not do away with the feeling that forms like θρέψω and τρέφω belong to each other" (Curtius, as translated and quoted in DAWE95, p. 23). Because of the latinised renderings "bh", "ph" etc, the scholars (not only Catholic) of the 19th century felt free to treat the single sounds of Hindi and Greek as juxtaposition of a plosive with an h and to move the h around wherever they saw fit, which they considered as a prerequisite for explaining common features of Sanskrit and ancient Greek, specifically Grassmann’s law. The first observation is that, besides Grassmann (which developed separately in Sanskrit and Greek), there is little similarity of behaviour between Greek and Sanskrit "aspitates", other phonotactics rules, such as Bartholomae's law for Sanskrit and ΦΘΧ2 for Greek, not corresponding to each other (cf. DAWE95, p. 23ff). Furthermore, the current interpretation of Bartholomae's law relies on the consideration of the distinctive property of the indian "voiced aspirates" not as voice+"aspiration", but as "murmur", a type of phonation that accompanies the consonant or the consonant cluster in its entire existence. Need I point out how much in line this is with my proposal that the ancient-greek "aspiration" was a property rather than a sound? Sadly, the majority of modern scholarship still abides by the 19th-century practice of considering "aspiration" as an h-appendix that can be severed from the rest of the sound (of which it is an integral part) and move around or spontaneously reproduce.

Before leaving South Asia, it may be worth mentioning some more direct evidence submitted by Sturtevant: "Indian coins of the second and first centuries B.C. show th, ph, and kh for θ, φ, and χ" (STUR20, p. 178). Disregarding the facts that decipherment of inscriptions on indian coins is not always trustworthy and that the identity of the phonetic values between modern indian languages and the Prakrit of the coins has not been established, the transliteration should not come as a surprise to the Catholics, who similarly explain the transliteration of latin F (which they wrongly equate with [f]) by the greek Φ (which they take to be [pʰ]): "The use of φ to represent Latin f (for example, Φουνδάνιος) is not surprising; this was the nearest possible approach to the foreign sound" (STUR20, p. 178); "The fact, on the other hand, that e.g. Latin Fabius is rendered in Greek as Φαβιος is no counterindication even for the period of such transcriptions; for Greek had no other way in which to represent the Latin f, and in such circumstances it would be quite normal to represent it by the symbol for the nearest available sound in Greek, even though this were still a plosive [ph]" (ALLE87, p. 22). If the symbols of the Kharoṣṭhī script had the values commonly accepted, then there was no symbol for the exact value [f], [θ] or [x] (the orthodox values that Φ, Θ, Χ possibly had at that time) and an approximation by the symbols for [pʰ], [tʰ] and [kʰ] would be reasonable. Thus, even if we accept the unwarranted assumptions upon which it is based, this argument has absolutely no evidential value for the alleged "aspirated" character of Φ, Θ, Χ.

Greeks Like It Rough?

Convinced that the indian values (yet not quite those exactly) survived until the Roman period and that the "modern" values of Φ, Θ, Χ did not appear until II AD, many Catholics see a verification of their theories in the descriptions of the Greek Grammarians. To do away with this illusion, I need only mention the observation made by Chatzidakis: "Οὐδαμοῦ βεβαίως λέγουσιν οἱ Γραμματικοὶ ὅτι τα δασέα λεγόμενα συνέκειντο ἐκ τῶν ψιλῶν σὺν τῷ δασεῖ πνεύματι, οὐδὲ ποιοῦνται τὸ παράπαν μνείαν τινὰ ὅτι μετὰ τὰ κλειστὰ π, κ ἐξεφωνεῖτο δασὺ πνεῦμα|Nowhere of course do the Grammarians say that the so-called aspirates were comprised of the tenues plus the rough breathing, neither do they at all make any reference that after the stops π, χ a rough breathing was pronounced" (CHAT02, p. 449).In fact, Chatzidakis accepts a(n almost) "modern" value of the aspiratae since before the time of the Grammarians, as evident from his subsequent statement "ἐπ’ αὐτῶν τὰ δασέα δὲν ἦσαν πλέον κλειστὰ+δασεῖ|at their time the aspiratae were no longer stops+rough [breathing]". This admission by the Greek Pope (or rather the Pope’s legate in Greece) should be enough to end all illusion about confirmation of the Catholic theory by the Grammarians. Nevertheless, some modern Catholics still interpret the words of the Grammarians as the fancy takes them. We have seen an example of how an unclear elementary fragment of description by Halicarnasseus is interpreted as indication of lack of allophony for the velars. And more recently we have seen how the term "πνεῦμα" has been thoughtlessly interpreted as referring to the consonant [h], whereas it in all certainty meant the airstream that is present in all pulmonic sounds and most prominent in the continuants. The various catholic treatises are filled with similar misunderstandings caused by careless and possibly thoughtless reading. For every statement of the Grammarians that the Catholics read, they made two or three further (unwarranted) assumptions.

An argument that we have already seen relies on the characterisation of ζ, ξ, ψ, λ, μ, ν, ρ, σ as "ἡμίφωνα" and of β γ δ κ π τ θ φ χ as "ἄφωνα". It does not need to be repeated here that neither the names per se nor the descriptions of the Grammarians corroborate the view that the former are the continuants and the latter are the stops. The reasons for this partition seem to be anything but what is usually assumed; for example, Thrax considers the ἄφωνα cacophonous and the ἡμίφωνα unintelligible sounds.There are several alternative explanations that one can come up with. The most reasonable is that the ἄφωνα are three series of phonemes that are related to each other through assimilation, which turns a ψιλόν to a δασύ(ε-λειπ-θην→ἐλήφθην) or to a μέσον (ἑπτά↔ἕβδομος) and vice versa (γε-γραφ-ται→γέγραπται, τε-τριβ-ται→τέτριπται), i.e., they form a natural class; on the other hand, the ἡμίφωνα comprise the (not-interrelated) sibilants (σ, ζ, ξ, ψ) and "liquids" (μ, ν, ρ, λ), the latter sometimes behaving metrically not as consonants (in the case of "correptio attica"; cf. ALLE87, p. 107), which may justify the term ἡμίφωνα. An explanation apparently promoted by Dawes is tradition, namely repetition of an older theory that was developed at a time of a different phonological model and was maintained or misapplied by the Grammarians; it appears that in an english textbook of 1892 "f, v, th are classed with p, k, t under the big heading 'Mutes,' and are called aspirated mutes or aspirates" (DAWE95, p. 33). But Allen sees a confirmation "In Aristotle, Poetics, I456b[, wherein] the latter[=ἄφωνα] are described as 'having contact' (μετὰ προσβολῆς) like the former[=ἡμίφωνα], but as not being pronounceable without a vowel" (ALLE87, p. 19). The exact wording of the published passage is: "ἔστιν δὲ ταῦτα φωνῆεν μὲν <τὸ> ἄνευ προσβολῆς ἔχον φωνὴν ἀκουστήν, ἡμίφωνον δὲ τὸ μετὰ προσβολῆς ἔχον φωνὴν ἀκουστήν, οἷον τὸ Σ καὶ τὸ Ρ, ἄφωνον δὲ τὸ μετὰ προσβολῆς καθ᾽ αὑτὸ μὲν οὐδεμίαν ἔχον φωνήν, μετὰ δὲ τῶν ἐχόντων τινὰ φωνὴν γινόμενον ἀκουστόν, οἷον τὸ Γ καὶ τὸ Δ|A vowel is that which without any addition has an audible sound; a semivowel needs the addition of another letter to give it audible sound, for instance Σ and Ρ; a mute is that which with addition has no sound of its own but becomes audible when combined with some of the letters which have a sound. Examples of mutes are Γ and Δ [Fyfe’s translation]". In true Aristotelian fashion, the choice of words is so perverse that the passage almost makes no sense: the ἡμίφωνα have sound (φωνή) of their own with προσβολή and the ἄφωνα have no sound of their own with προσβολή, but they become audible with those that have some sound? Irrespective of the meaning assigned to the terms φωνή and προσβολή, it is not clear what the difference between the ἄφωνα and the ἡμίφωνα is. According to Fyfe’s translation, both types of consonants need another letter (presumably a vowel) to become audible, so that Allen’s assessment is not correct. Furthermore, Allen’s translation of the term "προσβολή" as "contact" has no basis either in the text or in the usual meaning of the term (why would Aristotle refrain from using the term "ἐπαφή" or some of the other common terms for describing a contact?). Contact of which two things? It appears as an attempt to suggest an obstruction of the vocal tract (which is usually achieved with contact of the lips or of the tongue and the palate), thus misleading the reader to conclude a stop pronunciation; however, "μετὰ προσβολῆς" is a feature that characterises both types of consonants and, if it comprised the concept of contact, it would be difficult to explain how it applied to Σ (which is explicitly mentioned as an example). Instead, Fyfe’s interpretation of the term as "addition" (of another letter, in the sense of bringing one towards the other) is more reasonable in this context and consistent with the already mentioned corresponding description of Halicarnasseus (ROBE10, p. 140), which describes the vowels as stand-alone sounds and both types of consonants as non-independent sounds. At any rate, the meaning of "προσβολή" is irrelevant, because it is mentioned in conjunction with both ἄφωνα and ἡμίφωνα and cannot, therefore, be considered the distinguishing feature between the two. What is more important, neither passage describes that the ἡμίφωνα are sounds that can be sustained and the ἀφωνα cannot, which would be the case, if the catholic interpretation were correct. The identification of the term ἡμίφωνα with the continuants and of ἄφωνα with the stops has, therefore, no basis in the teaching of the Grammarians.

A further argument relies on the grouping of the ἄφωνα according to place and manner of articulation. We have seen that, according to the first criterion, they are classified as labial Π, Φ, Β ("ἀπὸ τῶν χειλῶν ἄκρων, ὅταν τοῦ στόματος πιεσθέντος τὸ προβαλλόμενον ἐκ τῆς ἀρτηρίας πνεῦμα λύσῃ τὸν δεσμὸν αὐτοῦ|from the edge of the lips, when the mouth is compressed and the breath, being driven forward from the windpipe, breaks through the obstruction [Roberts’ translation]"), dental Τ, Θ, Δ ("τῆς γλώττης ἄκρῳ τῷ στόματι προσερειδομένης κατὰ τοὺς μετεώρους ὀδόντας, ἔπειθ’ ὑπὸ τοῦ πνεύματος ἀπορριπιζομένης καὶ τὴν διέξοδον αὐτῷ κάτω περὶ τοὺς ὀδόντας ἀποδιδούσης|by the tongue being pressed hard against the extremity of the mouth near the upper teeth, then being blown back by the breath, and affording it an outlet downwards round the teeth [Roberts’ translation]") and "guttural" Κ, Χ, Γ ("τῆς γλώττης ἀνισταμένης πρὸς τὸν οὐρανὸν ἐγγὺς τοῦ φάρυγγος καὶ τῆς ἀρτηρίας ὑπηχούσης τῷ πνεύματι|with the tongue rising to the palate near the throat, and the windpipe echoing to the breath [Roberts’ translation]"). It should be again pointed out that the terms "labial" etc are not used by the Grammarians (cf. ROBE10, p. 149, n. 15), but are our inferences based on their descriptions (within parentheses for Halicarnasseus) and the modern phonology. Thus, the argument that if Φ had the modern labiodental value [f], it would not be grouped together with the (bi)labial Π, is fundamentally wrong: the Grammarians do not use a common term to refer to Π, Β, Φ (much less the term "(bi)labials"), but only provide a coarse description of the pronunciation, specifically that there is an involvement of the lips; the latter is true irrespective of whether Φ is a bilabial stop or a bilabial or labiodental fricative. The classification of the nine ἄφωνα in these three groups according to the (approximate) place of pronunciation makes sense also for the current values of these letters; the (more accurate) description of [f] as labiodental is meaningful for the IPA, which must distinguish between the sounds of all languages, but not within a single language, like (modern) Greek, which does not have any other competing voiceless labial fricative. Furthermore, the relationship between the present (labiodental) value of Φ and the present (bilabial) value of Π is established by their behaviour in the cases of dissimilation (wherein a stop is converted to a fricative or vice versa, in order to avoid sequences of two stops or two fricatives), which is a "modern" development (i.e., not inherited from a previous state of the language when the pronunciation might have been different); thus, the stop Π=[p] is converted to the fricative Φ=[f] in ΠΤ→ΦΤ (e.g., επτά→εφτά), i.e., not to [ɸ], which would be the bilabial equivalent. Thus, [p] and [f] do form a natural class ("voiceless labials") in Greek, even though they differ in the passive articulator.The same is true for many semitic languages, where [p] and [f] (not [ɸ]) are allophones of the phonologically "stop" phoneme /p/ (כ); this holds true not only for Hebrew (where a Yiddish, i.e. German influence might be claimed for the modern language), but also for Aramaic (where one also finds the pairs [b]/[v], [t]/[θ], [d]/[ð], which are also not exactly corresponding), while in Arabic [f] (not [ɸ]) is a lenited version of "An original */p/". Similarly in Icelandic, [pʰ] and [p] alternate with [f], not [ɸ] or any other bilabial.

At this juncture, it is probably meaningful to diverge a little and discuss the distinction between the bilabial fricative [ɸ] and the labiodental one [f]. The latter is usually described as the present value of Φ, while the former is often postulated as a transitional value, before Φ assumed its "modern" value. This distinction is immaterial. As already mentioned, the acoustic difference of the two fricatives is so trifle that it takes a lot of training to distinguish it (cf. STUR20, p. 178). Moreover, the former is much rarer than the latter; in some languages its existence is due to special (phonological) circumstances (e.g., japanese [ɸ] is only encountered as [ɸu], as in Fukushima, Fuji, which phonologically is /hu/); there are only a few languages that employ both of them and, in order to distinguish between them, the features of one of them are exaggerated (e.g., in Ewe, "The f and v are stronger than in most languages, [f͈] and [v͈], with the upper lip noticeably raised"). In languages that do not possess both, the corresponding phoneme (be it /ɸ/ or /f/) may be realised in either mode without noticeable impact. In the case of English, the description of Lagefoged is most telling: "In English there is no phonemic distinction between [ɸ] and [f] or between [β] and [v]. Consequently, there is no need to maintain a perceptual distinction between these two possibilities. In saying a word such as „fin", the upper lip is seldom lifted out of the way. The friction is often formed between the lower lip and both the upper teeth and the upper lip" (KEHR02. p. 61). I can confirm the same about my pronunciation of greek Φ: in the course of normal speaking, my lower lip is raised towards the upper teeth, but usually its inner surface comes into contact with the teeth, the extremity of the lip is forming a narrow passage with that of the upper lip and the air escapes both through the interstices of the teeth and through this narrow passage. This articulation could be described either as bilabial or labiodental, both descriptions being (partially) correct. I will, therefore, never consider the two fricatives as distinct; if there are indications about one, they will be also valid for the other. The only difference is that the labiodental pronunciation is more stable (hence more common) and the bilabial one is weaker (which i.a. accounts for its disappearance in medieval Spanish; cf. LLOY87, p. 322ff).

Going back to the evidence provided by the Grammarians, which the Catholics attempt to read as a confirmation of the aspirated pronunciation, a further grouping is based on the manner of pronunciation of the ἄφωνα: "τῶν δὲ καλουμένων ἀφώνων ἐννέα ὄντων τρία μέν ἐστι ψιλά, τρία δὲ δασέα, τρία δὲ μεταξὺ τούτων· ψιλὰ μὲν τὸ κ καὶ τὸ π καὶ τὸ τ, δασέα δὲ τὸ θ καὶ τὸ φ καὶ τὸ χ, κοινὰ δὲ ἀμφοῖν τὸ β καὶ τὸ γ καὶ τὸ δ|Of the so-called 'voiceless letters,' which are nine in number, three are smooth, three rough, and three between these. The smooth are κ, π, τ; the rough θ, φ, χ; the intermediate, β, γ, δ. [Roberts' translation]" (ROBE10, p. 148). The simplistic catholic interpretation is that, since Φ, Θ, Χ are described as "δασέα" and the "aspiration" as δασεῖα or "πνεῦμα δασύ", there must have been some "aspiration" (i.e., [h], as they believe) in Φ, Θ, Χ. However, even if "aspiration" were [h] (which is highly unlikely), the term δασέα for Φ, Θ, Χ would not imply that they comprise a δασεῖα (as the term δασυνόμενα would probably do), but it might very well mean that they resemble the δασεῖα; since [h] is commonly (albeit not entirely correctly) considered a "fricative", it would not be a surprise (even for Catholics, who believe that ῾=[h]) if a similar term were used for fricative Φ, Θ, Χ.

Unfortunately, the Grammarians do not tell us how the δασέα were pronounced. Halicarnasseus provides useless circular definitions: "καὶ ψιλὸν μέν ἐστιν αὐτῶν τὸ π, δασὺ δὲ τὸ φ, μέσον δὲ ἀμφοῖν τὸ β|Among these π is smooth, φ rough, and β comes between the two [Roberts' translation]", "ψιλὸν μὲν γὰρ αὐτῶν ἐστι τὸ τ, δασὺ δὲ τὸ θ, μέσον δὲ καὶ ἐπίκοινον τὸ δ|τ being the smoothest of them, θ the roughest, and δ medial or common [Roberts' translation]", "οὐδὲν οὐδὲ ταῦτα διαφέροντα τῷ σχήματι ἀλλήλων, πλὴν ὅτι τὸ μὲν κ ψιλῶς λέγεται, τὸ δὲ χ δασέως, τὸ δὲ γ μετρίως καὶ μεταξὺ ἀμφοῖν|These, again, differ in no way from one another as regards formation; but κ is pronounced smoothly, χ roughly, γ moderately and between the two [Roberts' translation]" (ROBE10, pp. 148-151); it is evident that one is taught how to pronounce the δασέα Φ, Θ, Χ, only if one already knows what it means to pronounce a consonant "δασέως"! Nevertheless, some Catholics isolate the first part of the last quote ("οὐδὲν οὐδὲ ταῦτα διαφέροντα τῷ σχήματι ἀλλήλων|These, again, differ in no way from one another as regards formation") and claim that, since the δασέα do not differ from the ψιλά, they must have been stops with the addition of some further feature, which they specify as "aspiration" [h]. This is yet another biased interpretation, which disregards the fact that in modern-Greek phonology the difference between the voiceless stops (Π, Τ, Κ) and the corresponding fricatives (Φ, Θ, Χ) is merely one of stricture: once the occlusion of the stop is relaxed, the airstream is afforded a narrow passage and the corresponding fricative is produced. This is more evident in the case of Κ and Χ, to which the cited passage actually relates.

The description of Halicarnasseus concludes: "τούτων κράτιστα μέν ἐστιν ὅσα τῷ πνεύματι πολλῷ λέγεται, δεύτερα δὲ ὅσα μέσῳ, κάκιστα δὲ ὅσα ψιλῷ· ταῦτα μὲν γὰρ τὴν αὑτῶν δύναμιν ἔχει μόνην, τὰ δὲ δασέα καὶ τὴν τοῦ πνεύματος προσθήκην, ὥστ’ ἐγγύς που τελειότερα εἶναι ἐκείνων|Of these the best are those which are uttered with a full breath; next those with moderate breath; worst those with smooth breath, since they have their own force alone, while the rough letters have the breath also added, so that they are somewhere nearer perfection than the others [Roberts' translation]" (ROBE10, p. 150). Faithful to the catholic practices, Allen isolates four words from this passage ("τὴν τοῦ πνεύματος προσθήκην|the addition of the breath", emphasis in the original) and declares that "Such a terminology would be eminently appropriate to the opposition of aspirated and unaspirated consonants, but hardly to the distinction between fricative and plosive, i.e. between incomplete and complete closure of the organs" (ALLE87, p. 19). One can only wonder whether here Allen deliberately tries to mislead the reader or honestly believes what he writes, because he is so biased that he overlooks basic facts of phonology. The difference between a fricative like [x] and a stop like [k] is merely the feature [+cont], which characterises the former and means that "air passes through [the vocal tract] in a continuous stream"; in other words, [x] is [k] with the presence of (a noticeable) airstream (cf. also the remark of Papadimitrakopoulos on Blass' similar misinterpretation of the expression "τῷ πνεύματι πολλῷ|with a full breath"). It is probably because of his failure to realise that the term "πνεῦμα" (as used throughout the descriptions of Halicarnasseus) refers to the airstream, that prevents Allen from considering this possibility. He further seems to interpret the term "προσθήκη" not as "addition" (its normal meaning), but evidently as "concatenation" or "appending"; how erroneous this interpretation may be is illustrated by, e.g, the consideration that [b] can be described as [p] "with the addition of voice", without meaning that "voice" is a segment that follows [p]. The description of Halicarnasseus may, therefore, well refer to a stop/continuant opposition, contrary to Allen’s assertion. The same holds true for the passage of Pseudo-Aristotle "δασεῖαι δ’ εἰσὶ τῶν φωνῶν ὅσαις ἔσωθεν τὸ πνεῦμα εὐθέως συνεκβάλλομεν μετὰ τῶν φθόγγων, ψιλαὶ δὲ εἰσὶ τοὐναντίον ὅσαι γίγνονται χωρὶς τῆς τοῦ πνεύματος ἐκβολῆς|rough/thick/full are those of the voices in which we expel the breath immediately with the sounds, smooth/thin/bare are on the contrary those which are produced without the expulsion of breath" cited by Allen (ALLE87, pp. 15, 18 and 162). Allen himself (!) admits (ALLE87, p. 18, n. 14) that the expression "μετὰ τῶν φθόγγων", where "μετὰ" means "with" rather than "after" the sound, "might be interpreted as implying simultaneous breath, i.e. friction", but then he discards this possibility because he believes that the word "εὐθέως|immediately, as soon as" is incompatible with his (!) interpretation of "μετὰ". However, in yet another brilliant moment of scientific deception, Allen refrains from citing the entire passage (he only quotes the second part, about "ψιλαὶ", and the words "εὐθέως μετὰ τῶν φθόγγων" from the first part) and understandably so, because the word "συνεκβάλλομεν", which he conveniently omits, leaves no doubt that the breath (i.e., airstream) is sent out together with the sound. In other words, the "improbable" interpretation of the terms "δασύς/δασεῖα/δασύ" is the "aspirated" realisation and the most likely is a realisation that involves the simultaneous expulsion of the airstream, like a... fricative! If I were a Catholic, I would vehemently claim that "De Audibilibus" is the work of a "byzantine" author.

This brings us to the main catholic misconception, namely the meaning of the terms "δασύς/δασεῖα/δασύ" and "ψιλός/ψιλή/ψιλόν". Allen argues (ALLE87, pp. 18-19) that they relate to a "binary opposition" and, quoting three passages of Herodotus (iv. 175, iii. 32, iii. 108), concludes that it is "a 'privative' opposition, contrasting the presence with the absence of an additional discrete feature, rather than one inherent quality with another". One can only be awed by the mastery with which Allen twists the meaning of these words and presents them as something they are not, namely an opposition of the yes/no type and still not a general binary distinction (for it would also be compatible with the orthodox specification [+cont]/[-cont]), but one that refers to "an additional discrete feature", which he apparently identifies with the sound [h] that he appends to [p], [t], [k] to (misleadingly/wrongly) represent the aspirated stops. He also refers to the use of the terms δασύ and ψιλόν to refer to the πνεῦμα δασύ and πνεῦμα ψιλόν, where he declares "there is no doubt that this is a privative opposition of the aspirate [h] to zero". In all this reasoning, Allen is trying too hard to prove something he has decided in advance and resorts to all the usual tricks of catholic scholarship:

Nice try, Syd!

I believe that the reason behind this misinterpretation is that the issue is looked at from the wrong perspective. It is also exacerbated by the misconception that πνεῦμα=[h]. If we are to look at the two terms ψιλός and δασύς without prejudice that they relate to "aspiration", i.e., [h], we can, in general, adopt Sturtevant’s definition of ψιλός αs "'bare, without something which might naturally accompany it'. When contrasted with δασύς it must mean 'without that whose presence is denoted by δασύς'" (STUR37, p. 117). But this only means that the consonants called ψιλά are "plain vanilla" (as we have already concluded) and does not tell us what was that which characterised the δασέα. As regards the meaning of δασύς, we can see that its most usual meaning is "thick". The other meanings (including "hairy, shaggy", which is one that cannot be easily associated with some phonetic feature) appear to be non-original secondary developments. This is more evident if one takes into account the latin cognate "densus|dense", which is more likely to be its original meaning.cf., e.g., the common modern term δασύτριχος (from Koine "δασύθριξ"), which would be a pleonasm (like "shaggy-haired"), if the main meaning of δασύς were "hairy". Instead, it is clearly used here with the meaning "dense". It is, therefore, not unreasonable to argue that a consonant characterised as "δασύ" with respect to a "ψιλόν" one, is one that is thicker or denser, i.e., fuller than the latter. Hence, to consider, under the orthodox phonological model, a fricative as a fuller sound than a stop is an acceptable interpretation of the terminology of the Grammarians. This does not mean that the term δασύς necessarily describes a fricative. It would also seem reasonable for the Grammarians to have referred to (pre- or post-)aspirated stops as "fuller" than the plain ones, as the airstream is more noticeable in them; however, since in pronouncing (pre- or post-)aspirated stops the airstream is (at some point) choked by the occlusion, this possibility seems to be less likely than a pronunciation involving an unobstructed airstream.

It is also worth examining the most common translation of δασύς among the Catholics, namely "rough"; e.g., Allen states about the voiceless mutes that "The two varieties were categorized by the Greek grammarians as (γράμμα) ψιλό ('smooth, plain', i.e. unaspirated) and δασύ ('rough', i.e. aspirated). The expected Latin translation of these terms would be (littera) lenis and aspera [...] But in fact the Latin terms,as found e.g. in Priscian, are tenuis and aspirata" (ALLE87, p. 15). It is not clear how the meaning "rough" has come about, but the reason seems to be the latin term "asper", the origins of which are obscure (perhaps by false etymology from "a(d)spirare|to breathe (to/at)"[LAT]?). Otherwise, the Greeks do not use any other term meaning "rough" (such as "τραχύς"), but appear to consistently use "δασύς".The first instance of the term "τραχύς" used for the aspiratae must be the description of Aristides Quintilianus: "καὶ ἔστι λίαν τραχέα|and are very rough" (BLAS90, p. 101, n. 1). It is not clear whether he reflects an older tradition or, "writing in the third century", he translates the (by then established) latin term "asper". It is, therefore, unlikely that the intended meaning of "δασύς" in connection with the letters Φ, Θ, Χ is "rough". In that respect, note also the aforementioned declaration of Halicarnasseus: "κράτιστα μέν ἐστιν ὅσα τῷ πνεύματι πολλῷ λέγεται, δεύτερα δὲ ὅσα μέσῳ, κάκιστα δὲ ὅσα ψιλῷ|the best are those which are uttered with a full breath; next those with moderate breath; worst those with smooth breath [Roberts' translation]". How can anything "rough" be considered "best"? The characterisation of δασέα as "best", is understandable if δασύ is considered to mean "full" or "dense" (cf. Roberts' translation "with a full breath") and little if it means "rough". Again, we are not informed whether the presence of full breath refers to the airstream accompanying the fricatives or that following the (post-)aspirated stops (as per Catholicism); however, it is clear that if one is to be favoured more than the other, it should be the former.

The Catholics had better take Chatzidakis’ advice: there is no verification of an aspirated value in the ancient texts.

Fiction or Friction?

We have seen that, contrary to the widespread (catholic) belief, there is no conclusive evidence in favour of aspirated values for Φ, Θ, Χ. In fact, fricative values of the δασέα are at least as compatible with the data presented so far as the aspirated values. Thus, in accordance with the Presumed-Innocent Principle, both catholic and orthodox values should be considered as potentially valid. But do we have any evidence that might tip the scale in the orthodox direction? In accordance with the Principle of Tradition set out in this work, we do not need to demonstrate the validity of the present values; the onus is on the Catholics to provide evidence of their incompatibility with the evidence and we have seen that they have not done a good job at it. Nevertheless, one can find further evidence suggestive of the orthodox values and not easily explainable under the catholic model.

Progressing Backwards

Let us try to work our way from the edge of Antiquity towards the Classical time. Allen concedes that "fricative pronunciation was well established in the Byzantine period" (ALLE87, p. 25). The term "Byzantine" is (intentionally?) vague, as it spans a period of 12 centuries, but what Allen has in mind is probably the beginning of the "Byzantine era". Indeed, "In the fourth century A.D. Latin f was the regular transcription of φ; and in the same century Ulfilas represented θ and φ by Gothic ϸ and f; for example Ϸomas=Θωμᾶς, Filippus=Φίλιππος" (STUR20, p. 181). We therefore, have the consent of the Catholics to consider IV AD the time of generalised fricative pronunciation of Φ, Θ, Χ.

It is, nevertheless, strange that the Catholics advertise IV AD as the start of "fricative times" and not II AD, which has been identified above as the start of the generalised Φ↔F transliterations. The further evidence we have provided (based on the correct interpretation of ΦΘΧ12) suggest that Φ was [f] also in the times before II AD (a convergence of Φ and F towards [f] down different paths to arrive at the final value [f] at the same time is rather unlikely). It would, therefore, be reasonable to claim that the spirant pronunciation was effective throughout the imperial times.

The (northwest) semitic languages provide a good reference point for verifying fricative values, as they were affected by the begadkefat, namely the fricativization of "original" stops /b/, /g/, /d/, /k/, /p/, /t/. This phenomenon must have been so intense in III and II BC, when the Septuagint was written, that (according to Sáenz-Badillos) "The bgdkpt consonants had just one realization in Greek, namely, as fricatives", i.e., as Φ, Θ, Χ (e.g., יוסף→ΙΩΣΗΦ, יפת→ΙΑΦΕΘ, כלב→ΧΑΛΕΒcf. also the names of the letters of Hebrew in the Book of Lamentations, i.e., Θ for ת (ΒΗΘ-בית, ΔΑΛΕΘ-דלת, ΗΘ-חית, ΤΗΘ-טית, ΘΑΥ-תו), Φ for פ/ף (ΑΛΕΦ-אלף, ΧΑΦ-כף, ΦΗ-פא, ΚΩΦ-קוף), Χ for כ/ך (ΧΑΦ-כף, ΣΑΜΕΧ-סמך).). That the indicated values were fricatives (and not "aspirated" consonants) is also suggested by the regular transcription of the uvular/pharyngeal fricative ח as Χ (רחל→ΡΑΧΗΛ, אחאב→ΑΧΑΑΒ). Of course, the Catholics (who do not even mention this in their books) will contend that Greek did not have the fricative values in their inventory, so they used the closest sounds (i.e., the "aspirated" ones). In addition to not being clear why [kʰ] and not [h] is the closest to [χ]/[ħ] and why [s] was not considered for [θ] (cf. the arguments on the latin transcriptions), it should be noted that the transcriptions of פ ,ת ,כ as Φ, Θ, Χ (and PH, TH, CH) have a history of at least seven centuries, extending into the times when even the Catholics concede a "modern" pronunciation of the Greek aspiratae (and their latin equivalents). Indeed, they are used in late Punic of I BC, along with their latin equivalents PH, TH, CH almost since the time of their conception,In the latino-punic inscriptions, starting from the destruction of Carthage (145 BC) until IV AD, "Punic /k/" was mostly represented by CH, "Punic /p/" almost always by F (only once as PH, which however is described as homophone of F "in North Africa in the period under discussion") and "Punich /t/" by TH in most words. Remember that the digraphs PH, TH, CH were introduced in the second half of II BC (the beginning of the neo-Punic era) and their preferred use for indicating the punic fricatives ("lenited stops") [f], [θ], [x] would be difficult to explain if they were not designed to represent fricative values in Greek, as well. but also later, e.g., systematic use of Φ, Θ, Χ in the Secunda of III AD and of PH, TH, CH by Jerome in IV AD. The consistent use of Φ, Θ, Χ (and their latin equivalents) for semitic fricatives over a period of seven centuries does not fit well with the catholic timeline (aspirates before II AD and fricatives afterwards). What a curious coincidence that the graphemes (both in Greek and Latin) chosen to represent the semitic fricatives, despite not being fricatives themselves (according to the Catholics), came to represent exactly those values! If we are to avoid five layers of assumptions for explaining away the transliteration of Semitic, we can pretty safely state that the fricative values are the reasonable realisations of Φ, Θ, Χ in republican and hellenistic times, too.

(ΦΘΧ14) Φ, Θ and Χ regularly represent NW-semitic פ/ף(=[f]), ת(=[θ]) and כ/ך(=[x])

One can also find similar evidence in transcriptions of the classical times. Chatzidakis (yes, the Greek "Pope") is compelled to concede an orthodox value for Θ (and by analogy for Φ and Χ) in Attic (and Ionic) of IV BC, because "τὸ ἀρχ. Περσ. καὶ Ἰταλικὸν θr πρῶτον μὲν παριστᾶτο διὰ τοῦ τρ, ἔπειτα δὲ διὰ τοῦ δρ καὶ θρ|the ancient persian and italian θr was initially represented by τρ and subsequently by δρ and θρ" (CHAT02, p. 456). His "italian evidence" seem to be confined to the single word "λίτρα ἀπὸ τοῦ Ἰταλικοῦ liθra ὅθεν ἔπειτα libra|λίτρα from the italian liθra whence later [came] libra", the transmission time of which he declares as unknown, yet "οὐχὶ πρὸ τῆς Η᾽[π.Χ. ἐκατοντ.]|not before the 8th [century BC]"; due to its singularity and uncertainty, the significance of this "evidence" is next to none. The Persian evidence is, however, more abundant and, hence, more important. Chatzidakis sees an indication of phonological change in the variation of the transliterations of persian [θr], as he notes the use of ΤΡ in Aeschylus and Herodotus, but of ΘΡ in Ctesias and some inscriptions of Asia Minor in IV BC. The latter is certainly of some weight: why use an aspirated stop for a fricative, particularly in a position where any "aspiration" would be least audible (i.e., [tr] and [tʰr] would be much less distinguishable than, say, [ta] and [tʰa])?The catholic counterargument may be that the ability of the ancients to distinguish between stops and "aspirates" in that position (i.e., /_[r]) is illustrated by minimal pairs such as Κρόνος|Saturn/χρόνος|time. I would say that the very existence of such minimal pairs rather speaks strongly of a phonetic difference between Κ and Χ larger than a mere (appended) exhalation. The former might be suggestive of a non-fricative value for Θ, but this does not necessarily follow. First of all, the time interval between Aeschylus/Herodotus and Ctesias is too short (they were practically contemporaries) to justify an abrupt change from a predominantly stop to a predominantly fricative value. Moreover, the source of the earlier instances need not be directly persian; it could have been an intermediary third language (Lydian? Aramaic?), which did not distinguish between [θ] and [t] or at least [θr] and [tr]; the later, more accurate forms may have been the result of direct contact with native speakers of Persian (in fact, Ctesias did spend several years with Persians). I am not aware of any more thorough study of Greek-Persian transliterations, which could corroborate this conclusion. Furthermore, I have not been able to find evidence establishing the time of emergence of persian fricatives from "original" *p, *t, *k (which is the source of persian [θr]), either; although these are commonly held to have been a feature of Proto-Iranian, it appears (p. 10) that "we do not have certain evidence that the evolution K → X /_C took place already in Proto-Iranian"; as such, it is also likely that the fricativization of stops gradually occurred during the times of the Achaemenid Empire, perhaps under the influence of Aramaic, which underwent a similar sound change about that time. Whatever the case with persian [f], [θ], [x] < *p, *t, *k /_C may be, we do have also evidence of Φ, Θ, Χ being used for fricatives of different (older) origins, such as Ἀρταφέρνης for Artaphernes. The persian transliterations, thus, bring the orthodox pronunciation of Φ, Θ, Χ in the heart of the classical period.

It can, therefore, be seen that evidence from the Middle East is rather compatible with the historical (=orthodox) pronunciation of the aspiratae over a period that extends all the way inside the classical age. I have not seen any transliterations that could confirm the catholic thesis. Conveniently, there is no language with which Greek came into contact (other than the Prakrit of I BC in remote Bactria) that could have possessed aspirated stops.

Update 10/2018: Lately I revisited Maurice Pope's excellent book on the decryption of ancient scripts (POPE99) and studied more closely the decipherment of the Persian Cuneiform. Pope describes (pp. 99-102) the method adopted by Grotefend, who identified the names of two kings and matched their signs to "Darius" and "Xerxes". From the latter, he managed to identify 𐎧𐏁 with the Greek letter Ξ, assigning (?) to the first symbol the value [k] and to the second one the value [s]. Pope goes on (p. 105) to explain how Lassen refined Grotefend's association of symbols and sounds. In the case of Xerxes' name, "The first sign 𐎧, which was also the first sign in the word for king, was clearly a 'k' sound. The question was whether it should be aspirated". Now, that was a surprise for me, because I did not expect the term "aspirated " to have been used in 1836 with its modern sense (it appears that the values [pʰ], etc for the Greek ψιλά were established in the second half of the 19th century). Bypassing the question of what Lassen might have meant with this term,Even after having studied the relevant part of Lassen's book, I am still not sure what Lassen had in mind. He says that Grotefend identified 𐎧 and 𐏁 with (German) ch ([x]) and sch ([ʃ]) and states "ich schreibe dafür kʿ und sʿ womit ich dieselben Laute meine|I write instead kʿ and sʿ, with which I mean the same sounds", which only implies a change of notation. Then he associates the first of these symbols to the kʿ of "Zend" (a language that had no aspirates, but at the time kʿ could have been understood as [kʰ]), as well as the Greek χ (which was probably still identified with the German ch). Unfortunately, the texts of 19th-century linguists (as well as those of many moderns of obsolete mentality) are a phonological mess. I realised that Pope's intention was to refer to [kʰ], as he concludes that "it should be aspirated - for not only is Avestan [where the word for king starts with kh] elsewhere closer to Old Persian than Sanskrit [where the word for warrior starts with k] but the same sign occurs in the second place in the word for 'Achaemenid' where the Greeks transliterated it with a chi". So, we do have an important (pre)classical association of Greek Χ with an aspirated sound (𐎧) of another language, which may strongly suggest that Χ had its Catholic value in classical times or right before it. Right? WRONG! It results that Old Persian did not have aspirated consonants and the value of 𐎧 is... [x]! I suspect that the mapping of Greek ψιλά to Old-Persian fricatives is confirmed by further examples (like Φραόρτης and the already mentioned Ἀρταφέρνης), which compels us to acknowledge that the real evidence points to fricative ψιλά in the classical era, too, and that no evidence for "aspirated" values are to be found outside the catholic imagination. Dear Catholics, wake up and smell the... friction.I find it dishonest that modern scholars are willing to revise the phonological models of all other languages but Greek. In the case of Avestan and Old Persian, the existence of both voiceless and voiced aspirates was postulated in the 19th century, just like for Greek, undoubtedly due to the widespread infatuation with Sanskrit. This theory is now abandoned and all Iranian is reconstructed to include plenty of fricatives, but the values of the Greek phonemes linked to those fricatives in several instances have not even been doubted, much less revised. Pope is an example of such dishonesty, as he assumes an aspirated value for the Greek ψιλά, e.g., for explaining Lassen's method, but fails to point out that the initially assumed aspirated values turned out to be fricatives. Neither can it be argued that Pope was not aware of recent developments. On p. 109, he provides a "Comparative Table" of the evolution of the decipherment, wherein after listing the values "kh", "k‛", "k’h" (!!) and "kha/khu" proposed up to 1846 for 𐎧, he mentions the value arrived at in "an accepted modern system of transliteration (from Kent 1950)": "xᵃ"!

Coughing CHasm

As observed above, we cannot rely on evidence from other languages for establishing the pre-classical values of Φ, Θ, Χ; the literary evidence is scarce (if any), our knowledge of foreign phonologies limited (if not speculative) and the linguistic theory self-proving (i.e., circular). Therefore, we are virtually in the dark. However, having tradition as our guide, a hint might be provided by the internal evidence of the language. One such type of evidence is onomatopoeia.

I was explaining to some Spanish friends of mine that the origin of the word caos|chaos[SPA] is the Greek word χάος, which is pronounced [xaos], not [kaos], as it is in Spanish. I also explained that its original meaning is essentially "void", which is reflected in the word’s stem ("χα"=[xa]). Later, I realised that, in providing the onomatopoetic etymology, I used the present-Greek articulation, whereas the word is a very old one (attested in so old a text as Hesiod’s "Theogony", VIII-VII BC) and the linguists tell us that the pronunciation of Greek Χ in ancient times was [kh], or better [kʰ], so I was probably wrong. But my explanation, based on the "modern Greek pronunciation", makes much more sense than any explanation based on Χ being an "aspirated Κ". For, no matter how much "aspiration" one adds to (the obstruent) Κ, the "reconstructed" pronunciation of χάος as [kʰaos] does not give the impression of a "vast chasm, void" (as is the definition of "chaos"). In fact, "chasm" also comes from Greek χάσμα, which only when pronounced the "modern" way can be associated with "space", "hollow" or... "open mouth" (as is the English rendering of "chasm"). The same can possibly be said about Charybdis=Χάρυβδις. All these words relate to the notion of a gaping space and their first syllable χα conveys that notion only in its orthodox pronunciation [xa], [x] and [a] being the most "open" of the greek consonants and vowels respectively (the voice of [ɣ] giving the impression of a "fuller" sound, the sibilants providing a different effect and all other continuants involving some kind of contact). On the other hand the catholic pronunciation of the syllable as [kʰa], where χ is a plosive, rather conveys the impression of an explosion; this is very unlikely to relate to onomatopoeia or etymology, unless Hesiod had heard of the... Big-Bang theory! It is customary for the Catholics to argue that, as fricatives ([x] in this case) were (allegedly) absent from ancient-Greek phonology, aspirated equivalents ([kʰ] in this case) were their best approximation; however, this is not a case of needing to approximate the sound [x], but to find the right phonemes that imitate a "vast chasm" or "open mouth"; if the catholic model were correct, the best choice for this purpose would be ἄος, ἄσμα, or even better ἅος, ἅσμα (with the addition of an initial "aspiration", which was allegedly equivalent to [h], the most "open" consonant); why then add a plosive that creates the exact opposite impression from the intended meaning?

Another word that is most likely onomatopoeic, is the verb φυσάω-φυσῶ; the effect of the act of blowing, i.e., letting a continuous stream of air escape the mouth, is described by this word only under the orthodox model Φ=[f] (or at most Φ=[ɸ]); the modern (greek) rendering of the act is through the syllable φου-φου, that is [fu-fu], and it is reasonable to assume that the ancients coined this word to imitate the exact same sound (remember that Υ's "original" value was almost certainly [u]). Under the catholic model, on the other hand, the pronunciation of this syllable, i.e. [pʰu], is rather suggestive of... spitting, a notion that in ancient Greek is better served with a combination of (actual) plosives, namely πτύω.

A related comment that is suggestive of a fricative pronunciation in classical Athens is provided by Plato (in the words of Socrates), who describes the letters Φ, Ψ, Σ, Ζ as "πνευματώδη|pronounced with much breath [Fowler's translation]" and argues that the (imaginary) "namegiver" uses them "ὅταν που τὸ φυσῶδες μιμῆται|Whenever he imitates that which resembles blowing [Fowler's translation]" (Plat. Crat. 427α).The evidence from Cratylus has often been ridiculed by the Catholics, due to the fallacy of the etymologies provided by Socrates (e.g., CHAT98, p. 375). However, this is just another case of fighting the straw-man: it is not the etymologies that constitute the evidence, but rather the fact that they would not be credible or plausible, if they were not backed by the pronunciation, specifically a pronunciation along "modern" lines (cf. GELD70, p. 16: "In the Cratylus of Plato, the obviously false etymology of Δημήτηρ from δίδωμι and μήτηρ, derives all its little plausibility from the resemblance between δη- and δι-"). It is difficult to explain why Φ is classed with Σ, much less why it is called "πνευματῶδες". Blass has bypassed the problem by translating "πνευματώδη" as "letters with a strong breathing" (BLAS90, p. 105), which allowed him to simultaneously refer to the release of [pʰ] and the spirantic character of [s], while Allen makes a vague reference to the passage asserting that Plato "is here mainly concerned with the needs of his 'gestural' theory of the origin of language" (ALLE87, p. 23) and points out that the use of the term "πνευματώδη" for both aspirates and fricatives is paralleled by the similar practice of the indian Grammarians. With respect to the latter, it has already been pointed out that the like treatment of both aspirates and fricatives by the indian Grammarians is rather a reason to seek a divergent (with respect to their modern values) pronunciation of the indian aspirates in ancient times. The former is a common catholic trick for distracting the reader's attention from the actual evidence; the purpose for which Plato has written the passage (whether for establishing a "gestural" theory or for proving the existence of extra-terrestrials) is irrelevant; the description applies irrespective of the purpose and would not be used as an argument by "Socrates" if it were blatantly wrong. Furthermore, it is not correct to isolate the term "πνευματώδη" and suggest that it refers to "letters with a strong breathing"; all it takes is to read the last part of the sentence, where they are all described as suitable to imitate the "φυσῶδες",In that respect, Dawes has already commented on Blass' thesis: "There can, however, be no doubt that Plato did not intend to convey any such notion by his word πνευματώδη, if the passage is read in connection with the context, and especially with the next sentence, which says 'these letters are used to imitate what is windy;' hence πνευματώδη must mean, as Jowett says, that they require a great deal of breath. Placed in juxtaposition with simple and compound sibilants as it is here, and then further being designated as one of the letters which is used to imitate what is 'windy,' certainly seems as if it would answer better to this description if =f than if it were =p+h, and it would also more fitly be ranked with sibilants" (DAWE95, p. 50). which would not make sense if Φ were primarily plosive.The Catholics may object that Ψ is also partly plosive, but this rests on the common prejudice that Ψ=[ps], which is not necessarily true for all dialects of Greek. The value that "Socrates" had in mind might have been the original ionic (possibly a single sibilant) or his contemporary attic equivalent (possibly [fs]=ΦΣ). This passage from Cratylus is, therefore, more than just about a mere classing-together of Φ, Ψ, Σ and Ζ; it incontrovertibly describes all of them as fricatives. The only viable defence line for the Catholics is to declare this work of Plato's a fake or at least that passage a "byzantine" addition or alternatively to claim that Greek entirely lacked fricatives (by assigning an affricate value [t͡s] to Σ) and could only use affricates and aspirates to convey the notion of a stream of air ("φυσῶδες").

CHTHonic (di)PHTHongs

Another case of internal evidence relating to the perennial values of the δασέα in Greek (i.e., since Greek has been documented) is the phonetic realisation of the clusters φθ, χθ. These consonant sequences are very common in Greek, not only as results of (the aforementioned) assimilation, but also in the native vocabulary, often as word-initial, e.g., ἰχθύς, φθείρω. The difficulty of phonetically rendering χθ, φθ under the catholic model (ΘC, ΦC, ΧC) has led some scholars (as described in BLAS90, p. 105; DAWE95, p. 27) to theorise that the orthography was at variance with the actual pronunciation (which was, allegedly, the easier to pronounce πθ, κθ). The simple observation that the assumed "actual pronunciation" would have been reflected at least in a number of "misspellings" (i.e., πθ, κθ)Dawes has a nice way of expressing this: "if φθ, χθ were practically pronounced as πθ, κθ, we ought to have, not only half-a-dozen, but many dozen instances of such miswritings. It appears somewhat incredible that every stonemason should have a sufficiently critical eye to discern that it would look incorrect, or that he should remember his schooling so well as never to forget that it would be inconsistent with the law of assimilation to write a tenuis before an aspirate, and therefore always carefully wrote two aspirates, even though he pronounced the first of the two exactly, or very nearly, as a tenuis" (DAWE95, p. 28). is enough to invalidate these theories. The Catholics, therefore, consider that the consonant clusters χθ, φθ corresponded to a sequence of two aspirates (which they render with their latin spelling as phth, khth, even though they actually mean [pʰtʰ], [kʰtʰ]).Allen declares (ALLE87, p. 27) that a similar pronunciation (i.e., [kʰpʰ]) should apply for "the additional combination χφ in e.g. εχ φυλεσ and compound εχφο[ρησαντι]" caused by the "assimilation" ἐκ→ἐχ. This conclusion is obviously based on the typical dogma that spelling always corresponds to pronunciation. But, if that were so, the conclusion should also be extended to the combination of the same "assimilated" form ἐχ with a word-initial χ, such as ἐχ Χαλκίδος, as well as to the "geminated" φφ, θθ, χχ of Σαφφώ, κάθθανε, Βάχχος (DAWE95, p. 16), which, however he explains otherwise (n. 33).

However, starting from this clarification, some Orthodox have declared the catholic pronunciation of χθ and φθ impossible. Thus: "Combinations like φθόνος, συμφθείρω, χθών, Σαπφώ, Σαφφώ, Ἀτθίς, Ἀθθίς, ἤρχθην, ἠλέγχθην, ἐκάμφθην, ἐχθρός, that is, p-h+t-h-όνος, συμp-h+ t-hείρω, k-h+t-hών, Σαp+p-hώ, Σαp+h-p-hώ, Ἀτt+hίς, Ἀt+h-t+hίς, ἤρk-h+t-hην, ἠλέγk-h-t-hην, etc. are not only incompatible with Greek phonology [...], but constitute a physiological impossibility in any actual language" (JANN97, p. 58), blatantly copied (but not acknowledged as such) in CARA95, n. 68. This view is lambasted by Allen (ALLE87, p. 27) as "a priory dogma". Make a note of the expression, as it will be a recurrent theme in the religious-style dispute we are examining!

Specifically, Allen states that this view "has no basis whatever in reality", because "one can hear them as a normal feature of a number of living languages-as e.g. Armenian alotʻkʻ [ayothkh] 'prayer', or Georgian pʻkʻvili 'flour', tʻitʻkʻmis 'almost', or Abaza (N.W. Caucasian) apʻqʻa 'in front'". He then mumbles something about "a rule in Georgian", which is reminiscent of the assimilations π-θ→φθ, κ-θ→χθ (in ἐλήφθην, ἐδέρχθην) of Greek, yet not fully, since assimilation is imperative only if the second consonant is "located further back in the mouth" (thus, corresponding to φθ, but not to χθ). He finally submits that modern indian languages feature "sequences of aspirated followed by unaspirated plosive [...], e.g. in Hindi participial forms such as likhtā 'writing', ūbhtā 'rising'". He concludes that "There is thus no phonetic improbability whatever about the first consonant of the groups φθ and χθ being aspirated as well as the second" (emphasis in the original; lack of it mine).

At first sight, Allen seems to have a point. However, he is guilty of misrepresentation, as he does not quote the passage of Jannaris in its entirety (see above), but he leaves out the parts where Jannaris analyses φθόνος as "p-h+t-h-όνος", etc, in accordance with the catholic narrative of his time, i.e., that φ=p+h, θ=t+h, χ=k+h (in the strict interpretation of ΦΘΧ1-ΦΘΧ3). Thus, while Jannaris argues against the pronounceability of [phth] and [khth], Allen is claiming the existence of [pʰtʰ] and [kʰtʰ] in "a number of living languages". This is a typical straw-man argument, as none of the examples put forward by Allen comprises the four-consonant sequences [phth] and [khth] declared as "a physiological impossibility" by Jannaris.

Irrespective of what Jannaris claimed, it is worth examining Allen's arguments, for it seems that he is not so successful in attacking the straw man either. To begin with, none of his examples comprises the sought sequences [pʰtʰ] and [kʰtʰ]! However, these examples may serve to demonstrate that sequences of aspirated consonants are not "impossible" (a thesis that he, strangely, does not take on, but opts to argue against their "improbability").

The Hindi participles are interesting, as they illustrate the possibility of pronouncing an affricated plosive before another plosive. However, I suspect that the two plosives in these cases are on different sides of the syllable boundary, which certainly makes the pronunciation easier, whereas in Greek φθ, χθ belong to the same syllable (more on this later), as can be seen by the fact that they often introduce words (χθών, φθόνος, etc); I am not aware of any indian words that may start with some similar combination of consonants.

Georgian is a more promising language in that respect, as its vocabulary comprises many words beginning with an abnormal (for other languages) number of consonants, such as "/prckvna/ 'to peel', /mc’vrtneli/ 'trainer', /brt’χ’eli/ 'flat'", yet "native speakers have no difficulty in pronouncing words like the above, with all the consonants" (BUTS02, p. 1). Curious to hear how a cluster of aspirates, such as the "pʻkʻvili 'flour'" (i.e., ფქვილი) mentioned by Allen, sounds, I asked two native Georgian speakers what the georgian word for "flour" is; I was surprised when I did not hear the faintest trace of "aspiration" and, after having them repeat it 3-4 times, they offered to transcribe it for me in Greek: ΠΚΟΥΙΛΙ; that is [pkwili] or rather [pkʷili], as they pronounced it. I was intrigued by this "discovery" and decided to investigate the matter further. It turns out that it is not at all certain that the georgian "aspirates" are aspirated consonants. Most likely, Allen's illusion that some georgian plosives are "aspirated" is due to the latin transcription (pʻ, tʻ, kʻ) that he uses and possibly due to the fact that "Some phonetic studies suggest that the voiceless consonants are aspirated" (BUTS02, p. 85), i.e., may be based on a fancy theory that even native scholars cannot verify!Even when the normal voiceless plosives are analysed as "aspirated", clusters of such stops ("harmonic clusters") are reported to "consist of a single release: [d˺g, t˺kʰ, t˺kʼ]" (KEHR02, p. 156), which means that no independent "aspirated" release of each consonant takes place (Kehrein argues that there is independent release, but this is based on a quote from a work several decades prior to the ones that reach the above conclusion). If that were the case in ancient Greek, i.e., if there were a single release, then the spellings πθ, κθ would better represent such a single release, whereas Georgian has no other way of indicating a voiceless unaspirated stop, as the other two available homorganic stops are (allegedly) "ejective" and voiced, respectively. Georgians themselves transliterate their "aspirates" ფ, თ, ქ as P, T, K in latin characters (e.g., in road signs) and do not distinguish them therefrom (BUTS02, p. 77). The notation pʻ, tʻ, kʻ is reserved for the so-called "ejectives", which are distinguished from the normal plosives p, t, k, but this is a completely different matter (they are "glottalised"). Furthermore, there appears to be no connection between the "aspirates" and the phoneme /h/, which "is restricted to word-initial position" and "never occurs in consonant sequences" (BUTS02, p. 87).It may be that [h], the voiceless glottal fricative of English and German, never actually existed as a native sound in Georgian. It is submitted (BUTS02, p. 87) that "it tends to delete in Modern Georgian" and that "Most of the words beginning with the consonant /h/ are loan words, e.g. from Greek: /hipoteza/ 'hypothesis', /harmonia/ 'harmony', etc". As we do not know how the character ჰ, which corresponds to the "phoneme" /h/, sounded in Old Georgian or what it stood for and since the foreign (greek, etc) words that do have an /h/ are clearly learned words with a spelling pronunciation (based on the pronunciation of h in western Europe), the existence of a native sound [h] in Georgian cannot be established with certainty (cf. also KEHR02, p. 159, n. 96: "[h] neither occurs here nor in any other cluster of the language"). So much for Georgian "aspirates" and "aspirated clusters".

I have not been able to verify the other "evidence" produced by Allen. Armenian is a more popular example for arguing the possibility of clustering aspirates together (e.g., STUR20, p. 182; DAWE95, p. 28). Yet, none of the references to Armenian appears to be based on first-hand experience, but rather on hearsay or on the works of others and are often misunderstandings caused by faulty notation; as the above example of Georgian teaches, sometimes scholars (even native ones) report features that may not be there, while notation may be misleading (e.g., ʻ for Allen denotes "aspiration", but it is usually applied to "ejectives" in Georgian). It is not easy to find a native speaker of Armenian, in order to listen to the alleged clusters of "aspirates". Based on the experience of Georgian, I was skeptical about the characterisation of the consonants transcribed pʻ, tʻ, kʻ or p’, t’, k’ as "aspirated" for two reasons: a) they correspond (e.g., alphabetically) to georgian "aspirates", which are not exactly "aspirated" and b) the division of the armenian plosives into voiced, voiceless and aspirates may be a mere technicality, called for out of the need to distinguish between three kinds of armenian stops, "aspirated" being merely a convenient, but not necessarily accurate term.The Voice-Onset Time theory is a tempting solution to the problem of aspirated clusters. If I understand it correctly, it advocates three types of plosives, namely voiced - tenues (voiceless unaspirated) - aspirated, according to the relative position of the onset of voicing and the release of the plosive (before - simultaneously - after, respectively). In the case of the last two types, which are unvoiced, the voicing refers to that of a following sonorant. Under this theory, in the case of words such as ε-λειπ-θην→ἐλήφθην, if θ is "aspirated", the voicing of η would start (more than a perceptible threshold) after θ's release. Then the temporal distance between the release of the previous plosive π and the beginning of voicing (always of η) would be even greater (than that between the release of θ and the voicing of η) and π would have to count as "aspirated", i.e., φ. However, the application of this theory would not explain why the first of a cluster of plosives (e.g., κτείνω, πτύω) remains a tenuis despite its release being separated from the onset of voicing by a greater distance than for a simple aspirated plosive (e.g., φύω), much less why in voiceless-mute geminates the first element is always a tenuis (e.g., ἵππος, Σαπφώ) and not an aspirated plosive. Furthermore, this theory would place the tenues "between" the aspirated and the voiced plosives and would not explain why the latter (instead of the former) were perceived by the Grammarians as "mediae". Nevertheless, one can find some recorded Armenian samples (albeit of awful quality) and it is quite evident (even to my untrained ear) that there is some difference between normal voiceless and aspirated armenian stops, which involves some degree of aspiration similar to that of word-initial english voiceless stops. For the question at hand, namely how aspirated clusters sound, there is but one example: պէտք (bɛtʰkʰ). I was not able to discern how the pronounciation of the cluster /tʰkʰ/ in this particular recording differs from my own pronunciation of /tk/, e.g., in "Atkins" (i.e., [tk]). Unfortunately, I could not locate any word ending in the corresponding unaspirated cluster /tk/, so that the two be compared. Most importantly, it is not at all certain whether the armenian aspirated clusters are comparable to the ancient-greek ones, as the former appear to be exclusively in coda position, whereas the latter are only found in onsets (cf. KEHR02, p. 87); it is not as easy to pronounce a cluster of affricates when a (inherently voiced) vowel follows in the same prosodic unit, as (I guess) it would be in pausa. The only cited armenian example appearing to involve an aspirated cluster in an onset is "tʻkʻanem" submitted by Sturtevant (STUR20, p. 182); however, the actual (modern) pronunciation of the related թքել (tʿkʿel) is [tʰəˈkʰɛl], i.e., a "schwa" appears to be necessary in order to ensure the pronunciation of both as aspirates; in other words, the only attested pronunciation is not one of an aspirated cluster (the assumed old-armenian pronunciation is only a conjecture, possibly based on the spelling).

Native speakers of Abaza are even harder to find. In this case, it appears that Allen did have first-hand experience with the language, albeit only "a hundred hours work with a single expatriate informant" (LYON06, p. 14). Nevertheless, it is not clarified whether the "aspirated cluster" in the single example provided ("apʻqʻa") is tautosyllabic (as in Greek) or heterosyllabic (the syllable boundary providing the necessary pause to pronounce the two "aspirated" consonants individually).

Whatever the case may be, it is true that humans can be trained to pronounce practically any weird sound and any complex combination of sounds. Furthermore, the phonological features of the world's languages are so rich that one can probably find examples to justify any exotic theory. So, strictly speaking, clusters of aspirated stops (or even the clusters [phth] and [khth] objected to by Jannaris) are not "impossible" to pronounce. Nevertheless, the existence of a certain sound, feature or pattern in one language does not render them equally probable for another language. For instance, the Khoisan family makes use of clicks as consonants, but their existence in an Indo-European language, like Latin or ancient Greek, is rather improbable. Similarly, even if the existence of aspirate clusters can be established in some Caucasian languages (South for Georgian, N.W. for Abaza; Armenian probably being the most "Caucasian" of the Indo-European languages), this does not speak against its "phonetic improbability" (not "impossibility") in a language like ancient Greek, given its past (no contact with Caucasian, no verifiable Caucasian substrate, no similar feature in any of the related languages that did not have contact with Caucasian) and its present (i.e., the phonology of the surviving Greek dialects).

In view of the above, as well as of Kehrein's admission that "Ancient Greek is the only case of 'aspirate assimilation' I know of" (KEHR02,p. 180), we can safely state that

(ΦΘΧ15) Tautosyllabic [pʰtʰ] and [kʰtʰ] are not reported for any living language

On the other hand, the orthodox pronunciation of φθ, χθ as "fricative" clusters is certainly easier to pronounceThis may not be evident for word-initial clusters to an Englishman, who pronounces, e.g., "chthonic" and "phthisis" with "silent" ch and ph. However, Englishmen have certainly an... issue with word initial clusters, like "psychology", "xenon", "tsar", "pterodactyl", "ctenoid", "mnemonic", "pneumatic", even native words like "knee", the only "pronounceable" word-initial clusters comprising a sibilant and/or a liquid (tr, pl, fr, spr, sl, etc). and consistent with the past and present circumstances of the language.

The Writing on the Wall

If one is to deduce the pronunciation of a certain symbol at a certain time period, the contemporary evidence of that period should bear more weight. We have seen that the only direct evidence of the ancient times concerns the surviving inscriptions. Nevertheless, in the particular case of Χ, Φ, Θ, the inscriptional evidence is of little help. The reason has already been explained: variations of two (or more) graphemes in inscriptions is caused by uncertainty about which of the phonetically equivalent graphemes to use. Two (or more) graphemes are phonetically equivalent when they represent the exact same sound. In the history of Greek, Φ, Θ, Χ never coincided phonetically with other letters.As evident from Tsakonian, the phoneme represented by mainstream-Greek Θ ended up having the value [s] in many instances. However, the corresponding words are always spelled with Σ, not Θ and, although the phenomenon is probably an ancient one, there is no evidence (e.g., Θ where Σ would be expected) that Θ was ever used to denote the value of Σ in Laconian or its offshoot. There is also the case of diphthongs ΑΥ and ΕΥ, the second element of which ended up having the value of Φ before voiceless consonants. However, this development appears to be post-christian, i.e., outside our main period of interest. It, therefore, seems pointless to seek inscriptional evidence that would establish the phonology of the aspiratae, as there is no phonemic merger to confirm.

This has not deterred Teodorsson from providing inscriptional "evidence" that the aspiratae had not assumed their orthodox values in Attica of IV BC. He focuses on variations of the aspiratae Φ, Θ, Χ with their corresponding voiceless stops Π, Τ, Κ, as well as with letters that he thinks would make better candidates (e.g., Σ, old-Attic Η, also the voiced counterparts Β, Δ, Γ), if the aspiratae were fricatives. Based on the fact that the former are more numerous than the latter, he concludes that the aspiratae were more stops than fricatives (TEOD74, pp. 220-225)! Here lies Teodorsson's biggest mistake (he appears to be intending to appease the Catholics by granting them the consonants, after having denied them the vowels). He claims phonetic similarity as the reason behind the alternative spellings, but this is rarely the case. We have identified three possible causes of orthographic "variations": A) Pure Chance, B) Alternative Pronunciation, C) Genuine Confusion.

A prerequisite for the latter is phonetic identity; however, the phonemic independence of the aspiratae is established for the entire (written) history of Greek, so that no Greek (modern or ancient) would confuse Φ, Θ, Χ with some other letter due to phonetic reasons.There is one source of confusion, namely the development of ΑΥ, ΕΥ to [af], [ef] before voiceless consonants (resulting in the phonetic merger of, e.g., ευφορία and εφορία). In fact, Teodorsson does provide a couple of instances of confusion of ΕΥΦ and ΕΦ, but he judges them to be too few to substantiate a fricative pronunciation of Φ (p. 220). Nevertheless, the dearth of such evidence may at most suggest that the Υ of ΑΥ, ΕΥ and Φ had not coincided yet, but not what their values were. Thus, if Φ were always a fricative, the lack of (numerous) confusions may merely mean that the pronunciation of ΑΥ, ΕΥ as [af], [ef] (from an almost certain original [aw], [ew]) had not been widespread yet. Only in some special cases may some instances be due to (conditional) phonetic developments. For one, Teodorsson proposes the surprising development "[tʰ] > [t] /[s]_" for Θ, based on 6 instances of ΣΤ for normative ΣΘ. This conclusion is surprising, because it would correspond to an aberrant development of Θ's phoneme. These specific instances are rather better explained under the orthodox model: if Θ was (always) a fricative, the emergence of Τ in lieu of Θ after Σ (corresponding to the change [θ] > [t] /[s]_) is probably the first manifestation of dissimilation, a frequent phenomenon in modern Greek, wherein a sequence of two fricatives is avoided by turning the second one into the corresponding stop. There are some further instances of suspected dissimilation, such as ΕΥΤΥΜΙΔΕΣ for Εὐθυμίδης (p. 133), wherein the first Υ (of the diphthong ΕΥ) is a continuant consonant (irrespective of whether it was the suspected original [w] or the modern [f] or the assumed intermediate [v]) and could have induced the transformation of the fricative Θ τo the corresponding stop.

For the majority of the cases, though, one cannot resort to phonetic identity for explaining the variations, because there was never any identity of sound for the aspiratae. We, therefore, need to seek other (non-phonetic) factors that might have caused the misspellings or, if we consider phonetic similarity as the reason, provide a plausible explanation why two similar, yet distinct phonemes would be confused. One alternative explanation is provided by the former cause (pure chance); however, in order to evaluate this possibility properly it is necessary to look at information about the context of the misspellings, its number as compared to that of patently non-phonetic misspellings (e.g., Φ↔Λ), etc; unfortunately, such information is not available, as it is filtered out by Teodorsson and his likes;In an exceptional case, Teodorsson presents three instances of using Τ for normative Σ, which he cannot explain phonetically (pp. 222-223). It is evident that it is more reasonable to ascribe these cases to pure chance, rather than to phonetic factors, which would require a series of unverifiable assumptions. thus, the influence of random factors cannot be assessed (but still remains a possibility). A non-phonetic factor that might contribute to random misspellings is graphemic similarity. Several instances have been attributed to such similarities (e.g., on p. 133, Teodorsson assumes ΜΙΣΤΟΟΔΙΚΟΣ to stand for ΜΙΣΤΘΟΔΙΚΟΣ, Θ's dot or dash having been left out). It is, therefore, possible that some of the Χ↔Κ variations are due to imperfect or careless rendering of the corresponding graphemes.

The second possible cause (alternative pronunciation) is plausible for phonetically similar phonemes only under certain circumstances. One is false etymology, where a word is pronounced with some sounds slightly modified, because it is (wrongly) assumed to derive from a word comprising those sounds. We have seen Blass resort to false etymology to justify the variation ἐντελέχεια↔ἐνδελεχής. However, it is wrong to conclude that alternation of Φ, Θ, Χ with Π, Τ, Κ due to false etymology is an indication of a stop quality in the former. Take for example the word πολυθρόνα, which is considered to derive from poltrona|armchair[ITA] under the influence of its association with πολύ|very and θρόνος|throne; the Θ in the greek word has emerged from an original /t/; as the derivation is post-medieval, there is no doubt that the value of Θ was a fricative ([θ]). It is, therefore, possible to have alternations of fricative and stop values as a result of false etymology.

Other than the above, the only case when two phonetically similar, yet not identical phonemes might be confused is by a non-native speaker. Ruijgh (among numerous other Catholics) is playing the "alien" card to explain the orthographic variations studied by Teodorsson: specifically, he submits that a non-native speaker would project the vocalic system of Attic onto the one of his own language, thus potentially confusing "ε/ι/υ, η/ει/ι/υ, ω/ου" if the vowels of his own language are "/a i u ā ī ū/"; similarly, he claims that the substitution of Φ, Θ, Χ by Π, Τ, Κ is due to the foreigner's language lacking aspirated stops (RUIJ78, p. 84). Here, there is a need to distinguish between the two types of misspellings. As far as the interchanges of vowels are concerned, mispronunciation by foreigners is not the only possible (phonetic) explanation; since the vowels concerned have historically coincided, (early) phonetic identity is also a plausible explanation (actually, more likely, given their numerous instances). The consonants, on the other hand, have never coincided, so that the only plausible, phonetically relevant reason can be confusion by foreigners. Thus, if we are to associate the relatively (to the vowels) less frequent interchanges of the aspiratae with other consonants to phonology, then we have to assume that they are due to some foreigner substituting the sounds of Φ, Θ, Χ, which did not exist in his native tongue, by some more familiar sound; the reverse would be a case of hypercorrection, while confusion between sounds that were both absent from his native phonology (like Φ and Θ) would be due to phonetic uncertainty (caused by unfamiliarity). However, it is not certain which familiar sound a non-native speaker would select as the "closest" approximation of Φ, Θ, Χ. As a matter of fact, there is no way of knowing the nationality and, hence, the native phonology of the person committing those errors, much less determine which of the native sounds would seem to him as "closer". It is, therefore, evident that the substitution of Τ for Θ and vice versa does not suggest anything particular for the value of Θ, as [t] would be a fairly acceptable pronunciation for either [tʰ] or [θ].One argument put forward by the Catholics is that, if Θ were the fricative it is today, it would be confused with Σ rather than with Τ. This argument is evidently of germanic origin (used also by Teodorsson), since [s] is their usual realisation of English th(=/θ/). However, the german practice is not a law and not everyone thinks like a German. In that respect, I have observed the pronunciation of an Argentinian colleague and was surprised to notice that he was pronouncing the [θ] of English as [t], e.g., [tæŋk ju] for "thank you". In other words, despite his native phonology comprising the fricative [s] and despite himself (like all Argentinians) observing the seseo (i.e., the realisation of general-Spanish /θ/ as [s]), he felt it more proper to approximate the English fricative [θ] by the stop [t] rather than the fricative [s] (advertised as "closer" by the Catholics). It is, therefore, incorrect to conclude that the interchange of the tenues and the aspiratae confirms the catholic model.

Thus, as far as Teodorsson's cases of tenuis↔aspirata are concerned, some may be results of genuine conditional phonetic change of a general fricative pronunciation (e.g., [θ] > [t] /[s]_ in ΚΛΕΣΤΕΝΟΣ for Κλεισθένους; also ΕΥΤΥΦΡΩΝ for Εὐθύφρων), some are consequences of obvious psilosis, i.e., omission/suppression of a δασεῖα that cannot act on a previous tenuis (e.g., ΚΑΤΕΜΕΡΑΝ for καθ'ἡμέραν←κατὰ ἡμέραν), some due to a change of attitude towards gemination (e.g., ΒΑΧΧΕ for Βάκχε, which incidentally is incompatible with the catholic model ΧΧ=[kʰkʰ]), some are potential graphemic confusions (e.g., ΧΑΛΟΣ for καλός) and some have to be products of random mistakes. Most importantly, even under the assumption that the reason is phonetic similarity, it is incorrect to assume that the pure stops would be perceived as approximations of aspirated stops but not of the associated fricatives; it is reasonable for phonemes of the same natural class to be interchanged, particularly if the class relates to the same articulator, and this is true for any (voiceless) variation of a voiceless unaspirated stop. As a consequence, the relatively few instances cited by Teodorsson can be due to a number of factors, none of which necessitates the catholic values of the aspiratae.

On the other hand, there is an interesting "variation" that is not easily explainable under the catholic model. It concerns the interchange of Φ and Θ in some old (attic and non-attic) inscriptions, probably first noticed and presented by Arena, who claims that they "nur unter Voraussetzung einer spirantischen Aussprache der betreffenden Laute ϑ, φ erläutert werden kann|can only be explained under the assumption of a spirantic pronunciation of the concerned sounds ϑ, φ" (AREN66, p. 14). Langdon attempted to counter the evidence "by rejecting two of the inscriptions cited by Arena, which have been misread, and by giving a simpler explanation for the use of phi for theta in the others" (LANG79, p. 180). Nevertheless, his main contribution lies with the first part, as he does provide clear photographs of an inscription that has been wrongly reported as having Φ for Θ and submits the same for a second inscription, but concedes that "The majority [=all but two] of the inscriptions discussed by Arena, on the other hand, clearly do contain phi where theta is expected, or vice versa" (p. 181); in other words, the rejection of two readings is a diversion that does not change the fact that Φ and Θ are sometimes confused. As regards the explaining away, Langdon ascribes it to "the one [explanation] most often given, namely spelling mistakes by the inscriber", i.a. due to "the natural tendency to mix up two letters that are similar in appearance" (p. 181); naturally, Langdon does not bother explain/investigate whether mixup of other similar graphemes provides comparably numerous instances, but is content with such an assertion. It is already evident from the title of the article that Langdon is on the wrong track: φ≠Θ is a superfluous inequation, as neither Arena nor anyone else can claim that the two phonemes ever were phonetically identical. Langdon also points to "the fact that in Greek inscriptions tau and theta are frequently confused, which makes sense if theta were an aspirate but not if it were a spirant" (p. 180). The arbitrariness of this conclusion and the plausibility of Τ in lieu of a fricative Θ have been demonstrated above: if Τ↔Θ is due to phonetics, Τ=[t] is a good approximation equally for Θ=[tʰ] and Θ=[θ]. On the contrary, if the reason behind the variation Φ↔Θ is phonetic, then it is more unusual for stops of different articulators (e.g., Φ=[pʰ] and Θ=[tʰ]) to be confused and one is compelled to assume not only fricative pronunciations of Φ and Θ, but specifically labiodental (Φ=[f]) and interdental (Θ=[θ]) articulations, which are sufficiently close to cause confusion in several contexts, from babyspeak (e.g., my son's [fanasis] for Θανάσης) to Russian (cf. Феодор for Θεόδωρος). And if the inscriptions in question are works of non-native speakers, as e.g. posited (cf. AREN66, p. 15) for IG XII,5 97 (on which Langdon writes an irrelevant and uninteresting "Note"), then the phonetic explanation is the most promising (cf. also AREN66, p. 17).

Teodorsson discards the entire corpus cited by Arena, claiming that "the graphemes <Φ> and <Θ> each had a number of allographs" and specifically that they were allographs of each other (TEOD74, p. 222, n. 181). To back this claim, he cites a few instances where the two letters appear in different forms. It is not clear how we can be sure that these instances (as far as the graphemic interchange Φ↔Θ is concerned, possibly including the "cartwheel" shape of Θ) are allographs (i.e., established variants of the normative forms) or genuine interchanges of Φ and Θ; in fact, no overlap is reported by Jeffery among the various glyph forms of Θ and Φ, so that Teodorsson's claim is not justified. It is, therefore, unfair to exclude the Φ↔Θ variations based on the claim that they are due to non-phonetic factors, but include all the other variations of tenuis↔aspirata despite the patent possibility of non-phonetic causes. All instances should have been included and, if a phonetic conclusion is sought, it should be explained that the orthodox model accounts for both kinds of variations, whereas the catholic one provides no satisfactory phonetic explanation for the interchanges Φ↔Θ. Having said that, it should be stressed that it is wrong to consider this (as well as the other evidence) a proof of the orthodox pronunciation; but it is equally wrong to consider the alternative (non-phonetic) explanations a certainty; thus, to the extent that a phonetic factor is to blame, odds are with the Orthodox rather than with the Catholics. In accordance with the Aggregation Principle, the inscriptional evidence tips the scale towards Orthodoxy (as far as the pronunciation of the aspiratae is concerned).

There could be only one way to verify the catholic thesis that the δασέα were composite sounds comprising a stop followed by "aspiration", namely that Φ=Πh, Θ=Τh, Χ=Κh: if the composition was so evident that it led the Romans to use the spellings ph, th, ch and the Therans and Meleans to resort to ΠΗ, ΚΗ, a spelling comprising their two individual parts should occasionally pop up in lieu of the established (allegedly shortcut) graphemes Φ, Θ, Χ. This would not be possible in the Ionic alphabet, which had no symbol for the "aspiration", but it is a valid expectation for the alphabets that had retained the original function of Η as the δασεῖα, such as the old-attic alphabet. Similarly to Ζ, which is also a composite sound in the catholic universe, but never encountered as a composite grapheme, Θ is never spelled ΤΗ, while ΠΗ and ΚΗ only appear in the alphabets of Thera and Melos, which lacked symbols for Φ and Θ, but never in alphabets that possessed such symbols.

(ΦΘΧ16) Θ never appears as ΤΗ; Φ, Χ and ΠΗ, ΚΗ never appear together

As observed by Dawes, "how is it possible to explain that never once in Attica before the time of Eukleides whilst Η still = h, nor in any other dialect, as for instance in Herculaneum where they had a special sign ˫ for 'h,' do we find on an inscription of any kind πh, κh, or τh written in elisions, erases and compound words for φ, χ or θ (excepting of course the cases of Thera and Melos)?" (DAWE95, p. 98). Juxtaposed to the superficial interpretation of the interchanges Τ↔Θ and Κ↔Χ by Teodorsson as evidence of the "preservation" of aspirated values for the δασέα in classical Attic, ΦΘΧ16 indicates that there is no positive indication of the catholic values in the ancient inscriptions, certainly not in the attic ones.

Mopping Up

So far, we have addressed all the major arguments put forward by the Catholics for "proving" that Φ, Θ, Χ were aspirated stops and explained why the submitted evidence is also compatible with the orthodox values. We have also presented further evidence indicative of fricative rather than aspirated values for Φ, Θ, Χ. Before proceeding to the conclusions, it is necessary to deal with a few peripheral arguments that bear less importance.

Ashes and Embers

A favourite hallucination of the Catholics is that "further confirmation of the plosive value of φ in classical times is perhaps provided [...] by the surely deliberate use of π and φ in Pindar's description of a volcano" (ALLE87, p. 22). We have already discussed the relevant passage, which reads "ἀλλ' ἐν ὄρφναισιν πέτρας φοίνισσα κυλινδομένα φλὸξ ἐς βαθεῖαν φέρει πόντου πλάκα σὺν πατάγῳ|but in the darkness of night the crimson flame hurls rocks down to the deep plain of the sea with a crashing roar [Svarlien's translation]". What Allen means with the statement that the use of π and φ was deliberate is not clear, but he probably implies that, since a volcano is explosive by nature, so must be both π and φ. Such a conclusion is (per se) arbitrary: why does φ have to be a plosive sound denoting the volcano's explosive character? if Pindar indeed intended to give the impression of an erupting volcano by means of phonetics, was π not enough? after all, ν, λ and (especially) σ are each repeated in the passage more often than φ, without being (or anyone claiming that they are) plosive. What is usually claimed is that an alliterative effect is produced because most words of the passage begin with a labial (bringing in also β of "βαθεῖαν") and that the alliteration is indicative of the volcanic activity only if all labials are plosive.

Indeed, if we pronounce π and φ alike (which a catholic Englishman, who always aspirates his p's, would certainly do), there is an apparent alliteration; however, such an alliterative effect would rely on our (or the Englishman's) inability to distinguish [p] and [pʰ], whereas it is evident that φ (as well as β) was never identical to π, but it was sufficiently different therefrom to constitute a different phoneme. Since the words in question begin with different, sufficiently distinct phonemes, there is no question of a single alliteration spanning the entire passage. It is evident that, if alliteration was indeed intentional, there are two alliterations: one involving the use of π mainly in the last part of the sentence ("πέτρας ... πόντου πλάκα σὺν πατάγῳ") and one involving the use of φ mainly in the first part of the sentence ("ὄρφναισιν ... φοίνισσα κυλινδομένα φλὸξ"). It does not follow that the two alliterations are related, much less that they both relate to the explosive nature of the volcano. Only the one involving π can be considered to have a "bursty" character, as it refers to hurled stones ("πέτρας"). The one involving φ relates to the crimson rolling flame ("φοίνισσα κυλινδομένα φλὸξ") and it is patent that the orthodox value (Φ=[f]) is reasonable (and probably makes more sense) for describing the flowing flame!The other "further confirmation of the plosive value of φ" submitted by Allen, namely "the presumably onomatopoeic πομφόλυξ, πομφολύζειν, for the sound of bubbling" is not serious. It again relies on the pronunciation of φ as π and on the forced association with the English stem bubbl-; if the intension were to denote a plosive quality, the mere use of π would do, whereas the use of an aspirated plosive (as φ is considered by the Catholics to have been) adds exactly nothing in that respect. Furthermore, Allen's translation is misleading: πομφόλυξ is the bubble itself, not any sound associated with it (specifically, "πομφόλυγες are the constituent parts of ἀφρός"; incidentally, ἀφρός is foam); the associated verb πομφολύζειν has the meaning "bubble or boil up" and it is incorrect (if not dishonest) for Allen to allude to the sound of bubbling, whereas the verb merely refers to the creation of bubbles (πομφόλυγες), not to any bursting thereof. The implausibility of "πομφόλυξ" having any plosive etymology is evident from its synonyms "φυσαλίδα", "φούσκα", "φουσκάλα", all of which can be traced back to the already discussed verb "φυσάω, -ῶ".


A "scientific" verification of the catholic values of the aspiratae is often seen in the way ν behaves before them. As explained by Allen, "When in Attic the nasal ν was followed by the fricative σ, the nasal was generally lost or assimilated to the fricative - thus e.g. συν + σιτεῖν → συσσιτεῖν, συν + στέλλειν → συστέλειν. [...] This, however, does not occur before φ, θ, χ, but the ν is either retained or changed in type [...] in the same way as before an unaspirated plosive. [...] This treatment contrasts with that of modern Greek, where before the now fricative φ, θ, χ a final ν is lost in the same way as before σ and other continuants" (ALLE87, pp. 21-22). This is a tricky argument, because it may sound reasonable at first sight (which is as far as most Catholics go), but upon close inspection it boils down to comparing two different things.

The modern phenomenon is indeed one of loss of word-final ν when followed by a continuant. However, the ancient one relates not to loss, but to assimilation: ν assimilates to the following sibilant (ν+σ→σσ), liquid (ν+ρ→ρρ, ν+λ→λλ) or nasal (ν+μ→μμ); the loss of the assimilated consonant in συν + στέλλειν → συστέλειν (and before all words starting with σ+consonant) is a consequence of the fact that geminate consonants stand their ground in isolation, but not in combination with other consonants (i.e., σστ→στ). In general, the ancient "phonetic law" can be defined as an assimilation of final ν to the class of the following consonant, i.e., "sibilant", "liquid", "labial", "dental", "velar"; for the latter three, it was converted to the corresponding (labial, dental or velar) nasal; for the former two, it was converted to the corresponding consonant, as there is no "sibilant" or "liquid" nasal. The two phenomena (the ancient and the modern) relate to different treatments of the word-final ν and, being unrelated, they cannot be used to draw any conclusions about the phonology of the aspiratae.

There is a second fallacy in Allen's argument, which relates to the ancient phenomenon alone. He observes that ν is assimilated before σ (as explained, it is not correct to say that it is "lost") but is "either retained or changed in type" before φ, θ, χ, as it does before κ, π, τ. Again, ν's behaviour in the latter case is misrepresented, as the "retention" relates to ν remaining [n] before θ, τ and the "change in type" relates to it becoming [m] before φ, π and [ŋ] before χ, κ; in other words, it still assimilates and the only difference is that it is no longer a nasal before σ (or ρ, λ), which is reasonable, since there are no sibilant (or liquid) nasals in Greek.

Even if Allen's description of the phenomenon were accurate, it would not be awkward for labial, dental and velar fricatives (as φ, θ, χ are under the orthodox model) to class with the other labials, dentals or velars respectively, rather than with all other fricatives in (a certain rule of) a language's phonotactics. For instance, σ is voiced before μ in modern Greek (κόσμος=[kozmos]), but χ and θ are not (αιχμή=[exmi] not [eɣmi], σταθμός=[staθmos] not [staðmos]); however, this does not make them less "fricative" than σ.

What Allen also fails to realise is that the modern language has not been kind to pre-consonantal nasals in general, i.e., irrespective of the consonant's nature. Before the continuants (β, γ, δ, ζ?, θ, λ, μ, ν, ρ, σ, φ, χ) nasals usually delete, i.e., they are not pronounced. Before the stops κ, π, τ (also ξ, ψ), the nasal has induced voice to the following plosive, turning it to [g/ɟ], [b], [d], respectively; at the same time, the corresponding clusters γκ, μπ, ντ attained phonemic status and are (almost) indifferently realised as [ŋg/ɲɟ], [mb], [nd] or [g/ɟ], [b], [d], respectively, the reason being either the orthography's inability to distinguish between the two variants (the digraphs γκ, μπ, ντ being used to graphically render foreign [g], [b], [d]) or new phonotactics. Whatever the case, the fact is that most modern speakers use the pronunciation [g/ɟ], [b], [d] (without the nasal) as the rule (thus, Άγγελος|Angelos (given name)=[ˈAɟelos], κομπιούτερ|computer=[koˈbʝuter], ζόμπι|zombie=[ˈzobi] instead of the "correct" [ˈAɲɟelos], [komˈpçuter], [ˈzombi]) and the pronunciation [ŋg/ɲɟ], [mb], [nd] as the exception, often the result of hypercorrection (thus, χοτ ντογκ|hot dog may be pronounced [xot doŋg] by those not familiar with its English spelling). In other words, the pronunciation of the nasal before stops is entirely optional, as it has no phonemic significance, thus leading to the awkward spelling νγκ for [ŋg] (e.g., πάρκινγκ|parking, Ένγκελς|Engels), wherein the superfluous ν is used in order to force pronunciation of the nasal. It is, therefore, pointless to seek evidence in the modern treatment of pre-consonantal nasals, as their deletion concerns all consonants, whereas the ancient language at most assimilated the nasal before the ἡμίφωνα.

It would probably also be worth discussing Sturtevant's odd arguments for establishing the aspirated character of the δασέα based on the conversion of π, τ, κ to φ, θ, χ and vice versa. He claims (STUR20, p. 175) that inscriptional instances like hέχει, hέχον, Ηισθμοῖ, ἱχθύς, Ἀνθίλοχος, φαρθένος, Διοφείθης (his notation) are due to the fact that "A feature of Attic Greek was the rather extensive assimilation of aspiration". The first observation to be made is that, in most of the cited examples, this extensive feature of Attic violates Grassmann's law (ΦΘΧ6), which Sturtevant uses earlier (p. 174) as another "proof" that δασέα=plosive+h. Furthermore, his explanation that "only a true aspirate could induce initial aspiration, as in the first four examples" is assuming that the script faithfully represents the actual speech; however, these examples (as well as appearances of unetymological δασεῖα which is not followed by an aspirata in the next syllable; see TEOD74, pp. 144-145) are rather indications of uncertainty about the presence or not of a δασεῖα, i.e., that δασεῖα had lost its phonetic value, at least for some speakers. The other cases of π→φ etc are most likely mere misspellings, not necessarily results of assimilation (which, in any case, was not as "extensive" as advertised by Sturtevant). If there is assimilation involved, it concerned the replication of a property, not an individual sound (I am not aware of any case of assimilation of the latter kind; the only known assimilations, e.g., for modern Greek, involve a property like voice, place and manner of articulation); this property could be whatever characterised the δασέα from the ψιλά, be it stricture as in modern Greek, [spread] as considered by Kehrein, or any other feature. Such an "assimilation of aspiration" would confirm the catholic thesis that δασέα=ψιλά+[h] as much as modern φούχτα from (attic) πύκτα (evidently after the dissimilation κτ→χτ) indicates aspirated φ and χ in the late Middle Ages! The claim that there was a "metathesis of aspiration in Ionic κίθων" etc (an argument apparently introduced by Curtius), which allegedly confirmed the catholic values of the δασέα, had already been addressed by Papadimitrakopoulos by reference to modern βαθρακός from βάτραχος, etc (PAPA89, p. 623) where there can certainly be no transfer of [h].

Faulty Body Doubles

An anomaly that is often taken by the Catholics as proof that there is "a stop element" in the aspiratae is that geminated Φ, Θ, Χ appear as ΠΦ, ΤΘ, ΚΧ. Allen states that such a spelling "would not be appropriate if the original sound were a fricative but entirely so if it were a plosive: thus [ph, th, kh] → [pph, tth, kkh]". It is interesting how the representation of a single sound as a sequence of two distinct sounds is conveniently employed to mathematically justify the conclusion. Irrespective of that, a stop articulation with an aspirated release would be a plausible (yet not the only) explanation of this irregular gemination; the explanation becomes problematic once further phenomena are taken into account, particularly the already mentioned clusters ΦΘ, ΧΘ; why would ΠΘ, ΚΘ assimilate to ΦΘ, ΧΘ, i.e. develop an aspirated release between the two "stop elements", but ΦΦ, ΘΘ, ΧΧ lose the inherent aspirated release of the first element? It is often attempted to imply that the pronunciation of ΧΧ, ΘΘ (when Χ=[pʰ], etc) is more difficult than that of ΧΘ, because the former are homorganic clusters, whereas the latter are not and are easier to pronounce (e.g., BLAS90, p. 105). This argument may seem convincing when these clusters are looked at in isolation, but if one looks at the context in which these clusters appear, its fallacy becomes clear: the geminated clusters always span two syllables and are (or may be) thus pronounced in isolation, providing the opportunity for clear "aspiration" of the first consonant (if those were indeed aspirated); on the other hand, the clusters ΧΘ, ΦΘ appear in the same syllable (indubitable when they are word initial) and must be pronounced together, thus making the pronunciation of an intermediate "aspiration" more difficult. In other words, if the first consonants of the aforementioned clusters were aspirated stops and one had to give, it would be rather the first consonant of the intra-syllabic clusters ΦΘ, ΧΘ and not that of the inter-syllabic clusters ΦΦ, ΘΘ, ΧΧ, which is exactly the opposite of what we observe.

Another riddle is the occasional appearance of the geminated aspiratae as ΦΦ, ΘΘ, ΧΧ, which is hard to explain under the catholic model, as correctly pointed out by Jannaris and conceded by Blass: "khkh [is] impossible, since here the organ is the same" (BLAS90, p. 105). Allen explains them as "graphic doubling[s] after the analogy of other (unaspirated) geminated forms" (ALLE87, p. 27, n. 33), i.e., as a graphemic analogy (!!), the actual pronunciation apparently being at odds with the spelling. Teodorsson reports four instances of ΧΧ for (normative) ΚΧ in attic inscriptions (TEOD74, p. 135), as suspected cases of (re-?)assimilation! Why squeeze an aspirated release between two identical stops, when the reason for leaving it out in the first place (allegedly) was the impossibility of pronunciation?

All the above receive a satisfactory explanation under the orthodox model, when one takes into account the established principle that phonotactics (or sound change in certain environments) in a language may change without necessitating a change of phonology. In modern Greek, there is a propensity for dissimilation of sequences of stops (ΚΤ→ΧΤ, ΠΤ→ΦΤ) or of continuants (ΧΘ→ΧΤ, ΦΘ→ΦΤ). From the former (i.e., the stops which are considered to have always been pronounced the same), it is evident that this is not due to a change of pronunciation, but to an aversion to continuity that the language has developed in the course of time. The opposite tendency is observed in ancient Greek, which seems to have liked sequences of similar sounds: ΠΘ, ΚΘ assimilate to ΧΘ, ΦΘ (next to ΦΤ→ΠΤ, ΧΤ→ΚΤ). That the assimilated feature was stricture (i.e., fricative before a fricative), is suggested by the fact that "Before a /tʰ/ (aorist passive stem), velars become [kʰ], labials become [pʰ], and dentals become [s]" (in non-prejudiced terms: before θ of the aorist passive stem, κ, γ become χ and π, β become φ, while τ, δ become σ); the result of the assimilation of dentals (i.a. τ) before θ was a fricative (σ) and only if the result (φ and χ) of the same kind of assimilation for the other tenues (i.a. π and κ) were also fricatives would the analogy make sense. Furthermore, the attic convention of rendering ξ and ψ as ΧΣ and ΦΣ, if interpreted as indication of the alternative pronunciations ΧΣalt and ΦΣalt, appears to suggest that Attic went one step further in assimilation, by extending it to the original clusters /ks/ and /ps/.

If assimilation was the dominant tendency in ancient Greek, how is the gemination of the aspiratae Φ, Θ, Χ as ΠΦ, ΤΘ, ΚΧ to be explained? Why did the anticipated sequences ΦΦ, ΘΘ, ΧΧ dissimilate to ΠΦ, ΤΘ, ΚΧ, contrary to the general trend of assimilation? One important reason must have been that, contrary to all the above cases of assimilation, the two consonants in these clusters were on different sides of a syllable boundary, leaving the first one exposed (being the coda of the preceding syllable) to any kind of dissimilative sound change. Besides the already mentioned influences of suspected phonotactics, such as OCP-like constraints (Grassmann's law) and Choeroboscus' "rule" of no aspiratae in coda (which work for any value of Φ, Θ, Χ), there is a purely orthodox explanation: if Φ, Θ, Χ were fricatives (hence, continuants), then the acoustic difference between the single and the geminated letters would have been rather a fine one; the conversion of the first of the ΦΦ, ΘΘ, ΧΧ to its corresponding (voiceless) stop would make this difference acoustically more salient.An alternative explanation would be based on the theory outlined in the next chapter, which identifies gemination with a sustained occlusion: if Φ, Θ. Χ were fricatives, the attempt to (create and) maintain an occlusion would result in the homorganic stop (Π, Τ, Κ) being produced before the actual release of the fricative. Thus, fricative values of Φ, Θ, Χ not only make the pronunciation of ΧΘ, ΦΘ easier, but they may explain the appearance of the geminated aspiratae as ΠΦ, ΤΘ, ΚΧ better than the Catholics think.

The Jimmy Hoffas

If one accepts (for the sake of argument) the catholic thesis that Φ, Θ, Χ were aspirated stops, one may naturally wonder what happened to those sounds. One certain fact is that they have left no aspirated traces in any greek dialect, be it predecessor, sibling or offshoot of Koine/Attic. The offsprings of Φ, Θ, Χ are universally fricatives (the reported aspirates of Tsakonian are developments of consonant clusters, never of original Φ, Θ, Χ, which have remained [f], [θ], [x] or at most [s]).

Not only are there no relics of aspirated stops, but there is no evidence of any change from aspirated to fricative pronunciation either. It is as if the change happened overnight, since no-one reports (and nothing suggests) a parallel existence of the (alleged) "old", aspirated pronunciation and the (alleged) "new" fricative one (the imaginative "verifications" of such existence will be addressed later). Although the Catholics have not felt obliged to provide evidence of a transition from "old" to "new" pronunciation, they have not been shy about the particulars of such a transition: "we may be sure that Greek φ, θ, and χ have developed into the modern voiceless spirants through the intermediate stage of affricates" (STUR20, p. 183)! Unfortunately, in the catholic world it costs nothing to speculate; at most, if there is strong evidence against the speculation, it will be "amended" by further speculation. Without providing any supporting evidence, this unfounded scenario is... a legend in his (=Sturtevant's) own mind and is of absolutely no value.

Allen, having convinced himself that the aspirated plosives were a certainty in classical times, considers evidence "which would suggest that the beginnings of such a change could be traced to the 2 c. B.C" (ALLE87, p. 23), but finds it inconclusive and examines evidence of a later dating. Nevertheless, the latter "evidence" relate to unequivocal verification of the fricative values (e.g. latin and gothic transcriptions, classification by the Grammarians, spelling mistakes), not of the alleged transitional period. Of the produced evidence, one worth discussing is probably the claim that "the labial φ may have developed its fricative pronunciation earlier than θ or χ" (p. 24); although the evidence is refuted twice (once because in Latin of II-III AD there was only one corresponding fricative, f) by Allen himself, the claim is maintained as a possibility because in Ossetic "Old Iranian t and k have developed initially to the aspirated plosives [th] and [kh]; but Old Iranian p has gone beyond the [ph] stage to give a fricative [f]"; apart from the impossibility to verify the Ossetic claim (which furthermore appears not to hold water), any such development in another language is not comparable to Greek: the sounds of Φ, Θ, Χ constituted a natural class throughout the entire history of Greek (i.e., both in "ancient" and "modern" Greek) and it is, therefore, very unlikely that they developed separately, yet stayed phonologically related and ended up phonetically corresponding (as they were before the alleged change).

All in all, the transitional period necessary for a change to a new pronunciation remains more elusive than the corpse of Jimmy Hoffa.Chatzidakis, who in view of some of the above considerations is compelled to concede that "τὰ δασέα κατὰ τὴν Ε᾽ ἑκατοντ. συνετάκησαν εὶς ἕνα φθόγγον ἁπλοῦν καὶ δασύν, ἄγνωστον ἀκριβῶς ποῖον|in the 5th cent., the aspiratae were fused in a single aspirated sound, not being known which one exactly" (CHAT02, p. 460), i.e., diverges from the official catholic timeline, cannot escape the doctrine he was repeatedly taught in Leipzig that an affricate stage (Πφ Κχ in his notation) must have preceded the fricative one and sees a verification of this in the "σποραδικῶς απαντῶσαι γραφαί, οἷον ὄκχος, ὀκχέω, ἐκχθέματα|occasionally encountered writings, like ὄκχος, ὀκχέω, ἐκχθέματα" (p. 461). In addition to not being clear whether the first two are not instances of (irregular) gemination and the third one confusion about the fate of the (morpheme-internally) unusual cluster κθ resulting from composition, in order to "read" the above spellings as verification of the affricate value [k͡x], we need to assume that χ=[x], which would mean that χ was already a fricative!

What Are You Sinking About?

The Catholics pretend that they have discovered evidence of the spread of linguistic innovation in non-Attic dialects. In a recurring pattern, whenever there is clear evidence of "modern" pronunciation in peripheral dialects, the Catholics confine the feature to these particular dialects, which they identify as the initiators of the respective "innovation". While the gradual spread of sound change from one dialect to the other may work for decentralised languages, like pre-Lutheran German, the analogy does not apply in Greek. Attic (and to a lesser extend Ionic), being adopted as official by the Macedonians and spread to the end of the (then) known world, was the basis for the common language and the other dialects had a marginal role, if at all. Thus, the claim that a marginal dialect, like Boeotian for the monophthongisation of the diphthongs and Laconian for the fricativization of the aspiratae, set the trend for mainstream Greek is one that must be strongly substantiated; however, no Catholic felt the need to go beyond a mere assertion of the influence that the (heavily outnumbered, strongly marginalised and utterly insignificant) Boeotians and Laconians would never have on the Koine. The developments in these dialects can, therefore, have nothing to do with those in the Koine and, should the thesis hold true, it would mean that the same sound change occurred independently both in the dialects and in the common language, as if sound change of (the assumed) affricates to fricatives is a one-way street. If there were a gradual sound change, it would have spread not from one (distinct) dialect to another, but from one place in the domain of the same dialect (Koine or Attic) to another (cf., e.g., the spread of [f]→[h]→[Ø] in medieval Castilian from north to south; LLOY87, pp. 322-326). It is, therefore, evident that the probability of such independent, yet identical development is much much lower than that of the orthodox thesis that the aspiratae were universally fricatives throughout the existence of Greek as a separate language (or language family) or at least since the adoption of the alphabet.

In the specific case of the aspiratae, the catholic claim is that the aspiratae had become fricatives in Laconic (and possibly in other dialects) since V BC or earlier. The reason is the occasional representation of laconic θ as σ "in Attic writers - e.g. ναὶ τὼ σιώ, παρσένε in Aristophanes, σύματος in Thucydides. In the 4 c. B.C. spellings of this kind appear inscriptionally at Sparta" (ALLE87, p. 26). The mere fact that this is put forward as an indication of fricative pronunciation of Θ in Laconia, but not in Attica makes us wonder: What are they sinking about? As demonstrated in the ingenious commercial, only a German would make this connection. Why would an Englishman subscribe to this narrow-minded interpretation of the evidence? As a matter of fact, Allen does rightfully ponder "whether the σ in these cases represents a dental [θ] or whether in fact this had already changed in Laconian to the alveolar [s] which seems to be attested in its modern descendant Tsaconian". It appears that only their prejudice has prevented the Catholics from putting two and two together.

Doric dialects, particularly Laconian, suffered assibilation of the dental fricatives. This has been established for the voiced fricative (Δ→Ζ), whereas it is suggested for the voiceless fricative (Θ→Σ) by the present state of Tsakonian (as also pointed out by Allen). The simple interpretation of the evidence is, therefore, that the renderings of Laconian by the attic writers (σιώ, παρσένε, σύματος, etc) are the first indications of "convergence" of Θ and Σ (still distinct, but not enough for the Athenians to distinguish), which must have been completed in IV BC, as suggested by the graphic confusion in Laconia proper (otherwise the native speakers would not confuse two distinct phonemes). The catholic explanation that Θ was [θ] in Laconian, but the Athenians (because they pronounced their Θ differently) perceived it as [s], which happened to be the value that it verifiably attained in Laconian, is unnecessarily complicated.

Apparitians|typical armenian ending; cf. The Kardashians and Haunted Schools

As a result of the catholic dogma that Φ, Θ, Χ did not always stand for [f], [θ], [x] in Greek, sightings of evidence confirming the "preservation" of the (alleged) old pronunciation as [pʰ], [tʰ], [kʰ] are sometimes reported in scholarly works, in order to explain superficially studied transliterations. These invariably involve the determination that a foreign sound, which is (today) a fricative (usually velar [x]) was not represented by the greek letter (usually Χ) that is now pronounced as the same fricative; this is surprisingly interpreted unidirectionally as an indication that it was the Greek letter that had not yet attained its "modern" value, not the foreign one. Once more, the present pronunciation of every other language is used as the arbiter for determining the pronunciation of Greek. Furthermore, it is weird that the same people who deem that Η (the old symbol of δασεῖα, which they claim to have been [h]) originated from ח (today [ħ], [χ] or [x]) and not from ﬣ (today [h]), do not see that sometimes an established phoneme-grapheme mapping may not be maintained during alphabet transmission.

The most extravagant such claim must be Trubetzkoy's suggestion that Greek χ was still an aspirated stop in IX AD! His argument is based on the existence of an extremely rare (two instances!) grapheme that was "used for the guttural spirant x" alongside a symbol for what he considers the "native Slavic phoneme" for [x]. Based on their relative order in two abecedaria, he concludes that the former was used to represent the greek sound, which did not exist in Slavic (and was, he says, still [kʰ]), while the latter was alien to Greek and a new grapheme had to be introduced in Glagolitic. Other than the relative order, there is only one specimen for each letter: "xerovĭskǫ"/"xerovimĭskǫjǫ" for the former from the borrowed word for cherub (which, incidentally, is not actually greek) and "xvalǫ"/"xvalami" for the latter. To base such an unlikely theory on such meagre evidence is not very scientific (but, unfortunately, very common in official linguistics). There are several plausible explanations that do not necessitate asserting that the speakers of Greek still... coughed their χ in the late Middle Ages:

The last two alternatives are summarily dismissed based on arguments (addressed in the notes above) that at most establish that there was no strong need for representing two related sounds. However, one should not forget that the twin representation was only initial and transitory (almost elusive in written evidence), the use of a single symbol being established relatively early. This could only mean that the Slavs realised that there were no strong enough reasons for maintaining both symbols. In other words, the rationale behind the adoption of two letters must have anyway been week, otherwise the double representation would not have been given up so quickly. This explanation would then make much more sense than claiming that the Slavs were observing the pronunciation of the native Greeks and (in a manner similar to the catholic narrative for the transcription of φ and latin f) eliminated the respective letter from their alphabet, as soon as the alleged [kʰ] pronunciation had become obsolete.

Trubetzkoy is not to blame for this naive claim though. Apparitions of evidence of a non-fricative value of χ appear in many scholarly works, wherein transliterations that contravene one's expectations are usually resolved to the detriment of greek phonology. This theory about the peculiarities of the Glagolitic alphabet is developed along the lines of similar conclusions derived from "its Coptic, Armenian, and Georgian counterparts". All these cases (including Glagolitic, albeit with reservation) are mentioned by Allen as evidence "that a scholarly pronunciation of φ, θ, χ as plosives continued for some time in the schools" (ALLE87, p. 25).

The earliest of these must be the evidence from Egyptian. We have already mentioned that the evidence from a (single) demotic text of II AD, where "Greek φ and χ there represented Egyptian ph and kh, and not the fricatives f and ḫ", is superficial. Furthermore, the lack of value of egyptian transliterations is argued by none other than Chatzidakis: "ἐν Αἰγύπτῳ τὸ ἡμέτερον κα ἄλλοτε μὲν ἐγράφετο διὰ τοῦ ka ἄλλοτε διὰ τοῦ ga ὑπὸ τῶν Αἰγυπτίων, καὶ τὸ ἡμέτερον γ ἄλλοτε μὲν διὰ τοῦ g, συχνότερον δ’ ὅμως διὰ τοῦ ng ὑπ’ αὐτῶν|in Egypt, our κα was sometimes written by means of ka and others by means of ga by the Egyptians, and our γ sometimes by means of g, more frequently though by means of ng by them" (CHAT02, p. 349); it is evident that, if we are to be consistent in our interpretation of the evidence, we should conclude that Κ (at least before Α) had a second articulation besides the voiceless plosive one (not necessarily a voiced one though, as Chatzidakis hastily asserts; cf. PEUS99, p. 80). At any rate, the expected transliterations are to be found "in Coptic where φ and χ are occasionally rendered by specific Coptic letters for labial and velar fricatives respectively, e.g. σχῆμα 'form' > CϩΗΜΑ and CϢΗΜΑ (besides CΧΗΜΑ)" (PEUS99, p. 90); it is hard to imagine that, if Φ and Χ were aspirated stops, they would be represented by continuants.

Nevertheless, the most important argument, probably used for the first time by Blass, is based on the composition of the coptic alphabet, which was established in III AD: "The Egyptian Christians, when they devised a new alphabet, mainly borrowed from the Greek, for their national language, employed the symbols Θ Φ Χ for the real aspirates which are found in Egyptian; on the other hand for the sounds f and ch, which they likewise possessed, they adopted peculiar symbols which were annexed to the Greek alphabet" (BLAS90, p. 106); this argument is repeated by Sturtevant (STUR20, p. 179) and Allen (ALLE87, p. 25). It is reasonable to ask why two different symbols were invented for each of the labial and velar fricatives, if their sounds in Greek and Egyptian coincided. It is also reasonable to suspect that the answer lies in a phonetic difference; even a small one would do. The Catholics insist that (in such late an era) Φ and Χ were (still) aspirated stops, but no such assumption is necessary, as they have neglected to establish the exact value of the egyptian fricatives. To begin with, (pre-Coptic) Egyptian is considered to have lacked a velar fricative; ḫ is "reconstructed" as the uvular fricative, a sound sufficiently different from [x] of Greek to justify a graphic distinction.Note that this is not a mere case of transliteration where a sufficiently similar sound could have been used for approximately representing a non-native one, but a convention arising from the need to accurately represent two audibly different sounds that existed at the same time in Egypt. Things do not look good for egyptian "f" either: "The nature of <f> and <b> is somewhat less clear" (PEUS99, p. 133); this means that the value [f] accepted by most has not been actually proven (in fact, "an emphatic bilabial plosive /ṗ/" value has even been suggested).Just like Latin, a voiceless labiodental fricative [f] would be an outlier in egyptian phonology without having a voiced counterpart; only when the latter gap was filled (surprise surprise by Β=[v]), does the phoneme appear to fully correspond to greek Φ. The two symbols did attain the values [f] and [x] in Coptic, the confusion with Φ and Χ becoming more frequent, but there is nothing to exclude the possibility that it was the greek sounds that stood their ground, the egyptian ones moving towards them; it is more reasonable to assume that this "adjustment" was made under the influence of Greek, which in post-Alexandrine Egypt was an important factor to reckon with; furthermore, the fricatives of Egyptian (as well as Latin and other languages) appear to be products of uncoordinated development and have no counterparts in at least one of the three dimensions of The Matrix, whereas the greek fricatives are well integrated into a fully symmetrical phonological system.

Of somewhat different nature is the evidence from Armenian. Here, the greek δασέα correspond to aspirated stops (փ, թ, ք), both in alphabetic position and in transliterations (from V AD on): "The Armenian aspirates, tʻ, pʻ, and kʻ, regularly represent the Greek rough mutes, as atʻɫestaikʻ=ἀθληταί, pʻaɫang=φάλαγξ, kʻart=χάρτης" (STUR20, p. 179). The important fact is that, unlike Coptic and Slavic, the armenian counterparts of the greek δασέα form a complete series firmly integrated in the phonological system, just like the greek series. This fact alone should be enough to end all discussion about the reasons behind the association between the greek δασέα and the armenian "aspirates"; after all, the Catholics themselves accept that aspirated stops are good approximations of fricatives, when the latter are not part of the native phonology (cf. the "evidence" from Prakrit discussed above). Nevertheless, a recurrent argument is used in this case too: "both the Armenian and Georgian alphabets, formed around the 5 c. A.D., use symbols based on Greek χ to represent their aspirated plosive kʻ [kh] and not their fricative x [x]; moreover, Greek words borrowed early into Armenian also show kʻ and not x for χ (e.g. kʻart = χάρτης); only after the 10 c. does Armenian x or š begin to appear for Greek χ" (ALLE87, p. 25). However, just like Egyptian, Classical-Armenian "x" is considered to have been an uvular fricative [χ], a sound that the native speakers of Armenian would certainly construe as different from greek [x] much better than modern western scholars do.

But this is not all. The crucial question that is not answered in this case either is how we know that armenian թ, փ, ք were "aspirates, tʻ, pʻ, and kʻ" in Classical Armenian (corresponding to the time of the introduction of the armenian alphabet and the aforementioned transliterations). It appears that their classical values have been decided based on the modern values of those letters. Is it possible that their classical values were [θ], [f], [x], later changing to their modern aspirated values [tʰ], [pʰ], [kʰ] (possibly under caucasian, e.g., georgian, influence)? Before rushing to deny this thesis, a number of facts testifying in favour of its plausibility should be considered. In evolutionary linguistics, it is considered that Armenian's development from PIE is "paralleled by Grimm's Law in the Germanic languages": a series like (reconstructed) *t, *d, *dh has developed into *t', *t, *d in Armenian and into *θ, *t, *d in Germanic. Is it unreasonable to assume that the voiceless plosives also changed in accordance with the germanic model *t→þ [θ] and then turned to aspirates, possibly under the influence of Kartvelian? Internal developments in Armenian suggest that kartvelian influence was post-classical: "by the 10th century, spoken Armenian had changed drastically from the Classical standard, growing increasingly agglutinative", a feature attributed to Armenian's northern neighbours. A possible sound shift from fricative to aspirated pronunciation around X AD is probably also suggested by the introduction of ֆ "in the 11th century" to represent foreign [f], which was until then rendered as փ (modern pʻ). All in all, the evidence is probably better explained by assuming a diverging Classical-Armenian pronunciation, than a "retained" late-archaic greek one.

Apart from the above-mentioned considerations, the greatest concern against adopting the catholic interpretation of "preservation" of an aspirate pronunciation should be the repercussions it would have on the catholic timeline. We have seen that (based i.a. on latin and gothic transliterations) the Catholics accept that the fricative pronunciation of Φ, Θ, Χ was established by IV AD; the actual evidence that the Catholics condescend to submit is rather suggestive of II AD. How, then, can this simple truth be reconciled with the claim that an aspirated pronunciation is reflected in the later evidence from Coptic, Armenian and Slavic? Taking Gothic vs. Armenian as an example, we would have to assume that the δασέα were already fricatives in IV AD, when Ulfilas invented the gothic alphabet, but relapsed to aspirated stops in V AD, when Mesrop came up with the armenian one. The catholic heads have found the solution in that "a scholarly pronunciation of φ, θ, χ as plosives continued for some time in the schools" (ALLE87, p. 25). Did Ulfilas and Mesrop (even St. Cyril!) go to greek schools that were so much different in teaching? And, if the taught pronunciation were at variance with everyday (or at least prevalent) speech, where is the (direct) evidence of this scholarly tradition? In the countless works produced in this prolific period there is not the faintest hint that the δασέα were not pronounced properly or that some different "scholarly pronunciation" was recommendable. To claim that the... spirit of the aspiratae was lingering in the schools of the late roman era up until IX AD without leaving a trace is yet another case of catholic science fiction.

Cutting A Long Story Short

After the overwhelmingly exhaustive analysis of all the presented evidence, it is time to evaluate the available data.


Before deciding what the δασέα were, we need to determine what the δασεῖα was. I am afraid we cannot have a definite answer to the latter question, at least one that would satisfy either side. The universally accepted value of the δασεῖα (not only by the Catholics, but also by most Orthodox, at least for a very old time) is [h]. There is no doubt that the conviction about the correctness of this value comes from the correspondence of δασεῖα with the latin H (ΦΘΧ1) and (since latin H has left no traces in Romance) through the value that this letter (not phoneme) has in Germanic. What everyone is disregarding though, is that δασεῖα is sufficiently "transparent" to allow interaction (elision and crasis) of vowels standing on its sides (ΦΘΧ7), yet not disappear, but remain behind after the interaction to influence a preceding stop (ΦΘΧ2). This behaviour is not typical of a consonantal segment, not even [h]. By applying Sturtevant's test for the πνεύματα, we can declare the catholic value C invalid, at least until we "find a language in which elision actually occurs in that position".

Unfortunately, no-one has felt obliged to produce such evidence, probably because everyone considers the (axiomatic) identity ῾=h=[h] so certain that the possibility of elision and crasis before [h] is proven by... ancient Greek! However, the evidence produced in support of this identity is anything but conclusive. To begin with the main misconception, the value of latin H, which appears to have been very similar (if not identical) to the δασεῖα, is established not based on facts, but rather on (germanic) bias. It is true that in the germanic alphabets modelled largely after the latin one, the grapheme <H> stands for [h], a degenerate [x] (or [ç] or [χ]), but neither are these languages scions of Latin, nor can we claim a surviving pronunciation of latin H at the time these alphabets were devised; as such, the value of germanic H(=[h]) may have as much to do with that of latin H(=?) as cyrillic Н(=[n]) with greek Η(=[i]) in the Middle Ages (or both of those with latin H). In fact, since any phonetic value of latin H has disappeared from the face of the earth and (more importantly) since latin H also satisfied ΦΘΧ7, it is reasonable to assume that it too was something more elusive than a consonantal segment.

The other evidence presented as verifications of the (a priori) accepted value ῾=[h] are typical cases of hallucination: they only "confirm" the catholic value, if one has already accepted it as true; e.g., PIE *s or *j considered to be the ancestors of δασεῖα (ΦΘΧ10) could have developed into anything (or even nothing), but the Catholics see only the possibility *s/*j→[h]. In fact, some "proofs" are so flawed that they actually work against the catholic thesis, the most prominent example being the consideration that the term πνεῦμα (as well as the latin "aspiratio") refers to [h], something that is patently wrong (ΦΘΧ8). Against the assumed value speaks the origin of δασεῖα's original grapheme, which is based not on ה (as would be expected if it were [h]), but on ח, which is the first fricative of the NW-semitic alphabet (ΦΘΧ9). Evidence from transliterations between Greek and other languages that might have possessed a /h/ phoneme are little researched and inconclusive and, if they have any phonetic significance, they may call for a revision of the phonology of those languages as much as they do for Greek. All things considered, the additional evidence put forward in favour of C are not enough to mitigate the effect of ΦΘΧ7, which decisively speaks against the identification of δασεῖα with the segment [h].

Things do not look good for the Orthodox either. Specifically, the δασεῖα may have been "transparent", but it was not nonexistent, as it had enough "substance" to turn a tenuis to an aspirata after the vowel interaction (to which it was transparent) took place. Although the "h aspiré" of French is a good example of a symbol that has a functional, but no phonetic value, it is difficult to maintain that something similar happened with δασεῖα. Specifically, the theory that it merely stood for a deleted consonant does not explain what caused the conversion of a tenuis to aspirata; the theory that it was just a reminder for turning a preceding tenuis to an aspirata also fails to account for the mechanism that would cause the conversion and, most importantly, must be based on the assumption that the tradition was established before the adoption of the alphabet and was maintained without the help of written records. It is too precarious to subscribe to any of those theories.

As for the actual phonetic significance of the δασεῖα, our best guess is that it was some property (alt) that accompanied the vowel throughout its duration or at least for its initial part. This would mean that the grapheme <Η> was an inline diacritic comparable to the IPA symbol for stress rather than a segment, a fact that would also explain the ease with which Η was abandoned in 403 BC as a symbol for the δασεῖα (even though the latter did not cease to have a phonetic significance), as well as its re-introduction in the Koine as a symbol (˫ or ῾) that was written above the δασυνόμενον vowel. Unfortunately, we cannot be more precise, as neither is tradition a useful guide nor is there any reported phonetic property (that I know of) that exhibits the same characteristics. We will come back to this later, after we evaluate the evidence on the aspiratae.


Up until the middle of the 19th century the traditional fricative values of Φ, Θ, Χ were not seriously doubted. Even their emergence (ΦΘΧ2) from Π, Τ, Κ before the δασεῖα (which was always considered to stand for the segment [h]) did not seem so unnatural, but was considered a peculiarity of Greek. It appears that the establishment of the aspirated values (in which Curtius must have played an important role) was mainly due to a) the obsession of the (19th-century) linguists with Sanskrit and b) a prejudice against non-sibilant fricatives as an integrated part of a language's phonology. The latter is a consequence of the fact that most important western-European languages lack the phonological symmetry exhibited by Greek, in which fricatives play an important role; with the exception of Spanish (where the symmetry is accidental, being a result of "desibilation"), no other western language (that I know) has a complete fricative series (Romance having only labials resulting from fricativization of labio-velars, English having strangely done away with the germanic velar fricatives and German undergoing its own version of seseo). Symmetry in those languages is a matter of stops and indian phonology provided the model for stop-only series.

One fact that is not sufficiently emphasised is that Φ, Θ, Χ correspond (in terms of phonological origin) not to the indian aspirated voiceless stops, but to the breathy-voiced stops [b̤], [d̤], [ɡ̈] (ΦΘΧ13), which were conveniently represented as "bh", "dh", "gh", giving the wrong impression that the assumed voiceless aspirates "ph", "th", "kh" of Greek were a mere product of devoicing of the first element. After two centuries of studying Sanskrit from books of (bad) latin transliterations, it appears that scholars have finally discovered the true nature of the indian "voiced aspirates" (as something different from the "voiceless aspirates"), but have not adjusted their theories on the ancient-Greek δασέα, which are still considered to be descendants of "original *bh, *dh, *gh".It is beyond doubt that the most reasonable explanation of the correspondence between greek Φ, Θ, Χ and sanskrit "bh", "dh", "gh" is that they are descendants of the same phonemes. It is also acceptable to represent those phonemes as *bh, *dh, *gh (for lack of better symbol, even though something like *B, *D, *G would also do the trick). However, the exact phonetic value of those phonemes in PIE can only be speculated (to suit one's theories). To base the "reconstructed" pronunciation of Φ, Θ, Χ on the certainty that the typographical conventions *bh, *dh, *gh indicate "voiced aspirates" would be comparable to the misconception that class-2 stops of Egyptian were voiced because of their transcriptions as b, d, g (PEUS99, p. 80). In fact, the only point of similarity between Greek and Sanskrit is Grassmann's law (ΦΘΧ6), which must have been obsolete already in classical times.It is violated in the case of word-initial Υ (e.g., "ὑφίστημι", "ὑφή", "ὑφέν", "ὕθλος"), which was always δασυνόμενον (ΦΘΧ11), as well as in compounds like αχθοφόρος or ὅθεν. Inscriptional violations are also reported (e.g., TEOD74, p. 221, n. 179: "The vase painter Cachrylion almost exclusively writes his name Χαχρυλίον"). According to Chatzidakis (CHAT02, p. 448), Grassmann's law was active before the split of Greek into the ancient dialects, but it was inactive already before Homer, as there are numerous violations in the epics. However, Grassmann's law is an OCP-like constraint that merely tells us that a certain property (δασύτης) was not tolerated in two consecutive syllables, but not what the property involved; neither can it be concluded that the property was the same in Greek and Sanskrit (where the law developed independently). That the nature of the involved property could have been different in Greek and Sanskrit is suggested by Bartholomae's law, which states that voice assimilates (in both directions) in consonant clusters, but "aspiration" does not;KEHR02, p. 180: "Sanskrit is especially illuminating in this respect because the language has laryngeal agreement of 'breathy voice' ([spread, voice]), but only [voice] truly assimilates while [spread] is realized with the release of the second stop: [b.dʱ]". this contrasts with the greek rules of assimilation, where i.a. a δασύ induces δασύτης to an immediately preceding mute (κθ/γθ→χθ, etc). It is, thus, fair to say that, besides Grassmann's law, there is nothing common between Sanskrit and Greek. Finally, it is not even certain that the phonemes transliterated as "bh, dh, gh" had their modern indian values in Sanskrit, the descriptions of the Indian Grammarians (who apply the same term to them and to the fricatives) leaving open the possibility of them being fricatives.

Another point that adds to the catholic certainty that Φ (and accordingly Θ, Χ) was not a fricative until the christian era is the avoidance of latin F for transliterating greek Φ, which was represented by the digraph PH (ΦΘΧ1), and Quintilian's testimony that Greeks could not pronounce F. Crucially, the identity F=[f] cannot be proven until the time F and Φ (are considered to have) coincided. In fact, the manner in which Latin Grammarians prescribe spelling rules (ΦΘΧ12) makes it clear that the confusion was between F and PH, i.e., an internal orthographical issue of Latin, and it was these graphemes that came to represent the same sound. A careful evaluation of the evidence leads to the reasonable conclusion that F stood for a sound different from [f] (for which the special digraph PH had been devised) until it converged to it in I BC-II AD. This assumption would explain many anomalies, like the unusually isolated place of [f] in the catholic phonology of Latin, the unpaired phoneme /w/ that was fricativized around the same time, Quintilian's description of F as inhuman and the development F→H→Ø in Spanish without further assumptions.

As concerns the testimony of the Grammarians, who lived at Roman times, the only certainty is the verdict of Chatzidakis that nowhere do they describe the aspiratae as tenues+δασεῖα(=[h] according to the Catholics). They merely use the recursive definition that they were pronounced δασέως, which does not reveal anything about the nature of δασύτης. Neither does the division of the consonants between ἄφωνα and ἡμίφωνα tell us anything about the reasons behind it. The passages used by the Catholics as confirmation of an aspirated pronunciation are misquoted, mutilated, twisted and interpreted as something they are not. In fact, the fondness of Halicarnasseus (and the subsequent Grammarians) for the δασέα can hardly be justified, if they were plosives followed by an exhalation.

Once these strong prejudices have been overcome, the importance of the evidence confirming the fricative nature of Φ, Θ, Χ cannot be overlooked. The transliterations of Gothic, Semitic (ΦΘΧ14) and Persian progressively establish the validity of the orthodox values in the imperial, hellenistic and late-classical age, a fact acknowledged only by Chatzidakis among the Catholics. The traditional (fricative) values are also compliant with onomatopoeic words like χάος/χάσμα and φυσῶ, which make little sense under the catholic model, as well as with Φ's and Σ's description by Plato as "πνευματώδη" indicating the "φυσῶδες". The clusters ΧΘ, ΦΘ are particularly problematic under the catholic model, as [pʰtʰ] and [kʰtʰ] are not reported in any language (ΦΘΧ15); their emergence from the assimilation of a preceding tenuis by a following aspirata (i.e., ΚΘ→ΧΘ, ΠΘ→ΦΘ) would also make Greek unique, if ΘC, ΦC, ΧC were true, in that an exhalation ([ʰ]) would have to be inserted where it would only make pronunciation more difficult (cf. the aforementioned behaviour of Sanskrit, where the exhalation is maintained at or transferred to the end of the cluster). The ancient inscriptions cannot and do not confirm the catholic thesis; contrary to the catholic assertion, the interchanges Π↔Φ, Τ↔Θ, Κ↔Χ are as normal for fricative Φ, Θ, Χ as they are for aspirated ones; on the other hand, the (few but important) interchanges Φ↔Θ are reasonable only for the modern values of the two letters. Finally, as regards the geographically limited spellings ΠΗ, ΚΗ, ΦΘΧ3 is to be interpreted (in accordance with Allen's verdict in ALLE87, p. 29, n. 36, based i.a. on ΦΘΧ4 and ΦΘΧ5) as a case of a digraph being devised for a phoneme not covered by the available graphemes (paralleled by the etruscan FH for later 𐌚); this is consolidated by the observation that not once does a spelling ΠΗ or ΚΗ occur where the graphemes Φ and Χ were available (ΦΘΧ16).

It is, therefore, fair to say that ΘO, ΦO, ΧO not only fare at least as well with the evidence thrown against them, but that they also explain further (silenced) data better than ΘC, ΦC, ΧC do. The beauty of the orthodox consonantal model will be more apparent after the presentation of the mediae.

Joint Venture

There is one point that still remains a riddle, particularly under the orthodox model, namely the relation between δασέα and δασεῖα and the emergence of the former under the influence of the latter on the ψιλά (ΦΘΧ2). The δασεῖα appears to have been active at least until christian times (cf., e.g., THUM88). Admittedly, there are several instances of omission or wrong insertion of the δασεῖα in ancient inscriptions (e.g., ALLE87, p. 52), there are violations of ΦΘΧ2 (e.g., TEOD74, p. 133), but this only testifies to the elusive nature of δασεῖα (a property by all odds) and it would be too precarious to profess that δασεῖα and its effects were a distant memory; even in the unlikely case that it ceased to exist before classical times, ΦΘΧ2 must have been active at least at a very old time and needs to be explained (as far as possible).

Essentially what is needed is to identify a property of a (word-initial) vowel that can induce another property on a preceding (plosive) consonant. The latter should be the feature that characterises the δασέα with respect to the ψιλά. Under the orthodox model this feature is one of stricture, where ψιλά (voiceless stops) are [-cont] and δασέα (voiceless fricatives) are [+cont]. Obviously, the same distinction cannot be claimed for vowels (which are by definition [+cont]), so that we cannot identify δασεῖα with [+cont] (and even less ψιλή with [-cont]). Nevertheless, there is no need for the vowel δασύτης to be the same as the consonant δασύτης. It very well possible that a vowel property induce a consonant property of a different kind. For example, in Icelandic, stress (which is not a laryngeal property) can affect the ordering of (pre- vs. post-)aspiration (KEHR02, p. 111) or even eliminate it in some dialects (n. 46). Under this consideration, it is evident that, although there is no straightforward solution to the riddle of induction of δασύτης from a vowel to a consonant, there are several indirect possibilities. Unfortunately, we do not have enough data for the nature of vowel δασύτης to make an educated guess and it seems that there is no modern parallel of a property of a vowel affecting stricture of a neighbouring consonantI firmly believe that the change [k]→[ʃ]/_[a] (e.g., castellum|castle[LAT]→chastel→chateau[FRA]) in French comprised the intermediate stages [k]→[x]→[ç]→[ʃ], where the fricativization of the first change was caused by the "openness" of [a] (and the second change is probably supported by the southern-Dutch variant [ç] for ch). I do not have a definitive proof for it, but the official thesis (e.g., GRAN07, p. 142, §341) that /k/ was "palatalised" before [a] cannot hold water, for [a] is not a front vowel. In any event, if we were to use this (hypothetical) development as our guide to explain ΦΘΧ2 under the orthodox model, we would have to identify δασύτης with "openness"; but this could not be the case, because δασύτης can characterise all vowels and it is not easy to maintain that each vowel had an "open" (ἁ) and a "closed" (ἀ) variant. (but, keep reading). Under these circumstances, we can only describe ΦΘΧ2 (i.e., state it), but not explain it, a stance we had also take in the case of metre.

The Catholics believe that their values (ΘC, ΦC, ΧC) explain ΦΘΧ2, because they treat the δασεῖα as a segment that can be transferred from the left side of the vowel to the right side of the consonant. However, this assumption violates the more important restriction ΦΘΧ7, which is almost always a prerequisite for ΦΘΧ2 to take effect. In addition to relying on the untenable thesis ῾=[h], the catholic interpretation does require (under ΦΘΧ4 and ΦΘΧ5) some fusion of two distinct phonemes ([p/t/k] and [h]), which could have any outcome, including less stricture (the [+cont] feature of the fricatives). The only assumption that would save the catholic thesis would be to identify δασεῖα with [spread] and claim that it... spreads to the preceding consonant(s) of the onset, although there is no other example of similar behaviour (KEHR02, p. 180, as presented above). In fact, [spread] is a nice candidate for δασύτης, as it is a laryngeal feature that can accompany both vowels and consonants; it, therefore, deserves some more investigation.

There are two possibilities for a vowel articulation to include the feature [spread]. One is for the vowel to be voiceless ([ḁ]). However, the status of "voiceless vowels" is hazy: the only example I keep seeing is Japanese, where voicelessness is restricted to the close vowels. Furthermore, judging from the claim that there are voiceless vowels in English, with the second letter of "peculiar" and "potato" as mentioned examples (!), it appears that what are reported as "voiceless" are actually "ultra short" vowels. If my understanding is correct, the δασυνόμενα of ancient Greek could not be "voiceless vowels", as they also include undeniably long vowels (e.g., ἥλιος|sun).

The second and most reasonable possibility is for the vowel to be breathy ([a̤])Kehrein treats "breathy" and "voiceless" vowels as phonologically equivalent, albeit phonetically different (KEHR02, p. 82)., which would actually identify δασύτης with [murmur]. Contrastive (i.e., non-allophonic) murmured vowels are not very common. A prominent example is Gujarati, which also comprises breathy consonants. If we are to interpret ΦΘΧ2 strictly as transfer of the property δασύτης from the vowel to the preceding consonant(s), then we would expect murmured vowels in Gujarati to induce murmur to a preceding stop. However, this is not the case, as it is reported that Gujarati contrasts "/baɾ/ 'twelve', /ba̤ɾ/ 'outside', /bʱaɾ/ 'burden'"; had it been true that the murmured vowel [a̤] causes the plosive [b] to become breathy voiced ("voiced aspirate" [bʱ]), the last two examples would be equivalent. On the other hand, the surprising effect that vowel murmur has on (some) plosives is to turn them into... spirants! Indeed, we are informed that "with murmuring of vowels, the voiced aspirated stops /ɡʱ, d̪ʱ, bʱ/ have voiced spirant allophones [ɣ, ð, β]", a behaviour that also applies to voiceless plosives, "including /pʰ/ being usually realized as [f] in the standard dialect". I believe that this example is enough to end any illusion one may have that ΦΘΧ2 speaks decisively against orthodoxy and for catholicism.

Murmur is a strong candidate for δασύτης; it might also explain its identification with germanic [h] through the association of [ha] with [a̤]~[ɦa], it would exactly correspond to the indian transliterations (if they are reported correctly) and would not get in the way of elision and crasis (probably paralleled by a similar behaviour in Gujarati, at least as far as crasis is concerned); most importantly, it would explain the terms δασεῖα/δασέα, i.e., dense, used by the ancient Grammarians and Plato's πνευματώδη, i.e., breathy! However, [murmur] is not the only candidate. There are many kinds of concomitant articulatory features (phonation, nasality, etc) which could be considered. As I am not familiar with all of them, I cannot conduct in-depth research for all possibilities. A thorough investigation of all possibilities would require further research and... imagination.

Even without a definite answer to the question of δασύτης, the situation with murmur in Gujarati indicates that the orthodox values of Φ, Θ, Χ may well be compatible with ΦΘΧ2. Furthermore, these values are sufficiently well established for the ancient language based on the other evidence, which they explain better than the catholic ones, tradition heavily favouring their fricative nature and providing no traces of aspirated values or of their alleged transition to fricatives. It is time to see how the last item of discord on ancient-Greek consonants, namely the mediae, fits into this model.


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