Where did one speak luwili? Geographic and linguistic diversity of Luwian cuneiform texts

More intriguing is the non-trivial structural resemblance between the rituals attributed to Kuwattalla and Silalluhi, on the one hand, and the Mastigga tradition, on the other hand. Mas- tigga was known as “woman from Kizzuwadna”, and the Kizzuwadna features of the Mastigga tradition received extensive coverage in Miller 2004. We argue in detail in our forthcoming edition of the Kuwattalla tradition that the absolute majority of rites characterising Mas- tigga's ritual against domestic quarrels (CTH 404.1) find their counterparts in this corpus. Here we will perform the opposite test, namely a brief analysis of CTH 761.1 under the prism of its possible parallels with CTH 404.1 (its division into paragraphs adopted in this paper follows Miller 2004).

The identified rites of CTH 761.1 include the manipulations with a symbol of miasma made of paste (KUB 32.10+), substitution rites involving a sheep and two additional animals (KUB 32.9(+) rev. 1'-14', KUB 35.24+ obv. 1-36", and KBo 29.15), and the presentation of a pot with vegetable soup (KUB 32.9(+) obv. 19-34). They are punctuated by recurrent purification rites involving manipulations with a lump of dough and ablution with water (for the sequence, see KUB 32.9(+) rev. 15'-34'). All these building blocks of the “Great Ritual” have likely counterparts in CTH 404.1, especially if one interprets them with the help of the later versions of the Kuwattalla tradition.

The initial part of this rite (§§ 8-10, 15-17) features manipulations with human figurines, symbolising sources of witchcraft, as well as hands and tongues made of paste, symbolising its impact. Although the type of the paste object in KUB 32.10+ remains unclear, both the anthropomorphic figurines and body parts made of paste are deployed in the same functions in the later versions of the Kuwattalla tradition. The mammals used as substitutes in CTH 404.1 include sheep (§§ 20-21), black sheep (§§ 22-23), piglet (§§ 24-25) and puppy (§ 30). While the identification of animals in CTH 761.1 present difficulties, we learn from (4-5) that the Kuwat- talla tradition availed itself of the white and black sheep, while Bo 4388 adds the piglet and puppy to the ritual inventory. As mentioned earlier in this subsection, the formula “it will not become seed” bridges the presentation of pots in the two tradition. In both cases the seeds' failure to realise their procreative function is presumably compared with the failure of witchcraft. But the most significant parallel arguably involves the purification rites: the combined purification with water and dough is likewise repeatedly used in CTH 404.1 (§§ 19, 31).

When assessing the similarities between the Kuwattalla and Mastigga traditions, it is important to keep in mind that they primarily concern the Hittite frame of the rituals. This is in stark contrast to the parallels between the Kuwattalla and Tunnawiya traditions, which primarily involve Luwian incantations. One way to interpret this discrepancy in line with our previous findings is to assume that Kuwattalla's “Great Ritual” was tailor-made in Hattusa to satisfy the taste of her royal patrons and perhaps other members of the local elites. This implied minimal interference with the subject matter of the incantations, especially given the fact that the Luwian language may not have been commonly understood at the time. The main focus must have been on the adjustment of the ritual's subject matter and its performative aspects. Therefore, if our hypothesis holds, the Luwian insertions reflecting the best practices of the Lower Land could coexist with Kizzuwadna templates within the Kuwattalla tradition from the beginning of its written attestation. Going back to the nakkussi-rite, its presence in CTH 404.1 enhances the probability that its counterpart was also present in CTH 761.1. This is not, however, the same thing as to claim that the earliest version of the Kuwattalla tradition featured the reflexes of Hurr. *nakkosse `release'. As a parallel, one can consider the Arzawa rituals, which frequently feature scapegoats but refer to them without resorting to Hurrian loanwords.

One item that remains unexplained by the proposed scenario is /xalal(i)-/ `pure', a West Semitic loanword in Luwian (del Olmo Lete and Sanmartin 2015: 354-355). This adjective is common to the texts of the Kuwattalla and Puriyanni traditions, including their oldest versions. One has no doubt that this lexeme was well integrated in the Luwian dialect under discussion, in particular because of the attested derivative /xalalanussa-/ `to purify'. It is, however, possible to question the relevance of this lexeme for determining the geographic background of the respective rituals. In KBo 11.2 i 10, we also find ha-la-li-en-zi as a foreign word in a Hittite text, endowed with the characteristic Empire Luwian ending acc.sg /-ntsi/ (Yakubovich 2010: 30). Therefore, in the second millennium BCE this Luwian lexeme probably functioned as the standard (pan-dialectal) equivalent of Hittite parkui- `pure'. The way the West Semitic adjective hl found its way into the Luwian language remains to be investigated, but this problem should be kept apart from Hurrian borrowings into a specific Luwian dialect.

Summing up, the presence of Hurrian loanwords and other Kizzuwadna features in the Kuwattalla tradition is undeniable, but most of them can be accounted for within the context of its evolution at the court of Hattusa. These findings need not contradict the hypothesis of inherited similarity between the rituals attributed to Kuwattalla and those from the Lower Land, which primarily manifests itself at the level of Luwian incantations.

2. The Tradition of Taurisa

2.1 The Taurisa Triad

The idea that certain Luwian incantations are connected with the area of Taurisa is in itself not new. Thus, Taracha (2009: 100) observes: “Central Anatolia was inhabited ... by Luwian population groups which had gained dominance in some centers, thus prompting changes in local beliefs. Among the gods of Taurisa, a town which ... should be located in the Zuliya/Зekerek basin, there are the Luwian Sun-god Tiwad and Kamrusepa appearing as the parents of the local LAMMA god with the Luwian epithet /wasxatsa-/ `(most) precious'.” Nonetheless, the implications of this observation for the spread of the Luwian language appear not to have been systematically pursued. This is rather unfortunate, because the valley of Zuliya/Зekerek is located to the east/northeast of Hattusa and to the north of the upper valley of Kizilirmak known as the Upper Land in Hittite sources (Corti 2017a: 237). Whether or not one may wish to refer to this area as Central, Northern, or Eastern Anatolia, this is clearly not a region that is prototypically associated with the Luwians, and yet it emerges as the likely birthplace of a group of Luwian incantations.

It is appropriate to review the facts that offer the philological justification of this conclusion. The theonym ^URUTa-u-ri-si-iz-za-as `(god) of Taurisa' (KUB 35.107 iii 10') appears immediately above the Luwian narrative about the diseases, who were not invited to a banquet organised by the Sun-god and took offense (CTH 764). A gloss written above the line defines Ta-u-ri- si-iz-za-as as wa-as-ha-az<-za-as><sup>D</sup>LAMMA-as <<za-as>>`Most Precious Tutelary God' (vel sim.). The interpretation of Luw./wasxatsa-/ as `(most) precious' follows from the functional identity of Luw.in- za-ga-a-an wa-as-ha KUB 35.54 ii 32' and KЩ.BABBAR-an KЩ.GI-a[n] `silver (and) gold' in KUB 35.52+:9'. As became increasingly clear with the join KBo 29.2 + KUB 35.52 (Sasseville 2021: 553-554), one is dealing here with two paral-lel versions of the Puriyanni ritual for the purification of the house (CTH 758), which sometimes diverge in their grammar but not in their substance. We interpret Luw. /wasxa-/ as `treasure' and /wasxatsa-/ as its derived adjective, possibly with elative connotations, for which see Yakubovich 2013b. The earlier interpretation of /wasxatsa-/ as `sanc-tified, holy' is reflected in the citation from Taracha 2009 immediately above, cf. also the translation `patron' (Mel- chert 2015: 410). The choice between these two alternatives does not impact the main claims of the present paper.Due to the join made by Jared Miller we have learned that the fragment KUB 35.107 belongs to the same tablet as the matching Hittite narrative about the offended deity (KBo 43.223 + KBo 9.127 + KUB 36.41). The main difference of the Hittite account is that in this case the offended deity is the “Great Deity” (DINGIR-LUM RA-BU-U). But the Protective god of Taurisa is also mentioned in the introduction to the Hittite account, and in fact the respective Hittite and Lu- wian sentences display a close match.

Several more fragments mention the Tutelary God in the Luwian context. The most informative one among them for our purposes is KBo 8.130 + KBo 29.25, classified now as CTH 766 following the suggestion of David Sasseville, who is also responsible for the join. There we find the dative phrase KBo 8.130+ rev. 9' [^URU] Ta-ы-ri-si-iz-za^DLAMMA-ya `to the Tutelary God of Taurisa', as well as the collocation KUB 8.130+ obv. 17 [...] <sup>D</sup>LAMMA-ya-as <sup>MUNUS</sup>AMA-ni <sup>D</sup>Kam-r[u-se-pa-i] `the Tutelary God to Kamrusepa, (his) mother', which introduces the third member of the Taurisa triad. But the occurrences of KUB 35.103(+) iii 4' [...] ^DLAMMA-ya in the pregnancy incantation (CTH 766) and KBo 13.260 i 35 [...] ^DLAMMA-ya-a[s] in the incantation for a sick child (CTH 765) are no less important, because they provide arguments for linking the respective large fragments to the Taurisa tradition. Since no specific protective gods other than that of Taurisa appear in Luwian cuneiform incantations, one can make an educated guess that the fragmentary attestations of ^DLAMMA in CTH 764-766 all refer to the same deity, while several forms of the adjective Ta-ы-ri-si-iz-za-are probably lost in the lacunae.

The occurrences of the goddess Kamrusepa in Luwian passages are likewise compatible with the assumption that she is invariably mentioned there as the patron goddess of Taurisa. In addition to the fragments discussed above it also occurs in KUB 35.108(+) (CTH 766) and KUB 35.88 (CTH 765). The fragment KUB 35.108(+) does not contain a single complete line, but the juxtaposition of ^DTi-wa-d[a-...] and ^DKam-ru-se-pa-as-si-is in lines 5' and 6' suggests that the two deities are mentioned together as heads of the Taurisa pantheon. The juxtaposition of ^DKam-ru-si-pa-as in KUB 35.88 iii 9' and ^DZu-li-ya-ya-a[n] in KUB 35.88 iii 10' is no less telling. The deity Zuliya occurs several times in the description of the rites for the Tutelary God of Taurisa in a version of the AN.TAH.SUM Festival (Galmarini 2015: 53). In fact, it is almost certain that Zuliya is a river goddess (cf. Haas 1994: 452), which brings us back to the assumed location of Taurisa on the river Zuliya. If the Luwian form ^DZu-li-ya-ya-a[n] does not contain a dittography, it may well represent a substantivised possessive adjective in /-ja/i-/ derived from the hydronym.

Furthermore, the references to Kamrusepa in Hittite texts containing Luwian fragments also form a uniform cluster, which is equally compatible with the Taurisa connection. Before the story of the angry Great Deity is told in Hittite in CTH 464.I.A, the Sun-god addresses Kamrusepa a question i-ni-wa ku-it `What is that?' (KBo 43.223+ i 19', Steitler 2017: 388). In KBo 12.89 iii 9' (CTH 765), it is Kamrusepa who sees something from the sky and then poses the same question i-ni-ma-wa ku-it `But what is that?', which is answered by another narrative about the divine banquet and an offended supernatural being. In KBo 12.100 iii 12-13 (CTH 765) Kamrusepa apparently sees again something from the sky, after which one can restore the sentence wa-as-ha-i[s^DLAMMA ^URUTA-U-RI-SA] A-NA^DUTU-wa tar-kum-mi-ya-[u-wa-an-zi ti-ya-at] `The Most Precious Tutelary God of Taurisa began to explain to the Sun-god' (cf. already Hutter 2003: 257). Needless to say, the sentence thus restored represents a paraphrase of (14). External parallels to these formulaic passages are addressed below in 3.3.Finally, KUB 35.90 is too fragmentary for any restorations, but even here the mention of Kamrusepa in line 5' is followed by the possessive adjective ^DUTU-sa-an-za-a[n] in line 7'.

It emerges from this discussion that the bulk of the fragments traditionally listed under CTH 764-766 form a closely-knit group, which exhibits connection with a particular divine triad. This fact was not always emphasised in the previous Hittitological discourse. Thus Hutter (2003: 231) addresses the function of Kamrusepa in the passages mentioned above together with the rituals collected on the SammeltafelKUB 7.1 + KBo 3.8, where she is mentioned in a company of the goddess Maliya and the god Pirwa. Making an additional step in the same direction, Yakubovich (2010: 23) views the possible Nesite affinities of Kamrusepa as a potential obstacle to localising the origin of CTH 764. Yet, once one steps back from attempts to generalise over isolated divine names and pays due attention to the systemic similarities, the segregation of texts connected with the triad of the Sun-god, Kamrusepa, and the Tutelary god of Taurisa ceases to cause difficulties (cf. already Starke 1985: 203). Another potential difficulty mentioned in Yakubovich 2010: 23 is the occurrence of DNu-ы-pa-ti-ga-asin KUB 35.108(+) iv 12' (CTH 764). While there is no doubt that the god Nubadig has the Hurrian origin, the occur-rence of this theonym in what apparently represents a formulaic divine list may well represent secondary influence. One can compare an equally isolated mention of hurlas DInar `Inar of the Hurrians' in the Istanuwa tradition (KUB 35.135 iv 16'). For Hurrian influence on the religion of Hattusa in the Early New Kingdom period, cf. also 2.2 above. As we shall see below, there are independent reasons to think that Kamrusepa has nothing to do with Nesa but represents an avatar of the Hattian goddess Katahzifuri in the texts under discussion.

The content of the fragments listed under CTH 764-766 is sharply different from the ritualistic traditions addressed in Section Two. Here we are mainly dealing not with rituals as such, but rather with incantations presented separately from the description of non-verbal acts, e.g. offerings, whether or not they had to be accompanied by such. Accordingly, they are introduced as Hitt. hukmais `conjuration' or as its Akkadographic equivalent SIPAT, but not as SISKUR `ritual'. Furthermore, we do not have a single mention of a specific performer in connection with this group of texts: apparently the incantations mentioning members of the Tau- risa trial were treated as folklore and did not require fixed authorship. As for their function, most of the texts grouped under CTH 764-766 concentrate on ensuring successful childbirth and fighting children's diseases. The common designation of patients in these incantations is DUMU.NAM.LЬ.U19.LU `human child' / `human being'. With regard to their form, several of them contain historiolae, the best understood of which are the above-mentioned narratives of the divine banquet. The combination of these features clearly sets CTH 764-766 apart from the rest of the Luwian corpus, a conclusion that is anticipated in Starke 1985. This generalisation is not meant to undermine the hypothesis that some of the incantations grouped under CTH 767 ultimately belong to the same tradition as CTH 764-766. The texts of this group, which certainly require fur-ther study, contain isolated Luwian code-switches, whereas our present paper focuses on longer Luwian utterances. It is also worth mentioning that some of the texts currently grouped under CTH 770 (unidentified Luwian fragments) can in fact be assigned to CTH 764-766, as is, for example, the case of KUB 35.90. An example of a Hittite text that is

likely related to the Taurisa tradition is KUB 12.26 (CTH 441.1), representing a ritual for the reconciliation between mother and child, where the Sun-God and Kamrusepa act as protagonists in a historiola (cf. Watkins 2010: 358-359).

2.2 Anchoring the Taurisa tradition

But does this grouping guarantee that the texts mentioning the Taurisa triad are necessarily connected with Taurisa? The question is not as trivial as it may seem: one would, for example, hardly claim that all the texts mentioning the Storm-god of Zippalanda and his divine parents are necessarily connected with the town of Zippalanda, because the veneration of the Storm- god of Zippalanda was deeply integrated in the state cult of Hattusa. The cult expansion of the Goddess of the Night, documented in Miller 2004: 259-439, provides an illustration of how provincial pantheons could undergo changes in the Kingdom of Hattusa in the historical period. On the other hand, one could argue that the use of Luwian in CTH 764-766 reflects fairly recent demographic changes in Taurisa. The combination of these potential problems arguably contributed to underestimating the relevance of the “Taurisa connection” in the recent discussions of Luwian historical geography.

To begin with the geographic reality behind the cult of the Tutelary god of Taurisa, its welcome confirmation comes from a recent study of textual variation in the AN.TAH.SUM festival (Galmarini 2015). It is undertaken against the background of a general observation that “LAMMA of Taurisa rarely appears in the Hittite religious texts” (p. 49). The author's philological analysis is conducive to identifying two traditions of celebrating the great spring festival, only one of which includes the veneration of the Tutelary god of Taurisa. The places where the king administers his cult are variously called ^GISTIR ^URUTAURISA `forest of Taurisa', <sup>GIS</sup>KIRI6 harwasiyas `garden of secrecy', and Й ^DLAMMA ^URUTAURISA `temple of the Tutelary god of Taurisa'. Tentatively but quite plausibly, he connects this variety with the changing itineraries of the AN.TAH.SUM festival: while originally the king made a detour to the Taurisa area, by the Late Empire period it became more practical to administer the same rites in a special sanctuary built in Hattusa for the Tutelary god of Taurisa (pp. 53-54). The precise date of cult adlocation naturally cannot be established with certainty, but the only occurrence of the reference to the temple of the Tutelary god of Taurisa belongs to the fragment KBo 45.16+, which is dated as Late New Script on palaeographic grounds. We are grateful to Dr. Susanne Gцrke and the Project “Hethitische Festrituale” of the Mainz Academy of Sciences and Literature for confirming this information.

The facts and interpretations offered by Galmarini flesh out the picture of the Tutelary God of Taurisa as a provincial deity, which may have undergone adlocation to Hattusa at some point in time but still remained on the periphery of the imperial pantheon. It is potentially compatible with two hypotheses: either the Luwian incantations collected under CTH 764-766 reflect the Taurisa tradition, or they were collected in Hattusa after the adlocation. It is, however, impossible to identify the dialect of the conjurations under discussion with Empire Luwian. The archaic accusative plural forms of the common gender, such as KBo 43.223+ iii 12' a-li-in-zaHUR.SAG^HВ-ti-in-za, KBo 43.223+ iii 23' za-ar-pi-in-za, KBo 8.130+ rev. 12' a-pi-in-za or KBo 13.260 ii 7 pu-u-sa-an-ni-in-za, provide the most solid argument against such an assumption. The merger of nom.pl.c and acc.pl.c yielding the nom.-acc.pl.c ending /-ntsi/ represents the most obvious common innovation of the “Glossenkeil” words and Luwian hieroglyphic inscriptions of the Iron Age (Yakubovich 2010: 26-38). The fact that a number of texts belonging to the tradition (notably CTH 764) are recorded in Middle Script likewise strongly speaks against their written fixation after the adlocation of the cult to Hattusa.

Once we assume that the Luwian incantations referring to the cult of Taurisa also reflect the local variety of speech, identifying its dialectal features becomes a sensible task. If the Luwian presence in Taurisa represented a result of recent migrations to the area, the local dialect would be likely to bear resemblance to the dialect of the area where the relevant migrations started. If, on the other hand, Luwian were spoken in the area of Taurisa for a long period of time, the local dialect would acquire distinct features of its own, perhaps displaying shared isoglosses with the geographically adjacent forms of Luwian.

A linguistic feature that appears to be closely aligned with the texts of this corpus concerns the second-position clitics. It is well known that some Luwian dialects, although not all of them, feature the particle /=wa/ as part of the Wackernagel clitic chain. This morpheme is absolutely pervasive in Late Luwian, where it can be best described as a clause-demarcational clitic. It is attested only once in the Zarpiya ritual (KUB 9.31 ii 33 with dupl.) and lacks assured attestations in the texts belonging to the Puriyanni and Kuwattalla traditions. The sequence KUB 35.55:8' [...]x-as-wa-du-wa-at-ta (CTH 758, Starke 1985: 71) is probably to be interpreted as [ka]r-as<<-wa>>-du-wa-at-ta ~ /karstu(w)=ad(a)=tta/ `let him cut it away', on the assumption of an anticipation error. We are grateful to H. Craig Melchert for the suggestion of such an analysis. In the Luwian texts belonging to CTH 764-766 it is reasonably common but does not appear in every sentence. In a Luwian version of the banquet narrative (KBo 43.223+, CTH 764.I.A), it seems to behave as a particle introducing direct speech, which supports its etymological connection with the Hittite clitic particle =wa(r) having the same meaning. Nevertheless, if one factors in the rest of the Taurisa tradition, the semantics of /=wa/ cannot be reduced to that of an ordinary quotation particle. One can contrast, for example, the passage KBo 13.260 iii 24-29, where the particles /=wa/ and /=ba/ alternate in wish formulae with parallel syntactic structures.

A salient peculiarity of a relatively small group of Luwian texts is the presence of the particle =ku-wa /=g^wa/. Melchert (1993: 105) lists its occurrences under =ku on the assumption that we are dealing with a combination of the two clitics. This is not, however, the optimal solution, because the particle *=ku does not appear without the extension -wa in Luwian. One doubtful instance of the unextended =ku is cited in Melchert 1993 is KUB 35.133(+) ii 3'. The inspection of the photograph shows, however, that the sequence traditionally read as ku-i-pa-ku-ni-ya-as in this line has a more likely reading ku-i-pa-ku-i-ya-as. If so, it can be analysed as /kwi-ba-kwi(j)=as/, where /kwi-ba-kwi/ is a free choice indefinite pronoun (cf. Sideltsev and Yakubovich 2016: 91-92). What contributed to the confusion between the particles /=wa/ and /=g^wa/ was their shared slot in the idealised maximal clitic chain, situated between the discourse particles /=ba/, /=ya/ and the pronominal clitics. The examples such as KUB 35.103(+) ii 15' [a]n-ni-is-ku-wa-ti, KUB 35.103(+) iii 4 za-am-pa-ku-wa, KBo 8.130+ rev. iii? 14' [...(-)]x-an-za-pa-ku-wa, and KUB 35.128 iii? 10' a-ku- wa-an will suffice to illustrate this distribution. The plausible semantic interpretation of /=g^wa/, reached in Simon 2020, is the successive-adversative particle `in turn'. The following example

(1) features variation between /=g^wa/ and /=wa/ in two subsequent clauses.

The majority of the texts featuring /=g^wa/ can be assuredly or tentatively assigned to CTH 764-766 on independent grounds. The attribution of KUB 35.103(+), KBo 8.130+, KUB 35.90, and KBo 13.260 has already been discussed above. The fragment KUB 35.99 mentions the Luwian words for `snake' and `wolf', which otherwise only occur in the Taurisa incantations. The rare word wa-lu-ti-in of unknown meaning draws a bridge between KUB 35.99 and KUB 35.128, which are in addition written in the same hand. The fragments KUB 35.109 and KUB 35.79 probably belong to the same composition (see e.g. the rare shared word ku-li-ma-as-si-), We interpret Luw. /kulimass(i)-/ as `type of enclosure', a cognate of Palaic kuwalima- `id.', for which see Sasseville and Yakubovich 2018, forthcoming. Note that this lexeme is attested next to the designations of domestic animals in KUB 35.109.while the reference to DUMU.NAM.LЬ.U19.LU `human child' in KUB 35.79 iv? 7' suggests that we are dealing with yet another incantation meant for fighting children's diseases. The enumeration of landscape features, including ha-a-pi-in-ni-in-za `little rivers, brooks' places KBo 7.68(+) next to KUB 35.89 and KBo 43.223+, both assured representatives of the Taurisa tradition. Only in the instance of KUB 35.100 and KBo 29.38 no independent link with the texts grouped under CTH 764-766 imposes itself, but the latter of these two Luwian fragments mentions the Kaska-people, which independently vindicates its connection to the northern part of Asia Minor.

The findings regarding the particles of the Taurisa tradition can now be placed in a historical-geographic perspective. The absolute productivity of /=wa/ in Late Luwian is foreshadowed by a similar process in Empire Luwian, as should already be clear from the preliminary edition of YALBURT and EMОRGAZО inscriptions in Hawkins 1995. The implementation of the same tendency on a more limited scale in the corpus under discussion is compatible with placing its source in the vicinity of Hattusa, the cradle of the Empire Luwian koine. It is worth mentioning in this connection that the use of the quotative particle =wa(r) in Hittite was likewise not uniform but depended on a text genre. While the texts of administrative or official nature, such as annals, treaties or official letters, normally deploy =wa(r) in every clause of the quoted speech, certain other texts, primarily rituals and myths, use the less consistent pattern, which gravitates towards the use of =wa(r) only in the first clause of the quoted passage (Fortson 1998: 22-24, 27). The refinement of Fortson's syntactic observations is now provided in Sideltsev 2020. As for /=g^wa/, it can be analysed as a result of the fusion between *=gu< PIE *=kwe (cf. Hittite =kku `now, even, and', Palaic =ku `now, further') and the particle /=wa/ discussed immediately above. In other words, the analysis of Melchert 1993: 105, while unfounded from the synchronic viewpoint, still holds as a plausible diachronic explanation. Naturally, for the proposed fusion to take place, the original quotative meaning of /=wa/ must have been sufficiently bleached, which again suggests a transition to the state of affairs attested in Hattusa Luwian. Summing up, it is fair to say that the linguistic features of the Taurisa tradition support rather than contradict the localisation of the relevant dialect in a region of North Anatolia, adjacent to but not identical with the area of Hattusa. In this sense, the default localisation of this tradition in the Зekerek valley perfectly fits the bill. The two particles addressed above were singled out as dialectal isoglosses because Luw. /=wa/ is extremely common, while Luw. /=gwa/ is restricted to the Taurisa tradition. Nevertheless, there are additional linguistic differences within the corpus of Luwian cuneiform texts, which deserve a separate brief mention. Thus, the texts of the Taurisa tradition commonly feature the forms /kwadi/ `how', /abadi/ `thus', whereas those belonging to the traditions of Puriyanni and Kuwattalla use alongside them the extended variants /kwadin/ and /abadin/. The stem /xwidumar-/ `life' occurs in KBo 13.260 iii 18' (CTH 765), while the Kuwattalla tradition features the cognate /xwidwalahid-/in the same meaning. The form a-ad-du-wa-an-za `evil' occurring in KBo 13.260 iii 13' reflects the sound change /-lts-/ > /-nts-/, which also characterises the dialect of Hattusa (cf. Yakubovich 2013/2014: 285-286 with a similar historical analysis but without separating the Taurisa tradition).

2.3 A Rhetorical Figure

There is an additional level of contact that can strengthen the proposed localisation of the dialect underlying CTH 764-766. Certain formulae occurring in the Taurisa incantations find direct counterparts in other texts emanating from the northern part of Asia Minor. Thus Klinger (1996: 158) stresses the fact that the clause “Kamrusepa saw it (looking) down from Heaven” occurs in both CTH 727 (the Hittite and Hattian bilingual myth about the Moon that fell down from the sky) and in the texts booked under CTH 765. Furthermore, as was already mentioned in Subsection 3.1, Kamrusepa typically issues a cry of amazement or indignation immediately upon checking the situation on earth. The bilingual CTH 727 clearly contains an original Hat- tian narrative featuring the goddess Katahzifuri, which only secondarily underwent syncretism with Kamrusepa. In the Hattian version of CTH 727 the goddess is called Katahzifuri, but she apparently undertakes essentially the same action: “Es wirkte stдndig? erbarmungsvoll die (Gцttin) Katahziwuri hier vom strahlenden (Himmel) aus. Dann <sah sie es> (und rief) folgendermassen: “Was (ist) in dieser Weise (geschehen)?”” (Schuster 2002: 386). Another fragmentary Hittite-Luwian passage KBo 12.100 obv. 12-13 (CTH 765) also features Kamrusepa, who looks down from the sky but apparently suppresses her cry of indignation. The same is apparently true of KUB 17.8+ iv1-2 (CTH 457.1.A). But if the formula under discussion reflects the Hattian narrative of Katahzifuri and not Nesite folklore in (10), the same conclusion can also be extended to (11). The ability of the goddess Katahzifuri to cross ethnic boundaries is independently confirmed through the direct use of her name in Palaic texts, where she likewise occupies the second position in the local pantheon (cf. Taracha 2009: 58).

But Kamrusepa (=Katahzifuri) is not the only deity to ask perplexed or angry rhetorical questions in texts belonging to the Taurisa tradition. In CTH 764.I, it comes from the Sun-god Tiwad (12), apparently after he sees that the gods vomited or trampled everything three times (Steitler 2017: 397, but cf. van den Hout 1994: 315-316 and Mouton 2007: 276 on /tarsija-/ `to trample'). His question prompts the narrative about a divine feast to which various diseases have not been invited (Steitler 2017: 393). The parallel Hittite story found on the same tablet contains the identical emotional query (13), whose motivation is unfortunately lost in a lacuna and whose answer consists of a similar narrative about the neglected Great Deity (Steitler2017: 389). But the Palaic Sun-god Tiyat (whose name is cognate with that of the Luwian Sun- god Tiwad) also asks the same question in a different text CTH 752, apparently upon learning that the divine guests eat and drink but cannot quench their hunger and thirst (14). The following Palaic narrative represents a version of the Anatolian myth of a disappearing deity, which can only loosely be compared with the narratives about neglected deities preserved in Hittite and Luwian transmissions in CTH 764. But the juxtaposition of all the three questions with mythological narratives can be regarded as a significant parallel in itself, beyond the sheer similarity among (12-14).

The parallels adduced above need not be taken as a testimony of direct influence of Palaic mythological narratives upon the Taurisa tradition or vice versa. All the passages mentioned here are ultimately steeped in the oral folklore of North Central Anatolia, the fragments of which are transmitted in Hattian, Hittite, Palaic, and also Luwian. One of the recurrent themes there is the conflict between one deity or a group of related deities and the rest of the pantheon, which results in the disruption of the natural world order. It is important to observe that CTH 752 features not only Palaic but also Luwian incantations (Yakubovich 2010: 256257). Furthermore, the most famous Hittite narrative of a disappearing deity, namely the Myth of Telepinu (CTH 324), contains non-trivial loanwords from Luwian, such as the combination of auspicious terms salhiyanti- `growth' and mannitti- `proliferation (?)' (cf. Yakubovich 2010: 235-236 and Rieken, forthcoming). Goedegebuure 2008 offers an elaborate structural argument in favour of the hypothesis that an Anatolian Indo-European language had been spoken in north-central Anatolia already in the early second millennium BCE and functioned as a substrate for the non-Indo-European Hattian language. She justly describes Hittite as an unlikely candidate for such a substrate, because the indigenous name of the language (Nesite) is consistent with their localisation in the area of Nesa along the southern bend of the Kizilirmak River before the conquests of Anitta. She objects to the substrate role of Palaic on the grounds that this language is “too peripheral” (ibid.: 171) and opts for Luwian as the most likely candidate, referring to the demonstrable Luwian migrations in pre-historic period. Whatever is said in this chapter about the Luwian dialect of Taurisa broadly supports Goedegebuure's claim. We see, however, no logical necessity to assume that Hattian was impacted by just one Anatolian language or that Palaic was as peripheral in the early second millennium BCE as it was half a millennium later. Since Palaic famously shows more traces of interference with Hattian than any other Indo-European Anatolian language, it remains perfectly possible that Hattian and Palaic speakers lived side by side on a large territory, and the areas of Palaic and Luwian substrate effectively bordered each other in the North of Asia Minor.

The goal of Section Three was to demonstrate that the Luwian texts from the area of Tau- risa can be set aside from the rest of the Luwian texts based on their function, pantheon, linguistic features, and formulaic repertoire. There is every reason to treat the Taurisa corpus (CTH 764-766) as a cohesive group, on a par with the Luwian corpora associated with Hat- tusa, Istanuwa, or the Lower Land / Kizzuwadna.

3. 

3. Back to the Broader Picture

We have seen that the study of individual traditions associated with the use of the Luwian incantations is conducive to revealing additional fine differences among the regional varieties of Luwian. This empirical conclusion agrees well with what one expects on general sociolinguis- tic grounds: in the absence of an overarching written norm, enforced by scribal training, the ongoing differentiation among the spoken Luwian dialects was faithfully transmitted in writing. The new results, however, prompt the refinement of Luwian dialectal geography vis-а-vis the results reached in Yakubovich 2010.

The largest dialectal corpus of Luwian cuneiform texts available to us appears now to be associated with the Lower Land, while the tradition connected with the town of Taurisa situated to the northeast of Hattusa emerges as the close second. The Songs of Istanuwa and the incantations embedded in Kizzuwadna ritual texts all yield corpora of more modest dimensions. Although the term Kizzuwadna Luwian, was used very broadly in the last ten years, in the narrow sense, it can now be restricted to the incantations of the Zarpiya ritual (CTH 757) and perhaps to KUB 35.8 with related fragments (see Section One). In a broad sense, it can still be applied to a linguistic continuum stretching from Kizzuwadna proper to the Lower Land, with a caveat that fine linguistic differences between the Luwian dialects used in both regions deserve further study. In contrast, the dialect of Taurisa both emerges as linguistically distinct from the dialect of Kizzuwatna / Lower Land, even in the first approximation, and cannot be treated as part of a continuum in view of its remote geographic location. Finally, Hurrian influence can no longer be used as a decisive argument in distinguishing between Luwian dialects, because our study of CTH 759-763 suggests that it can increase over time within the same tradition.

The expanded dialectal landscape of Luwian cuneiform texts has repercussions for revisiting the sociolinguistic situation in the western Anatolian region of Arzawa. There is no doubt that Arzawa elites had Luwic personal names, and the Arzawa ritualistic traditions show similarities with those of Kizzuwadna and the Lower Land, but no Luwian incantations have been found thus far in the Arzawa rituals. Two different explanations of this seeming discrepancy were advanced in recent years. According to Melchert (2013: 170) “the lack of any Luvian incantations and rarity even of isolated Luvian technical terms in “Arzawan” rituals reflects that knowledge in Hattusa of the ritual practices of Arzawa was very indirect”. Archi (2015: 291) prefers a different explanation: the Arzawa rituals were collected “in the field” by Hattusa scribes, while the Kizzuwadna rituals “were obtained ... probably (at least in part) from the writing school of that kingdom”.

The disadvantage of Melchert's hypothesis lies in failing to address the content of the Ar- zawa compositions. As cogently argued by Archi, most of them are rituals against epidemics and based on their palaeographic history, many of them were likely recorded in response to the epidemic that decimated Hattusa at the time of Mursili II. If so, collecting these texts was not an idle intellectual enterprise, but rather an emergency measure, and therefore the scribes had every reason to accurately learn and record all the best practices attributed to the respected western ritualists. After the annexation of Arzawa by Hattusa this was not a logistically difficult task. Archi's proposal, on the other hand, crucially depends on the assumed dichotomy between the pre-existing literacy in Kizzuwadna, with its own scribal traditions and attitudes, and the lack of such in Arzawa. If our proposal of Luwian texts associated with the Lower Land and Taurisa holds water, they must have been recorded by the scribes trained in the Kingdom of Hattusa, who also happened to be responsible for collecting the Arzawa ritualsaccording to Archi's views. Under such circumstances, it is not obvious why the Hattusa scribes would make efforts to record the original Luwian incantations from the Taurisa performers but not from the Arzawa practitioners.

Therefore, one has to look for alternative solutions. One hypothesis that still awaits its refutation attributes the lack of code-switching in Arzawa rituals to “the inability of Hattusa scribes, many of whom were Hittite and Luvian bilinguals, to understand the native language of Arzawa ritualists” (Yakubovich 2013a: 109). Given the distance between Hattusa and Ar- zawa, there is nothing counterintuitive about the assumption that the local dialects were situated at the opposite ends of the Luwic dialectal continuum. The objections of Archi (2015: 283a) to this hypothesis do not really go to the heart of the matter: Archi merely stresses how little we know about the language(s) of Arzawa. The burden of proof normally lies upon those who wish to demonstrate that the two languages or dialects are mutually understandable, rather than those who claim the opposite.

Returning to the attested dialects that belong to Luwian in the narrow sense, the findings of the present paper are conducive to revisiting some of their peculiarities. An isogloss whose description stood well the test of time is the merger of the nominative and accusative plural of the common gender in the dialect of Hattusa. We have seen in Section Three that the dialect of the Taurisa tradition, which presumably developed to the northeast of Hattusa, does not share this innovation, featuring the archaic accusative plural forms in /-nts(a)/ in lieu of /-ntsi/. The dialects of Kizzuwadna and the Lower Land, spoken to the south of Hattusa, preserve the same archaic ending. This conforms to the hypothesis that the case merger in Hattusa postdates the initial Luwian migrations and constitutes the defining feature of what is now called Empire Luwian. It also strengthens the case for the connection of this process with the merger of nom.pl.c and acc.pl.c in Late Hittite (cf. already Rieken 2006: 274-275). The direction of the influence probably was from Luwian to Hittite, because the result of the merger was always /-ntsi/ in Empire Luwian, whereas Late Hittite displays a complex distribution of nom.-acc.pl.c endings depending on a stem type (Yakubovich 2010: 337-345).

The situation becomes more intricate once one turns to the development of the Anatolian genitive case in Luwian. The survival of this category was traditionally described as an archaism of “Hieroglyphic Luwian”, whereas their (near-)replacement with possessive adjectives in “Cuneiform Luwian” was viewed as an innovation (cf. Melchert 2003: 171). For a recent discussion of the distribution between the genitives and possessive adjectives in Hieroglyphic texts, see Bauer 2014: 169-186. A number of Luwian forms attested in cuneiform transmission were analysed as relics of the genitives in /-assa/ or /-assi/ in Yakubovich 2010: 38-45. The alleged genitives in -as-sa were, however, pro-vided with an alternative interpretation as a particular class of possessive adjectives in Melchert 2012. The few possi-ble genitives in -as-si were explained with reference to a possible interference of Hattusa scribes in Yakubovich 2010. The same analysis was essentially maintained in Yakubovich 2010, except that the terms “Cuneiform Luwian” and “Hieroglyphic Luwian” were replaced there with Kizzuwadna Luwian and Empire Lu- wian. The refined analysis of Luwian traditions makes it now clear that the disappearance of genitives represents a common feature of at least two distinct Luwian dialectal corpora, namely the Lower Land traditions, explored in Section Two, and the Taurisa tradition, addressed in Section Three. The preservation of genitives as a category in the dialect of Hattusa alone, but not in the Luwian dialects flanking the capital from various sides, while not impossible, requires explanation.

A possible solution emerges once we take into consideration external evidence. As shown in Adiego 2010, there are two classes of possessive adjectives in Lycian (A), the best-studied member of the Luwic group besides Luwian. The possessive adjectives derived from the appellatives belong to the common e/i-declension type, which they share with the majority of other Lycian adjectives. Those derived from proper nouns show, on the contrary, an unusual declension pattern: nom.sg -h, acc.sg. -hn, loc.sg. -he, shared only with the otherwise problematic s-stems. Adiego plausibly concludes that the origin of the second pattern must be analogical but does not specify its ultimate source. Now the scrutiny of the Luwian hieroglyphic texts reveals a discrepancy in the proportion of proper nouns among the genitive case forms vs. possessive adjectives. About two thirds of the genitives are derived from proper nouns, while in the instance of the possessive adjectives this number is slightly more than one third. The difference is significant, and when contrasted with the Lycian data is conducive to formulating a hypothesis about the exclusive or statistical association of genitives with proper nouns in Proto-Luwic. Presumably the Lycian paradigm of analogical possessive adjectives derived from proper nouns reflects the second wave of case attraction in possessive constructions. For the mechanism of case attraction in the Anatolian languages, see Yakubovich 2008: 196-202. In order to make the proposed Lycian analogy work one has to assume that the declension of secondary possessive adjectives in Lycian was modelled after the paradigm of the consonantal stems after the genitives in *-s > -h were reanalysed as nominatives. This implies that either some of the attested Lycian s-stems continue the inherited consonantal stems, or additional consonantal stems had existed in Proto-Lycian but were subsequently remodelled to a more productive type. The choice between these two alternatives naturally pertains to the domain of Lycian historical morphology and goes beyond the scope of the present paper.

If one reconstructs the genitives as a residual category that was also restricted to or statistically aligned with proper nouns for Proto-Luwian, then its gradual disappearance in the majority of the Luwian dialects would appear logical. The declensional pattern of the proper nouns could easily have been levelled to that of the appellatives. This said, we have limited opportunities to judge how pervasive this levelling happened to be, because the possessive forms of proper nouns are genuinely rare in cuneiform texts. Thus, the attestation of possessive adjectives within the Puriyanni and Kuwattalla-Silalluhi traditions appears to be limited to the following six examples: KUB 35.54 ii 14', iii 7, KBo 29.4(+):8' DIM-as-sa-an-za; KUB 35.71(+) ii 7 [U]RUNi-nu-wa-wa-an-na-as-sa-ti; KUB 35.71(+) ii 8 DlSTAR-as-sa-a-an-za-[ti], KUB 35.82 i 7' DSa-us-qa-a-as- si-in-z[i]. Therefore, one cannot absolutely exclude that the avoidance of genitives of in the Lower Land incanta-tions was merely a matter of elevated register. The situation in the Luwian dialect of Hattusa was different for a good reason. The widespread Hittite-Luwian bilingualism in the capital, which was postulated on independent grounds, could lead to the retention of the Luwian genitives and the relaxation of their association with proper nouns, since the Hittite language had no such association. At the same time, new possessive adjectives continued to be formed in Empire Luwian via the mechanism of case attraction, but unlike Lycian, the Luwian language did not perpetuate any formal distinction between their primary and secondary varieties. This eventually led to a complex pattern of coexistence between genitives and possessive adjectival forms, which can be observed in Late Luwian.