Other Non-Attic Characters

Home > Greek > Unicode > Characters Outside Unicode

There are three characters which were innovated in epichoric scripts, and are described in Jeffery (1990:38-40). All three of them have a claim to characterhood, but that claim is weak enough that they tend not to appear in editions as anything but normalised characters—in the last character's case, with a diacritic if you're lucky.

1. Archaic Sampi

U+0372 Greek Capital Letter Archaic Sampi [Ͳ], U+0373 Greek Small Letter Archaic Sampi [ͳ]

The first character is the sampi [disigma], as it was used (briefly) in the Ionic alphabet as a sibilant (see also discussion at www.parthia.com). The first question to answer is whether it should be separated from the numerical sampi at all. If it is, it is for rationales similar to those for koppa—the letter sampi looks little like the numeral, is identified with it only tentatively, and there is the potential for plaintext Greek differentiating the numeral from the letter. In the one instance Jeffery (1990:420) does use sampi in transcription (an abecedary from Samos), sampi has its archaic look: αβγδεϝζ[η]θ̣κλμ[ν]ξοπϙρστ̣υφχψω .

The second question is what the phonetic value of sampi was. It turns up in contexts where Greek would normally write σσ or ξ. Jeffery (1990:39), who already suspected Ionic Greek of using ξ for [ʃ], also suspects that sampi was originally borrowed from Carian, and used to express the Carian sibilant in loanwords.

The third question is whether there is any use of sampi in transcription at all. In Jeffery, there is; the following from Ephesus is the only instance she provides a transcription for:

τεαραϙοντα μνεαι το πρω[τον] εσταθ[ησ]αν εκ ττων δω[ρων] χρυσο (Jeffery 1990:414)

four hundred mnas [unit of weight] were the first of the gifts of gold.

Where Ephesus has τεαραϙοντα, Classical Greek would normally have τεσσεράκοντα. Buck (1955:18) also notes the Ionic sampi, transcribing it as capital T:

T, occuring also in the Carian alphabet and perhaps ultimately another modification of the san, is used for usual σσ = Att. ττ as Halicarnassus (ἈλικαρναTέ(ω)ν beside Ἀλικαρνασσέων), Teos (θαλάTης beside θάλασσαν), Ephesus (τέTαρες, etc.) and elsewhere.

Unlike Corinthian EI, this transcription is not confined to discussions of the alphabet, it routinely appears in published texts as a distinct letter, rather than as normalised σσ or ξ—aided, I surmise, by the suspicion that it represents a distinct phone. Often the required glyph has not been available typographically, and the letter is represented with a capital tau in a case-mixing context (τέTαρες), which has been assumed to disambiguate the letter. Since ϡ is never used to transliterate sampi as a letter, and the letter is usually not normalised, there is a strong case for disunifying the two sampis.

If there is such disunification, however, it is worth noting that Carian is already roadmapped for U+10B00 - U+10B1F. If a distinct sampi is to be encoded, it may make sense for transcriptions of Ionic to borrow the Carian codepoint. (I must admit, the gifs attached to the Indo-European Database site page on Anatolian alphabets don't exactly reassure me that a Carian proposal is going to come any time soon.)

Note also that although it is not included in the Old Italic block, there was some use of sampi in writing Celtic in Greek script (Markey & Mees 2003:123), for Proto-Indo-European *st, Celtic *ts. (Mees has told me that he has flirted with using ϡ, and decided against it, since in his line of business scholars don't profilerate but transliterate.)

1.1. Pamphylian sampi

The alphabet of the deviant Pamphylian dialect includes a letter like a square psi, , which had the phonetic values /s/, /ss/ and /ps/ (Brixhe 1976:7-9). (Note that there is a distinct glyph in the Pamphylian epichoric alphabet that corresponds to psi proper, so this is a letter distinct from psi in Pamphylian.) The identification of the Pamphylian letter with Ionian sampi is not definitive historically: the Pamphylian character could have been an independent loan from Carian, from the Cypriot syllabary, or directly from Semitic scripts. But although the Ionian and Pamphylian glyphs are not identical, their phonetics overlaps enough that they can usefully be conflated as the one codepoint.

Archaic sampi was included in a proposal I submitted to the UTC in 2005, L2/05-003 Proposal to add Greek epigraphical letters (see also L2/04-389). It was accepted in Unicode 5.1.0, April 2008. The codepoint was intended to encompass Pamphylian sampi as a glyph variant.

2. Zigzag Iota

The second letter is still poorly understood, and is shared between archaic Greek scripts and other languages. Jeffery (1990:39) has a summary discussion of it:

The way Jeffery reports it, we don't know exactly what the Eteocretan instance sounded like—and if it is an iota, why they had two of them; we don't know if the Eteocretan instance has anything to do with the Sikiniot chi; we don't know what the Lemnian instance has to do with either; nor what the Phrygian z has to do with all three. So we ain't going anywhere with this soon.

So the best thing to go with is what scholars do. They transliterate, of course; and they don't transliterate this character as anything distinct. In fact, the last fifty years of research has left nothing of the zigzag. In particular:


The latest and fullest treatment of Eteocretan is Duhoux (1982). He argues convincingly (pp. 167-171) that Jeffery is wrong about the glyph, and it really is iota. He's not too worried about the fact that in the Dreros 1 inscription there would thus be two glyphs for iota, 8 and : the early alphabets were in flux, and the Dreros 1 inscription also has two forms of mu. In fact, even the second half of the Dreros 1 inscription (lines 3-5), which is in Greek, has both iotas. Nor is he concerned that Sikinos turned the Cretan ᛇ glyph into chi; Crete didn't have a chi in its script, and in it had an available spare letter.

Moreover, there are good internal grounds for calling this iota rather than chi. I wouldn't know Eteocretan if it hit me (though my ancestors are from Eteocretan country), but even I can tell that Jeffery's transcription does look like too many consonants in a row for the character to be truly a consonant—and an aspirate at that. The glyph appears once as a correction of alpha—which is likelier if it is a vowel. If isn't iota, then the Praesos 1 inscription has no iotas at all, which is inconsistent with the other inscriptions. And an iota reading leads to the recurrence of the morphemes -ιναι- and -μιτ- in Eteocretan—and we need all the recurring morphemes we can get. So that settles that.

To illustrate the foregoing, I give the Dreros 1 inscription (last seen conscripted by the Italians during World War II to support a church wall :-( -- Duhoux 1982:36-37), and the Praisos 1 inscriptions, as edited by Duhoux; to indicate the glyph, I use ι̱.

Dreros 1

[±2–4]y?ρμαϝ ‖ ετ ‖ ισαλβρετκομν
[±2–4]·δ ‖ μεν ‖ ι̱ναι ‖ ισαλυρια ‖ λμο vacat
ετυρο·[±3–4]μυναοδ̣μ̣εν· ο̣?ι̣?υ̣?[
[±3–5]ματρι̱ταια [± 7]

Praesos 1

ος ‖ βαρζε ‖ α[±1]·ο[
]αρκ·αγ̣σετ ‖̣ μεδ·
αρκρκοκλες̣ ‖ δε·[

Note in the Dreros inscription Duhoux's typically Gallic rendering of yot. This is Duhoux's tentative interpretation of the character, which he thinks looks more like the Phoenecian yot than iota did; other scholars have thought it an epsilon or an alpha.

If the glyph is chi in Sikinos, then of course there is no controversy about transcribing it as chi.


The glyph's value has been settled since 1969 (although Jeffery's work hasn't been updated to reflect this): it's a jot, and TITUS' online Old Phrygian corpus is accordingly replete with the letter in their Greek script versions of Phrygian. (Brixhe & Lejeune 1984, whose publication TITUS reproduces, transliterate into Roman, which means they Gallicly use y.) So inscription W-01c reads in Latin transcription ataniyen kuryaneyon tanegertoy (Brixhe & Lejeune 1984:42), which in Greek script would be something like ατανιϳεν κυρϳανεϳον τανεγερτοϳ.


We have a notion that the Lemnian glyph represents, as Duhoux (1982:168) puts it, "a sibilant or affricate phoneme whose precise value has not yet been determined (ζ, ś, etc.)" Trombetti (1928:188) in his original publication renders it as z; e.g. holaiez naφoθ ziazi maraz m-av, as does André Charbonet. Etrusca Philologia uses s: holaies naφoθ siasi maras mav. ("[To] Holaies grandson of Sias, {Maras Mav/Maras, however}".)

And there you have it. Zigzag iota is not a distinct letter under any scheme of transliteration that has persisted, and does not need allocation. QED.

This is an excellent illustration of Bunz's "Don't Proliferate, Transliterate" point: in 1950 there was a strong case for a distinct zigzag codepoint, but by 1985 that case had evaporated. If the decision to adopt the codepoint had been made in 1950, Unicode would have been stuck with it in perpetuity. This is the kind of misstep philologists want to avoid.

3. Tsan

U+0376 Greek Capital Letter Pamphylian Digamma [Ͷ], U+0377 Greek Small Letter Pamphylian Digamma [ͷ]

The third letter is an innovation restricted to Mantinea in Arcadia, in the fifth century BC; it is shaped like И. Jeffery (1990:40) describes it as occuring "for the dental τ < *qu̯, with several other abnormal letter-forms; and I should guess it to be short-lived local invention." Jeffery (1990:212-213) discusses it further, as do Buck (1955:18, 62) and Bubenik (1983:82); there is a recent article specifically about it by Yves Duhoux (2006): La lettre ͷ et quelques problèmes connexes en arcadien archaïque (IG V 2.262). Kadmos 45: 20-68.

We can note the following:

Jeffery does not cite the Mantinean inscription. Buck does (1955:198), transliterating tsan as σ̱:

εἴ σ̱ις ἰν το(ῖ) ἰεροῖ το̄͂ν τότ[ε ἀπυθανόντο̄ν] φονέ̄ς ἐστι, εἴσ̱' αὐτός εἴσ̱ε [το̄͂ν ἐσγόνο̄ν] σ̱ις κὰ τὀ̄ρρέντερον, εἴσ̱ε τ[ο̄͂ν ἀνδρο̄͂ν] εἴσ̱ε τᾶς φαρθένο̄, ἰνμενφ[ὲς ἐ̄͂ναι κᾶ] τὸ χρε̄στέ̄ριον· εἰ δὲ μὲ̄, ἴλαο[ν ἐ̄͂ναι].

If anyone (present) in the temple is a murderer of those who perished at that time, either himself or any of his descendents in the male line (that is, if any one of these is present in the temple), (a murderer) of either the men or the maiden, it shall be impious in the eyes of the oracle; if not, it shall be propitious.

(Attic: εἴ τις ἐν τῷ ἱερῷ τῶν τότε ἀποθανόντων φονεύς ἐστι, εἴτ' αὐτός εἴτε τῶν ἐγγόνων τις κατὰ τὸ ἀρρέντερον, εἴτε τῶν ἀνδρῶν εἴτε τῆς παρθένου, ἐμμεμφὲς εἶναι κατὰ τὸ χρηστήριον· εἰ δὲ μή, ἵλεων εἶναι.) τις < *kʷis, εἴτε < *eːkʷe.

The transliteration <σ̱> is most usual; it appears in Liddell-Scott-Jones, the authoritative Greek-English dictionary (e.g. the entry for τις—the version online at Perseus has skipped the minus sign below diacritic. Schwyzer chose <ś> instead (Schwyzer, E. 1923. Dialectorum Graecarum Exempla Epigraphica Potiora. Leipzig: Teubner p. 319). In his linguistic treatise, however, Bubeník uses the tsan glyph:

/ts/ could be spelled in Arcadian with a special letter И as in Иις /tsis/ 'who' (= τίς), οИεοι /otseoi/ 'to whomever' (cf. Homeric ὅτεῳ) (Bubeník, V. 1983. The phonological interpretation of Ancient Greek: A pandialectal analysis. Toronto: University of Toronto Press. p. 82)

And the tsan is reasonably frequent in linguistic discussion of Arcadian.

Despite the importance of tsan for Greek historical linguistics, it seems unwarranted to add a Unicode codepoint for just 13 tokens on a single inscription. Even if that inscription is like something out of Jacobean tragedy. ("Judgement against certain persons guilty of sacrilege toward Athena Alea, whose temple had been made the scene of a bloody fray.")

3.1. Pamphylian digamma

The alphabet of the deviant Greek dialect of Pamphylia (Brixhe 1976:5–6) used three glyphs to represent /w/: digamma, beta, and the glyph И -- the same glyph we have just seen used for tsan. The latter glyph also appears where Greek would normally have a beta; and a similar glyph appears as a form of digamma or beta in other epichoric alphabets (Crete, Melos: Jeffery 1990:308). Brixhe (1976:6) concludes that the glyph И is a local development from Phoenecian waw, while the standard digamma was later imported from other Greek alphabets. (For this reason, he notes, И is unrelated to tsan, which developed from tsade.)

Brixhe (1976:46–57) discusses the distribution of И and Ϝ, and find that in early inscriptions И appears after vowels—either in diphthongs ([aw, ew]), or in glides near back vowels ([sbaluwau])—while the normal digamma appears elsewhere. He concludes that the Pamphylians imported the normal digamma, in order to be used alongside the local form, motivated by a phonetic split: /w/ had gained the allophone [v], represented by Ϝ, while И was kept for [w]:


"Épitaphe de Miaklis, fils de Varnis, de Meaklis, fils de Apélauruwis, et de Korvalina, fille de Varnis." (Brixhe 1976:232)

As you would expect, Brixhe transliterates the normal digamma as <v>, and the Pamphylian variant as <w>. Before Brixhe had reached this conclusion, Buck (1955:47) had concluded that the И glyph represents the early development of a [v] pronunciation for /b/ and /w/, and recommended that it "is best transcribed as v (w in German publications)." He thus gets the phonetic value of И wrong, according to Brixhe: it was Ϝ, not И, that was pronounced [v]. But since Ϝ as /w/ was a widespread feature of Greek alphabets, and И was not, it was expedient to assign И the "odd" transcription, <v>.

That said, the distribution of the indigenous and imported digamma was inconsistent even in the earliest surviving inscriptions—e.g. И remained used in [v] contexts in archaic spellings like Иανάᛉα /wanassa/ [vanassa] "queen", referring to the goddess Artemis. [Note the Pamphylian sampi.] And as far as we can tell, [v] was only an allophone of /w/ (as well as of /b/), and not a distinct phoneme—so that normally it would not be accorded its own letter in an alphabet. As a result, on the one hand, publications of Pamphylian texts keep the normal digamma and the indigenous digamma distinct, and do not normalise И to the more common Ϝ. On the other hand, И and Ϝ are treated as the same underlying letter where normalisation is appropriate.

Dspite these instances of conflation, publications of Pamphylian as text consistently differentiate the two forms of digamma, and the distinction is not predictable from context. So they should be represented as distinct codepoints.

This letter was included as Pamphylian Digamma in a proposal I have submitted to the UTC, L2/05-003 Proposal to add Greek epigraphical letters (see also L2/04-389). I argued that Tsan should also be encoded with this codepoint, for expediency. The letter was accepted in Unicode 5.1.0, April 2008.

Nick Nicholas, opoudjis [AT] optusnet . com . au
Created: 2003-09-07; Last revision: 2008-05-14
URL: http://www.opoudjis.net/unicode/other_nonattic.html