The Tshal pa and the *Zha lu ma manuscripts are the twin fonts from which most of the later standard editions of the bKa' 'gyur appear to flow, hence the division of the bKa' 'gyur tradition as we now know it40 into what have been called the "Eastern" and "Western" branches. Identifying this bifurcation, and making a start at sorting out the twists and turns on both sides of the tradition has been the major achievement of recent bKa' 'gyur scholarship, above all that of Eimer (see especially Eimer, 1992), followed more recently by several other scholars. This scholarship brings three basic methods to bear on the problem of determining the affinities of the various accessible editions. The first is to examine Tibetan histories, biographies and the catalogues of these editions (dkar chag; see Martin, in this volume) for information relating to their creation; the second is to note carefully the order of sections and individual titles within the editions, since this can also indicate affinities; and the third is to apply classical text-critical technique to the problem, by editing individual texts, i.e., collating as many editions as possible and noting patterns of variants. Given the vastness of the bKa' 'gyur tradition, it is little wonder that these methods have not yet yielded all the answers, and that many problems remain unsolved. At the same time, some progress has been made. What follows is, I hope, a reasonably accurate and reliable reflection of our present state of knowledge.

On the so-called "Western" side of the picture the *Zha lu ma passes from the realm of hypothesis into that of historical fact in the form of the manuscript bKa' 'gyur which was made in 1431 on the order of the ruler Rab brtan Kun bzang 'phags pa (1389-1442) and deposited in the dPal 'khor chos sde Monastery at rGyal rtse.41 This is known as the Them spangs ma Manuscript. Complete in 111 volumes, it did not include the rNying rgyud collection. There is no doubt that some of its sections were edited by Bu ston and[page 81] his successors at Zha lu, but the provenance of others is not yet known. Whether the original still exists is a matter of some uncertainty, but there are still several old manuscripts at rGyal rtse, and one of these could be it. The Them spangs ma is extremely important, for it was much copied; during the reign of the fifth Dalai Lama alone (1617-1682), over a hundred copies were made. One such copy was presented to the Mongols in 1671, and now rests in the State Library at Ulan Bator.42 Another was made during the years 1858-1878 and later donated to the Japanese monk and traveller Kawaguchi Ekai; this is now in the possession of the Tōyō Bunko, Tokyo. These are two recognized copies of the Them spangs ma, but we also have to reckon with the many others which were made, and the copies which were made from them. Into this category fall the London Manuscript bKa' 'gyur, which derives from a manuscript held at Shel dkar chos sde,43 and the sTog Palace bKa' 'gyur, which was copied from a Bhutanese exemplar (Skorupski).44 No doubt many more of these copies will eventually come to light. The best general term for all these manuscripts is "the Them spangs ma tradition."

On the other ("Eastern") side of the picture the Tshal pa manuscript provided the basis for the first xylographic or woodblock print of the bKa' 'gyur, the Yongle edition made in Beijing in 1410. At this point the printing technology first invented by the Chinese largely for the purposes of propagating Buddhist literature was enthusiastically adopted by the Tibetans, who were to continue to use it up to the twentieth century, not least to produce ever more editions of the bKa' 'gyur (cf. Snellgrove and Richardson: 160). In Beijing new impressions continued to be taken from the Yongle blocks, and when they wore out, new blocks were prepared, using prints struck from the old blocks as masters. Minor alterations were sometimes made when this was done. In this way were produced the Wanli impression of 1605, the Kangxi impressions of 1684/92, those of 1700, 1717-1720, the Qianlong impression of 1737, and at least one further impression after 1765.45 But these are not the only offspring of the Tshal pa, for a copy of it kept at the castle of 'Phying ba sTag rtse in 'Phyong rgyas, a copy which must have received further editorial attention, was the basis for the 'Jang Sa tham or Li thang edition in 110 volumes of 1609-1614, which has only recently become available in the West.46 The same 'Phying ba sTag rtse Manuscript must also have been the basis for some of the sNar thang blockprint of 1730-1732 (on[page 82] which see below).47 The Li thang was in its turn the basis for the Co ne edition (107 volumes) of 1721-1731. A convenient term for all these editions is "the Tshal pa tradition."

So far all this looks relatively neat, but in fact we have as yet made no mention of the whole question of what is technically known as "contamination." Contamination occurs when one text is not copied from another in a simple linear progression, but instead mixes readings from two or more exemplars, or "conflates" them. In such a situation parentage is often difficult to trace. The later bKa' 'gyur tradition is in fact bedeviled by contamination, due in part to the great pains the compilers of new editions took to ensure that their text was as sound as possible, which they did by consulting as many reputable old editions as they could lay their hands on. Thus the block-print edition in 104 volumes produced in 1733 at the Sa skya pa monastery of sDe dge, which took as its base text the Li thang, also borrowed readings from the lHo rdzong bKa' 'gyur, a descendant of the Them spangs ma, as well as from a bKa' 'gyur produced by A gnyen pa kshi. The sDe dge xylograph thus represents a conflation of the two main branches of the tradition, as do its later offshoots, the Ra rgya (1814-1820), the Urga (1908-1910) and the Wa ra editions (twentieth century).48 Similarly, later reprints of the Peking edition often altered the text of the blocks with reference to the Li thang, while the modern Lhasa edition, produced in 1934, is widely known to be a conflation of sDe dge and sNar thang.49 The sNar thang blockprint edition of 1730-1732, however, is the most unusual case of mixed parentage, since although it takes its texts from at least two separate editions, it does not apparently conflate their readings: text by text, it seems to follow one edition or the other scrupulously. Text-critical research by Eimer and others has only recently enabled us to identify the sNar thang xylograph's two sources: one of them is the 'Phying ba sTag rtse manuscript of the Tshal pa edition,50 and the other is the Shel dkar copy of the Them spangs ma, on which the London Manuscript was based.51 What remains to be worked out is which texts it took from which sources, and whether we can identify the point where it switched from one to the other. At this stage it appears that the 'Dul ba section follows the Them spangs ma, while most of the mDo follows the Tshal pa (making the sNar thang in this respect a sister of the Li thang). Evidence for the rGyud section is sparse. We should note, however, that the sNar thang follows the basic order of the Tshal pa editions. The way in which[page 83] this edition was produced is a good illustration of the care the Tibetan editors took over their work, and of the sophistication of their approach. The same is true of sDe dge. Using these bKa' 'gyurs to edit texts ourselves, we are impressed by the extremely small number of errors which they introduced into the tradition, even though they have complicated our task somewhat by conflating their sources. One other point which needs to be noted in connection with these later printed editions is that the Tibetan canon was never entirely "closed," and that editors of the bKa' 'gyur seem to have had few qualms about adding recently translated or discovered works to existing editions. Texts were still being translated in the seventeenth and eighteenth centuries, albeit not at the prodigious rate of earlier periods.

This picture of the history of the bKa' 'gyur, as complex as it is, may soon need to be revised and elaborated. First of all, new bKa' 'gyurs continue to come to light, some of which do not fit at all well into this scheme. This is, for example, the case with the most recent arrival in the West, the Phug brag (also spelled Phu brag, sPu brag, sPud tra, etc.).52 In terms of organization this edition, produced ca. 1700, follows neither the Thems spang ma nor the Tshal pa traditions, it contains texts found in no other bKa' 'gyur, and it carries multiple translations of works. Since it has only recently become available, not much text-critical work on individual titles within this collection has been done, but what little research there is suggests an independent tradition, which is sometimes closer to the Them spangs ma, sometimes to the Tshal pa editions.53 In the second place, studies of the Tibetan sūtra translations found at Dunhuang, which date from the eighth to the eleventh centuries, are showing us that at its very beginnings the tradition which was later to become known as the bKa' 'gyur was not at all uniform, but highly contaminated. The Dunhuang collection is in fact a confusing mixture of crude archaic versions and later revised translations, together with texts standing somewhere in between, which must be either half-revised versions or conflations of old and new. If the situation frozen in time by the virtual sealing off of the Dunhuang collection was repeated at other Tibetan book repositories, then it would be surprising if the later history of the bKa' 'gyur did not turn out to be vastly more complicated than this survey might suggest. After all, we must remember that from the earliest times most Buddhist monasteries in Tibet would have possessed their own collections of scriptures, their own Sher[page 84] phyins, mDo mangs, rGyud 'bums, 'Dul bas, and so on, and that eventually many of these collections must have interacted in one way or another with the systematized bKa' 'gyur tradition sketched in this paper, which was itself derived from various monastic holdings of this type. The resulting pattern of criss-cross lines of descent, mutual influence and exchange is undoubtedly complex in the extreme.54