I have worked on manuscript P.3720 from the Pelliot collection at the Bibliothèque nationale (BnF) as an example of a composite manuscript. This is a scroll originating from the Dunhuang cave library discovered at the beginning of the 20th century. It is a collation of different texts, including appointment decrees, religious poetry, a funerary inscription, another inscription about the beginnings of the Mogao caves, etc. In addition, some of the texts come from different sources, having been written at different times by different hands on different paper, before they were all joined together into a single scroll. Thus physically the manuscript is also a composite object, consisting of separate sheets of paper glued together sometime during the early half of the 10th century. A remarkable aspect of the arrangement is that some of the texts are dated and the dates range from 851 to 934, thus there are 83 years between the earliest and latest sources. My goal was to reconstruct the process and motivation behind assembling this manuscript. An important question is whether the component texts were copied at different times, as their dates imply, or they were copied — together with the dates — from earlier sources.
Now I had a chance to look at this manuscript in person and this was when the idea of working on it came to mind. However, during the process of doing research, I was no longer in Paris and had to work with digital images from the IDP website. But when I counted the sheets of paper from which the manuscript was assembled, no matter how carefully I tried to do this, I always ended up with 9 sheets on the recto and 6 on the verso. This discrepancy was quite disturbing as I did not understand what was going on. In addition, there is an important text on the verso, called the Mogaoku ji 莫高窟記 (Record of the Mogao Caves) which talks about the establishment of the cave complex at Mogao during the 4th century AD. According to the digital photographs, this text was at the beginning of the scroll on the verso, but upside down. I thought I understood where this text was in terms of the whole scroll, including what was on its recto.
The surprise came when I finally had a second chance to look at the manuscript in person. I wanted to produce an accurate count of sheets and determine the original components of this composite object. I simply could not understand how could there be 9 sheets on one side and 6 on the other. But when I actually examined the scroll at the BnF, I realized that it consisted of 12 sheets, which was, of course, the same for both sides. The digital images, as well as the printed images in the Shanghai guji chubanshe edition, were unable to show that some of the sheets that looked continuous were in reality composed of two components. Only by turning the scroll towards the light was I able to see where some of the components were glued together.
The second surprise was that the back of the scroll was incomplete on the digital images. There were sheets missing from the beginning of the scroll, and the sheet with the Mogaoku ji text, which was shown as being upside down at the beginning of the scroll, in reality was right side up and about a third into the scroll. This, of course, changed lots of things because I realized that originally I had been wrong about every single sheet when pairing the texts on the two sides. The relationships changed. And, since most of the texts were dated, the dates changed, too.
All this is just an example of how looking at an original manuscript can clarify a number of issues that appear odd or mysterious on a reproduction. Of course, in most cases there is no difficulty in understanding the structure of a manuscript, which in most cases is actually quite simple, but there are exceptional cases of very complex structures (e.g. P.3720) which require an examination in person.
Imre Galambos, “Manuscript Copies of Stone Inscriptions in the Dunhuang Corpus: Issues of Dating and Provenance.” Asiatische Studien/Études Asiatiques LXIII 4 (2009): 809-826.