A preliminary survey of some early Buddhist manuscripts recently
acquired by the British Library

Richard Salomon

The Journal of the American Oriental Society
Vol.117 No.2 (April-June 1997)

COPYRIGHT 1997 American Oriental Society

            1. Introduction; general description of the manuscripts 
            The Oriental and India Office Collections of the British Library 
            have recently acquired, with the assistance of an anonymous 
            benefactor, a substantial collection of early Buddhist texts written 
            on birch-bark scrolls in the Gandhari or Northwestern Prakrit 
            language and the Kharosthi script. The original provenance of the 
            manuscripts is not known, but may be Afghanistan, in view of certain 
            resemblances (discussed below) to other materials previously found 
            The manuscripts comprise thirteen rolls of birch bark which had been 
            removed from their original container. According to verbal reports, 
            they were originally found inside one of a group of five large clay 
            pots, each bearing a Kharosthi dedicatory inscription, which have 
            also been acquired by the Library. The bark rolls are extremely 
            fragile and, in fact, had already been seriously damaged, in that 
            substantial portions of one vertical edge of most of the manuscripts 
            had been destroyed. When acquired by the Library, the scrolls were 
            still in their original rolled-up state, and the exceedingly 
            delicate task of unrolling them was successfully carried out by the 
            conservation staff of the British Library. This has now made it 
            possible to prepare preliminary photographs of the manuscripts, an 
            example which is shown in figure 1, and to conduct a provisional 
            survey of their contents. 
            The scrolls proved to consist of birch-bark strips, typically about 
            five to nine inches in width, on which the texts were written in 
            black ink. The long scrolls were built up out of shorter strips, 
            apparently around twelve to eighteen inches long, which were 
            overlapped and glued together, as shown by blank spaces in several 
            fragments in which the original strips have separated. The scrolls 
            were reinforced by a thread sewn along both margins. In a few cases 
            traces of the original thread are preserved, and in many places the 
            needle holes along the margins are still visible. 
            Typically, the scribes began writing at the top of the recto, 
            continued to the bottom of the recto, and then reversed the scroll 
            both from top to bottom and from front to back and continued writing 
            from the bottom edge of the verso back to the top of the scroll. 
            This means that the texts both began and ended at the top of the 
            scroll, which would be on the outside when it was rolled up from the 
            bottom. But this is precisely the part of the scroll that is most 
            subject to wear and tear, especially in the case of a fragile 
            material like birch bark, which becomes extremely brittle when it 
            dries out. The unfortunate result is that, but for one fragmentary 
            exception, we do not have the beginning or end of any scroll, or the 
            label or colophon that might have accompanied it. Virtually all the 
            surviving material, in other words, is from the middle and bottom of 
            the original scrolls. This situation is apparently not due to damage 
            inflicted since they were recently rediscovered, but probably 
            reflects their already imperfect condition when they were interred 
            in antiquity (as discussed in part 2). The surviving sections of the 
            scrolls range in length from mere fragments of a few lines or even a 
            few letters to substantial, though still incomplete, portions of 
            complete scrolls. The longest intact section of a single scroll is 
            about eighty-four inches long. 
            For all these reasons, the condition of the manuscripts is only fair 
            at best, and often much worse than that. All are incomplete, and 
            many are mere fragments. Moreover, in most cases the delicate 
            surface of the bark is peeled, faded, discolored, or otherwise 
            damaged, so that it can be difficult or, not infrequently, nearly 
            impossible to decipher the texts. Even where the texts are more or 
            less legible, they contain, almost without exception, frequent and 
            sizable lacunae. 
            2. Constitution, disposition, and affiliation of the manuscripts 
            It has already become clear in the course of the preliminary 
            cataloguing of the manuscripts that the original thirteen rolls do 
            not all constitute single texts or scrolls. Although some of them do 
            contain the remnants of a single scroll, several proved to contain 
            fragments, of widely varying size, of two, three, or even more 
            originally separate scrolls. In several cases it was also noticed 
            that separate fragments of the same text, and presumably of the same 
            original scroll, were found in two or more of the thirteen rolls. 
            And in at least one case, a scroll was broken in half lengthwise and 
            the two long narrow halves of the text were placed in different 
            rolls. This lengthwise splitting of the original scroll probably 
            resulted from its having been bound by a string or ribbon and left 
            untouched for a long period in antiquity, with the result that, as 
            the bark became dry and brittle, the binding cord cut through and 
            divided it in half. 
            These peculiarities of the condition and disposition of the texts 
            all point to the conclusion that these manuscripts were already in 
            fragmentary or damaged condition in antiquity, before they were 
            interred in the clay jar in which they were reportedly discovered. 
            This implies that they were discarded worn-out texts, an impression 
            which is confirmed by the observation that five of them have 
            secondary interlinear notations, in hands clearly different from 
            those of the original scribes, reading likhidago, "[it has been] 
            written," likhidago sarva "[it has] all [been] written," and the 
            like (see fig. 2). These interlineations seem to be notations by 
            later copyists who had rewritten the texts onto new manuscripts and 
            marked the old ones as "copied," i.e., as ready to be discarded. 
            Such discards were then rolled up together, apparently more or less 
            at random, placed inside clay pots, and buried, perhaps in a small 
            stupa within the precincts of the monastery to which they belonged. 
            Such a practice is attested by earlier discoveries, such as those at 
            Hadda, in eastern Afghanistan, where Barthoux (1933: 60) found 
            similar clay pots containing, in some cases, fragmentary remains of 
            birch-bark manuscripts, and in others, pieces of human bone. It thus 
            appears that the relics of venerable monks and of Buddhist texts 
            were conceived and treated similarly as sacred objects deserving of 
            ritual interment. 
               What we have in this new collection, in other words, is, in all 
            likelihood, something roughly analogous to the genizah of Jewish 
            tradition, that is, a collection of discarded documents for which 
            religious law or custom required a ritual interment. The source of 
            these discarded texts was no doubt the library, or perhaps rather 
            the scriptorium, of a Gandharan Buddhist monastery, probably an 
            establishment of the Dharmaguptaka sect. This affiliation is 
            indicated by the inscription on the jar in which the scrolls were 
            reportedly found, which records its dedication to members of that 
            sect (dhamauteana parigrahami, "in the possession of the 
            Dharmaguptakas"). Although this sect has hitherto been only very 
            sparsely attested in the northwest, this and several other recent 
            discoveries, including several that have not yet been published, of 
            Kharosthi inscriptions recording donations to the Dharmaguptakas 
            indicate that they were a major sect in that region, particularly in 
            All in all, the preliminary survey revealed that the thirteen 
            original rolls of manuscript material contained thirty-two separate 
            "fragments," a fragment being here defined as a piece, of any size, 
            of an originally separate scroll. However, it was further 
            determined, by connecting separated fragments on the grounds of 
            similar handwritings and contents, that these thirty-two "fragments" 
            actually stemmed from about twenty-two different original scrolls. 
            The number of separate texts, however, is larger, probably about 
            twenty-six, because some scrolls contain two, and possibly even more 
            than two, separate and apparently unrelated texts. In many such 
            cases, it appears that the first scribe used only the recto, which 
            was apparently the preferred writing surface, and ended at or near 
            the bottom. Another scribe, perhaps at a later date seems to have 
            used the empty surface at the bottom of the recto and the completely 
            blank verso to record another text. 
               These figures, at this point, are only provisional, and will 
            almost certainly have to be adjusted as a result of the more 
            detailed studies of the manuscripts, but they are sufficiently 
            secure to give a general idea of the extent of the collection. 
            Although this is presumably only a small fraction of the total 
            amount of literature in the monastery's library, it should prove to 
            be enough to give at least a partial view of the contents of such a 
            3. Identification and classification of the texts 
            Identification and classification of the texts is still for the most 
            part at a preliminary stage, and only a few of them have been 
            positively identified with parallel texts in Pali, Sanskrit, 
            Chinese, or Tibetan. But the major genres of Buddhist canonical and 
            paracanonical literature represented by this collection have become 
            clear, at least in general outline. Most of the texts which are 
            sufficiently legible to be analyzed in the preliminary survey seem 
            to fall into the following categories: 
            1. Didactic or popular poetry, such as a Gandhari version of the 
            "Rhinoceros' Horn Sutra," a well-known poem otherwise preserved in 
            Pali as the Khagga-visana-sutta of the Sutta-nipata, and in Sanskrit 
            incorporated into the Mahavastu (ed. E. Senart, 1.307-9). 
            2. Avadana texts describing the past lives and karmic background of 
            various Buddhist personages, for example, a collection of stories 
            describing the previous incarnations (provayoge = Sanskrit 
            purvayogah) of Ajnata-kaundinya, Ananda, and the Buddha himself. 
            3. Canonical sutra texts and commentaries thereon, for instance, a 
            Gandhari version of the Sangiti-sutra (also extant in Pali, 
            Sanskrit, and Chinese) with an unidentified commentary. 
            4. Abhidharma texts, as yet unidentified. 
            5. Stotra text (only one fragment). 
            It may seem surprising that no Vinaya material at all has been found 
            in this substantial body of manuscripts. BUt a similar lacuna has 
            been noted among the oldest of the Central Asian Sanskrit 
            manuscripts, and Sander (1991: 141-42) has plausibly hypothesized 
            that the Vinaya texts were preserved by oral recitation and not 
            normally set down in writing in early times. It is possible, of 
            course, that the absence of Vinaya texts among these new manuscripts 
            is merely coincidental, "the luck of the draw," as it were, but I 
            think it more likely that there were few if any Vinaya manuscripts 
            in our hypothetical complete monastic library, for reasons similar 
            to those adduced by Sander. 
            4. Date of the manuscripts 
            Certain considerations point to a possible date for the manuscripts 
            as early as the first half of the first century A.D. The first of 
            these is a clear reference, though in an uncertain context in a 
            fragmentary text, to jihonige mahaksatra . . . (see fig. 3). Here a 
            reconstruction such as mahaksatra(pe*) 'great satrap' is obvious, 
            and there can be little doubt that the reference is to the 
            Indo-Scythian satrap Jihonika, who is known from his coins and from 
            the Taxila silver-vase inscription (Konow 1929:81-82), and who is 
            likely to have ruled around the fourth decade of the first century 
            A.D. (MacDowall 1973: 229). Of course, this reference to a 
            contemporary historical figure, which is a (pleasant) surprise 
            though not completely without parallel in Buddhist tradition, only 
            establishes that the text in question was originally composed during 
            or after the reign of Jihonika, but not necessarily that our actual 
            manuscript was written in or around his time. 
               But the dedicatory inscription on one of the clay pots associated 
            with the manuscripts (though not, apparently, the one in which they 
            were found) also points to a date in the early first century A.D. 
            This inscription records its donation by a woman named Vasavadatta, 
            the wife of Susoma or Suhasoma (. . . deyadharme vasavadatae 
            susomabharyae. . . . svamiasa suhasomasa sammepratyasae . . . 
            bhavatu, ". . . the pious gift of Vasavadatta, wife of Susoma . . . 
            . May it be for the principal share [of merit] for [her] husband 
            Suhasoma"). Both of these names match with ones known from other 
            inscriptions datable to the early first century. Vasavadatta is 
            given as the name of the sister of the Apraca prince Indravarman in 
            his reliquary inscription of the Azes year 63 =6 A.D. (Salomon and 
            Schopen 1984: 108-9). Suhasoma appears as the name of a royal 
            kinsman (anakaena) and official (asmanakarena) of King Senavarman of 
            Odi in his gold-leaf inscription (Salomon 1986: 265), which is 
            undated but attributable from its reference to the Kusana overlord 
            Kujula Kadphises to the first half of the first century. Of these 
            two names, the second in particular, Suhasoma, is an unusual one and 
            therefore very likely to refer to the same person in the two 
            inscriptions where it occurs. 
            Unfortunately, the chronological significance of this inscription on 
            jar A is vitiated by the lack of any reliable evidence as to the 
            archaeological relationship of that jar with the jar (D) in which 
            the scrolls were found. While there is reason to believe that both 
            may have come from the same site, and hence may be more or less 
            contemporary, there is no way to establish this. Other criteria, 
            such as paleographic and linguistic features, indicate a dating 
            range for the manuscripts from the beginning of the first century to 
            the first half of the second century A.D. Thus, although it cannot 
            be proven at this point, there is some reason to think that they 
            date from the earlier part of the range, i.e., from the first half 
            of the first century A.D. The possibility of such a date for this 
            group of relics has been confirmed, or at least not contradicted, by 
            thermoluminescence testing of the clay pots, which indicated a 
            dating range from the first to the eighth centuries A.D., with a 10% 
            margin of variation and no weighting implications for any period 
            within this broad span. 
            5. Relationships to previous discoveries 
            Though unprecedented, the discovery of a large corpus of Buddhist 
            texts written on birch-bark scrolls in the Gandhari language and 
            Kharosthi script is not entirely unexpected. Only one more-or-less 
            intact manuscript of this type has previously come to light, namely 
            the "Gandhari Dharmapada," definitively published in Brough 1962, 
            which was discovered in 1892 near Khotan (now in Xinjiang Autonomous 
            Region, China). The new manuscripts are broadly similar in form, 
            age, and contents to the Gandhara Dharmapada, though there are some 
            significant differences in the details of such features as language, 
            orthography, and arrangement of the text. 
               But besides the well-known case of the Gandhari Dharmapada, there 
            have apparently been several other examples of similar materials 
            found in Afghanistan and Pakistan, though none of these has ever 
            been properly published; therefore, they have gone almost entirely 
            unnoticed in scholarship on the relevant fields. Thus, the early 
            archaeological explorer of the northwest, Charles Masson (in Wilson 
            1841: 59-60), reported that some of the Buddhist reliquaries which 
            he found in eastern Afghanistan were "accompanied by twists of 
            tuz-leaves, inscribed internally with characters"; Wilson, in an 
            editor's note (p. 60, n. 1), explained that "it seems likely that 
            what Mr. Masson denominates 'tuz-leaves' is the inner bark of the 
            Bhurj or birch tree, which was very commonly used for writing upon." 
            Barthoux, in his excavations at Hadda (1933:60-61), discovered 
            numerous Kharosthi texts on birch bark, including some contained in 
            clay pots like the newly discovered manuscripts. The fate of these 
            manuscripts is described in his own words: "ces fragments, tres 
            fragiles, etaient deja broyes par les decombres, et en les retirant, 
            malgre toutes les precautions prises, l'on achevait de les detruire" 
            (p. 61), and this explains why these, and probably the other similar 
            discoveries as well, were never properly published. 
            6. Implications for the study of Buddhist literature and canons 
            Mainly on the basis of the evidence of the Gandhari Dharmapada, and 
            secondarily on the grounds of inscriptional testimony of what seem 
            to be Gandhari versions of Buddhist texts (Brough 1962: 42) and of 
            evidence from the Chinese Buddhist tradition, it has been proposed 
            that there may have existed a Buddhist canon in Gandhari, of which, 
            until now, only a few fragments have survived. Thus Brough 
            concluded, with due caution, that "the existence of this [Gandhari] 
            Dharmapada does imply the existence of a canon of which it formed a 
            part" (p. 43). The new discovery thus confirms what already seemed 
            likely, namely that the Gandharan Buddhists in the early centuries 
            of the Christian era did have a substantial corpus of written 
            scriptures in the Gandhari language, comprising a considerable 
            variety of genres ranging from didactic poetry to scholastic 
            As to the contents, arrangement, and affiliations of this canon, it 
            would be premature to make anything more than some very partial and 
            provisional observations. Broadly speaking, it appears to represent 
            early northern Indian Buddhist teaching and practice; nothing has 
            been found in the texts to suggest anything like Mahayana doctrinal 
            developments. This is in accord with the apparent connection of the 
            scrolls with an establishment of the Dharmaguptakas (see sec. 2), a 
            non-Mahayana sect generally understood to be affiliated with the 
            Some, though by no means all, of the texts have either direct 
            parallels or partial similarities to portions of the Pali canon or 
            to Chinese translations of northern Indian Buddhist texts. Of 
            special interest is an apparent concentration of texts parallel or 
            related to various parts of the Pali Khuddaka-nikaya, and especially 
            the Sutta-nipata. These include the aforementioned Gandhari version 
            of the "Rhinoceros' Horn Sutra" (Khagga-visana-sutta), which appears 
            as the third sutra in the Pali Sutta-nipata, as well as a commentary 
            on a sequence of verses, most of which correspond to passages from 
            various sections of the Sutta-nipata. Another scroll preserves a 
            small fragment that matches well with the concluding portion of the 
            Bhiksu-varga of the Gandhari Dharmapada, which in turn closely 
            resembles the Uraga-sutta, the first sutta of the Pali Sutta-nipata. 
            This pattern of close association with the Sutta-nipata is of 
            special interest because the Sutta-nipata generally, and certain 
            parts of it in particular (including the Khaggavisana-sutta and the 
            Uraga-sutta), have long been felt by Buddhist scholars to represent 
            one of the earliest strata of the Pali canon, on the grounds of 
            their numerous linguistic, stylistic, and doctrinal archaisms. The 
            apparent concentration of Sutta-nipata-related texts in the new 
            Gandhari corpus thus is likely not only to confirm the long-standing 
            hypothesis of the antiquity and importance of this collection, but 
            also to illuminate its textual history and role in the propagation 
            of early Buddhism. This, of course, is only one example of the many 
            contributions which the new documents can be expected to make to the 
            study of Indian Buddhist textual and doctrinal history. 
            7. Linguistic and paleographical features 
            The new documents should also prove to be highly useful for 
            linguistic and paleographic studies. As might be expected, the 
            various manuscripts show considerable divergences and 
            inconsistencies in their renderings of the Gandhari language, 
            reinforcing the impression gained from the previously known 
            specimens, mostly epigraphical, that the language was never fully 
            standardized or regularized. However, these differences, though 
            considerable, are likely to be more on the level of orthography than 
            of actual dialectal or chronological variation. Examples of notable 
            orthographic or dialectal peculiarities which were previously 
            attested only sporadically, if at all, in Gandhari documents include 
            the replacement, in one set of texts in the same hand, of g by gh in 
            all cases. We also find in several documents the use of the 
            subscript pre-consonantal form of r to denote, apparently, a 
            geminate consonant; for instance, in the Rhinoceros' Horn Sutra MS 
            the word for "rhinoceros" (Sanskrit khadga) is regularly spelled 
            (according to conventional transcription) kharga, which seems to 
            reflect the pronunciation khagga. 
            Also worthy of note is the absence of certain dialectal/orthographic 
            features peculiar to the Gandhari Dharmapada, such as its 
            distinctive treatment of combinations of nasal + homorganic stop of 
            the type vinadi = Sanskrit vindati (Brough 1962: 98-99). These 
            contrasts make it clear that the linguistic and orthographic 
            peculiarities of the Dharmapada text do not represent a simple 
            contrast between literary and epigraphic Gandhari, as it might have 
            seemed until now. Detailed linguistic and paleographic study of the 
            new documents should gradually clarify the complex patterns of 
            development of Gandhari as a literary language. 
            8. Plans for study and publication 
            In 1996, the British Library and the University of Washington 
            entered into a formal cooperative agreement in order to facilitate 
            the efficient and systematic study and publication of this new 
            collection of early Buddhist manuscripts, once again with the 
            assistance of an anonymous donor. The goal of the project is to 
            coordinate the preparation of a series of volumes, to be published 
            by the British Library, containing editions, translations, and 
            studies of the texts. An initial introductory volume containing a 
            detailed description and survey of the collection is currently under 
            preparation and is to be published as soon as possible. This is to 
            be followed by the first text volume, which will present the 
            "Rhinoceros' Horn Sutra" and associated texts. Plans for further 
            volumes, including a projected facsimile edition of the manuscripts, 
            are currently under discussion. 
            Barthoux, J. 1933. Les Fouilles de Hadda, I: Stupas et Sites. Text 
            et Dessins. Memoires de la Delegation archeologique francaise en 
            Afghanistan 4. Paris: Les Editions d'art et d'histoire. 
            Brough, John. 1962, The Gandhari Dharmapada. London Oriental Series 
            7. London: Oxford University. 
            Konow, Sten. 1929. Kharoshthi Inscriptions with the Exception of 
            Those of Asoka. Corpus Inscriptionum Indicarum, vol. II, part 1. 
            Calcutta: Government of India. 
            MacDowall, David W. 1973. "The Azes hoard from Shaikan-Dheri: Fresh 
            Evidence for the Context of Jihonika." In South Asian Archaeology: 
            Papers from the First International Conference of South Asian 
            Archaeologists Held in the University of Cambridge, ed. Norman 
            Hammond. Pp. 215-30. London: Duckworth. 
            Salomon, Richard. 1986. "The Inscription of Senavarma, King of Odi." 
            Indo-Iranian Journal 29: 261-93. 
            Salomon, Richard, and Gregory Schopen. 1984. "The Indravarman 
            (Avaca) Casket Inscription Reconsidered: Further Evidence for 
            Canonical Passages in Buddhist Inscriptions." Journal of the 
            International Association of Buddhist Studies 7: 107-23. 
            Sander, Lore. 1991. "The Earliest Manuscripts from Central Asia and 
            the Sarvastivada Mission." In Corolla Iranica: Papers in honour of 
            Prof. Dr. David Neil MacKenzie on the occasion of his 65th birthday 
            on April 8th, 1991, ed. Ronald E. Emmerick and Dieter Weber. Pp. 
            133-50. Frankfurt am Main: Peter Lang. 
            Wilson, H. H. 1841. Ariana Antiqua: A Descriptive Account of the 
            Antiquities and Coins of Afghanistan. . . London: East India