The decipherment of the script

3The Carian script was deciphered in the 1980s and 1990s. After the isolated, rather cumbersome, but in many ways prophetic attempt of Kowalski in 19751, it was the British Egyptologist John D. Ray who made the definitive breakthrough, with his proposal of new values for many Carian letters in a series of papers published since 19812. Nevertheless, his system was not entirely satisfactory, and the final decipherment – in the sense of an almost complete list of correct sound values for Carian letters – was the result of the independent and complementary efforts of Diether Schürr and myself in the years that followed3. Leaving aside some minor corrections introduced later, particularly after the discovery of the Kaunos bilingual, the table included in my 1993 book Studia Carica offers a decipherment system, now known as the ‘Ray-Schürr-Adiego system’, which remains valid and has obtained general acceptance. In 1996 the discovery of the Kaunos Carian-Greek bilingual inscription offered totally external support for the validity of our new proposal4.

4This is not the place for a detailed history of the decipherment5. It will suffice to note that the decipherment of the script originated from the existence of bilingual texts – mostly in Egyptian and Carian – in which proper names could be recognized in Carian and Egyptian (or Greek) spelling. This is a logical procedure that has been used in other celebrated decipherments; the surprising thing is that it had not been attempted much earlier in the case of Carian, given that most of the Egyptian-Carian bilinguals had been known since the beginning of the history of the research. The main reason for this delay, as I have explained elsewhere, lies in the aberrant nature of the Carian alphabet: most of the letters resemble Greek letters but represent different sounds.

5As already said, the decipherment of the Carian script is now almost complete. It is true that some signs remain undeciphered, but they are in general letters of a very limited use, in most cases present only in specific local varieties. The only letter that appears in a wide range of sites and about whose value nothing can be said is H, present in the Carian alphabet of Egypt and in several local alphabets of Caria proper. The rest of the undeciphered signs are present only in the alphabet attested in Kaunos. Certainly, the values proposed for some other letters are somewhat hypothetical (, for example), but this is due mainly to their low frequency. For the more usual signs, we are able to offer a satisfactory sound value. Of course, it is difficult in some cases to attain phonetic precision, but this is inevitable in ancient languages for which only written records exist. For example, the exact value of a letter such as ś is not clear, but we are at least able to affirm that it is sibilant, different from s (very probably/s/) and š (very probably/š/). The existence of controversy among scholars about the precise sound value of this and other Carian letters should not mislead colleagues in other disciplines into thinking that the decipherment is incomplete or deficient: the overall validity of the decipherment is accepted and these points of disagreement belong to the realm of more specialized studies. After many years of research, the phonetics of Lycian z or the vocalic digraphs in Gothic or the laryngeal ḫ in Hittite, to mention only three typical examples from different ancient languages, are still being disputed though there is a consensus that the respective writing systems are deciphered and fully intelligible. So controversies of this type are bound to surround a recently deciphered language such as Carian for some time to come.

The Carian alphabet

6Parallel to the decipherment of Carian, our knowledge of the alphabet has increased dramatically in the last years. This progress is due not only to the decipherment itself, but also to the discovery of long texts in Mylasa and Hyllarima that provide us with almost complete inventories for the alphabetic varieties used in these sites. Admittedly, we do not yet have a complete panorama of the diverse local alphabets, but we are at least able to establish a series of conclusions that would have been unimaginable only a few years ago. I will try to summarize them briefly:

71) A distinction should be made between the situation in Egypt and the situation in Caria proper. In Egypt, we find basically a single alphabetic variant attested in different Egyptian sites and at different times (Sais, Saqqâra, Abu Simbel, Thebes, etc.). It is true that some variation did exist there, but it was not especially marked, and the image of a single model spread across the Egyptian soil prevails clearly in spite of the differences of detail. Therefore, the alphabet best known from Egypt, that of the Saqqâra necropolis of Memphis, serves as a model for all the Egyptian documentation; we can ignore the small divergences found in certain places (see fig. 1)6.

82) The situation in Caria is very different: here we find strongly differentiated local varieties. In fact, with the sole exceptions of Sinuri and Kildara, which offer a basically identical variety, and the inscription of Krya in Lycia, close to the Kaunos alphabet, each Carian site seems to have had its own alphabet. Of course, one finds some similarities between alphabets of neighbouring places, but it is not possible to establish – despite previous efforts – rigid groupings of alphabetic variants on geographical bases. Figure 2 shows some of these varieties, together with the alphabet of Saqqâra, for the sake of comparison.

93) None of these well known alphabets from Caria proper shows a close similarity to the alphabet used in Egypt. This suggests that the Carian alphabet of Egypt does not come from any of these Carian sites. The only Carian inscriptions apparently from Caria proper that show a clear connection to the Caro-Egyptian alphabet are three inscriptions on vessels. Of these, two are of unknown origin, but the remaining one has been attributed (though not with total certainty) to Halikarnassos. This is, therefore, at present, the only hypothesis (and a fragile one at that) that we can formulate for the precise origin of the Carian alphabet of Egypt.

104) A very interesting consequence of the improvement in our knowledge of Carian alphabetic variants is that the number of different letters found in each alphabetic variety barely exceeds 30: the highest number documented for a specific alphabet is currently 31, in Memphis. In Kaunos we find 29 letters. For the rest of Carian alphabets from Caria proper, the available inventories are incomplete, but the total number of different letters must have been very similar. The old image of a writing system with a surprisingly high number of letters, which even led some scholars to assume that the writing was semisyllabic, has proved to be wrong. So in this regard the Carian alphabet is not particularly different from (for instance) the Lycian alphabet (29 letters).

115) The most remarkable peculiarity of the Carian alphabet is undoubtedly the aberrant nature of the sound values attributed to many letters of an apparent Greek origin: F is/r/, is /t/, N is /m/, is /m/, Ψ is /n/, etc. Moreover, basic Greek letters, such as Κ or Π are absent, and it is clear that the Carian letter T does not represent the basic sound /t/, as would be expected from a Greek point of view: this letter, still undeciphered, has a very limited use, and the value /t/ is clearly represented by the letter with the shape of qoppa (and with a sign with the shape of rho P in the Kaunian alphabet).

126) This aberrant attribution of sound values to letters of a clear Greek form is a strong argument in favour of a monogenesis of the Carian alphabet: despite the differences between the diverse Carian local scripts, all of them share the same anomalies in the sound values. For instance, ny is always m, qoppa is always t, etc., regardless of the local variety concerned. Moreover, in a very surprising way, we find that the variation across local scripts in the shape of some letters exactly mirrors the variation found in the diverse modalities of the Greek alphabet: the archaic digamma C and the more standard digamma F are used in their respective Carian local varieties for the same aberrant sound value /r/, and the same can be said of the two variants of the letter theta, ⴲ and ʘ (both used to denote q). We can therefore speak of a ‘consistent aberration’ in the use of the Carian letters.

The decipherment of the language

The indirect sources

13As in other ancient fragmentary languages, the glosses (lexical items with their meaning) and the proper names transmitted by other languages as Carian or in contexts that point to their Carian character must be taken in account. However, the value of each of these two indirect sources is radically different in the case of Carian: while the Carian glosses are constituted only by a handful of words, none of which is directly attested in the present corpus of Carian inscriptions7, Carian proper names in classical sources are relatively abundant, and the identification of a good number of them in Carian texts is one of the most striking achievements of the decipherment. To understand the importance of onomastics in analysing Carian inscriptions we should bear in mind that a high percentage of these contain proper names, due to their typological characteristics (funerary texts dealing with the deceased and his/her genealogy, graffiti consisting of ‘signatures’ of Carian visitors, votive inscriptions including the name of the donor, lists of priests, and so on).

14In principle, the contribution of onomastic analysis to the decipherment of the language should be treated with some reticence. The names do not mean anything: they are used only to designate and differentiate individuals, not to represent objects or actions. Moreover, onomastics is not a sure indication of genetic affiliation of the language where the names are used, because names of very different origin may be used (this is so in our modern languages). However, in the case of Carian onomastics, the evidence points to a clear link between Carian and the Anatolian languages: thanks to painstaking research over the past hundred years, the panorama of Anatolian onomastics now allows us to identify many indigenous names as typically Anatolian, including a considerable number of Carian ones. To quote only a few examples, in msnord-/ Μασσανωραδα we recognize the ‘Luwic’8 word for ‘god’, CLuw. māssan(i)-, Lyc. maha(na)-, Mil. masa-; similarly, the forms para-/ Παρα-, šar-/Σαρ-, which appear as the first elements in some names, strongly recall the adverbial stems Hitt. para, CLuw. parī, HLuw. pa+ ra/i ‘forth, away’, Lyc. pri ‘forth, in front’ and Hitt. šara, šēr, Luw. šarri, šarri, Lyc. hri ‘above, up’, respectively.

15Fig. 3 displays some instances of onomastic identifications obtained by the new decipherment.

The bilingual inscriptions

16For the linguistic interpretation of an unknown language, the study of bilingual texts is usually the first step. Though in the case of Carian some exist, the results that we can draw from them are unfortunately very limited: the amount of bilingual documents (in Greek and Carian, and in Egyptian and Carian) is very small, and, examined one by one, the information they provide is not particularly significant, due to their singular characteristics.

17The Egypto-Carian bilinguals consist mostly of simple onomastic formulae where the correspondences are limited to proper names. They represent a powerful instrument for the decipherment of the script, as already mentioned, but they are of little help in the linguistic interpretation of Carian. They cannot even identify the Carian word for ‘son’, which has been reliably established using other procedures (see below). In the few examples where the word is used in the Egyptian text as a part of the onomastic formula, no corresponding word appears in Carian; it only serves to aid our analysis of the relation between two proper names, corroborating the interpretation of -ś as a genitive mark. The only exception to these very poor results is the possibility of identifying the Carian word for ‘interpreter, dragoman’: in the inscription E. Me 8, the Carian phrase paraeym armon-i seems to correspond literally to Egyptian Prjm p3wḥm ‘Prjm the interpreter’9; therefore, armon seems to be the Carian word for ‘interpreter’. No satisfactory etymological explanation has been given for this Carian word in the framework of Anatolian10, but it is worth mentioning the striking similarity with Greek ἑρμηνεύς (Dorian ἑρμᾱνεύς) ‘interpreter’, a word of unknown etymology, but for which an origin of Asia Minor has been proposed (see DELG s. v.). Could ἑρμηνεύς be a Carian loanword in Greek, or, at least, could ἑρμηνεύς and armon-have a common origin from a third, presumably Aegean-Anatolian language ? Both hypotheses are very attractive, though neither can be demonstrated at present.

18As for the bilinguals from outside Egypt, perhaps the clearest document is the regrettably very brief Greek-Carian bilingual funerary inscription from Athens (G. 1). I have repeatedly suggested that there is a literal correspondence between the first Greek line and the sole line in Carian11:

19Therefore, śjas would be a word meaning ‘funerary monument’, while san would be a demonstrative referring to śjas, in a relation identical to Greek τόδε regarding σῆμα. Both identifications, translated to other contexts, have produced very consistent results: a very similar form to śjas, śas, appears in a brief inscription from Euromos (C. Eu 1), followed by a typical onomastic formula ‘X son of X’. The funerary character of the inscription seems very probable – although the stone was found reused:

20As for san as demonstrative pronoun, in an article published in 1993, Melchert connected this form with the word snn which appears in two inscriptions on phiales (C. xx 1, C. Ha 1), in both cases accompanying a word orkn whose simplest interpretation is a noun referring to the phiale, so that snn orkn, orkn... snn admits a good translation as ‘this phiale’ (Melchert 1993: 79-80; see below). san, snn can also be connected with another form: sa, documented in a funerary stele from Saqqâra. In the Saqqâra corpus, upe/upa seems to be the word for ‘(funerary) stele’, and in one case it appears as upesa (E. Me 26), so that an interpretation of this later sequence as upe + sa ‘this stele’ makes good sense.

21Only in one other case do we know for sure that we are dealing with a true bilingual inscription, i. e. two versions, one in Carian and one in Greek, of the same text. This is the well-known Greek-Carian bilingual inscription of Kaunos, found in 1996 (C. Ka 5):

22The two Athenian citizens mentioned in the Greek part – both honoured as proxenoi and euergetai of the Kaunians – also appear in the Carian part. Moreover, the beginning of the Greek text, with the typical formula ‘it seemed good to the Kaunians’ (ἔδοξε Καυν[ί]οις) finds a clear parallel in the presence of the indigenous name of Kaunos, kbid-in the first line of the Carian part12. These patent correspondences between the texts were greeted with enthusiasm by scholars who realized the implications of this text for a better understanding of Carian when it appeared in Christian Marek and Peter Frei’s exemplary publication13. Once the initial fervour had died down, the Kaunos bilingual has turned out to be a disappointingly modest tool for the decipherment of the Carian language. The reasons for these relatively poor results are simple: (1) the use of scriptio continua, together with a (typically for Carian) defective notation of vowels, seriously hampers the word segmentation and subsequent morphosyntactic analysis; (2) our practically total ignorance of Carian verb conjugation prevents us from identifying the verbs in the text: we are not even able to ascertain whether the Carian part reflects literally the structure impersonal verb + infinitive clause of the Greek section, or whether a different construction has been employed; (3) the Greek text is broken immediately after the proclamation of the two honourees as proxenoi and euergetes, so that it is impossible to take the comparison of both texts further; (4) the central part of the Carian text is damaged; (5) a high number of words in the Carian part are hapax legomena, so that no parallels can be traced with other Carian texts.

23These difficulties suppose that at the present the only clear parallelisms that are commonly accepted, apart from the onomastic equivalences, are the equations sarniš (or perhaps the entire sequences sarniš mdoτ₂ or sarniš mdοτ₂un) = προξένους, and unδo[ ]tλš (or simply δo[ ]τλš) = εὐεργέτας. Two other equivalences are less widely accepted, but certain scholars support them: i[ ]inis δrual as the equivalence to Greek ἐπὶ δημιο[υ]ργοῦ Ἱπποσθένους ‘under the demiurge Hipposthenes’, and the word otrš as corresponding to Greek [αὐτο] ὺς ‘themselves’. Both claims are backed by attractive etymological connections: in i[ ]inis δrual, δ-, a sound that in Carian commonly represents an older consonantal cluster * -nd-, recalls Lycian ñte /nde/ ‘in’, so that i[ ]inis δ=rual would mean ‘in (δ =) the magistracy of Hipposthenes’. As for otrš, it could be the Carian counterpart of Lycian atra-/atla- ‘person, self’. It is therefore probable that the sequence otrš sb at[ms]kmt is equivalent to [αὐτο]ὺς καὶ ἐκγόνους ‘themselves and (their) descendants’ in the Greek part (on the integration of at[ms]kmt, see below). There are also other, much more fragile, proposals of interpretation of this bilingual text which need not detain us here.

24Despite the paucity of the results, the Kaunos bilingual inscription provides us with some very interesting morphological information: we are able to identify an acc. sg. ending – n (already recognized prior to the discovery of the bilingual), a gen. sg. ending (or possessive suffix) – s(-?), an acc. pl. ending – š and a suffix for ethnonyms – yn- (in kbd-yn-š ‘Kaunians’). All these morphological elements are consistent with a close relationship of Carian to the so-called ‘Luwic’ dialects of the Anatolian Indo-European linguistic group (Hieroglyphic and Cuneiform Luwian, Lycian and Milyan). While the endings – n – š and – s are not specifically Luwic – they could be interpreted as generically Anatolian or even Indo-European – the suffix – yn- is clearly the Carian counterpart of CLuw. – wanni-, HLuw. – wani-, Lyc. – ñni, Mil. – wñni-, a specific Luwic suffix. Also important is the confirmation that sb is a coordinative conjunction (‘and’= Greek καί)14. In this case there are good Luwic parallels as well: Lycian se and particularly Milyan sebe (both ‘and’).

25The rest of the Carian-Greek bilinguals offer even more limited results. From the outset, we should discount an inscription from Stratonikeia, where the Greek and Carian texts are presumably of different times and show no connection, and an excessively fragmentary inscription from Kindye. The remaining bilinguals are those from the sanctuary of Sinuri, from Hyllarima and from Kildara. I will discuss them briefly now.

26In the case of the Sinuri inscription (C.Si 2), the fragment of the stone shows the last lines – only partially conserved – of a Greek text, followed by the beginning of a long text in Carian whose four first lines are quite readable, but the following ones are badly damaged. In principle, no possible connection between the two texts can be envisaged, due to the very scarce remains of the Greek, and to the fact that we have the end of a text and the beginning of another, meaning that the parts that could be related to each other are missing. However, thanks to Diether Schürr, we can go a little further in our analysis: some years ago, Schürr proposed that another stone fragment from the same site and containing two Greek texts enacted by the Carian dynasts Idrieus and Ada may have been part of the same stele as the Carian-Greek bilingual15. This hypothesis cannot be confirmed, because the bilingual fragment is now lost and the stones therefore cannot be physically compared. The proposed join between the two pieces is not direct – we must assume that intermediate fragments are missing – but the hypothesis is very attractive because we can establish clear parallelisms between the texts, though practically only at the level of onomastic identifications.

(Greek text:)
Robert 1945, n. 75:
[- - - - - - - -Ἰδριέως Ἑκατόμνω]
[καὶ Ἄ] δας Ἑκατόμνω κα[
[- - ?Σ]υεννιτων ἱερείος [
[- - ]νος Πονμοοννου
ἀτέλ]ειαν πάντων ἐπ[

[Ἰδ]ριεὺς Ἑκατό[μνω καὶ Ἄδα]
Ἑκατόμνω ἔ[δωκαν ?
ἀτέλειαν[
συνγε[ν
τις ο[
ουγ.[

Robert 1945, n. 74 + C.Si 2:

27The Carian text, read following the Ray-Schürr-Adiego system, begins with the names and filiations of the two dynasts, as do the two Greek sections in the other stone: [--] ryin tmñoś sb ada tmñoś ‘Idrieus (son) of Hekatomnos and Ada (daughter) of Hekatomnos’. Note that the indigenous name of the dynast Idrieus seems to have been clearly different, and note also the presence of sb as coordinative conjunction (see above). More important than this parallelism, which does not necessarily mean that the texts have the same content (this formula can head any dynastic decree; in fact, the other stone fragment presents the formula twice, heading possibly different decrees) are the forms Πονμοοννου (in the first decree of the joined stone) and pñmnnśñ. It is clearly an indigenous proper name (*Πονμοοννος = pñmnn-) that appears in Greek as genitive, and in Carian as – apparently – a ‘genitive in accusative’ (pñmnn-ś-ñ), a construction that recalls the similar formation in Lycian (-hñ). The name * Πονμοοννος could be, as Robert already suggested, an older form of Πορμουνος, the name of the syngeneia that appears from the end of the fourth century onwards in the charge of the sanctuary of Sinuri instead of the previous syngeneia of Pelekos.

28The certain connections between the Greek and the Carian parts of this recomposed stele end here: the Greek texts are too fragmentary to allow us to decide on the aim of the dynasts’ decrees. Both mention an ἀτέλεια, ‘tax-exemption’, and it is reasonable to suppose that a similar concession was recorded in the Carian text. It is to be expected that the initial reference to the Carian dynasts – apparently in the nominative – was followed by a plural verb in the sense of ‘to give’, as probably happened in the second Greek text of the other fragment (‘Idrieus...and Ada... gave....’), but any attempt to recognize such a form in Carian at present would be merely conjectural16.

29The bilingual inscription of Hyllarima (C.Hy 1) is a very unusual text, both from the point of view of the circumstances of its discovery and from the point of view of the content and structure of the inscription. The stele consists of two fragments: one was found in 1933 and published by Laumonier, the other seventy-one years later ( !), in 2004, and edited very recently by Debord, Varinlioğlu and myself17. Unlike the Sinuri bilingual, the join of both fragments is beyond question, as they fit together perfectly, meaning that the Carian part of a text is complete and practically without any problems for reading – a circumstance that is absolutely exceptional for a text of this length.

30Concerning the structure and content, the Hyllarima stone is a rather curious mélange of inscriptions, in Carian and Greek, of different times and on different subjects, although more or less clear connections can be established between them. According to Pierre Debord, we can establish the following order of engraving of the texts:

• The oldest layer is constituted by the two first lines in Carian, practically uninterpretable (see below, however), by three further lines in column A, also in Carian, containing three onomastic formulae, and by the two first lines of column B, in Greek, consisting of a heading ‘priests of all the gods’ and a sole onomastic formula (rather surprisingly, given that the heading is in the plural). According to Debord, all these sections seem to be made by the same hand.

• A second layer is represented by the third onomastic formula in Carian.

• A third layer consists of the two final onomastic formulae in Carian in column A and of the heading ‘priest (singular) of all the gods’ plus an onomastic formula in Greek in column B. This third layer is, following Debord, of a single hand.

• From here on, the only language used is Greek. The fourth layer is a list of Apollo’s priests in column B, datable to 263/262 thanks to the reference to the joint rule of Antiochus I and his son Antiochus.

• The fifth layer consist of two sales of priesthoods on the same day, datable c. 197, engraved in column B and on the left side respectively.

• The sixth and last layer consist of two leases of land, probably at the same time as the fourth layer, or at least not much later.

31Therefore, only the first, second and third layers are relevant for the interpretation of this bilingual, although the information contained in the later sections may also be of interest. Here, only these initial layers are reproduced:

32If, as I assume, Debord’s hypothesis about the layering of Carian and Greek is right, this poses interesting questions about the use of the languages. It is regrettable that the two first Carian lines remain uninterpretable, and that even the exact order in which they should be read is unknown (are we dealing with two long lines of a single heading, or rather should we read column A first and column B second? And, in the latter case, does each column have an independent heading, or is the heading equally common for both?). Our ignorance of this clearly hampers any interpretation, but in any case, the succession of the three layers reveals a very particular use of Greek and Carian, regardless of the exact structure and reading order of the columns: given that there is no clear correspondence between the Greek and Carian sections, and that both languages were used in the same layers, we cannot speak here either of a single text in two languages, or of a linguistic replacement of Carian by Greek, but rather of a contemporary use of the two languages for different tasks.

33The two first lines in Carian are, as already said, practically unintelligible. Only two (very interesting) forms admit a possible interpretation: armotrqδosq and msoτylarmiτ. In the first case, we can recognize beyond question two typical Anatolian names of gods: armo = Arma, the moon-god, and trqδ = Tarhunt-, the Tempest-God. arma-trqδos, in my analysis, is a dvandva compound, ‘Arma+ Tarhunt’. In msoτylarmiτ, a reference to Hyllarima is evident in the second word, whereas for msoτ a more dubious connection to Luwian massana-, Lycian maha(n) -, Milyan masa-(the words for ‘god’ in Luwic dialects already mentioned) is also attractive. msoτ ylarmiτ could therefore be a reference to the ‘Hyllarimean gods’. Taking the interpretation further, I have suggested that the entire sequence molš msoτ ylarmiτ could mean ‘priests of the Hyllarimean gods’, but this is merely a hypothesis (on molš see below).

34On the last bilingual inscription of our list, the inscription from Kildara (C.Ki 1), little can be said. In this case, the evidence is rather negative: no apparent connection can be drawn between the Carian and the Greek texts.

[.......(.)]zoλbaa[..(.)] kiλ[
[...]uδa[...] trqδimr qrds tazomδ[
kiλaraδ[–]ybzsdmTnmkδa[–] aTuq[
iasoum

Greek text:

ἔδοξε Κιλδαρεῦσιν, ἐκκλησίας γενομένης· Υσσ[ωλλωι?]
Σαμωου εὐεργέτηι γενομένωι Κιλδαρέων ἀτέ[λειαν]
δοῦναι καὶ προεδρίαν καὶ ἐσαγωγὴν καὶ ἐξαγω[γὴν]
καὶ ἐν εἰρήνηι ἀσυλεὶ καὶ ἀσπονδεὶ καὶ αὐτῶι
καὶ ἐκγόνοις· καὶ Κιλδαρέας εἶναι ἂν θέ[λωσιν?]

35In the Greek part, an individual named Uss[(very probably Uss[ollos], as the different editors of the inscription have suggested), son of Samoos, is honoured as a benefactor (εὐεργέτης) by the community of Kildarans, and some typical privileges are conceded to him. Neither Uss[ollos] nor Samoos are recognizable in the extant text of the Carian inscription. Conversely, in this latter text, the theonym trqδ ‘Tarhunt’ (see above, on Hyllarima), or a proper name containing it, is mentioned, and no counterpart can be found in the Greek decree. Finally, no connection can be traced between this Carian text and the bilingual of Kaunos, despite the fact that a connection would be imaginable if the Carian part of Kildara had the same content as the Greek one, given the similarities between this latter text and the Kaunos bilingual, both being honorary decrees.

36The only element common to the Greek and the Carian sections is the reference to Kildara and the Kildarans: in the Carian text, this place name appears complete in one case (kiλara), while in the other case only the initial letters are conserved (kiλ[ ]; it is therefore impossible to ascertain in this latter case if the reference is to the place name or rather to an ethnic name derived from it. To the sequence qrds, which reappears in the longest inscription of Kaunos, I will refer later on in my discussion of this inscription.

37The value of the bilingual texts as a tool for deciphering the Carian language – not the script – is therefore very modest. The most usable document, the Kaunos bilingual, produces very limited results, but it offers at least some interesting indications that Carian belongs to the Luwic group of Anatolian languages.

Combinatory, contextual and etymological methods in the study of Carian

38Apart from the use of bilinguals, three other methods can be employed in the case of a fragmentary language like Carian, whose membership of a specific concrete linguistic family is quite well established: combinatory analysis, contextual analysis and etymological analysis. The possibilities and limitations of these methods in the study of fragmentary languages are well known and illustrated by the history of the research in many ancient languages. The case of Carian is no exception.

39Combinatory analysis involves the observation of recurrent words and structures in different texts in order to establish their meaning and function. It depends mostly on the amount of materials available: the higher the number of comparable and contrastable structures, the easier it is to identify and explain their constituents.

40Contextual analysis considers all the factors that may help to understand the overall sense of a text and to search for the elements that one might expect in a text of a particular kind: for instance, if the funerary character of an inscription is well established, one must assume that the name of the deceased will be mentioned, possibly together with other words belonging to the typical phraseology of epitaphs (reference to the tomb, genealogical information about the deceased, possibly accompanied by kinship terms, and so on).

41Etymological analysis starts from the hypothesis, supported by clear indices, that the language in question belongs to a certain linguistic family, and seeks to recognize words and morphosyntactic elements through comparison. While the etymological approach promises to offer more precise results than the other two methods, long experience in the study of fragmentary languages demonstrates that the complementary use (when possible) of combinatory and contextual methods provides the most reliable information. Anything but an extremely careful use of etymology risks forcing a specific interpretation of the text that may be based on false etymological connections. A good example of this kind of danger can be observed in Ševoroškin’s works prior to the definitive decipherment of Carian, particularly his later ones: he provided etymological explanations and translations of many words and structures resulting from his own decipherment, which afterwards proved to be wrong. Recourse to etymology is only acceptable when the semantic and functional results of the comparison proposed correspond closely to the prospects created by combinatory and contextual analysis.

42The requirement of a large quantity of texts for a combinatory analysis greatly limits the possibilities of interpreting Carian inscriptions. In fact, only the subcorpus of funerary inscriptions of Saqqâra (Memphis) offers at least a satisfactory number of texts with recurrent words and structures that can allow us to obtain some interesting results. It is nonetheless unfortunate that the Carians’ epitaphs were so inexpressive: in many cases, the funerary inscriptions of Memphis contain almost exclusively an onomastic formula, only occasionally accompanied by words from the common lexicon. No verbs have been detected in any funerary inscription from this subcorpus. Other subcorpora, like the funerary inscriptions of Caria, or the votive texts, are even fewer in number and so the possibilities of a combinatory analysis decrease dramatically. In general, with the current state of our research combinatory analysis serves basically to distinguish between proper names and the rest, thanks to the progress made in the study of Carian onomastics. This is a rather limited achievement, but it can at least guide our interpretation of the content and aim of certain texts. A good example can be observed in the new inscription from Mylasa (C.My 1, Blümel-Kızıl 2004):

43idrayridsemδbq mol ty[
tsial tusolś : moi m[–]sao[
βanol paruosś : p?au paryriś
qzali obrbiś : tsial obrbiś
βanol yrqsoś : paryri psoirś
[–]bδo pnuśoś : myze trdyś
arkbiom qzaliś : śumo kbdmuś
skduδrotozś : pau toiś
[–]qo idyriś : ksbo iduśolś
[–]obiokliś : toi yrqsoś

44The inscription consists of ten lines. Lines 2 to 10 contain exclusively onomastic formulae of identical structure: an individual’s name followed by the father’s name in genitive. Note how each onomastic formula is separated by means of interpunction or by line skips, creating a harmonious effect. Only the first line contains common vocabulary. Although this line remains enigmatic, it is at least obvious that we are dealing with a list of names, preceded by a heading and comparable to the typologically similar inscriptions widely attested in Greek. It is practically impossible, however, to determine the class of individuals recorded here, but the presence of the word molš is interesting: this is the word which, as we mentioned before, appears in the heading of the inscription of Hyllarima, a list of priests. One could think of a word meaning ‘priests’, as I suggested above, and assume that the text of Mylasa is also a list of priests. We can see therefore how the isolation of personal names permits at least several thoughts about the overall sense of a certain text18.

45In the following pages, I will comment briefly on different typological subcorpora, paying particular attention to the most important, that of Saqqâra. Given the limits of this paper, I will not offer a comprehensive study. I will summarize the main results obtained from the study of these subcorpora and I will comment on some particular texts for the sake of illustration.

The funerary inscriptions of Saqqâra

46As said above, the funerary inscriptions from Saqqâra (Memphis) are in general very brief, which accounts for the fact that a very high percentage of the words they contain are simply proper names. Those acquainted with the Lycian epitaphs, which normally use complete sentences about the construction of the tomb or about its uses, or with many Etruscan tituli sepulcrales, which add abundant information on the age and the cursus honorum of the deceased, will be disappointed by the austerity of the Carians: the only information relevant for the Memphis funerary steles is basically the mention of the deceased. This mention may appear directly, i. e. expressed in the nominative case, so that the deceased’s name is the title of the stele, or indirectly, in a genitive (characterized by the ending –ś) or the so-called ‘dative’ case in -s (more probably, in my opinion, a possessive ending). In these latter cases, it is the stele itself that, explicitly or elliptically, carries the nominative case, i. e. the stele is the title itself, on which the deceased’s name depends (‘stele of...’). In the construction with genitive, a word for ‘stele’ can appear (in the construction with ‘dative’, for which we have a few examples, the word for ‘stele’ is always omitted). The following examples show the different constructions:

(a) Name in nominative
pnuśoλ/ zmuś i (E. Me 19) ‘Pnuśoλ (= Πονυσσωλλος) the (son) of Zmu-’

(b) Name in genitive + word for ‘stele’
artayś: upe: [... (E. Me 22) ‘Of Artay (cf. Αρταος) the stele...’

(c) Name in genitive (ś-ending)
ttbazi[ś] | piub[a] ziś | aor[ś] (E. Me 1) ‘Of Ttbazi, of Piub[a]zi-, (nephew ?) of Aor.’.

(d) Name in ‘dative’ (s- ending)
ntokris | dwśoλś | mwdonś i (E. Me 35) ‘Of/For Ntokri (= Eg. Νιτωκρις), daughter of Dwśoλ- (= Υδυσσωλλος), the mwdon- (on mwdon-, see below)’.

47In example (b) we can see that the word for ‘stele’ is upe. This is not the only form found in the Saqqâra subcorpus: wpe, upa, and ue are also found. While wpe is apparently simply a different spelling of the word upe, and upa seems to point to a vocalic alternation a/e for phonetic reasons, the connection between all these forms with p and the ‘p-less’ form ue is much less clear. Schürr has also proposed a phonetic basis for the difference, by assuming a tendency of p to disappear19, but the evidence for this view is insufficient, and we cannot conclusively rule out the possibility that these are different words. Leaving aside this problem, upe shows a striking similarity to Lycian χupa/kupa/ ‘tomb’, but the connection encounters phonological and semantic problems: we would have to accept an ad hoc loss of k in Carian, and also a metonymy in Lycian or in Carian. In Carian, upe seems to mean ‘stele’, whereas in Lycian the meaning of χupa is ‘tomb’ in the sense of funerary building (χupa is usually employed as the direct object of the verb prñnawa-‘to build’); this implies a change from ‘stele’ to ‘funerary construction’ in Lycian or, conversely, from ‘funerary construction’ to ‘stele’ in Carian.

48Combinatory analysis has proved successful in identifying the particular use of two forms that seem to be constituents of the paradigm of a single word: mdayn-mdaýn (nominative) and mwdonś (genitive). These words form part of the onomastic formulae, but their recurrence is too high for them to be interpreted as simple personal names. The most accepted opinion is that they should be interpreted as a kind of ethnic designation. In favour of this hypothesis, the new evidence of the Kaunos bilingual has been adduced: as mentioned above, in this inscription a suffix -yn- for building ethnic names, with clear cognates in the rest of Luwic languages, has been recognized in the word kbd-yn-š ‘Kaunians’, and this suffix may also be present in mda-yn/ mda-ýn. However, it is not clear how to interpret the stem mda-/ mwd(o) -, and very divergent proposals have been formulated: it could be a generic term for ‘foreigner’, or it could be the designation of the ‘Caro-memphites’ or, more specifically, it could indicate a precise place of origin. For this latter solution, diverse attempts have been made to find place names that could be related formally to the stem mda-/mwd(o) -, the two clearest candidates proposed until now being the place name Μύνδος and the term denoting inhabitants, Μύνδον) eς (in the Athenian Tribute List; perhaps from a place name Μύνδον-of unknown location, if it is not a variant form of the Carian city name Amyzon [Αμυζων]). I would add another possible candidate: Mylasa [Μυλασα]. Carruba and Hawkins have recently argued in favour of recognizing the place name of Cuneiform sources Mutamutassa (phonologically/Mudamudassa/) as the ancestor of the classical Mylasa20. If this hypothesis is accepted, the connection between mda-/mwd(o) - and Mutamutassa-Mylasa would be worth considering. It is also interesting that Koray Konuk has been able to recognize the initial letters in Carian of the name Mylasa on some coins that he attributes to Hyssaldomos, the father of Hekatomnos21. These letters are NE, a beginning that recalls Nº in NCO mwdonś. It is true that and E are different letters in the Carian alphabet of Egypt, but it is not impossible that E here is a rotated form of . For this equation m(w)d(o)- = Mylasa, an obstacle is posed by the fact that the alphabet of Mylasa, now well known thanks to the new inscription mentioned above, is very different from that of Egypt. But it could be answered that (1) a Mylasan origin of many Egypto-Carians of Memphis does not necessarily imply that the alphabet used in Egypt was the particular variant used in Mylasa, and (2) the alphabet of Mylasa revealed by the new inscription may be the result of a later evolution. It is particularly relevant that the letter E that appears in the Mylasan coin of Hyssaldomos mentioned is absent from the – almost complete – inventory of letters found in the new inscription of Mylasa.

49The identification of mdayn/mdaýn/mwdon-as an ethnic name opens the way for the search for other possible ethnic names in the Saqqâra subcorpus. Janda22 drew our attention to the possibility that some Carian proper names in Saqqâra inscriptions, very similar to Carian Greek place names and generically interpreted as individual names23, could actually be ethnic names. This idea was revived by Melchert and I have expanded on it in recent studies24. The following is a list of these possible ethnic names:

50After carefully analysing all these forms and their respective contexts, it seems that this hypothesis may well be true. Apart from the clear similarities between these forms and a series of Carian place names, a good argument may also be found in their distributive properties: they never appear in the first position, i. e. as the name of the deceased, or in the second position, i. e. as the name of the father. In simple threefold structures, they occupy the third position, as can be seen in the following examples:

pikreś ue šarwljatś msnordś (E. Me 3) = ‘Of Pikre (= Πιγρης) the stele, (the son) of Šarwljat (= * Σαρ-+ Υλιατος), the Masanoradean’

arlišś: upe: arlio/[mś] i: yjas[iś] (E. Me 9) = ‘Of Arliš (= Αρλισσις) the stele, the (son) of Arlio[m] (= Αρλιωμος), the Iasean’

arlišś/ursleś/kiδbsiś (E. Me 15) = ‘Of Arliš, (the son) of Ursle, the Kindyan’

punwśoλś: somneś/qýblsiś i (E. Me 21) = ‘Of Punwśoλ (= Πονυσσωλλος), (the son) of Somne (= Σωμνης), the Kublissean’

šayriq | parpeymś i/yiasi (E. Me 25) = ‘Šayriq (= Σαυριγος), the (son) of Parpeym, the Iasean’

šaruśoλ/pλeqś i: śuγλiś (E. Me 30) = ‘Šaruśoλ (= Σαρυσσωλλος) the (son) of Pλeq (= Πελδηκος), the Suangelan’

lýsiś | upe | šrquqś i | ksolbś (E. Me 43a) = ‘Of Lýsi (= Λυξης) the stele, the (son) of Šrquq (= * Σαρ+ Γυγος), the Kasolabean’

apmen šrquqś kojoλ i (E. Me 44a) = ‘Apmen (son) of Šrquq (= * Σαρ+ Γυγος), the Koan’

[--] j [- ] ś/[- ] owtś/i: msn/ordś (E. Me 48) = ‘Of [--] j [- ], the (son) of [- ] owt, the Masanoradean’

51These distributive properties are identical to those of mwdon-. See the following examples:

irow | pikarmś | mwdon ^!ś (E. Me 14) = ‘Irow, (son) of Pikarm- (= Πιγραμος), the mwdon-)’

sanuqś | ue | pntmunś i/mwdonś i (E. Me 28) = ‘Of Sanuq the stele, the (son) of Pntmun-, the mwdon-’

s [--] etś | [ue] | ynemoriś | mwdonś (E. Me 29) = ‘Of S [--] et [the stele], (the son) of Ynemori, the mwdon-’

52Leaving aside these possible ethnic names and the words for stele, the funerary steles of Memphis also offer other little details about the common lexicon: a connecting particle already seen in the preceding examples (i) and some kinship terms.

53For i, a very good etymological connection has been proposed which allows us to clarify its original function, but does not explain all its actual uses: i seems to be the Carian counterpart of the Anatolian relative pronoun Hitt., CLuw. kuiš, HLuw. kuis, Lycian ti, Milyan ki25. But in most cases in Carian it seems to have become a particle introducing nominal complements, in a process that recalls the evolution of the Old Persian relative haya, taya, used already in this language not only as a true relative pronoun but also as a particle introducing nominal complements, in a role that in some cases is closer to that of an article. As is well known, from this Old Persian relative comes the so-called ezafe of the New Persian, a particle that serves to link nominal elements. Some uses of Carian i are similar to those extended uses of Old Persian relative and of the subsequent ezafe26. The problem with Carian i is that in many cases we are not sure which elements is it connecting, due to the morpho-syntactic ambiguity of the structure and to the fact that i seems to be dispensable27.

54As for the possible kinship terms, the clearest and most widespread (present not only in Saqqâra, but also in other Carian inscriptions) is mno-, whose meaning ‘son’ seems beyond doubt. In fact, the most easily analysed examples are from the subcorpus of funerary inscriptions from Caria proper. We have already seen the inscription śas: ktais idyriś: mn[ ‘Funerary monument of/for Ktai [= Ἑκαταῖος], son of Idyri’(where mno-appears incomplete), which can be directly compared to the Memphis text E. Me 47:

tqtes | paraiβreλś i | mn[o- ?] ‘Of/for Tqte, the son of Paraiβreλ (= * Παρα-Ιμβαρηλδος)’.

55However, in most examples of mno-in Saqqâra, the word does not appear accompanying the name in second position (the father’s name). This has led scholars to suggest a range of solutions28. But a careful analysis of all these inscriptions demonstrates that the meaning ‘son’ is also possible, and that the particular location of the word is determined by other factors29. Observe the following examples:

pjabrm | wśoλś | mwdonś i/kbjomś | m [noś] (E. Me 12)

irow | pikraś i/semwś | mnoś/mwdonś i (E. Me 16)

irowś: psTým[–]ś/ pttuś: mnoś (E. Me 27)

56Apparently, the relationship of mnoś to the deceased’s name seems to be more ‘remote’ than simply that of ‘son’, given that it does not appear immediately before or after, and referring to, the (supposed) father’s name. But several contextual details serve to clarify the interpretation and to rule out the difficulties: the stele containing the first text (E. Me 12) shows the prothesis of a woman, so that pjabrm must be a feminine name (fig. 10).

57As for the second and third examples, irow is a name of Egyptian origin, where it appears used as both a masculine and a feminine personal name. Assuming that here it is used for a woman in both examples, the three texts admit a similar explanation if we consider that the name in second position is not the father’s name but the husband’s name:

‘Pjabrm, (wife) of Uśoλ (= Υσσωλλος), the mwdon-, the son of Kbjom (= Κεβιωμος)’

‘Irow, (wife) of Pikra (= Πιγρης), the son of Semw, the mwdon-’

‘Of Irow, (wife) of PsTwm[-], the son of Pttu-’

58If this interpretation is right, the use of mno-would be caused by the female character of the deceased: in some cases – not necessarily always – it was preferred to stress the genealogy of the husband in detriment to that of the wife. In fact, these three inscriptions, analysed in this way, become a female noun followed by a typical threefold (in the two first examples) or twofold (in the last one) onomastic formula.

59No clear etymological connection can be drawn for mno- ‘son’. It recalls the Hieroglyphic Luwian name for ‘son’, nimuwiza, but for this connection to be possible, we would have to accept a different word formation and a sort of metathesis in Carian or, conversely, in HLuwian. Assuming this etymological connection for a moment, given that HLuw. nimuwiza is commonly interpreted as meaning ‘without force’, ‘power-less’ > ‘child, infant’ or sim. (ni ‘not’, CLuw., Hitt. muwa-‘power, might’) (cf. the parallelism with Lat. infans ‘unable to speak’= ‘infant’), the metathesis might be assigned to Carian (*nm^o> mnº). However, it cannot be ruled out that the sequence m-n- was older, and that nimuwiza was the result of a popular etymology based on ni + muwa-.

60Two other kinship terms have been recognized in the Saqqâra subcorpus: ted ‘father’, and en ‘mother’. These identifications, formulated independently by Ivo Hajnal and Diether Schürr30, are primarily based on good etymological evidence, but they fit the contexts where both words appear very well:

iturowś | kbjomś | i en | mw[d] onś i (E. Me 32)

šýinś | upe | arie^?ś i ted (E. Me 38)

61Both contexts are very similar: the deceased’s name in the genitive (in E. Me 38 governed by the word upe ‘stele’, elliptical in E. Me 32) accompanied by a personal name in the genitive and followed by the ‘relative’ i and the respective forms en and ted. The only notable difference is that in E. Me 32 the typical word mw[donś] plus i is added. Assuming here a true use of i as relative and the two corresponding meanings ‘mother’ and ‘father’, a good translation can be offered:

‘Of Iturow-, who (is) the mother (en) of Kbjom, the mw[d]on-’

‘Of Šýin- the stele, who (is) the father (ted) of Arie^?’

62This analysis implies that iturow- is a feminine name, a circumstance that does not pose problems: Iturow is the Carian adaptation of an Egyptian name Jr. t=w-r.r=w (DemNb: 70) *[jəturōw], Greek Ἰθορως, documented for both men and women.

63The etymological connections of en and ted are very convincing: for en, compare CLuw. anniš Lycian ẽni, Lydian ẽnaś, ‘mother’. For ted, compare CLuw. tāti-, Lyc. tedi-, Lyd. taada-

64As for the rest of subcorpora that can be established on the basis of their typology, the low number of inscriptions in each of them limits dramatically the possibility for a combinatory analysis. I will review some of them very briefly.

Funerary inscriptions from caria proper

65In Caria, too, some funerary inscriptions have been found, but far fewer than in the subcorpus of Saqqâra. We have already seen a funerary inscription from Euromos, where a form śas, comparable to the śjas in the Athens bilingual funerary text, appears as a reference to the tomb. In a recent article, Olivier Henry reviews most of the funerary inscriptions of Caria proper, and I refer the reader to his work for some important remarks31. Here I will merely point out the presence of a stem s(i)δi- that appears in almost all these inscriptions:

sδi amτ[/ pauś/ art{ } mon (C.Tr 1)

an siδi a/ rtmi pauś/ parŋaq^? (C.Tr 2)

sδi a[-]mob[C.Al 1 (C.Al 1)

sñis: sδisa/s: psuśoλś/ malś: mnoś (C. Ka 1)

qoτ₂omu sδisa/ m^?nś šoδubrś/ sb mnoś knor/ noril^?ams (C.Kr 1)

66Given this spread of the stem s(i)δi-, the simplest solution is to interpret it as a noun referring to the tomb, or, at least, as a formulaic expression required by the funerary character of the text. We also recognize the word mno-ś ‘son’ (in genitive), which appears in two inscriptions, and some clear personal names, as pau-ś (genitive), artmon (nominative), artmi (nominative or rather dative ?), psuśoλś (genitive), šoδubrś (genitive) and perhaps also mal-ś (genitive) and qτ₂omu (nominative). But none of these texts can be interpreted completely, due to their fragmentary nature (C.Tr 1, C.Al 1), the difficulties of reading (C.Tr 2, and particularly C.Kr 1), the presence of hapax legomena (parŋaq^?, sñis, knor norilams) and, in the case of C.Kr 1, the complexity of the structure. The clearest is C.Ka 1, where sδisas psuśoλś malś mnoś admits a translation like ‘tomb of Psuśoλ, son of Mal’ (without totally ruling out tomb of Psuśoλ (and) of Mal, (his) son, see Henry 2007), although the meaning and function of sñis remain unclear, as does the precise internal analysis of sδisas.

Votive inscriptions

67Within the typological category of votive inscriptions we should include some texts on Pharaonic objects from Egypt and others on vessels found in Caria or of unknown origin. I will leave aside those inscriptions that consist almost exclusively of personal names and I will focus my attention on other, longer texts that offer more interesting, albeit limited, information (I exclude C.xx 3, which remains completely impenetrable to me):

šarkbiom: zidks mδane: ýn-?-/mo | δen: tumn (E.Sa 1, reliquary for mummified reptiles; Egypto-Carian bilingual. The Egyptian text sounds thus: Jtm nṯr ‘3 dj ‘nḫ snb Š3rkbym ‘Atum the great god may give life and health to Š3kbym’)

ntros: prŋidas / orša /nu mδane: uksi wrmś (E.xx 7, bronze lion from Egypt; precise origin unknown)

smδýbrs | psnλo | mδ orkn týn | snn (C.Ha 1, bronze phiale, possibly from Halikarnassos)

?] areš | šanne mλne | siykloś | šann | trquδe | λmuδ [? (C.Ia 3, krater found in Iasos)

šrquq | qtblemś | ýbt | snn | orkn | ntro | pjdl^? (C.xx 1, bronze phiale of unknown origin, presumably from Caria proper)

ýśbiks not: alosλ arnosδ: jzpe mδane (C.xx 2, bronze dinos of unknown origin, presumably from Caria proper)

68The clearest text is C.xx 1, thanks to the interpretation proposed some years ago by Craig Melchert32. This interpretation was the result of a wise combination of the methodological approaches mentioned throughout this paper (combinatory, contextual and etymological) and permits an almost complete and reliable translation of the text, a very exceptional circumstance in Carian studies. Although Melchert himself has changed his interpretation in later works, in my view his first proposal remains valid.

69The text begins with a typical onomastic formula in the nominative Šrquq Qtblem-ś, Šrquq (= *Σαρ-+ Γυγος), (son) of Qtblem (= Κυτβελημις, Κοτβελημος), undoubtedly the donor’s name. A word ýbt follows, which Melchert interpreted as a verb, comparing it with Lycian ubete ‘he offered’. The comparison is very sound, both formally (a third person singular is expected here) and semantically (a verb ‘to offer’ is certainly adequate for a votive object). In snn orkn a nominal phrase in the accusative (marked by – n) would represent the direct object, i. e. the phiale offered by Šrquq. We have already seen that snn admits a good interpretation as a demonstrative pronoun, insofar as it seems to be formed on the same pronominal stem as san in the Athens bilingual and sa in upe sa. orkn snn would mean, therefore, ‘this bowl’, and ork- would be the Carian word for ‘bowl, phiale’ or sim. On this point, it is interesting to recall the connection proposed independently by Günter Neumann and Edwin Brown of ork-to Greek ὕρχη ‘jar’ and Lat. orca ‘butt, tun’. Also for two other Latin words that represent vessels, urceus, urna ‘water-pot’, a common origin with ὕρχη and orca has been proposed (see Chantraine DELG, Ernout-Meillet DELL ss. vv.). It seems that we are dealing with an originally unique stem (of a Mediterranean substrate?) that made generic reference to liquid containers.

70‘Šrquq (son) of Qtblem offered this bowl’: to whom? Independently, Melchert and Schürr connected the word ntro that immediately follows to the name of a god, isolable in the Lycian name Natrbbijẽmi33. This name appears in the Greek-Lycian-Aramaic trilingual of the Lethoon of Xanthos, where it is ‘translated’ in Greek as Apollodotos, so that natr- is undoubtedly an indigenous theonym assimilated to Apollo. According to Melchert’s analysis in his 1993 paper, ntro would therefore be a dative of this god name: ‘to Apollo’. Finally, pjdl is the most obscure word in the text, but Melchert was at least able to suggest an attractive connection to CLuw. piya-, Lyc. pije- ‘to give’ and to think of a noun with the meaning ‘gift’ and carrying out the function of apposition to orkn. The overall sense of the inscription would then be: ‘Šrquq (son) of Qtblem offered this bowl to Apollo as a gift’.

71Regrettably, none of the remaining inscriptions on objects permits a similarly thorough analysis and translation. For instance, the other inscription on a phiale (C.Ha 1) offers a clear parallel to the Šrquq-inscription: it is no coincidence that the nominal phrase snn orkn ‘this bowl’ can also be recognized here, given that we are dealing with the same kind of object. But the rest of the inscription is not so easily analysable: smδýbrs is clearly a personal noun, but the final – s is ambiguous: it may be a mark of ‘s-dative’, but also the final consonant of a s-stem in the nominative, and the two interpretations lead to very different analyses. On the other hand, the nominal phrase snn orkn in the accusative demands in principle a transitive verb, like ýbt in the Šrquq-text, but here there is no form comparable to ýbt, and it is impossible to decide between the three possible candidates for the verb: psnλo, mδ and týn.

72Another example of our difficulties in interpreting these inscriptions is the word mδane which appears in three of the six texts, without ruling out the possibility that mλne in the Iasos inscription C.Ia 3 may be a variant of the same word. A similar sequence also appears in some Theban graffiti. This concurrence has led several authors to suggest that mδane may be a verbal form, but this interpretation is seriously hampered by the fact that a sequence mδa, which seems to be closely connected to mδane, appears three times in the five first lines of the bilingual inscription of Sinuri: I think that this repetition of the same verb in the same text would be very strange34. All these difficulties mean that our analysis of this subcorpus is limited to the identification of proper names: in the first text, the presence of the name Šarkbiom is clear, thanks to the Egyptian version; it is also probable that tumn is related to the Egyptian god name Atum. In E.xx 7, the onomastic formula is uksi wrmś, whereas also the Carian Apollo, ntro-is present here in an s-form (ntro-s) and accompanied by a word apparently in agreement (prŋida-s) for which Schürr has suggested an attractive connection to Βρανχίδαι, the priestly family in charge of the sanctuary of Apollo in Didyma35. In the Iasos krater C.Ia 3, the personal names seem to be areš and siykloś (in genitive), and a god name can also be identified, as I proposed some years ago36: trquδe = Tarḫunt-, the Anatolian Tempest God, now attested also in Hyllarima, as we have already seen. Finally, in the dinos-inscription C.xx 2, I have proposed a personal name ýśbiks, also documented in a slightly different spelling, yiś{}biksś37 in a stele from Saqqâra (E.Me 46), and to connect the form alosδ arnosδ (also documented in Saqqâra under the form alos Karnos (E.Me 45) to the well known Carian place name Halikarnassos38.

The Kaunos bilingual inscription

73I will finish this brief overview of different types of Carian inscriptions by mentioning the longest text in Carian known until now: the Kaunos inscription C.Ka 2. This is a paradigmatic example of the practically insurmountable difficulties caused by long texts, but also of how new findings of Carian inscriptions can improve, albeit modestly, our knowledge of already known documents. Before the discovery of the Kaunian bilingual, scholars were unable to say anything about this inscription and it remained almost totally impenetrable. Thanks to the Kaunian bilingual, which provides some clear parallelisms to C.Ka 2, this long inscription is beginning to reveal some of its secrets, although we are still far from understanding the overall sense of the text.

74This is the text of C.Ka 2, according to the recent revision made by Frei and Marek39:

75The inscription is in scriptio continua. The segmentations proposed here are to a great extent purely hypothetical, but many of them are supported by the parallelisms that can be drawn with the Kaunos bilingual. In bold face, I mark the words that are common to C.Ka 2 and the Kaunos bilingual. Underlined is the sequence qrds, coincident – as already mentioned – with the Kildara inscription.

76The first word of the text, [u^?]omλŋ, immediately recalls the second word of the Kaunos bilingual (uiomλn). We are unaware of the precise function and meaning of this word, but in the Kaunos bilingual it seems to be closely linked to the place name kbidn ‘Kaunos’, and may therefore be a verb or a substantive that would be in some way equivalent to ἔδοξε ‘it seemed good’ in the Greek part of the bilingual. Although C.Ka 2 also comes from Kaunos, there is no mention of this place name here. This opens up the possibility of an interesting, but undemonstrable, hypothesis: that the word qrds could be used instead of the specific reference to the city of Kaunos. qrds could therefore be a noun referring to the community: city, assembly, or the like40. This hypothesis is even more attractive in view of the presence of the same word in the inscription of Kildara, apparently a public inscription, and also of the presence of a word qrdsoλš in the same inscription C-Ka 2.

77In lines 4 and 7, we have another word that is now known thanks to the Kaunos bilingual: šarniš. There it was considered as an accusative plural of the word equivalent to Greek προξένους. Its presence in C.Ka 2 is therefore of particular importance for ascertaining the type of text we are dealing with: it must also be a proxeny decree or, at least, an official document dealing with proxenoi (in plural). Equally significant is the identification of another sequence with a clear parallel in the Kaunos bilingual:

Kaunos bilingual:	otrš sb at[ms]/kmt
C.Ka 2:	otrš bi sb atmsk[m -]δ bi

78(Note that the integration is based precisely on this parallelism; in any case, it seems to be right.)

79We have already mentioned that the formula otrš sb at[ms]/kmt seems to correspond to Greek [αὐτο]ὺς καὶ ἐκγόνους ‘themselves and their descendants’(acc.) in the bilingual. If this equivalence is right, a similar reference to ‘themselves’– very probably the proxenoi – and their descendants must be assumed in C.Ka 2. However, the construction is not totally identical: in C.Ka 2 we see the repetition of bi, following each of the two members of the sequence. Its function is not clear: we could assume that they are either reinforcing particles built in correlation (in the sense of ‘as well as’, ‘both... and’, or similar) or postpositions (with the meaning ‘for’, for instance).

80The parallelisms between the inscriptions are therefore undeniable, and we can suspect that C.Ka 2 is also a proxeny decree. However, it is surprising that no onomastic formulae can be easily recognized here: we would expect at least two individuals (and, following the Carian practice, accompanied at least by their respective father’s names) to be mentioned, given that sarniš, proxenoi, is a plural noun. Following from the parallelisms between both inscriptions, these names would be in the accusative. None of the forms existing between the beginning [ui^?] omλŋ qrds and the first mention of sarniš meets all these requirements. A possible solution is to assume that the sequence [-]arλanoŋ sb zšariosŋ is formed by two personal names, with – ŋ as an alternative accusative mark (note that the precise value of this letter is controversial) and without mentioning their father’s names. But there is no explanation for these names from the point of view of Carian onomastics. An alternative explanation of the apparent lack of onomastic formulae would be to assume that the inscription C.Ka 2 is not properly a decree conceding proxeny to specific individuals, but rather a decree of a more general character, regulating proxeny privileges or the like.

Conclusions

81Though this latter analysis suggests that our present knowledge of Carian is very limited, I would not like to end this paper on a pessimistic note. A more balanced view must be drawn, taking account not only of the difficulties and uncertainties facing any linguistic interpretation, but also of what are, in my opinion, the very important results reached by the research on Carian in recent years:

• Despite some doubts about several letters, the Carian writing system can now be considered to be fully deciphered. We are in a position to transcribe Carian texts and to establish how Carian really sounded. If we compare the current situation with the state of the research thirty years ago, the progress in Carian studies has been dramatic.

• Parallel to the decipherment of the writing, we have attained a knowledge of the Carian alphabet, its functioning and its local varieties that was unimaginable some years ago.

• The decipherment of Carian challenges the views about the diffusion of the alphabet that have been accepted until now. Whatever the exact process of genesis of this aberrant alphabetic system, it was necessarily very different from other known processes.

• We are able to identify and in many cases analyse correctly the proper names that appear in the Carian inscriptions. Bearing in mind that the Carian texts are constituted in their majority by onomastic formulae, this means that we are able to interpret a substantial proportion of them, and, conversely, to localize sequences formed by the common lexicon.

• The analysis of proper names, both in direct and indirect sources, offers a consistent image of Carian as an Anatolian language, at least as regards the onomastic system.

• The morphological and lexical elements recognizable in the Carian texts, although limited, are enough to establish the genetic affiliation of Carian to the so-called ‘Luwic’ dialects (Luwian, Lycian, Milyan) of the Anatolian branch of the Indo-European linguistic family.

• The prospects for progressing further in the interpretation of Carian texts are seriously hampered by the scarcity of available materials. Whereas the etymological method can be used only with extreme caution for the reasons outlined above, and bilingual texts offer modest results, the use of combinatory and contextual tools is hindered by the low number of typologically related inscriptions. Only an increase in the documentation will allow us to make new breakthroughs in the linguistic decipherment of Carian.