PDF Version: ARIRIAB XIV (2011), 13-23 The ‘Split’ Collection of Kharoṣṭhī Texts Harry FALK (Berlin) 1. The name The local origins of the present collection are not clear. Several part of it were seen in Peshawar in 2004. According to usually reliable informants the collection of birch-barks was found in a stone case in the Pakistan-Afghanistan border area, comprising the Mohmand Agency and Bajaur. It was split on arrival and some parts are now in a Western collection, while others went to a Government agency and yet other parts may still be with the private owner who allowed me to work on five rolls. Hence the name, which should be regarded as provisional. No information is available on how many rolls there were in the beginning. The five rolls entrusted to me for study contain the following texts: 1. one fragment with a few stanzas from the aṭṭhakavagga of the Suttanipāta, 2. a metrical text on the life of the Buddha, 3. large parts of a Dharmapada, comprising 89 stanzas, 4. parts from a collection of avadānas, 5. a text called prajñāpāramitā containing textual parallels to chapters 1 and 5 of the Aṣṭasahasrikā Prajñāpāramitā, These texts will be presented below in detail. Every text was written by a different scribe. The writing styles vary so much that is seems logical to assume that they were written at times rather far apart from each other. Some hands appear so unique that it should be possible to link them to other manuscripts once they surface in other collections or in the market. 1. Stanzas from the Aṭṭhakavagga (figs. 1-2) The fragment measures only 1.8 + 11 cm and is inscribed on both sides. In width, the bark is almost completely preserved; only one mm or two are missing due to wear at the margins. On the recto, stanzas 841-844 of the Suttanipāta are found, and nos. 966-968 on the verso. Its original length must have been at least 45 cm, if the birch-bark contained nothing but stanzas 845 to 965 in between the preserved text. If this manuscript contained the whole of the Aṭṭhakavagga inscribed on both sides in an equal distribution, its length must have been at least 75 cm. 13 PDF Version: ARIRIAB XIV (2011), 13-23 A fully annotated edition will be published elsewhere; for a first impression the plain text may suffice. Side 1 reads (fig. 1): 1: .................[pi saṃño] [taśpa tuo] momuado ḍahas[i] O ° 2: samo viśeṣo uȿa va vihino yo mañati so vivaṯea keṉa tiṃsa viȿaṣo aviapamaṉa samo viśeṣo avi na taa bho[ṯ](i) 3: saco du so bramaṉa ki vaṯeha muṣo ṯi va a vivaṯea keṉa yasa samo visaṃmo avi [nasti] so kena vado paḍiȿajaṉeṇa ° // 4: ///[v]e mu[ṇi sa]tha[vaṇi ka]me[h](i) [rito apu]rekharaṇa kasa no vigiśa janena ka[r]ye//// Side two (fig. 2): 1: aṯagaphaṣeṣo kṣuȿae phuṭho ' śiṯo vatvana pi aȿivasaeha ° ede[ṣ]o [ph]u[x]o [bahu] 2: ȿo anvayo ° viryaparakrama driḍha karea O krosvaa manaa va[ṣa→śo] ma gache 3: ...................[piya va] 3:[adhva bha]ṯata aviȿabhunea O The orthography is very intricate and in some respects new, when compared to other Kharoṣṭhī mss known so far: - We see three sorts of letters used for ta: one “normal”, one under-barred (ṯa), and one “under-bent”, which looks like tva or ȿa (on this see below), and seems to be misread for ta in the case of uȿa (← uta) and for tu (in the case of vatvana (← v-atuna ← atuṇha). There are likewise three kinds of sa: - one normal; one of the corkscrew type, usually transcribed as a, either corresponding to an initial sa or wherever it stands for Skt. gen. -sya. The third variety is again “underbent”, transcribed here with a ȿa (Unicode U+023F). It occurs where Skt. would have a dha, as in viȿasu <≈ vidhāsu, aȿivasaeha <≈ adhivāsayeyya; kṣuȿaya <≈ kṣudhayā, bahuȿa <≈ bahudhā. Unfortunately, this application is not the only one; the under-bent ȿa also occurs where a saṃ is expected, as in pratiȿajaneṇa <≈ paṭisaṃyujeyya, and aviȿabhunea <≈ abhisaṃbhaveyya. There is no common logic apparent behind these two uses. The scribe obviously tried to differentiate the sounds he uttered when reciting the text. He found variants in pronunciation and tried to asign to them diacritic forms to express the differences he heard in writing. For us, the problems arising from this text are: - Was the scribe a pioneer in shaping under-bent variants, or did he have predecessors, each with their own systems of diacritics, which he mixed together? - How to represent under-bent sa? I suggest the ȿa, but am also aware of the danger inherent in introducing too many diacritic marks. 14 PDF Version: ARIRIAB XIV (2011), 13-23 No C14 date is available for this fragment. Compared to all other mss, this one appears to be relatively old on palaeographical grounds. The tradition to distinguish graphically between phonetical variants has a parallel in the Gomitra slab (Salomon 2009), where sa and da appear in two shapes. In the case of the sa, the straight line touches the lower end of the vertical in the middle, in the case of da it starts at the lower end and bends downward to the right. The sa with full underbar occurs only in dhamakaiḵa, Skt. dharmakathika, never for plain s(a) or even sy(a). The under-bent da is found in duvaḏaya, Skt. dvādaśa, in pratiṭhaviḏa, Skt. pratiṣṭhāpita, and may be nothing but a means to distinguish na and da, with no phonetical relevance at all. 2. A metrical text (fig. 3) These several fragments should once have formed a roll of about 26 cm in length. The preserved width is 16 cm on average. When intact, its width was about one third larger, so that we have to reckon with orginally ca. 21 cm. The intact margin was given a vertical line about 3 mm away from the border. No stitchings or glued overlaps are discernible. Apart from small stretches, the bark was inscribed on both sides. displaying an elegant and careful hand, not overly given to frills and footmarks. The sheet was turned lengthwise, when the first side was filled, so that on both sides the left margin is preserved and on both sides the beginning of every stanza is missing. The sa is the normal one, only Skt. genitive -sya is given the cork-screw -a shape. The lower part of the roll, possibly near the end of the composition, uses a different form of -sya, i.e. what I call the Wardak-sya, because nothing but this form is used on the two vases from Wardak (Falk 2008). Graphically, it consists of a standard sa with a standard ya-hook attached to its lower end to the right. This form is not found very frequently and its late appearance in the time of Huviṣka is conspicuous, whereas as a dissolvable ligature is should have stood at the beginning of all attempts to express sya. The text itself is metrical; every pāda is separated from the adjoining ones by a space with a clear circle in its centre. The metre is so far unattested in Gāndhārī, being predominantly Vegavatī, a variant of the more common Vaitālīya, well-known from Pali texts, and occurring here as well in decidedly fewer stanzas. Every line contains 4 pādas of stanzas, counting 14+16 / 14+16 mātrās. The difference to the Vaitālīya lies in the cadence, running -⏑⏑--, instead of -⏑-⏑-. Without the support of the metre, it would be even more difficult to reconstruct the text and try to understand it. The persons occurring are the Śākyamuni, 500 Arhats, Pūraṇakāśyapa makhaliputro, and some tīrthikas. Places are the Jetavana at Śrāvastī, the kingdom of Kosala; the terms used are ṛddhibala, jñānabala and vikurvāṇaḥ, usuallyj, but not exclusively, firm components of Mahāyāna texts. I present here a few lines which are characteristic in regard to style and contents, together with a Sanskrit chāyā, with “+” standing for an illegible akṣara, O for a large circle separating the pādas: 15 PDF Version: ARIRIAB XIV (2011), 13-23 e5: ///+ + + rthigaṇa-kuvida tada?ti?hi O paḍia + + + + nidro O jetavaṇe suadasa saaśe jetavane sugatasya sakāśe -⏑⏑-⏑⏑ -⏑⏑-- 16 e6: ///purano kaśava makhaliputro O bahunagaśata upaama O rayino edo gira kaayati purāṇakaśyapa-maskariputraḥ bahunāgaśatā upāgama rājñaḥ . . . .girā kathayaṃti ⏑-⏑-⏑⏑ ⏑⏑-⏑⏑ -⏑⏑ -⏑⏑-- - ⏑-⏑- vait. -v v - ⏑⏑-- e7: ///+ O + + vigurvaṇa irdhibalehi O uvaruaro irdhiviśeṣo O sadhakarom aa śakamunina vikurvāṇā ṛddhibalebhiḥ uparyupari ṛddhiviśeṣaḥ śraddhākarām astu śākyamunīnāṃ ⏑--+ -⏑⏑-- ⏑⏑-⏑⏑ -⏑⏑-- ⏑-⏑-⏑⏑ -⏑⏑-- Up to now, no coherent sense could be gained from the readings and the chāyā has to be taken cum grano salis. Nonetheless, this piece shows again the wide range of literary activity, exploiting standard topics, toying with tricky metres, and displaying a certain amount of ignorance: the otherwise widely known Pūraṇa Kāśyapa, who in most cases is mentioned immediately before Maskari Gośālīputra, occurs now as Kāśyapa Maskariputraḥ. Apart from paleographic information, this text poses just one general question which may fall into the cathegory of “problems”: is this an autograph or was this text transmitted for some time despite its rather limited intellectual appeal? Can we find mistakes arising from copying? If it was copied, would such a text influence discursive thinking, or was it copied simply for the sake of possessing just another piece of text? A hint towards copying can be seen in uvaruaro in line e7. The same pāda occurs just a few lines further up, where the scribe first wrote uvaroari, then changed the ri to ro, displaying again a certain ignorance, this time about the Sanskrit base of the term. We can assume that an exemplar read uvaroari, leading to another problem: can we collect evidence to show that this language tended to be less and less understood and that “thinking” copyists coined new words because of that? 3. Another Gāndhārī Dharmapada (fig 4) This roll was made of birch-bark and is ca. 50 cm long. The greatest width preserved is ca. 14 cm; in the best cases 6 to 7 letters are missing. Together with the very narrow margin only 2.5 cm have been lost, adding up to an original width of 16.5 cm. The roll has been inscribed only on one side. The verso is completely untouched. A good opportunity to measure the original size is provided by a repeated stanza (s. fig. 4). We see that here 9 to 10 akṣaras are missing to fill the line. That means the rolls were not bent in the middle, but lost some part because of friction on one end only. The differences between the two versions are slight, but show either that writing down 16 PDF Version: ARIRIAB XIV (2011), 13-23 from memory led to spontaneous variants, or that copying from two written versions simultaneously may have led to a rather inconsistent mixture of orthographical traditions. As it is, the fragments contain 87 stanzas from the Dharmapada covering several vargas. The end of an individual varga is indicated by a number, with no varga-title given. The chapters contain stanzas as follows: 8 preserved of an as yet unclear attribution 11 preserved of 11, elsewhere found in mārga/śīla/prakīrṇaka-varga 9 preserved of <9>, elsewhere found in ātma/arahanta/prakīrṇaka-varga 27 preserved of ‘29’, elsewhere found in jarā/yuga/anitya-varga 12 preserved of 12, everywhere else called malavarga 15 preserved of 15, everywhere else called puṣpavarga 5 ¾ preserved, elsewhere belonging to the sahasravarga The end comes abruptly, as if the scribe was prevented from continuing, or as if he did not resume the work after a break. For a preliminary guess at the school affiliation we can compare the Puṣpavarga in other traditions. The first row shows the order in our manuscript; then follow the Pali Dhammapada ed. Norman & von Hinüber, The Khotan Dhammapada ed. J. Brough, the Patna-Dhammapada ed. M. Cone, the Udānavarga ed. Bernhard, and finally the Udāna(varga) from Subashi, ed. H. Nakatani. Where the parallels use similar wording the parallel is printed in bold, identical succession of two stanzas is indicated by printing the first and the second as well in italics: 1= DhP 51/52 = Brough 290/91 = Patna 125/126 UdV Subashi Sub. 222 -------------------------------------------2= DhP 52 = Brough 291 = Patna 126 = UdV 18,7 = 3= DhP 53 = Brough 293 = Patna 130 = UdV 18,10 4= DhP 49 = Brough 292 = Patna 127 = UdV 18,8 = Sub. 224 5= DhP 50 = Brough 271 = Patna 309 = UdV 18,9 = Sub. 225 6= DhP 58 = Brough 303 = Patna 135 = UdV 18,12 = Sub. 226 7= DhP 59 = Brough 304 = Patna 136 = UdV 4,3 = Sub. 227 ------------end of puṣpavarga----------8= DhP 54 = Brough 295 = Patna 121 = UdV 6,16 9= DhP 55 = Brough 296 = Patna 122 = UdV 6,17 10 = DhP 57 = Brough 297 = Patna 124 = UdV 6,19 11 = DhP 48 = Brough 294 = Patna 128 = UdV 18,14 12 = DhP 47 = Brough 294 = Patna 129 = UdV 18,15 13 = DhP 46 = Brough 300 = Patna 134 = UdV 18,18 = Sub. 235 14 = DhP 44 = Brough 301 = Patna 131 = UdV 18,1 = Sub. 217 15 = DhP 45 = Brough 302 = Patna 132 = UdV 18,2 = Sub. 218 17 PDF Version: ARIRIAB XIV (2011), 13-23 We see instantly that Pali, Khotan and Patna end with the same stanza, whereas our ms. continues, as does the Udānavarga and Subashi. Only Subashi shows four stanzas in a row following the same sequence as our text. In short, our text belongs more to the Udānavarga- and the Subashi-tradition than to those of Pali, Khotan or Patna. For a first evaluation of the differences I present two stanzas: § 74 Brough: 303,18 Z 383 DhP 58 Patna 135 yaa saṃkara-uḍasvi yadha sagara-'uḍasa yathā yathā saṅkāra-dhānasmiṃ saṃkārakūṭamhi ujidasvi mahapase uidasa maha-pathe ujjhitasmiṃ mahāpathe ujjhitamhi mahāpathe paduma tatra jayea padumu tatra ja'e'a padumaṃ tattha jāyetha padumaṃ ubbhidaṃ assa suyigaṃdho mano/// suyiaa maṇoramu suci-gandhaṃ manoramaṃ śucigandhaṃ manoramaṃ. § 44 Brough 151,10 Z 198 Jāt. 6,28 vs. 118 UdV 1,7 sahi ege ṇa driśati sadi eki na diśadi sāyaṃ eke na dissanti sāyam eke na dṛśyante praṇe driṭha bahojaṇa pradu diṭho baho-jaṇo pāto diṭṭhā bahujjanā kālyaṃ dṛṣṭā mahājanāḥ / praṇi ege ṇa driśati pradu eki na diśadi pāto eke na dissanti kālyaṃ caike na dṛṣyante sa'i driṭha bahujaṇa sadi diṭha baho-jaṇo sāyaṃ diṭṭhā bahujjanā sāyaṃ dṛṣṭā mahājanāḥ The first line of parallels shows that Khotan and our ms display the usual range of variants we are used to from Gāndhārī. In this case, our ms replaces Skt. dh and tha by a; Khotan doesn’t. Our ms never drops intervocalic k/g or y – while Khotan does. With no example included in the above-given excerpt, Khotan together with our ms replaces ca by ya, while Pali and Patna don’t. Our ms contains stanzas which so far have not been met with in texts belonging to the Dharmapada collections. The new material can be traced back as follows: 1 from the Theragāthās 1 from the Therīgāthās 2 from the Aṅguttanikāya 6 from the Saṃyuttanikāya 11 from the Udānavarga 18 PDF Version: ARIRIAB XIV (2011), 13-23 12 from the Majjhimanikāya and 3 without known parallels A strong link to the Pali tradition seems obvious, as well as a prominent relationship with the Udānavarga. The untraced stanzas may be own productions. The most obvious advantage of this new text are the passages parallel to the Khotan Dhammapada, providing ample material for an attempt at defining Gāndhārī ‘dialects’. At the same time we also have a problem in deciding how far the “orthographical” differences represent differences in pronunciation or just scribal licences. 4. An Avadāna collection (fig. 5) Due to a very bad state of preservation, this ms now covers three glass frames holding ca. 300 fragments. The largest of them measures 5.5 × 14 cm (fig. 5), many have 2-6 cm2, many more not more than 1 cm2. One margin is preserved. There seems to be just one or two akṣaras missing at the right margin, so that here again we have a rather narrow strip of bark. The handwriting is clear with no footmarks. The style is remarkingly archaic. Preconsonantal r- is a hook, Aśokan style, anusvāras are left unwritten, śa and ya are clearly differenciated the old way. The ms is inscribed on both sides, with the birchbark turned lenghtwise, so that the left margin is left on both sides. As much as can be read, the Avadānas presented here contain stories involving a king in the fangs of the Ajīvikas. Buddhist sects are mentioned, i.e. the Dharmaguptakas, the Mahāsāṃghikas and the Seriyaputras. One of the persons occurring, Upatiśya, explains an Avadāna (uvadiśo avadana japati). In addition we have one cāṇḍālasuta and Aṅgulimāla, the thief, who gets advice from his wife. The place is Pāṭaliputra. Lots of (electronic) re-assembling of the fragments is necessary before a consistent picture will emerge. However, this collection appears to be rather old: the width of ca. 15 cm is decidedly less compared to the texts written in more modern style. So this compares to the Aṣṭakavarga fragment, which measures only ca. 11 cm across. In writing style, this is certainly the oldest one met with in the collection. With the chronological problem in view, a small fragment was subjected to a C14 test at the Leibnitz Labor at Kiel, Germany (no. KIA 32298, dated 2.11.2007). The result was a two sigma range “cal BC 184-46” with a probability of 95.4 %. The youngest peak is placed around 70 BC. At present, such an early date is difficult to digest. Either C14 from South-Asia reacts different from Western material, or there is something we do not know (yet). At the Atlanta Conference, R. Salomon presented a similar C14 result for an inscribed bark from a different collection. Since we now have no difficulty anymore to place everything pertaining to “Mahāyāna” Buddhism in the first century AD and even earlier, we might well be forced to reckon these Avadānas amongst the oldest Buddhist inscribed birchbark. 19 PDF Version: ARIRIAB XIV (2011), 13-23 5. A Prajñāpāramitā (fig. 6) This last birchbark once measured ca. 80 cm in length; the preserved width is ca. 15 cm. Both margins have suffered from wear and are not preserved anywhere. The verso is inscribed too for about 60 % and shows the text end together with a colophon. The letters are plain and simple, no frills, almost no footmarks and almost twice in height compared to the texts referred to above. No ‘modern’ train is seen; śa and ya are clearly separated, the preconsonantal r- is a hook and not a loop, the postconsonantal -r-bar straight, not curved. On palaeographic ground a date in the first century AD would not surprise. A C14 text was undertaken again at the Leibnitz Labor, Kiel, under number KIA 26906, dated 2.8.2005, with the result of a calibrated age of AD 74, won though two-sigma ranges from AD 25-43 (probability 14.3%) and AD 47-147 (probability 81.1%). The recto contains a text correlating with the first chapter of the Aṣṭasāhasrikā Prajñāpāramitā (APP), the text on the verso is directly related to the fifth chapter of the APP. A comparison with the Chinese translation of Lokakṣema, which is dated AD 179/180, and the classical Sanskrit version as translated by Kumārajīva clearly shows a development from a simple to a more developed text. The Gāndhārī text looks archaic and is less verbose than what Lokakṣema translated. It can be shown that his version was already slightly inflated by the insertion of stock phrases, appositions and synonyms. The Sanskrit version, finally, expanded still further. I would like to demonstrate this development through three examples. In an excerpt from the first chapter it is said in Gāndhārī (left) and in Sanskrit (right): cito na oliati ṇa viparaṭhi bhavati ṇa saṃtrāso avajati cittaṃ nāvalīyate na saṃlīyate na viṣīdati na viṣādam āpadyate nāsya vipṛṣṭhībhavati mānasaṃ na bhagnapṛṣṭhībhavati notrasyati na saṃtrasyati na saṃtrāsam āpadyate We see an expansion from three to nine phrases, where already the three expressions in the Gāndhārī text are nothing but repetitions of the same idea. Lokakṣema occupies a place in the middle, saying 菩薩聞是心不懈怠。不恐不怯不難不畏, that means he has just five phrases, which is an expansion seen from the Gāndhārī side, and which is shorter than Sanskrit with its nine phrases. A longer quotation from the first chapter provides the same picture. First comes the Gāndhārī text in italics, below it the parallel phrase from the standard Sanskrit version. In both versions only those parts of it that are found in Lokakṣema are highlighted in bold: 20 PDF Version: ARIRIAB XIV (2011), 13-23 saye hi bosisatvasa ° bodhisattvasya mahāsattvasya saced prañaparamida uadiśamanae ° evaṃ gambhīrāyāṃ prajñāpārāmatāyāṃ bhāṣyamāṇāyāṃ deśyamānāyām upadiśyamānāyā cito ṇa oli/// na viparapa/ṭhi bhavati cittaṃ nāvalīyate na saṃlīyate na viṣīditi na viṣādam āpadyate nāsya vipṛṣṭhībhavati mānasam na bhagnapṛṣṭhībhavati nottrasyati na saṃtrasyati ///satraso avajati ° [Lokakṣema with 5 negations] na saṃtrāsamāpadyate / adhimucyate ’dhyāśayena avirahito bodhisattvo mahasattvaḥ prajñāpāramitāyā veditavyaḥ thida bosisatvo ° niyuto ° avevaṭiae dhaṃdue ° sthito pun ’vinivartanīyāyāṃ bodhisattvabhūmau susthito ’sthānayogena avaro bhaṃte bhagava /// punar aparaṃ bhagavan bodhisattvena mahāsattvena /// paramidae caṃraṃtena ° prajñāpāramitāyāṃ caratā prajñāpāramitāṃ bhāvayatā rua ti na thadavo ° na rūpe sthātavyaṃ [ no parallel in Gāndhārī ] na vedanāyāṃ na saṃjñāyāṃ na saṃskāreṣu na vijñāne sthātavyam / tat kasya hetoḥ saye rua ti ṭhaveti rua [avisa....] /// saced rūpe tiṣṭhati rūpābhisaṃskāre carati na carati prajñāpāramitāyām / We see that Lokakṣema very often contains nothing but the contents found in Gāndhārī; in the last but one line there is a phrase in Sanskrit which Lokakṣema also had in his exemplar, but which is either not yet found or dropped in Gāndhārī. In one case there is a clear difference. While Gāndhārī speaks of a dhātu in: thida bosisatvo ° niyuto ° avevaṭiae dhaṃdue, the Sanskrit version avoids this term and presents a bodhisattvabhūmi instead. Lokakṣema’s Chinese translation (以入阿惟越致中悉了知。 不 可 復 退 ) reads as follows: “(Having heard of it, if a bodhisattva does not become slothful in mind, embarrassed, afraid, nor fearful,) then he will enter into (the state of) 21 PDF Version: ARIRIAB XIV (2011), 13-23 avivartika-(ship), understand completely and never retrogress.” Therefore, one can assume that the original text of the Chinese translation also did not read bodhisattvabhūmi, but it is unclear whether it contained dhātu or not. Another example from the fifth chapter compares only the Sanskrit version: ///sa ima yeva ° prañaparamida likhita uvaṇamea enāṃ prajñāpāramitāṃ likhitvā dadyāt upanāmayet atra prañaparamidae ///vajiśati śikṣi/// atraiva prajñāpāramitāyāṃ śikṣiṣyate yogam āpatsyate / eva sa prañaparamida bhuyasa matrae evam asyeyaṃ prajñāpāramitā bhūyasyā mātrayā bhavana parivuri gachiśati bhāvanāṃ vṛddhiṃ virūḍhiṃ vipulatāṃ paripūrīṃ gamiṣyatīti / aya te< ṇa purimake>ṇa ayaṃ kauśika tataḥ paurvakāt puñavisaṃkhareṇa bahudaro sakāśād kulaputrataḥ kuladuhitṛto vā puño prasavati bahutaraṃ puṇyaṃ prasavet / [ no parallel in Gāndhārī ] tat kasya hetoḥ / niyatam eṣo 'nuttaraṃ samyaksaṃbodhim abhisabudhya sattvānāṃ duḥkhasyāntaṃ kariṣyatīti // pun avaro kośiga . . . punar aparaṃ kauśika . . . Again we see additions of explanatory matter. Interesting is the change from an old Prakritic purimaka to Skt. paurvaka, where an inverse development is hard to imagine. The dogmatically difficult term puṇyavisaṃskāra, “dissolution (visaṃkhara) of puṇya”, is replaced by a plain “presence” (sakāśa), at least retaining a phonetical resemblance. These examples could be supplemented by many others. An edition combining the Gāndhārī readings from the birch-bark, the Chinese translatios by Lokakṣema, Kumārajīva etc. and the standard Sanskrit version, will be published in next year’s issue of ARIRIAB as a joint work of S. Karashima and myself. This will analyse and demonstrate the steps this text has taken from the so far oldest version in Gāndhārī, to Lokakṣema’s exemplar and further to the standard Sanskrit version. 22 PDF Version: ARIRIAB XIV (2011), 13-23 I conclude with the last lines of this text, with its colophon, reading: paḍhamae postae prañaparamidae budhamitra /// indraśavasa sadhaviharisa imena ca kuśalamūlena sarvasatvaṇa matrapitra (...) “In this first book of the prajñāpāramitā (of?) Buddhamitra (and NN?), the co-student of Indraśravas. By this root of bliss (may there be well-being?) for all people (and?) for mother and father (...).” This colophon shows that there were more books; it also shows that the title Prajñāpāramitā is not accompanied by a number, as are all the following versions. The role of Buddhamitra is unclear: is he author, scribe or sponsor? In epigraphical texts, kuśalamūla occurs only some time after the first century; here, this puṇya-related term is already used in a formulaic way. The greatest problem the “Split collection” poses is its nature: it can only be hoped that more parts can be found or given access to, to combine “Split-A” with “SplitB” etc, until it is split no more. References: Brough, John 1962 The Gandhari Dharmapada. London: Oxford University Press. Cone, Margaret 1989 “Patna Dharmapada, Part I: Text.” Journal of the Pali Text Society 13: 101-217. Falk, Harry 2008 “Another reliquary vase from Wardak and consecrating fire rites in Gandhāra.” Claudine BautzePicron (ed.), Religion and Art: New Issues in Indian Iconography and Iconology (Volume 1 of the proceedings of the 18th conference of the European Assiciation of South Asian Archaeologists, London 2005). London (The British Association for South Asian Studies): 63-80. Hinüber, Oskar von & K.R. Norman 1995 Dhammapada. Oxford (The Pali Text Society). Nakatani, Hideaki 1987 Udānavarga de Subaši: édition critique du ms. sanskrit sur bois provenant de Subaši; Bibliothèque National de Paris, Fonts Pelliot, tôme 1: texte et facsimilés (Publications de l’Institut de Civilisation Indienne, 53). Paris (de Boccard). Salomon, Richard 2009 “Observations on the Reliquary Slab Inscription of Gomitra.” ARIRIAB 12: 7-19. 23 PDF Version: ARIRIAB XIV (2011) H. Falk, “The ‘Split’ Collection of Kharoṣṭhī Texts,” pp. 13-23. PLATE 7 Figure 1: Aṭṭhakavagga, Side 1 Figure 2: Aṭṭhakavagga, Side 2 Figure 3: A metrical text Figure 4: Another Gāndhārī Dharmapada ..si ..ajara jivamaṇeṇa ḍaśamaṇeṇa ṇiv[uti] /// yovakṣemo aṇuta ajaro jivamaṇeṇa ḍaśamaṇeṇa ṇivuti ṇimesa paramo śodhi yoyakṣemo aṇutaro PDF Version: ARIRIAB XIV (2011) PLATE 8 Figure 5: An Avadāna collection Figure 6: A Prajñāpāramitā Figure 7: Colophon