Editing the Zoroastrian long liturgy

Editing the Zoroastrian long liturgy The Zoroastrian long liturgy in honour of the god Ahura Mazdā has been celebrated in the form as it appears in the manuscripts (or a similar form) since the Achaemenian times and continues being celebrated still today among the Parsis in India. During all these centuries it has been one of the most characteristic features of this religious community. Its origins go back to the Indo-Iranian sacrifice whose most significant historical witnesses are the Vedic sacrifice and the Zoroastrian long liturgy. The recitative of this ceremony was fixed at different stages in Eastern Iran before the Achaemenid times. The liturgy was then exported to Western Iran, the heartland of the Achaemenid power. The manuscripts reproduce the Western-Iranian recitation of this Easstern-Iranian text. In other areas, the liturgy was recited in a different way as shown through the discovery of a Sogdian version of one Zoroastrian prayer, the Aš əm Vohu (#1) (Gershevitch in Sims-Williams 1976). Today I m going to focus in two aspects of the edition of the Avestan long liturgy. I have choose them, because they suppose a change from the traditional way of editing Avestan texts and since they might shed some light over similar problems in others traditions. First, I ll show the necessity of editing these texts as liturgical texts and the consequences of this choice. And afterwards, I ll discuss the mutual influence between written and oral transmission and the problem of the almost infinite readings of each single word. The ritual character of the texts of the long liturgy The Pahlavi literature describes the Avesta as a collection of 21 books arranged scholastically. But this Great Avesta is lost. What we have in the manuscripts is basically two kind of ritual texts: 1. Descriptions of many variants of the long liturgy we have been speaking about 2. A collection of minor rituals However, the ritual character of these texts has often been overlooked. The texts contained in our manuscripts were considered fragments of the scholastic collection that only secondarily were used for ritual purposes. This has had important consequences for the interpretation of the Avestan texts and also for the way they have been edited. The manuscripts of the long liturgy are of two different types: - Liturgical manuscripts consist of complete descriptions of the ceremony including not only the recitative in Avestan, but also ritual directions in different languages (in Pahlavi and Farsi in Iran; in Gujarati and Pāzand in India). - Exegetical manuscripts consist of the text of the basic liturgy with a translation into Pahlavi or into Sanskrit. The exegetical manuscripts have been considered older than the liturgical ones and the origin of them. The liturgical manuscripts were obtained just through elimination of the Pahlavi translation. Today we know that exactly the contrary is true. First, the liturgy existed and then the translations were created, probably before the existence of any manuscripts. Then, the

liturgical and exegetical manuscripts were created directly from the oral tradition and independently the ones from the others. This does not exclude, of course, frequent collations of both traditions. The main consequence of the traditional view that did not recognize the liturgical character of the texts or considered it just secondary is the dependence of our editions from the exegetical manuscripts. This has important implications: 1. The ritual variety is not reflected in the standard editions of the long liturgy 2. The ritual directions included in the liturgical manuscripts, but not in the exegetical ones have never been edited either together with the Avestan recitative or separately. The long liturgy knows from old on the expected variety in a real ceremony. A Sasanian treatise on rituals, the Nerangestan, witnesses still a greater variety. The liturgical manuscripts reflect this variety. They attest following variants: - the Yasna ī Rapihin: A smaller ceremony than the standard daily liturgy celebrated only in the summer months - the standard morning daily ceremony or Yasna - the solemn ceremony or Wisperad that knows on its part different variants like the Wisperad ī Gāhānbar for the great seasonal festivals or the Dō- Homāst The Wisperad serves as the basis for more complex ceremonies: the intercalation ceremonies. The centre of the long liturgy is written in more archaic language that we call Old Avestan. This is the most sacred part of the liturgy during which in Sasanian times was still celebrated the animal sacrifice. In the intercalation ceremonies a Young Avestan text is divided in sections and intercalated between these Old Avestan texts. The manuscripts attest two such ceremonies: the Widēwdād and the Wištāsp Yašt. Furthermore, there are some mobile parts of each ceremony that change according to the day of celebration, the divinity of patronage, etc. The exegetical manuscripts do not reproduce this ritual variety. Only the standard version appears complete in the Pahlavi manuscripts. From the Wisperad, they include only a selection of fragments that need to be translated since they do not appear in the standard version. From the intercalation ceremonies, they contain only the intercalated texts. The Yasna ī Rapihwin is just totally missing in the exegetic manuscripts. Our editions of the Avesta are based on the exegetical manuscripts and accordingly they edit the texts as presented in the exegetical manuscripts. Therefore, only the standard morning liturgy is edited complete. From the Wisperad ceremony only the sections that appear in the exegetic manuscripts have been edited, but it must be emphasized that the Wispertad ceremony cannot be reconstructed on the basis of Geldner s edition of the Avesta. Actually, numerous sections of the Wisperad have never been edited, including some totally new texts. The same is true for the rest of liturgies, shorter and longer. A new edition of the long liturgy must be based mainly on the liturgical manuscripts and reproduce the ritual variety, including the different variants for different types of ceremonies, for different days and celebrations and indicating the different options for

the mobile sections that changed in every performance. It poses then a practical problem: how to present the different variants of each section in parallel so that every single liturgy can easily reconstructed and on the same time the parallels in the rest of variants are recognizable. I ve developed a system and new numbering in which each text that appears in different variants is quoted and gets independent numbering indicating in which ceremonies appears in this position (# 3, pdf of Y0). If in a ceremony several options are open, all of them are edited. Each variant have so a special numeration and is edited in such a way that the correspondence with the sections in other variants of the liturgy is recognizable. I have published on-line provisional versions of each variant of the liturgy (http://ada.usal.es/pages/completeceremonies) and created a synoptic table of the correspondences (http://ada.usal.es/pages/table). A further consequence of the dependence of the available editions from the Pahlavi manuscripts is that only the Avestan recitative is edited and the ritual instructions have been discarded. The liturgical manuscripts are guides for the right performance of the liturgy and include many indications about the ritual actions accompanying the recitation, the way how they have to be recited, how many times, etc. All these indications are missing in the actual editions of the Avestan texts. Let us compare in the edition of Geldner and the liturgical manuscripts the exchange of the bāǰ, an important ritual moment in the liturgy that allows the exchange of roles between the performers (#4): Geldner yaϑā. ahū. vairiiō. zaotā. frā. mē. mrūtē. aϑā. ratuš. aš āt cīt. haca. frā. ašạuua. vīδuuå. mraotū. aš əm. vohū. (si bār) yaϑā. ahū. vairiiō. (du bār) Translation of the text with ritual instructions in the ritual manuscripts If the zōt has made the preprations for the ceremony, (then he recites): yaϑā. ahū. vairiiō. zaōtā... If the raspīg has made the preparations, the zōt recites: yaϑā. ahū. vairiiō. yō. zaōtā. frā.mē. mrūtē. The raspīg (recites): aϑā. ratuš. aš āt.cīt... The ẕot goes to one side of the fire altar, the raspīg[?] to the other side and takes the centre of the fire altar. With three aš əm. vohū he cleans every side of the fire altar. Then he (recites) two yaϑā. ahū. vairiiō on the way. When (reciting) š iiaōθananąm he puts the right foot in the zōtgāh. Iranian liturgical manuscripts HT hm zwt NPŠE tn yšt sʾhtk YHWWNyt/ yaϑā. ahū. vairiiō. zaōtā. frā.mē. mrūtē. aϑā. ratuš. aš āt.cīt. haca. frā. ašạuua./ vīδuuå. mraōtū. HT lʾspyk yšt shtk YHWWNyt. zwt yaϑā. ahū. vairiiō. yō. zaōtā. frā.mē. mrūtē. lʾspyk aϑā. ratuš. aš āt.cīt. haca. frā. ašạuua. vīδuuå. mraōtū. zwt OL kwstk y TWḆ OZLWNtn mdyʾn ʾtyšt LAWHL/ OL kwstk TWḆ OZLWNtn mdyʾn ʾtyšt ḆYN OHDWNtn PWN aš əm. vohū. 3 ʾtyšt LAWHL OL kwstk TWḆ šwstn AMTš šwstn APš YDE PWN/ pʾtyʾp LAWHL krtn AHL yaθa. ahū. vairiiō. 2 ḆYN lʾs gwptn PWN š iiaōθananąm LGLE dšn ḆYN zwt gʾs HNHTWNtn

As you see, the ritual directions include essential information to know the course of the liturgy. The simple recitative of the Avestan text, as always edited, is deprived of its context and almost meaningless. A problem for the edition of ritual instructions is that the wording is not as homogeneous as in the case of the Avestan recitative. Although usually the same instructions appear at the same places, the wording changes a lot even among manuscripts of the same time and place. See a comparison between the directions of the beginning of the Yasna in different manuscripts (# 5) The best solution I can find for this problem is to edit consistently the directions of a chosen manuscript for each ceremony and to include the variants of other manuscripts in the apparatus. Same text, but infinite variant readings A further consequence of the interaction between ritual practice and manuscripts production is the continuous interference between oral and written transmission and the adaptation of the manuscripts to the ritual reality. For centuries, these ceremonies have been orally performed and transmitted. The priests knew them by heart, since the use of manuscripts was not allowed during the performance. The invention of the Avestan script provided a new, complementary tool for learning the recitatives and ritual directions. Our manuscripts do not derive from an archetype, but many priests created new manuscripts at different historical stages on the base of their own liturgical knowledge. Nonetheless, manuscripts are characterized for a great uniformity: all the manuscripts of the same ceremony contain basically the same text, with some differences (like addition of information about other variants and very different ways of abbreviating or some differences between the Indian and Iranian ritual practices). This is basically a consequence of the ritual uniformity and not of a careful process of copying from one single original. The direct influence of the ritual performance on the manuscripts is at the root of a very particularity of the Avestan transmission: viz, that it shows a great homogeneity regarding the text, but each single word of the texts appears in an almost infinite number of variant readings. Let see for example the variants of pari.yaoždaθəṇtəm in Y9.1 (#6). Even within the same manuscript each word is very often spelt differently in different attestations. Most of these variant readings are connected with two factors concerning the recitation in the ritual performance: 1. the recitation has kept changing till modern times and specially the recitation in India shows features differentiating it from the Iranian one 2. there is a certain instability in the right recitation, frequently with influences of the own language of the performers. Although the invention of the Avestan script has slowed down the phonetic evolution of the Avestan recitative of the long liturgy, it has never completely stopped it, and these changes are reflected in the manuscripts, especially in the liturgical but also in the exegetical ones. All editors have noticed that the text of the recitatives has been influenced by its ritual performance, but these influences were considered to be occasional and never systematic and therefore easy to be detected and eliminated.

Actually, we are not dealing only with occasional influences, but also with an evolution of the recitation that enters the written transmission. A clear example is the change of ū into ī in the Iranian manuscripts (PPP). The oldest liturgical Iranian manuscripts from the family of Marzbān, at the end of the 16 th and beginning of the 17 th century, still distinguish at least partly between long ū and long ī. However, from the middle of the 17 th century the Iranian manuscripts do use a single letter for both sounds. So Frēdōn Marzbān still uses two letters, although with some confusions, but Mihrābān Anōšagruwān, even if he is copying from a manuscript of Frēdōn Marzbān, uses only one letter. The same is true for all later Iranian manuscripts. The confusion between these two letters is an influence of the Yazdi dialect. In this dialect, ū is pronounced ī in several positions. The substitution is not occasional, but systematic. I have recently shown other instances of independent, regular phonetic evolutions in Indian and Iranian manuscripts (PPP). Thus, the ending ōiium is reduced in India in disyllabic words into ōīm, but in Iran remains as ōiium. This reduction is regular only in India and affects only disyllabic words, whereas longer words do not show this reduction. Furthermore, the acc.sg. of other uua- stems do also behave differently in the Indian and Iranian manuscripts. This research is at the very beginning, but it seems that such systematic differences between Indian and Iranian manuscripts are much more numerous. The Iranian manuscripts show then a different text than the Indian ones, but these differences have never been analyzed or taken into account. The standard editions are, actually, a totum revolutum of Iranian and Indian readings. Furthermore, there are instabilities in the recitation of the Avestan texts that are not systematic differences between manuscripts of different geographical origins. They have also never been analyzed, but it is likely that certain variations are more frequent in certain areas than in others and that also these occasional instabilities are consequence of the own language of the performers (#9). Thus in the Iranian manuscripts we find frequent confusions between ā and u. The Iranian manuscripts show often nj for nc, a typical development of the Persian language. A similar explanation applies for the use of z for j in Indian manuscripts. Therefore, it is of the greatest importance an analysis of phonetic and orthographic variation in the Avestan manuscripts. There are, besides, paleographic differences between different groups of manuscripts. Avestan paleography has only been analyzed concerning the original distribution of some letters that are, however, used indistinctly in the manuscripts. Nevertheless, the differences between Indian and Iranian manuscripts and some subgroups of manuscripts have never been analyzed. (#10) How to deal with this huge amount of readings has been traditionally one of the biggest problem of the editions of the Avestan texts. The first editions discarded most of these readings. Geldner increased drastically the number of readings quoted in the apparatus (#11) and this is one of the reasons why Geldner s edition became the canonic edition of the Avesta. Nevertheless, his apparatus is chaotic, far from being exhaustive, and it does not enable readers to know the exact form they will find in the quoted manuscripts (since he quotes together similar readings under just one of them). In an electronic edition dealing with such a great amount of variants should not be a problem, but in a

printed edition it is. Actually, I think that for both (electronic and print edition), the apparatus must clearly distinguish between real variant readings that are either possible real alternatives or imply textual accidents relevant for history of the text) and, on the other side, the variants that are purely due to different orthographical conventions or phonetic variants because of the influence of the local recitation. If we confine phonetica et orthographica to a separate apparatus the edition will be more handy. See my own proposal of edition of Y9.1 (PPP) This huge amount of variant readings poses not only a practical problem (how to deal with them), but also a real one: the choice of the right reading. The last 50 years of Avestan research have been dominated by the philological and linguistic work on Avestan leaded by K. Hoffmann at the University of Erlangen. Hoffman has rationalized older views of the transmission, has set up a clear goal for the editions and has developed a method for it. In his view, our manuscripts derive from a Sassanian original codex, the Sasanian archetype, through some hyparchetypes of the 10 th century in which some errors were already contained. The task of the editor of an Avestan text is to reconstruct the Sasanian hyparchetype. This archetype cannot be traced back by means of textual criticism, since textual criticism could bring us as far back as the hyparchetypes. Therefore, the reconstruction must be leaded by the linguistic analysis. Actually, in the very last years Hoffmann s views have been at least partially challenged. We know today that our manuscripts do not derive from the archetype and that the supposed hyparchetypes have never existed. Furthermore, the fact that the recitation has continued evolving till modern times and these changes have entered the manuscripts eliminating the evidence of older ways of reciting, prevent really a trustworthy reconstruction of the Sasanian archetype. Through Hoffmann s method we can know today better than never before some features of the Sasanian archetype, but we are still not able to reconstruct the whole shape of a supposed Sasanian codex. So in the case of the acc.sg. of the short u, long u and uua stems, the expected form is um, - ūm and and the evolution of uuam. In the Indian manuscripts all these endings appear as ūm. In the Iranian manuscripts, all appear as well as ūm, but some acc. of uuastems present um. The distribution between um/ūm in the Iranian manuscripts is clear: the long ūm appear only after two consonants. Under these circumstances, how it is possible to know the form attested in the Sasanian archetype? Was it the original distribution? Or has long ūm already extended to the short u stems or to some of the uua- stems? Every proposition would be simply a guess. Under these circumstances of an evolving recitation, the editor must set up clearly the chronologic and geographic scope of his edited text. A reconstruction of the Sasanian archetype that has probably never existed is for me a too risky enterprise. Even if the linguistic reconstruction would be possible, there has been ritual changes and innovations after the Sasanian time, as shows a simple comparison between our manuscripts and the Nērangestān. Therefore, we could perhaps put in a linguistic Sasanian form a liturgy that has probably never existed in this form. Therefore, it is more realistic to edit the liturgy as it was celebrated at a time we can reconstruct through textual criticism and linguistic analysis. In my opinion, the oldest version of the

liturgy we are able to edit is its celebration in Iran between the 10 th and the 17 th century. The data of the apparatus will allow us to follow up the ceremony s history in India and Iran until the 19 th century. Furthermore, we have some signs that lead us to believe that the oldest traceable Iranian liturgy and recitation were not very different from the Sasanian one, but not exactly the same. Sims-Williams, N. (1976). "The Sogdian fragments of the British library." Indo- Iranian Journal 18: 43-82.