CC BY-NC-ND
23
Ignasi-Xavier Adiego
University of Barcelona; ignasi.adiego@ub.edu
The longest Pisidian inscription (Kesme 2)
In this article, the author offers an analysis of the longest Pisidian inscription Kesme 2 (S 2), recently published by Claude Brixhe and Mehmet Ozsait. A segmentation of the scriptio continua is proposed by using a combinatory method. Some connections with the rest of Pisidian linguistic materials and also with other Luwic languages is suggested. However, the inscription continues to be a largely impenetrable text.
Keywords: Pisidian, Luwic dialects, Anatolian, Indo-European, Greek Epigraphy, Asia Minor.
§1. Recently, Claude Brixhe and Mehmet Ozsait have edited two Pisidian inscriptions from Asar Kale, a site on top of a hill very near Kesme (Brixhe-Ozsait 2013). Kesme is around 30 km NE from Selge, and at a similar distance SE from Adada. The ancient name of Asar Kale is unknown, but according to Drew-Bear and also to the editors, it may be the MouAaaa//a// (Zgusta KON §861-2 MouAaaaewv o 5^|aoç) mentioned in an inscription found near Kesme.
One of the two inscriptions (Kesme 1, now S 1 in Brixhe 2016) was already published by Brixhe and Drew-Bear, but that edition was "massacré par l'éditeur", according to Brixhe-Ozsait (2013). A new edition is proposed of this four-line, 34-letter text.
The other inscription, previously unpublished, is very impressive: it is a text of thirteen lines, complete, and thus constitutes the longest Pisidian text found to date (Kesme 2, now S 2)1.
This inscription, together with the other one from Kesme-Asar Kale and two other inscriptions re-edited in the same paper from roughly the same geographical area (the middle course of the river Eurymedon), give us a very new impression of Pisidian. The rest of the Pisidian corpus comprises basically very brief texts from the territory of Tymbriada which contain only personal names. These inscriptions from the middle Eurymedon area, and in particular the longest inscription, offer a different kind of text, which undoubtedly contain a common vocabulary. As we will see, although we might expect to find elements that would confirm the hypothesis that Pisidian is a Luwic dialect — a hypothesis based exclusively on personal names and on the presence of a sigmatic genitive — this new material is practically impenetrable and raises considerable doubts about the exact position of Pisidian among Ancient Asia Minor languages.
Kesme 2 is dated by the editors to the 2nd-3rd century A.D. In fact, this is the date they propose for all the corpus of Pisidian inscriptions (the four inscriptions edited in Brixhe-Ozsait and the brief epitaphs from Tymbriada).
The inscription is quite well preserved and Brixhe-Ozsait's edition and commentaries will serve as a good starting point. My aim in this brief paper is to try at least to segment the words (the text is in scriptio continua) to be able to recognize any recurrent elements and to suggest, in a very hypothetical way, some explanations for them. In this regard, I hope I will be able to go slightly further (though not much) than the editors.
1 For these two inscriptions see now also Brixhe (2016: 97-99).
Journal of Language Relationship • Вопросы языкового родства • 15/1 (2017) • Pp. 1-18 • © The authors, 2017
In order to analyse the text, for convenience I will use a Latin transliteration of the Greek alphabet used for Pisidian. I recognize that this is not the usual practice: the tradition in Pisidian studies, as it is with neo-Phrygian inscriptions, is to keep the text in the Greek alphabet. But I think that typographically it will be clearer if I use Latin, particularly in order to deal with the two different digammas present in the text (see immediately below).
My transliteration is conventional and should not offer problems. Note the following conventions: r\ = e, a = o, R is a variant of semicircular sigma, therefore = s. I do not transcribe the peculiar letter ] (a hapax which is difficult to interpret).
Apart from this last letter, perhaps the most notable feature of this inscription is the coexistence of two digamma letters: the common form F and the Pamphylian form «. The latter form is clearly differentiated from N n, so it must be a different letter. For a discussion of their value, see infra. Conventionally, I will transliterate « as <w> and F as <v>.
Here is Brixhe-Ozsait's edition:
H'EKA^PCiO OYA PFIA IOGA
oaaiamocotoctomaa TAKA \A-\ OPAL - r- O AIAPT11 nCXtT inAAOCTOTQKO K A \A-\T O TOA aOiA c O BIA Cm A POVC itokfaPOYA AT
^AP ANlNOY.Ti tY©l-rvtPEN
10
0ANUE \z 1/1 AP El l/I AT in A AO C T OKPAP OYAAC Ol AAI AtO
13 13 14 15
K AH-MAHOElEAPPFr-CClTokP
16
■D OVCWAMOEIFEIAI ^AAnAEIii^i UC-AIAPPI-B-OYEO P ELIEAPPAIA
17 IS
^COMA ATAC E OkigpAO YC OAC
1 MEKAOPErOOYAPriAIOEIA
2 OAAIAMOEOTOETOMAA
3 rAKAUHQPAC-r-OAl APni
4 nOCITiriAAOCTOTQKO
s KAMHTOTOAAOIAEOEIAEHA
6 PO Y E I TOKFAPOYAATI
7 t{?) A11ANINOYTIEYOHMEPEN s OANAEIEMAPEIMATIFIAAO
9 ETOKFAPOYAAEOI AAIAEO
10 K A MHM AMOElEAPPH -(;) I E -(E) T O K P
11 .(?) OYEMAMOEIEEIAIMEAAFIAEIAE
12 MEAI APPI-B-OYEOPEEIEAPPAI A ij .(?) EOMAAfAEEOKOflAOYCOAE
§2. The inscription poses several graphical problems which we must address before proceeding any further:
(1) the letter ] mentioned above: its value is not clear. Might it be a kind of z? Does it have a particular function? Might it be a variant of F? Brixhe-Ozsait (2013) considers this latter possibility but dismisses it. The question remains obscure.
(2) Some examples of o may be examples of Q. This is a typical crux in Greek epigraphy (and also in late Carian!): the difference between the letter omicron and the letter theta with central dot is not always clear. In this inscription, the editors express their doubts about the following cases: line 3 oaiarpi / Qaiarpi?; line 7 ...oemeren / Qemeren?; line 8 oan.../Qan...?. In an absolutely conventional way, I use <o> to reflect the possibility that the letter might be Q instead of o in the cases mentioned.
(3) Also problematic are the possible confusions between R = s and E = e. This affects line 1: mekloreg... / meklorsg... (But the editors clearly prefer mekloreg..., for contextual reasons). Line 1 ...eid / sid (but also here the reading e is preferred). For the first example, I agree with the
editors: a cluster srg seems highly improbable to me (and also all the possible segmentations if we have to deal with two words: s##rg, sr##g). In the other case, both readings may be acceptable. For this reason, I use a conventional transliteration, parallel to <o>: I transliterate this ambiguous e/s as <e>.
(4) At the beginnings of lines 11 and 13, the editors note the apparent traces of signs, but conclude that they are probably accidental marks. I accept this latter explanation and will ignore them.
(5) I also accept other solutions suggested by the editors, such as the reading of p in the last line.
(6) I also accept the presence of some ligatures, like of M-I = M + H = ne in lines 3, 5, 12 or H-M as H + « + E = eme in line 7 .
(7) One cannot be entirely sure that there are no abbreviations in this text. If there are, this would seriously hamper our task of segmentation. I assume, as the only way to begin to analyse the text, that there are no abbreviations.
§ 3. This is my transliteration:
The only thing that is certain about this inscription is that it contains numbers. The editors clearly identify two numerical expressions: in line 3, -g- is the Greek number '3'; in line 12, -bis '2'. Brixhe-Ozsait (2013) suggests that these numerals may accompany a personal name to express the second or the third person bearing the name, as is usual in Greek epigraphy.
Apart from these two examples, there is another numerical expression: in line 10, IE seems to appear between two horizontal traits, as do the other two numerals (although we must admit that the horizontal trait at the ending of the expression is not clear). Taken as a numerical expression, IE makes sense as 'fifteen'. The editors accept this only as a possibility, but I think that it is the simplest interpretation.
If this interpretation is correct, in this case at least it is hard to accept that this numeral was used in the sense proposed by the authors: '15' is very unlikely to have been used to refer to the repeated use of a name inside a family.
§ 4. The presence of these numerals is, as I mentioned above, the sole evidence that we can obtain for sure from this obscure text. In what follows, I will try to offer a possible segmentation of the text.
To carry out the segmentation, we have the following tools at our disposal:
mekloregoouarplioeid
oadiamosotostomla
gakaweoras -3 - oaiarpi
positipadostotoko
kawetotolaoiasoeiaswa
rousitokvaroudati
](?)apaninoutieuoemeren
oanaeiswareiwatipado
stokvaroudasoiadiaso
kawewawoeiearre -15- tokr
(-?)ouswawoeieeidiwedapaeias
wediarri -2- oueoresiearraia
(-?)somlagaseokoplousoas
1 2
3
4
5
6
7
8
9
10 11 12 13
1) The numerical expressions allow us to segment correctly before and after these marks. Unfortunately, this procedure offers very limited results, because only three numerical expressions are present; however, in combination with the other tools, it may become more useful.
2) Some elements are repeated along the course of the inscription. We can isolate them, at least in their initial part (the final part may present different endings, so the segmentation is less clear).
3) Inside the inscription we find many vowel clusters, some of them formed by four or even five or six vowels. It is logical to assume that they are the consequence of the meeting of two (or more?) different words, and so one can look for word boundaries inside them. However, as we will see below, this procedure is not without its difficulties.
4) The typology of syllabic structure can also help. This tool merits a further explanation. A simple look at the text suggests that Pisidian was characterized by a predominant presence of open syllables, i.e., syllables with a (C)V structure. Note, for instance, lines 4-5:
positipadostotoko kawetotolaoiasoeiaswa
In these two lines the only consonant clusters are -st- and -sw-. The rest of syllables follow the structure (C)V. As we will see below, this predominance of open syllables, and consequently the fact that the position of syllable end (coda) is limited to a few consonants, is present throughout the text and can be taken as a trait of the language encoded here.
§ 5. Thanks to the numerical expressions, we can recognize e i and s as possible word final sounds, and o (recall!: o / 9 ?), t, o, as possible initial word sounds. Look at the corresponding lines:
gakaweoras -3- oaiarpi kawewawoeiearre -15- tokr wediarri -2- oueoresiearraia
s as a word final sound is also guaranteed by the last word of the inscription: (-?)somlagaseokoplousoas
To these meagre results, we add that m can begin a word, as it appears at the very beginning of the text (meklorego...).
§ 6. Some elements are clearly repeated. Assuming that Pisidian was basically a suffix-inflected language, these repeated elements serve to establish boundaries in their initial part. Note the possible segmentations based on this principle:
mekloregoouarplioeidoadiamosotosto mlaga
kaweoras -3-
oaiarpiposi
tipadostotoko
kawetotolaoiasoeiaswarousito
kvaroudati+apaninoutieuoemerenoanaeiswareiwa
tipadosto
kvaroudasoiadiaso
kawe
wawoeie
arre -15-tokrous
wawoeieeidiwedapaeiaswedi
arri -2-
oueoresie
arraiaso
mlagaseokoplousoas
Nota bene:
1) In the case of tipadosto, as we will see below, it is possible that ti is (part of) a preceding word.
2) The segmentation of arr° may seem less sure, insofar as only three letters are implied. These first segmentations offer an interesting result: some repeated sequences appear in
immediate contact, which allows us to segment a complete word. This is the case of mlaga kawe, or tipadosto kvarouda... or wawoeie arre, or kawe wawoeie. Accepting these segmentations, we can refine our analysis:
mekloregoouarplioeidoadiamosotosto
mlaga
kawe
oras -3-
oaiarpiposi
tipadosto
toko
kawe
totolaoiasoeiaswarousito
kvaroudati+apaninoutieuoemerenoanaeiswareiwa
tipadosto
kvaroudasoiadiaso
kawe
wawoeie
arre -15-
tokrous
wawoeie
eidiwedapaeiaswedi arri -2-oueoresie arraiaso
mlagaseokoplousoas
A first conclusion can be drawn from this initial attempt to isolate sequences: the sound immediately preceding each of these possible word initial sequences is systematically a vowel or (in two cases) an s:
mlaga a
kawe e
oras -3-
oaiarpiposi
tipadosto
toko
kawe
totolaoiasoeiaswarousito
o
o
o
s
kvaroudati+apaninoutieuoemerenoanaeiswareiwa a
tipadosto
kvaroudasoiadiaso
kawe
wawoeie
arre -15-
tokrous
wawoeie
eidiwedapaeiaswedi arri -2-oueoresie arraiaso
mlagaseokoplousoas
e
e
i
i
o
o
e
e
o
s
This fact strengthens the impression mentioned above that this language favoured the existence of open syllables and drastically limited the presence of consonants at the end of a syllable and of a word.
§7. A trait of this inscription is the presence of vowel clusters, some of them of a considerable length. In principle, they could be used for establishing word boundaries assuming that we are dealing with the meeting of final and initial vowels of different words. But things are not so simple in Pisidian. Firstly, we must keep in mind that two of the most frequent vowels in these clusters are E <e> and O <o> which, in this inscription, can be easily misread instead of C <s> and Q <0> respectively. As we have seen, Brixhe-Ozsait (2013) admits these ambiguous readings in one instance of E/C and in three instances of O/Q, but in my opinion it is not entirely clear that all the rest of examples of these letters are reliable readings.
Secondly, the spelling practices in the age of this inscription favoured the use of clusters like O + vowel, OY + vowel, and O + vowel to represent /w+vowel/, as can be seen in the use of the Greek alphabet to reflect Anatolian proper names, or to write in Neo-Phrygian (see Brixhe-Ozsait 2013: 240); EI also represented i — as in the contemporary Greek — and intervocalic I could have represented a Pisidian /j/ sound. The use of O, OY to represent /w/ in our inscription2 is rather puzzling, since we already have two different digamma letters («, F) to represent this or a similar sound. But it is not phonologically impossible that in this text there may be a triple contrast, like for instance /w/, /v/ and /p/. Note the particularity that there is no letter B in this inscription.
These spelling practices and perhaps also the existence of internal vowel hiatuses may lead to the presence of such clusters inside words. This singularity of Pisidian was already observed by Ramsay, who portrayed it amusingly by saying that "The Pisidian Language seems to have delighted in vowels" (Ramsay 1883:74).
Let us look at the vowel clusters in this inscription. In order to avoid excessive speculation I examine only the ones recognized by Brixhe-Ozsait (2013) as ambiguous cases of els, old.
2 There are no instances of O before vowel in this inscription.
I present the examples with three or more vowels, and only with two vowels when the vowels involved are neither i nor u — i.e., when a hiatus is more probable.
1 ooua line 1 4 vowels
2 ioei (or: iosi)? line 1 4 vowels (doubtful!)
3 oa line 2 2 vowels
4 eo line 3 2 vowels
5 oaia (or: 9aia)? line 3 4 vowels (doubtful!) (if not, 3 vowels: aia)
6 aoia line 5 4 vowels
7 oeia line 5 4 vowels
8 ieuoe (or: ieu9e)? line 7 5 vowels (doubtful!) (if not, 3 vowels: ieu)
9 oa (or: 9a)? line 8 2 vowels
10 aei line 8 3 vowels
11 oia line 9 3 vowels
12 oeiea line 10 5 vowels
13 oeieei line 11 6 vowels
14 aeia line 11 4 vowels
15 oueo line 12 4 vowels
16 iea line 12 3 vowels
17 aia line 12 3 vowels
18 eo line 13 2 vowels
19 oa line 13 2 vowels
The cluster n. 15 in line 12, oueo, is a good example of 'delight in vowels': here the cluster appears immediately after the numerical expression (2), and so we are dealing with the beginning of a word. Although behind oueo... there may be two words (ou, or even o could have been independent words in Pisidian), oue- is an acceptable initial sequence in Pisidian, as is shown by the divine name (or epithet) Oueyeivoc; or Oueyeivac; (M^tqI 0£wv Oueyeiv^, in Tymbriada, SEG 55, 1447, 1448), the place name Oueppn (Zgusta KON §972) and the personal name OueAA II oc; II (Zgusta KPN §1151-2). If in all these examples oue represents /we/ or the like, /weo/ might be an acceptable word initial sequence even though it is not attested in the rest of the Pisidian documentation.
Previous analysis based on the recurrence of sequences allows us to resolve some of these clusters, at least partially. This is the case of 4 (kawe ## oras), 12 (wawoeie ## arre) and 13 (wawoeie ## eidiwedapeias...). In clusters 12 and 13 we still have a 4-vowel cluster at the end of the word (the same word: wawoeie) and the possibility of a further segmentation in two words (wawo ## eie, for instance) remains open.
Other clusters will be analysed later, in combination with the more speculative attempt to recognize endings.
§8. As for syllable structure, I have insisted repeatedly that this inscription seems to point to a high predominance of open syllables, and a clear limitation of sounds in syllable final and
consequently in word final positions. This statement can be ratified by the Pisidian inscriptions. A brief look at the subcorpus of brief epitaphs (see Adiego 2012) shows clearly that most of the syllables are open, and that practically only s and r can end a syllable or a word. The exceptions to this rule are some examples of geminations (for instance, eddi) and the letter ksi. It is not clear, however, that these clusters should be analysed as heterosyllabic. ksi could be a syllabic onset and dd may representing a sort of voiced stop (vs. the simple d representing possibly rather a fricative). Some examples of alternation -d-/-r-, i.e., rhotacism of -d- between vowels seems to point to a fricative articulation of this sound. Note that a similar rhotacism is present in neighbouring Pamphylian, where it is attributed to a substratum influence (see Brixhe 1976). Certainly, Pisidian onomastics in Greek sources offer a wider range of structures, but it is not clear that all these names, some of them found in contact zones with Phrygia, Lycaonia, etc., should be considered as strictly Pisidian.
In any case, the tendency to present open syllables and to limit the type of consonants in syllable final position in Pisidian suggests that in sequences such as oadiamosotostomla (second line of the inscription), segmentations like oad ## iam ## osot ## om ## la are highly improbable. Of course, this syllable typology allows us to say how the words are not separated, but it is less useful in a positive way: the sequence mentioned admits a great many different possibilities of segmentation even if one gives priority to parsing all the syllables as open: oadi ## mosoto.... vs. oa ## dimo ##soto, etc.
These probable restrictions on syllable finals, combined with the general principle of sonority sequencing in syllable structure, lead us to consider all the clusters of increasing sonority as tautosyllabic. These are the sequences involved and the lines where they appear:
-kl-, -pl- 1
-ml- 2
-kv- 6
-kv- 9
-kr- 10
-ml-, -pl- 13
In the cases of -ml-, -kv- clusters, this analysis is consistent with the segmentation proposed above on the basis of repeated sequences, as they turn out to appear as possible wordinitial sequences. The rest of the sequences constitute negative evidence: they tell us where the words are not cut, but it is impossible to establish whether or not they coincide with the beginning of a word.
The examples of clusters of decreasing sonority are dubious, for several reasons:
(1) the only possible example of n + obstruent depends on the reading of the second letter: 7-8 -no-, where a reading -nd- would make the sequence heterosyllabic. It is not clear to me whether n could really be a word final sound in Pisidian. There are no examples in the rest of the Pisidian inscriptions — although this may be a matter of chance, due to the scarcity of the corpus. Certainly, Pisidian onomastics in Greek sources show a few names ending in -n: personal names such as I|aav, Ei|aav, KaKKav (KPN+Zgusta 1970), Mavxouv, Taxiov, and a subgroup of names inflected in -wv, -wvog (ZopaAiwv, Ouppuwv [doubtful], KavSwv, Kepiwv, EaA|awv, Eipywv, Ta^wv, [..]ouAiwv; and place names like KeapeSiov, MopSiaiov or X£>|aa EaKqvov. But the genuine Pisidian character of some of these names is highly questionable: I|aav, Ei|aav is most probably a Phrygian name, given its frequent appearance in Phrygia; KaKKav appears in a boundary zone between Pisidia and Lycaonia; Taxiov is a widespread female name attested only once in Pisidia; Mavxouv is doubtful (it may simply be Greek:
see Zgusta KPN §868-1); the place names KeopeSiov, MopSidiov and EaK^vov are clearly adapted to the Greek inflection, an explanation that could also be envisaged for the curious subgroup of names inflected according to Greek -wv, -wvo; declination. Therefore, the possibility that the final -n was missing before the consonant, as happens in Pamphylian, ought to be taken into account; it would make a reading -nd- more unlikely. However, the reading 9 offers interesting results from the point of view of the interpretation of the sequence: see below § 10.
(2) the segmentation of the examples of s + stop (limited to st: lines 2, 4, 9) depends on whether Pisidian admitted syllable onsets such as st- sk-, sp-, sd-, sg, sb-, etc. which violate the sonority sequencing hierarchy but are present in many languages (for instance Latin, Greek or English). The rest of the Pisidian documentation shows very few examples of s + stop onsets:
In indirect sources, only a personal name and a place name begin with ot-: Exava^oac; (Zgusta 1970, §1472a) and Expou|a//a// respectively; and only a personal name begins with ok-: EKpaio;. The variants ExAeya, EoxAeya of the name of the well- known Pisidian city Selge (EeAyn) cannot be used as evidence for initial st- in Pisidian: as Brixhe rightly stated (Brixhe 1976:289); the original form must have been Selga/Salga or Slega/Slaga. The forms with ExAeya, EoxAeya, attested in coins, are the Pamphylian adaptation of the place name and x is easily explained here as an epenthetic sound (*slega > stlega). Note that Pamphylian was the language spoken in Selge despite its Pisidian location. Consequently, this may be an exclusively Pamphylian treatment.
In direct sources, the examples of st, sk, sp are also few and far between:
— In Brixhe's new corpus of Pisidian inscriptions (Brixhe 2016), the only example of initial st in an indigenous name is Staneis, Stanei in N 33, to be connected to the above-mentioned personal name Exava^oac;. In N 34 st appears in a purely Greek name, Stephanos. MOYOETOINA (N 32) must be segmented Mouos (genitive) Toina (Brixhe 2016: 90).
— There are no examples of sk sequences.
— Of the seven examples of sp sequences, four appear in the same inscription (N 37) and must be separated into two different words since the p is the initial of the name Piger-dotaris. Other example of sp appears in a new inscription (N 45) where it is clearly a word-medial cluster: Ospouna.
The only two possible examples of an initial sp cluster come from S 4: here a sequence spuadogwesi appears twice, which raises the possibility that this is in fact an initial cluster sp-
(3) The examples of sequences s + a voiced second element are equally scarce: there are no instances in the indirect sources, and the only possible examples in the direct ones are N10 OYANICBABOY and 32 nAnACTAAAOC. For N 10, Brixhe (2016) proposes a convincing segmentation /Oua Nis Babou/, and in the case of N 32, it is difficult to decide between a parsing /Papa Sgallos/ or an alternative parsing /Papas Gallos/. A sequence CA in S 3 appears in an impenetrable context.
(4) The clusters s + Pamphylian digamma (CM -sw-) in our inscription merit a chapter of their own. We find four examples (lines 5, 8, 11 and 11-12). If this represents a /sw/ sequence, there was no violation of sonority hierarchy, so that even if st, sp etc. clusters were not permitted in Pisidian, a /sw/ onset could be possible. In any case, the example in line 11 ouswawoeie... must be ruled out, as we have identified a sequence wawoeie that also appears in line 10. Two other examples coincide to offer a sequence swar: eiaswarousito and euoemerenoanaeiswaieiwa... This leads Brixhe-Ozsait (2013) to propose the isolation of a word beginning swar°. But as we will see below (§10), the options of segmenting s ## war... or sw ## ar here offer interesting results.
To sum up this discussion of clusters where s appears as the first element: the fact that -s can be a word-final sound, the fact that this ending could have a morphological value in Pis-idian (we know at least from the rest of the documentation that it served to express the singular genitive of proper names) and the fact that sC onsets do not seem to have been frequent in that language, makes a segmentation s ## C in principle preferable, though by no means certain. Further analysis is needed to qualify this statement.
§ 9. From here on, we enter a more speculative field. We must try to recognize some recurrent endings in order to identify other possible complete words. For this task, we will take into account the remarks on syllable structure and consonant clusters formulated above.
§ 9.1. -to is a clear word ending. It emerges naturally from the current state of analysis, as it appears in the segmented sequences:
mekloregoouarplioeidoadiamosotosto mlaga
oaiarpiposi tipadosto toko
eiaswarousito kvaroudati
tipadosto kvaroudaso
The other examples of to sequences as possible word endings are much less clear: in mekloregoouarplioeidoadiamosotosto ## mlaga, a segmentation ...to ## sto seems unlikely in view of the doubts about the existence of st onsets in Pisidian. In kawe totolaoiasoeiaswa, a toto laoiasoeiaswa segmentation would be acceptable but is unverifiable.
§ 9.2. Another possible ending is -so: it can be drawn from the segmentation of repeated elements in:
kvaroudasoiadiaso kawe
and in:
arraiaso
mlagaseokoplousoas
The first example is particularly interesting. In kvaroudasoiadiaso it is tempting to segment in turn kvaroudaso iadiaso, showing two words in agreement. Other possible though less clear examples are:
mekloregoouarplioeidoadiamoso tosto
totolaoiaso eiaswarousito
§ 9.3. A third recurrent element that might constitute a morphological ending is ti. It may be recognized in:
kvaroudati ]apaninouti euooemerenoanaeiswareiwati padosto where three words may be in agreement:
kvaroudati ]apaninouti euoemerenoanaeiswareiwati padosto
The first two examples seem quite likely. The third one is more doubtful: it clashes with the fact that tipadosto appears once more, which leads us to isolate a word tipadosto (see above):
oaiarpiposi tipadosto
But it is also possible that a word ending in ti might precede a word padosto. I will assume that both options are possible and I will notate this possibility with a hyphen: ti-padosto.
§ 9.4. -ie is another probable ending, which is obtained exclusively from the segmentation of repeated word beginnings and appears concentrated in lines 10-12 of the inscription:
kawe wawoeie arre -15-tokrous wawoeie
eidiwedapaeiaswedi arri -2-oueoresie
Note that the ending may be generically -el-e, and would include as possible words in agreement kawe and arre.
§ 9.5. Incorporating the analysis of the preceding possible endings, we can go further with the following (very hypothetical!) segmentation:
mekloregoouarplioeidoadiamoso tosto
mlaga kawe
oras -3-
oaiarpipositi-padosto toko
kawe
totolaoiaso eiaswarousito kvaroudati ]apaninouti
euoemerenoanaeiswareiwati-padosto
kvaroudaso
iadiaso
kawe wawoeie arre -15-tokrous
wawoeie
eidiwedapaeiaswedi arri -2-oueoresie arraiaso
mlagaseokoplousoas
§ 9.6. The segmentation in §9.5 begins to offer a series of possible individual words (or at least sequences comprising very short words): mlaga (2x), kawe (3x), oras, toko, tokrous, arrilare. In kvaroudati / kvaroudaso we recognize two clearly related forms, in terms of inflection or of derivation (see below § 10). As possible inflected words in -so, -to we can recognize arraiso, (ti)padosto, iadiaso and the kvaroudaso just mentioned.
Obviously, the remaining long chains must contain different words. I will propose some possible segmentations, but we are entering increasingly precarious terrain.
Let us start with the very beginning of the inscription:
mekloregoouarplioeidoadiamoso
We have already noted that a hiatus can be a clue for segmentation, but the "delight in vowels" of Pisidian advises caution. Here the first cluster ooua, with the repetition of o, suggests a segmentation meklorego ouarplioeidoadiamoso, where the initial <oua> may be a typical representation of /wa/. The rest of the vocalic hiatuses are less clear, but it is very tempting here to see three words ending in -o (and followed by a fourth one in -so):
meklorego ouarplio eido adiamoso
Here I will also use a hyphen to represent these very hypothetical segmentations: meklorego- ouarplio- eido- adiamoso
The other longest chain is:
euoemerenoanaeiswareiwati-padosto
Here the doubts about the exact character of o (= o?, 9?) hinder the analysis even more. I will return to this question later. On -sw-, see immediately below
euoemerenoanaeiswareiwati-padosto
Other sequences remain which are shorter but very possibly contain more than one word: oaiarpipositi padosto
totolaoiaso eiaswarousito
]apaninouti
eidiwedapaeiaswedi
mlagaseokoplousoas
For the first sequence (oaiarpipositi or simply oaiarpiposi), and for the second one (totalaoiaso) I cannot propose any solution.
In the third and fifth sequence, we once again find a cluster -sw-, as in euoemerenoan-aeiswareiwati-padosto. Here we are at an analytical crossroad: the three examples, compared one to one, offer two divergent solutions:
1) euoemerenoanaeiswaieiwati-padosto and eiaswarousito favour the isolation of a beginning of a word swar-
2) but eiaswarousito and eidiwedapaeiaswedi share a sequence eiasw, to be segmented eias w° (eia ## sw° seems less probable, but see below § 10).
To this dilemma, we should add that the remaining example of a -sw- is tokrouswawoeie, in which the cluster has to be separated s ## w, given that wawoeie is a clearly isolated word.
In order to reflect these different options, I use hyphens:
euoemerenoanaei-s-wareiwati-padosto
eia-s-warousito
eidiwedapa-eia-s-wedi
Further segmentations of the latter chain are complicated. We can envisage a segmentation eidi wedapa-eia-s-wedi and think of an agreement in -di. We can also speculate about the relationship between a wedapa-eia-s and wedi. eidi wedapa-eia-s wedi would be an interesting segmentation, but absolutely ad hoc. I prefer to leave the sequence without segmenting.
In ]apaninouti, the first sign remains a mystery. ]apaninouti seems to be an inflected form of a stem ]apaninou- or apaninout-, parallel to kvaroudati: see below §10. It is impossible to say whether there are one or two words behind ]apaninouti. It may even be a compound noun (or name) ]apa+ninouti.
The last sequence, mlagaseokoplousoas, begins with a word isolated as mlaga in lines 2-3. This would suggest a segmentation mlaga seokoplousoas, but it is also possible that here mlaga may represent another inflected form, and so mlagas eokoplousoas or mlagase okoplousoas (the latter supported by the presence of a hiatus) can be alternative solutions. I will represent these alternatives thus: mlaga-s-e-okoplousoas.
§ 10. After this analysis, we attain the following (very hypothetical!) segmentation:
meklorego-ouarplio-eido-adiamoso
tosto
mlaga
kawe
oras -3-
oaiarpipositi padosto
toko
kawe
totolaoiaso eia-s-warousito kvaroudati ]apaninouti
euoemerenoanaei-s-wareiwati padosto
kvaroudaso
iadiaso
kawe
wawoeie
arre -15-
tokrous
wawoeie
eidi-wedapa-eia-s-wedi
arri -2-
oueoresie
arraiaso
mlagas-e-okoplousoas
We now enter the most precarious terrain of all. How should we interpret all these possible words and endings?
§ 10.1. The first step is to look for personal names. This is the sensible decision taken by Brixhe-Ozsait (2013). Unfortunately, this inscription, unlike the brief epitaphs from Tymbriada, does not offer at first glance any tangible form to be identified as a personal name. Consequently, the connections with Pisidian onomastics are tenuous. This may seem surprising, because in a text of this length we would expect at least some proper names (personal names, place names, god names, and so on). But it is also important to note that Pisidian onomastics was undoubtedly very varied: a good example is the corpus of Greek inscriptions from Termessos IV, which contained unpublished inscriptions with a considerable number of new personal names, many of them difficult to connect, even partially, with previously known names.
This may explain why the results of Brixhe-Ozsait (2013) in this terrain are both very limited and also difficult to improve upon. In fact, the connection suggested by these authors, that I take as convincing, depends on the reading of one of the disputed letters in the inscription: in line 8, oanaeiswareiwatipado, with a reading <0> for o, offers a 9anaei... sequence that Brixhe-Ozsait try to connect with the name of the goddess Athena (Brixhe-Ozsait: 243). As these authors point out, the aphaeresis would be comparable to that found in the name of Athena attested in Lycian, Sidetic and Pamphylian. Although the interpretation as 'Athena' looks very attractive (see below § 10.4), it may in fact be a personal name based on the name of the goddess: 9anaei(s) can represent a Pisidian adaptation of the Greek female name AGnvaTg or the Greek male name AGvqaiog. For this latter adaptation, cf. that in Pamphylian the nominative singular -lq -£ig /i:s/ from -*iog (Brixhe 1976: 100), so *0dva£ig could be the Pamphylian form of AGvqaiog and the basis for a Pisidian 9anaei (cf. also Pamphylian 0avd5wpog = AGnvaSwpog for aphaeresis and vocalism).
Other reasonable connections proposed by Brixhe-Ozsait (2013) are (1) oras, a possible genitive of a personal name ora- = Qpag, Oupog < Luwic (and Hittite) ura- 'great' (but forms like Lycian Hura, where h probably < *s, complicate the dossier); and (2) ouarplio = Hittite warpalli- 'fort, puissant and OupnaAog (or OupnaAag), an indigenous name attested in Phrygia (Zgusta KPN § 1174). Further proposals seem to be more tenuous and remote (see Brixhe-Ozsait 2013: 247-248 for all these proposals).
In the following table I offer my own attempt to connect some sequences with Pisidian onomastics:
meklorego- ouarplio-êido- adiamoso Cf. OûaQTC£L|jLou (gen.) (LYC), apart from B.-O. connections Ano-aôi-ç, I5a-a§i-ç, Ki5aa-a§i-ç (PIS)
tosto
mlaga
kawë
oras -3-
ôaiarpipositi padosto Apniaç (PIS) (< ar+ Anatolian piia-) na5a-|jouQLç / ^napa^ouQLç1
toko
kawë cf. the Phrygian and Lycaonian place names Kd|3aÀa, KauaÀa and the Misian or Bithynian place name Kaur|
totolaoiaso Twxuv//a// Twxwv//La, place name PIS
eia-s-warousito Oupou-|3aig PIS (for the second element, cf. AAou-naig?? PIS) If swarou, cf. Oc|3apag PIS, Oc|3apa PIS
kvaroudati Kpapou-ng PAM
tapaninouti Nvog PIS, Noueiligll PIS
euoemeren oanaei-s-wareiwati padosto If Oanaeis = Aer|vaiog cf. supra. If swarei, cf. Oc|3apag, Oc|3apa naSa-|joupLg / "Hapaiaoupig1
kvaroudaso Kpapou-r|g name of a woman PIS (cf. supra kvaroudati)
iadiaso Ia-Ze^Lg (Moa-Zr||JLg), perhaps Ia+ a§i-g, cf. supra Ano-a8i-g, etc.?
kawe cf. supra
wawoeie cf. supra
arre -15-
tokrous
wawoeie cf. supra
eidi-wedapa-eia-s-wedi
arri -2-
oueoresie
arraiaso
mlagas-e-okoplousoas nAouai.v|xr|g or nAouai.v|xr|Tog (IS), nAoug (KAR); Eoag (PIS) Pisidian names in -oag
1 napa|joupiavog, napa|joupiavr|. Cf. also na5a|joupiavog, na5a|joupiavr|
These attempts at connections (all very tentative) do not necessarily imply that the sequences of the Pisidian text where they appear must be interpreted as proper names. Given that we are comparing stems rather than complete names, and as we cannot deduce from the context whether a proper name is being used, it is equally imaginable that these stems appear here as common lexical elements. Note, for instance, the form padosto (2x): the initial part can be compared with the first element of the name na5a-|aoupig, but it could be a word from the common lexicon (a verb? cf. infra) that shares the stem with a compound name. Incidentally, this is a good example of the difficulties of the comparison: besides na5a-|aoupig, there exists a variant napa-|aoupig. It is tempting to see in this latter name a dental rhotacism, which is well attested in Pamphylian (Brixhe 1976). However, there is an alternative explanation that destroys any connection with our inscription: na5a-|aoupig may be a dissimilation from napa-|aoupig, which would be the original form3.
Perhaps the most suggestive connection I am able to offer is the female name Kpapouqg (Zgusta 1970, §563a, attested in Cotenna) which seems to be closely related to kvaroudati and kvaroudaso. The coincidence of the six initial letters is unlikely to be a matter of chance. But how can we explain this connection? Are kvaroudati and kvaroudaso two differently inflected forms of a personal name? As we will see below, this possibility is not without its problems.
In any case, this connection is partial. The name Kpapouqg does not appear tel quel in the inscription, and this is precisely the clearest conclusion of the search of onomastic material in
3 napa- as a first element of compound names is well attested in Anatolian indigenous names, see Adiego 2007: 340 for Carian.
this long text: there are no examples of any known Pisidian proper names. I alluded above to this circumstance when I spoke of the lack of tangible onomastic identifications.
§ 10.2. The search for forms other than proper nouns is even more complicated. Today, we do not know any common Pisidian words, insofar as the brief epitaphs from Tymbriada contain exclusively personal names and the other inscriptions from the middle course of the Eurymedon are as impenetrable as our text. The only way to continue the search is to look for possible connections with the rest of Anatolian Luwic dialects (particularly the best known, Luwian and Lycian), but lacking a clear idea of the specific traits of Pisidian with respect to the other Luwic dialects makes any proposal highly speculative.
Brixhe-Ozsait (2013) ingeniously suggest that the word kawe, repeated three times, may be related to Luwian hawi-, Lycian xawa 'sheep'. If Pisidian was close to Lycian, the connection would be valid, because, as is well known, Lycian x represents a sound /k/. Recall also the Carian gloss kolov- npopaxov, in which kolov may come from *Kopi-ov = Luwian hawi-. If a reference to 'sheep' were present in the inscription, it would be tempting to see in wawoeie (or simply wawo), repeated twice, the Pisidian word for 'cow' corresponding to Lycian wawa 'cow, bovine'. The presence of the two words — probably in a sacrificial context — recalls the appearance of Lycian xawa- and wawa- side by side in inscription TL 149. It is also interesting that two instances of kawe and one instance of wawo(eie) occur in the proximity of numerical expressions (and the other instance of this latter word is not far away from another numerical expression):
kawe oras -3-
kawe wawoeie arre -15- tokrous wawoeie eidi-wedapa-eia-s-wedi arri -2-
Unfortunately, no other indices in the inscription invite us to think that these meanings can be right. Another totally different track to follow would be to connect kawe with Lydian kave- 'priest' (also attested in Greek inscriptions from Sardis as a loanword Kauav [acc.] 'priestess').
§ 10.3. Another terrain for speculation is the interpretation of the possible endings. We have recognized three repeated final sequences that might represent morphological marks: -so, -to and -ti (I leave aside -eie, which is less clear to me):
From a "Luwic" point of view, one might suggest linking them with possible "Luwic" morphemes:
1) -to could be a 3rd sg (or plural?) preterite ending, related etymologically to Lycian -te, Carian -t, Luwian -ta < *-to
2) -so could be a genitive singular ending = Lycian -he < *-so
3) -ti could be (a) a 3rd sg (or plural?) present ending = Lycian, Luwian -ti or (b) a dative of a -t- or -nt-stem (Cf. Lycian Trqqnt-i, CLuwian dUTU-ti-(i) (*Tiwat-i)
(1) and (2), if right, would be mutually consistent, insofar as they would coincide in showing the conservation of final *o where Lycian changes it to *e and Luwian to *a. The possible verbs tosto and padosto 2x) may recall Lycian 3rd preterite iteratives in °stte as astte, qastte, xistte.
A particular problem is posed by the forms kvaroudati and kvaroudaso. If both represent a personal name with different inflections, it is not easy to conciliate them: kvaroudaso would be a genitive in -so (= Lycian -he), of a stem kvarouda-, but kvaroudati would rather be a dative of a
dental or nt-stem (kvaroudat- or kvarouda(n)t-). If we do not accept a sort of heteroclitic inflection kvarouda- / kvaourdat- (or kvaroudant-) we may begin to speculate: kvaroudati could be a verb corresponding to a noun (not necessarily a personal name) kvarouda-. But this sort of speculation is worthless unless we first establish a firmer analysis of the words. In any case, it is puzzling (though also suggestive) to observe that a stem alternation kvarouda- / kvaroudat- recalls the élargissement in -t- of the vocalic stems of personal names in Greek, which was very commonly used in the incorporation of indigenous names (cf. in Pamphylian: Brixhe 1976: 104 ff.). So, in a Greek inscription from Pisidia, we could perfectly imagine a *Kpapou5axL as the dative of a personal name *Kpapou5aç. The same feature can be observed in ]apaninouti. If it is a personal name, it would be analysed in a Greek context as a simple dative of a name in -oûç, -OÛTOÇ (cf. Brixhe 1976: 106). Cf. datives as Aotcouqoutl TAM III, 1 222 from a Pisidian name Aonoupouç (Zgusta KPN §118) or Aààoutl (KPN §52-2) from AÀÀouç. Can we conclude that this élargissement penetrated into (some forms of the nominal paradigm of) Pisidian? Given the late date of the inscription, the prolonged contact with Greek would support this hypothesis.
§ 10.4. Finally, in an inscription of this length one might expect the presence of function words (pre- or postpositions, conjunctions, particles). I am afraid I am unable to offer any convincing suggestions. As a purely speculative exercise, I wonder whether the problematical sequence sw would be a copulative conjunction, etymologically equivalent to Carian sb, Milyan sebe 'and', in the following sequence:
euôëmeren ôanaei-s-wareiwati padosto
Once again assuming ô = Q, we can attempt the following segmentation:
euôëmeren Ganaei sw=areiwati padosto
Where Qanaei and areiwati could be datives coordinated by a conjunction sw. If Qanaei represents /Gana-i/, then this might be the dative of the goddess name Athena. As for areiwati, we could analyse it as another dative of a dental stem /areiwat-/ or /areiwant-/, which would probably be another theonym. Completing this highly speculative analysis, we might see an accusative in euôëmeren and a transitive verb in padosto (cf. supra our analysis as a possible preterite). If padosto had something to do with Lycian pddë 'place', it would be tempting to translate the entire sentence as "((s)he) placed the euôëmere- for Athena and Areiwa(n)t-. The word euôëmere- remains obscure. If ô must be read G here, one might think of a Greek loanword (euQëmere-), but I cannot suggest direct connections to the Greek lexicon4.
This analysis would imply that sw functions as a proclitic particle, in a way similar to Carian sb. Compare the Pisidian example with Carian:
Qanaei sw=areiwati
sarnajs II sb=taqbos (E.xx 6)
paraeym : sb=polo (E.Me 8)
This analysis may appear convincing at first glance, but I stress that it is only a possibility; we have no grounds for favouring it over alternative interpretations, such as the proposed analysis of Qanaeis as a genitive of a personal name.
4 The closest form I can give is the adjective au9r||j£Qog 'made or done on the very day' and the corresponding adverb au9r||J£Q6v 'on the very day, on the same day, immediately'.
§ 11. In a paper published in the Actas del IV Coloquio de lenguas y culturas paleohispánicas, and entitled "Gramática de los plomos ibéricos" my maestro Jürgen Untermann concluded his attempt to segment and analyse the long Iberian texts on lead with the following words: "I beg the readers to forget as soon as possible all my hypotheses regarding the meanings of the Iberian words or morphemes. These suggestions are completely subjective and provisional, and they should not restrain the imagination of researchers in their interest to penetrate the secrets that the Iberian leads preserve. Moreover, the author of these lines is always ready to abandon his own hypotheses when new interpretations or new findings reveal them to be unfounded. In my opinion, the indispensable basis of all studies of this kind is the careful segmentation of the texts, and my aim was to draw attention to some pathways along which we can progress in order to accomplish a task that is still very far from having attained satisfactory results" (Untermann 1985-86: 51). Simply replacing 'Iberian' by 'Pisidian', I can think of no better way to conclude my own paper.
References
Adiego, I.-X. 2012. Minima Pisidica: nota sobre la estructura de una inscripción pisidia de Timbriada. In: V. Orioles (ed.).
Per Roberto Gusmani. Linguistica storica e teorica. Studi in ricordo, 2/1: 17-26. Udine: Forum. Brixhe, C. 1976. Le dialecte grec de Pamphylie. Documents et grammaire. Paris: Adiren Maisonneuve. Brixhe, C. 2016. Stèles et langue de Pisidie. Nancy : A.D.R.A.
Brixhe, C. — Özsait, M. 2013. Cours moyen de l'Eurymédon: apparition du pisidien. In: H. Bru, G. Labarre (ed.),
L'Anatolie des peuples, des cités et des cultures (IIe millénaire av. J.-C. — Ve siècle ap. J.-C.): 231-250. Besançon: Presses Universitaires de Franche-Comté. Ramsay W. M. 1883. Notes and inscriptions from Asia Minor. Mittheilungen des Deutschen Archäologischen Institutes in Athen 8: 71-78.
Untermann, J. 1985-86. La gramática de los plomos ibéricos. Actas del IV Coloquio sobre lenguas y culturas
paleohispánicas = Veleia 2-3: 35-56. Zgusta KON = ZGUSTA, L. 1984. Kleinasiatische Ortsnamen. Heidelberg: Carl Winter.
Zgusta KPN = ZGUSTA L. 1964. Kleinasiatiasche Personennamen. Prag: Verlag der tschechoslowakischen Akademie der Wissenschaften.
Zgusta, L. 1970. Neue Beiträge zur kleinasiatischen Anthroponymie. Prag: Academia.
Игнази-Шабьер Адиего. Самая пространная писидийская надпись (Kesme 2)
В статье автор дает анализ самой пространной писидийской надписи Kesme 2, недавно опубликованной Клодом Бриксом и Мехметом Озсаитом. Предлагается членение scrip-tio continua при помощи комбинаторного метода. Прослеживаются определенные связи с прочим писидийским языковым материалом, а также с другими лувическими языками, однако надпись по-прежнему остается в целом недоступным текстом.
Ключевые слова: писидийский, лувические диалекты, анатолийские языки, индоевропейские языки, греческая эпиграфика, Малая Азия
