Verbal Art in Luwian and Hittite Incantations

1 Introduction

Luwian ritual incantations, or the Hittite ones that are obviously translated from Luwian and still show interference of that language, have enjoyed a good deal of attention in the past years. The interest was sparked not only by anticipated information on the religious background of the texts, but also by their linguistic form. After the earlier attempts in the last century by Watkins (1986; 1987) and Eichner (1993) to interpret certain passages as metric pieces of poetry, it was Melchert (2006) who, more generally, first drew attention to stylistic devices in Luwian incantations, such as alliteration, merism, repetition, parallelism, figura etymologica, or rhyming epithet and head noun. More recently, other articles by Watkins (2010), Bachvarova (2013), Francia (2013; 2014; 2016; 2018), and Mouton/Yakubovich (2019) followed with very useful observations on specific texts or expressions. There seems to have developed a consensus – although mostly implicit – that the incantations can be classified as ornamented prose.

In the present article, I shall evaluate these observations, add more evidence, and try to make some generalizations. The discussion will also contribute to the understanding of several of the relevant passages (Section 2). Based on these facts, a proposal is made regarding the technique of building the relevant sentences applied by the ritualists (Section 3). The final section will deal with the question of the origin of the stylistic devices under discussion (Section 4).

2 The Phenomenon

Multiple stylistic features are found in the coordinated pairs and triplets of example (1), as noted in Melchert (2006: 295–296):^[1]

(1) KUB 35.54 ii 48′–iii 5

[š]āndu=wa=ta parnantinzi

[ḫ]ūmmatiš ḫaššanittiš ḫuwaḫḫuršantinzi

t iyammiš t arušantiš

ad[duwalza utarša]

ḫallišša parattan[za]

puwatilza [n]anuntarriš[a]

irḫūwaššapa[r]ittaruwāšš[a]

ulantalliyan ḫuitw[aliyan]

“May the parts of the house let them go, (namely) the [p]recinct, hearth (and) ḫuwaḫḫuršant-, the ground (and) the beam: ev[il speech], impurity (and) defilement, past (or) [p]resent, internal (or) e[x]ternal, of the dead (or) the livi[ng]!”

Both [ḫ]ūmmatiš ḫaššanittiš ḫuwaḫḫuršantinzi and tiyammiš tarušantiš feature alliteration in coordinated lexemes. This is combined with Behaghel’s Law of Growing Members. In addition to the merisms in puwatilza [n]anuntarriš[a] “past (or) [p]resent” and ulantalliyan huitw[aliyan] “of the dead (or) the livi[ng]!”, Mouton/Yakubovich (2019) recently elucidated that of opaque irhūwašša pa[r]ittaruwāšš[a], now interpreted as “internal (or) e[x]ternal”.

Adding to the cases of alliteration of the initial consonant mentioned by Melchert (2006), more pertinent assonance is found between the verb and its object including not only the initial sound, but also the next vowel and the place of articulation of the following consonant; cf. the examples under (2a)–(2c), and (2d) with initial vowel:

(2a) KUB 35.43 ii 29

tapp ātta tāp aru [t]atarriyamman ḫīrūn

“s/he spat tāparu, [c]urse (and) perjury”

(2b) KUB 9.6 iii 28′′

wit panim=pa=an wid aindu

“may they smash him, (i.e., his) witpani-”^[2]

(2c) KUB 35.45 ii 26–27

dar awiddu tatar iyammanaššin ḫirutaššin en -an

“Let her hand over the lord of malediction and perjury!”

(2d) KUB 9.6 iii 26′′

ādd uwala ānn iti

“(who) did the evil”

Such stylistic features become even more frequent when one takes into account the phonetic readings of logograms. Thus, alliterations probably figure prominently also in example (3), where they contribute to the cohesion of the head noun and genitival attribute:

(3) KUB 9.4 iii 36–39 (Beckman 1990: 39)

t ar(a)šnaš t aškupiman

zi -aš (= ištanzanaš) impan

ni₃.te -aš=taš (= tuekkaš=taš) taššiyawar

“the crying of gorge, the burden of the soul, the pressure of your body”

The next passage in example (4) was analyzed by Francia (2013: 170–172; 2016; 2018: 86–89). It is part of a dialogue between the ritualist and the “evil tongues”, i.e., the malevolent speech:

(4) KUB 12.62 obv. 10′–12′

eme -aš (= lalaš) en-aš (= išḫaš) kuwapi pāši

kaskal -ši (= palši) karipuwan[zi pāimi]

ur.maḫ (= walwi) tarwauwanzi pāimi

alili waršuwanzi pāimi

dumu.⟨lu₂⟩.u₁₉.lu lala uwanzi pāimi

“‘O lord of tongue, where were you going?’ – ‘[I go] to the road for devouring (it), I go to the lion for fixing (it), I go to the flower for reaping (it), I go to the man for lala-ing’.”

Francia pointed out the abundance of stylistic devices in this short passage: the syntactic parallelism between the last four lines, the framing assonance lala- “tongue” in the first and the verb lala‑ in the final line (Francia 2018: 88 uses the term “ring‑composition”),^[3] the assonance between *walwi “to the lion” and tarwauwanzi “to fix” due to the repetition of the syllable wa/i in the central line as well as the assonance produced by the repetition of al in the three central lines as opposed to doubled la in the framing lines. Her findings show that not only alliteration in the strict sense of the word plays a role, but also assonances in word‑internal position using either the opening of a syllable (-CV-) or the coda of a closed syllable (-VC-).

In some cases, phonetic similarity goes further than alliteration and assonance when two juxtaposed words are largely identical. Melchert (2003: 297) mentions ipamanza(-) ḫuipamanza (KBo. 13.260 ii 28) “to the malicious (and) pernicious” (vel sim.). Also the well-known ai- wai-, aḫra- waḫra- and ašta- wašta- belong here (HED A, E and I 13–14). Even figura etymologica may be involved, as in ḫattayanza(-) ḫatta- (KBo. 13.260 ii 25–26) “violent (things) to the violent” (Melchert 2006: 297; see also Francia 2014: 8–9).

Notwithstanding the undisputable existence of stylistic devices in many of the Luwian and Luwian-based incantations, caution is in order. For instance, the evidence for rhyme that comes from those combinations that are syntactically coordinated and, necessarily, show the same ending is less cogent. Rhyme cannot be avoided under these circumstances. Therefore, only cumulative evidence together with other stylistic features proves that the phonetic effect is intentional.

In addition, it is hardly credible that in incantations such as example (1) the position of the object, the evils to be distanced, is “distanced as far as possible syntactically”, i.e., they are moved to the end of the sentence for the sake of iconicity, as maintained by Melchert (2006: 296). The fact that extra-position is found likewise for prospective pleasant states of the ritual client (cf. KBo. 13.260 iii 16′–20′) points into a different direction. It is rather linguistic factors such as the marking of an inferable topic or topic disambiguation (thus Sideltsev 2014: 827–828) or just the length of the list, i.e., the heaviness of the constituent, that induce the author to use this construction.

3 A Procedural Approach

Having seen that formal rhetorical figures play a significant role in the incantations, it is equally important to emphasize that, parallel to poetry in modern languages, alliteration, assonance and rhyme do not imply an etymological relationship between the relevant words. The examples under Section 2 provide ample evidence for this. Therefore, regarding the conclusions drawn alone from the co‑occurrence of lexemes within stylistic features by modern scholarship need independent confirmation.

For example, Taylor (2003: 81–113, non vidi, cited in Nikolaev 2019: 190–191 n. 49) suggests that, in the case of example (5), we are dealing not only with an accidental phonetic similarity put into use for assonance, but with the stylistic device of figura etymologica, which features the desired phonetic shape due to the common root of its components:

(5) KBo. 22.127++ iii 9′

zapp atta zamm anza utarša

“you cursed(?) the cursing speech”

Based on this assumption, he suggests the meaning ‘to curse’ for zapp-, and an etymological connection between the root zapp- and its putative derivative zamman‑, which, he assumes, comes from *zap-man- (cognate of Skt. śap- “to curse”, PIE *ḱep-).^[4] Consequently, he also posits a sound change PIE *-pm- > Luwian -mm-. As one can easily see, considerations on Luwian stylistics may have consequences even for neighboring fields such as Comparative Indo-European Linguistics, more specifically, in the case of zappatta zammanza, for sound change and etymology. The skepticism formulated here regarding Taylor’s proposal relies on the fact that the author’s choice of the sequence zappatta zammanza can easily be founded on their phonetic similarity, instead of their semantic and/or etymological contiguity. The verb zapp- likely designates some kind of destruction, but not necessarily cursing (see below).

In contrast with Taylor’s conception, Watkins (2010: 356–357) assumes an entirely different process, when he proposes that the Hittite authors created new words or modified existing ones in order to create a phonetic similarity. His example, a rhyme, is taken from the address of the servants of the Sun‑goddess within an Old Hittite prayer; cf. (6):

(6) KBo. 7.28 obv. 24–25

duwaddu taknaš ^d utu ‑waš arad.meš ‑šu šašnugatteni =ya=a[n kuiēš t]ašnukettani=ya=an kuiēš

“Mercy, o servants of the Sun‑goddess of the Earth, [who] both put her to bed and invigorate her!”

Of course, instead of “to invigorate her”, one expects something semantically different, like “to wake her up”, which, in Hittite, would be šarā tittanušketteni. Therefore, Watkins proposes that tittanušketteni was deprived of its preverb and reduplication, and its root ta- received the well‑known enlargement -š-, which then led to causative ta-š-nu-ške/a- “to make someone stand” in the sense of “to wake up”. This amounts to licensing a technique by which a word with appropriate semantics but inadequate phonetic shape is adjusted to the mold by being re-coined in a quite arbitrary manner (unnecessarily, since, as the anonymous reviewer notes, šarā tittanuškatteni is already phonologically close to šašnušgatteni). In addition, the resulting word tašnuške/a- turns out to be homonymous with an already existing word, i.e., tašnuške/a- “to make someone strong, to invigorate”, which is likely to cause misunderstandings, thus thwarting the desired effects of the forceful formal adaptation of the lexeme. It is needless to say that positing such putative and improbable neologisms demands stronger arguments than just an overly precise semantic prediction. Besides, they would constitute a serious challenge for etymological studies.

In sum, Taylor’s example of the figura etymologica assumes a correspondence between form and function of both words from the beginning, whereas Watkins supposes that the Hittite author deliberately adapted the form of the semantically appropriate word. Especially the latter assumption proves to be problematic. One may reasonably ask if the author did not choose a procedure of an intermediary kind, viz. to give preference to the phonetic shape and use an already available word that would produce the rhyme at the cost of making a compromise regarding its semantics. In the case cited above, šašnuške/a‑ “to put to bed” was the starting point, while tašnuške/a- “to make someone strong, to invigorate” was chosen for the sake of the rhyme. As a consequence, the semantic outcome was certainly not ideal, but the intended sense was at least still understood.

In fact, there seems to be the option of creating phonetic similarity by the deliberate choice of place names in a given context; cf. example (7) from the Tunnawiya ritual adduced by Bawanypeck (2005: 81):

(7) KUB 9.34 iii 28–19

šalliš=wa=kan dingir -lim-iš ^uru Lantaz unnieš laīš=wa takšani peran ḫuiyanteš inanaš lālaš kuišš=an weritenuer kueš=an=kan laḫlaḫḫinue〈〈nue〉〉r

“The great deity drove here from the city of Lanta- (and) lā(i)-ed (i.e., removed), those running midway in front, (the evils) of the illness and (the evils) of the tongue, who caused him to fear, who caused him to be worried.”

The passage deals with the purification from the evils of illness (and) slander (lāla-).^[5] The meaning of lā(i)- ‘to unbind, to take off, to dispel’ is well-known (CHD L–N 3) and describes exactly the process that is the goal of the ritual performance. The verb stands in the focus of the utterance, and its perception is further supported by the assonance of the city name ^uruLant(a)- (which, on the content level, may also contribute to the legitimation of the deity).

The same strategy of composition is found in example (8), another incantation attested in the ritual for the Tutelary Deity of the Hunting Bag:

(8) KBo. 12.96 i 14–15

(The dog barks, the pig grunts, and you, deity, shall not listen to (any noise) from anyone of them.) §

[mā]n ša^uruLalanda memai nu lalattaru [mā]n ša^uruWattarwa memai nu wattarittaru

“If (someone) from the city of Lalanda speaks, (it) shall be lala-ed, if (someone) from the city of Wattarwa speaks, (it) shall be wattari(ya)-ed.”

The aim of the incantation is that foreboding words shall remain unheard and ineffective (Neu 1968: 15–16). Therefore, although the respective meanings of the verbs lala- and wattari(ya)- are under discussion (see below), it is clear that they designate an action that destroys or hides the words in order to make them inefficient. Accordingly, most scholars who have dealt with the problem recently, agree that the names of the cities Lalanda and Wattarwa are chosen to match the verbs lala- and wattariya- (CHD L–N 26; Starke 1990: 565 n. 2104; Bawanypeck 2005: 80–81; Archi 2015: 289). Thus, the verbs with their specific meanings provide the starting points for the build‑up of their respective sentences, and the less meaningful city names are added for assonance. If anything, it is the general localization in or near the Arzawa lands that is relevant in the context. In more general terms, the task to achieve a structure with phonetic assonance is accomplished at the cost of a slightly less pointed semantic combination of lexemes.

The opposite assessment pronounced by Kronasser (1962–1966: 571), Kammenhuber (1985: 541), Tischler (HEG L–N 20–21), and more recently Francia (2016: 8), viz. that the verbs are neologisms derived from the city name just for this context and without a meaning more specific than ‘to act/speak in the manner of the people from the city of X’, not only flattens the semantic pointedness of the incantation, but even deprives it of any serious relationship to the ritual performance of counter‑acting the evil utterances.^[6] This would strain the efficiency of the spell too much and, therefore, reduces the plausibility of such an assumption.

Since the place name and verb are combined with the intention to produce assonance, Melchert (2006: 295) and Bachvarova (2013: 154) are obviously right when they point out that the phonetic effects must have been perceived as important for the efficacy of the rituals. Bachvarova even argues that, in several birth incantations translated from Luwian into Hittite, some Luwian lexemes remained untranslated in order to preserve the assonances. Her best example comes from the ritual of Pittei for the treatment of the bewitched son, parallel to (4) cited above; cf. example (9):

(9) KUB 44.4+ rev. 22–24 (Beckman 1983: 192–130; Giorgieri 2004: 410, 414–415; Bachvarova 2013: 137, 140)

eme.ḫi.a (= lala-) eme.ḫi.a (= lala-) kuwapi=wa paitte[ni]

^na₄ peruni palḫuna paiweni

xxx ana^na₄zu₂duwarnuma[nzi p]aiweni

ur.maḫ (= walwa‑) giš‑ruwanziki.min

ur.bar.ra patalhauna ki.min

zammanti dumu.nita lala una ki.min

“Tongues, tongues, where are y[ou] going? We are going to flatten out the rock. ... We are going to brea[k] the obsidian. Likewise to confine the lion. Likewise to fetter the wolf. Likewise to lala- the bewitched son.”

The relevant form here is again lalauna, a Luwian infinitive of the verb lala-, which, together with the initial lala- ‘tongue’ frames the passage (Francia 2018: 88). This is the crucial action performed on the patient, the son, with ill intent. These rituals are based on a very general binary opposition: between bound by an evil force, ill and out of order, on the one hand, and released from the evil force, healthy and orderly, on the other. At this point of the ritual, the bewitching tongues describe their evildoing, that is to throw nature into disorder by magic spell. In each case an object or animal is subjected to destruction or confinement that is thought to be impossible in the natural world, which indicates that they are “bound”. Likewise, the bewitched son is “bound”. Accordingly, no self-defense is possible for the bound objects and beings concerned. Only afterwards (in line 25), the ritualist turns the evildoing of the tongues against the sorcerer himself so that objects and beings, including the son, are released.

Under this interpretation of the overall passage, the understanding of lala- as a reduplicated form of lā(i)- ‘to release’ by Rosenkranz (1964: 247) and Yakubovich (2010: 404) is excluded and so is the proposal of Sommer/Falkenstein’s (1938: 136) ‘artikuliert reden’, where lala- is taken as a denominative verb derived from lala- ‘tongue’. The same word formation is either implicitly or explicitly assumed by Kronasser (1962–1966: 313), Starke (1982: 361), Beckman (1983: 193), Puhvel (HED L 43) and Francia (2018: 87), who propose a meaning ‘to speak, to throw a magic spell (on someone)’. Like in the case of tašnuške/a- (see above), this would demand the assumption of a neologism homonymous with a frequently used Luwian word (lala- ‘to take’), which would lead too easily to misunderstandings. Much more economical is simply to start with Luw. lala- ‘to take’ (Melchert in CLL 121).^[7] Its meaning lends itself most easily to the required sense of binding, illness and disorder: a child taken by an evil force is under control of that force. Thus, lala- ‘to take’ yields a plausible semantic interpretation that is close enough to the intended concept (without using the phonetically divergent technical term ḫamenk- ‘to fix, to bind onto’) and, at the same time, allows the assignment to a well-known verb.

Returning to the phonetic assonance between the passage-initial noun lala- ‘tongue’ and the final verb lala- ‘to take’, Bachvarova points out that, had the scribe decided to translate Luwian lala- into Hittite dā- ‘to take’, the framing effect would have been lost. Assonances like that of lala- ‘tongue’ and lala- ‘to take’ must have played a significant role for the performance of the ritual, simply because they made the composition of the spells much more difficult and prompted the readers to accept a certain degree of semantic infelicity. The additional effort in composition and parsing must have had a purpose.

How, then, did the ritualists proceed technically when they composed a new spell? A plausible scenario would allow us to understand the features of the spells described above as a necessary result of the very process that generated them. Following broadly Visser’s (1987) theory of the composition of Homeric verses,^[8] one may entertain the idea that the starting point for each spell is a word for its central concept, e.g., lala- ‘tongue’ in examples (7) and (9), the evil that has taken hold of the ritual client – in Visser’s terms, the determinant (Germ. Determinante). This is combined with a variable (Germ. Variable), such as lā(i)- ‘to release’ in example (7) or lala- ‘to take’ in example (9). Because it is not always easy to find a suitable term that fulfils the formal requirements of assonance, some allowances regarding the semantics may be licensed, e.g., tašnuške/a- ‘to invigorate’ instead of ‘to wake up’ as variable of the determinant šašnuške/a- ‘to put to bed’ in example (6). If further elements are needed, they may be filled in as necessary, e.g., unna- ‘to drive’ and pai- ‘to go’ in examples (7) and (9), respectively, or mema- ‘to speak’ in example (8).

It is not always easy to identify the determinant. In example (3) with head nouns and genitival attributes, these are probably the genitival attributes designating the members of the body to be healed (cf. 12 ^uzuur₂^ḫi.a-šu ‘his 12 body-parts’ a few lines earlier). With combinations of verb and direct object (cf. example 2), it is the direct objects which contain the information about the evil to be counteracted and that cannot easily be changed, while the verbs of performing or destroying the evil (‘to do’, ‘to smash’, ‘to deliver’, less so ‘to spit’) are relatively unspecific. The same applies to example (5), zappatta zammanza utarša, in which the object of the action, the bewitching or harmful speech, has to be remedied and is the center of the expression, i.e., the determinant. Consequently, the verb zappa-^(ti), being the variable, must only fit the context to a certain degree in that it denotes a kind of destruction, while the main reason for its choice is its assonance with zammanza. This does not preclude the figura etymologica suggested by Taylor (2003: 81–113), but it is not a reliable foundation for his far‑reaching phonological and etymological conclusions.^[9]

Summarizing the above view on the technique of composing spells, it is submitted here that functionally the most important lexeme of spell (the determinant) was pre-set. The second lexeme (the variable) was then chosen from a variety of options to show phonetic assonance with the determinant, even at the cost of slight semantic incoherence. Further lexemes and grammatical items could be added freely.

4 Origin

It is a remarkable fact that we find the vast majority of attestations for the stylistic devices treated above in rituals with obvious Luwian background. One possible explanation is the impact of the Hurro-Mesopotamian traditions, which naturally affected the Luwians in the South stronger than the Hittites living in the North. However, Schwemer (2013: 147) warns not to use cultural borrowing as the default explanation, but to exclude cross-cultural, typological similarities by tracing specific or unexpected parallels.

Regarding the stylistic devices treated here, a certain degree of specificity is provided by the phonological principles underlying the assonances. The Mesopotamian scholars applied them when ordering lexemes in several Sumero-Akkadian lexical lists and when relating Akkadian words to their assumed Sumerian roots in commentaries. Lambert (2001: 225–231)^[10] presents a number of examples which demonstrate that the Mesopotamian scholars recognized the natural classes of consonants that distinguish themselves by their common place of articulation or belong to a unified group of sibilants. Thus, in the lexical list MSL 9, 29–44, the verb nasāsu ‘to wail’ is followed not only by nissatu ‘grief’, but also by nesû ‘to be distant’, which is wholly unrelated. In MSL 17, 150–167, after qâšu ‘to give’ assonant qīštu “‘forest’ is listed, after kīsu ‘purse’ also qiššû ‘cucumber’. It is obvious that, according to the linguistic principles seen in these examples, phonetic similarity overrides semantic coherence (Lambert 2001: 231).

Moreover, Mesopotamian commentaries provide etymological interpretations as illustrated by example (10):

(10) BWL 72, 48 (cited after Lambert 2001: 229)

ak-k[a-an-nu: ...:] ana muḫḫikan: a-la-du

“‘wild ass’ [means ...], and derives from (Sum.) kan ‘beget’”

Lambert (2001: 229) points out that the etymologies define attributes or characteristics of the referents of the respective lemmata. Accordingly, they are probably thought to rely on an inherent truth. If this can be transferred to the wordplays and assonances found in incantations of Mesopotamian rituals, their relevance beyond the elevation of style becomes immediately clear. They generate linguistic sympathy and thus enhance the efficacy of the analogical patterns, e.g., kīma kasî liksûši kišpūša “Like the kasû plant, her witchcraft shall bind (liksûši) her!” (Schwemer 2007: 198; cf. also Schwemer 2019: 23–25 with other instances of the same phenomenon).

The following example (11) from the papilili-ritual from Ḫattuša brings us back to the rituals performed in the Hittite Empire. Beckman (2014: 67) considers the possibility that the ritualist is “playing on homophony.”^[11]

(11) KUB 39.71 iii 20–22 (Beckman 2014: 16, 37)

nu-kan₂ ^lu₂ša-ku-un-ni-eš a-na udu mun ka×u-i da-a-[i] ...

ta-pa-a-du li-ip-ḫu-ra-ki be-el-di

“the priest puts salt (Akk. ṭābtu) into the mouth of the sheep ...: ‘May good things (ṭābātu) gather around you, my lady!’”

The passage shows that also the Hittites farther to the North must have been familiar with the phenomenon of linguistic sympathy caused by the collocation of roots that feature different, but homorganic consonants.

Thus, in the pabilili ritual from Ḫattuša, it is the near-homonymy between the designation of a piece of the paraphernalia mentioned only in the ritual instruction and a blissful state or item solicited in the conjuration. By contrast, in the Boğazköy rituals with Luwian background, the same kind of assonance occurs in the expressions for the action and its object within a single sentence of the conjuration and is thus more explicit to the ritual client. Although it is clear that (a) in either case the assonance contributes to the success of the ritual, (b) the paraphernalia or their designations are chosen on account of their phonological shape overriding the function or the semantics, and (c) the limits of variability of the consonants are alike in Akkadian and Luwian, one has to admit that the difference in the way the assonance is put into use raises serious doubts as to the scenario of direct borrowing. Moreover, the phenomenon of linguistic sympathy is a wide-spread phenomenon. Therefore, due caution demands not to suspect that the Hurro-Mesopotamian scholars and ritualists provided anything more than the inspiration for the rules and licenses that governed the wording of the Luwian incantations. Since the Luwian spells were probably well understood by the ritual clients, it was attractive to transfer the figures of speech to a locus where they could be appreciated by their recipients, i.e., both of the assonant lexemes were accommodated in the incantation itself. This was done in several cases even at the cost of reduced semantic coherence. In the Hittite capital, too, the same principles were adhered to in the imported rituals when possible. Alternatively, the incantations were recited in Luwian, or important Luwian terms were left untranslated.

Abbreviations: The abbreviations follow those of the Reallexikon der Assyriologie und Vorderasiatischen Archäologie.