Skip to content

Response to Review of The Destruction of the Medieval Chinese Aristocracy

I would like to begin by thanking Michael Hoeckelmann for his generous review of my book. Here, I only want to make a few additional remarks on the use of digital tools for historical analysis, an issue briefly broached by the reviewer in his comments. My book was intended to revise in important ways our understanding of late Tang history (as explained in the review), but I also hope that my book will serve to encourage more work in pre-modern history using newly available technologies. Unfortunately, the use of these technologies shares many of the weaknesses of quantitative history – weaknesses that led to the complete rejection of quantitative methodologies by many historians working in the 1980s and 1990s. For example, one problem I frequently encounter when talking to the proponents of digital humanities involves their excitement over a new tool (be it ‘GIS’ or ‘SNA’) without a clear sense of how these tools might be used in a concrete way to advance scholarly knowledge. I hope that my book provides an example of how these tools might be used.

In his review of my book, Michael Hoecklemann also brings up the issue of sparse data. Most of the digital tools in question were designed for use by social scientists working on contemporary issues, who – in comparison to pre-modern historians – enjoy far more comprehensive data sets. In an appendix to my book, I estimated that epitaphs with biographical data survive for no more than 2.59 per cent of the approximately 37,500 male members of the ninth-century capital elite. This sample size is certainly sufficiently large for basic statistical comparisons, assuming one has a good sense of the biases inherent in the sample. However, I discovered that the small sample size created difficulties for network analysis of marriage ties, because one knows the marriage partner of so few individuals (most epitaphs do not identify unambiguously the wife or husband). My solution to the problem was to examine marriage ties between ‘patrilines’ (i.e., patrilineal clans) rather than between individuals, a solution that allowed me to create a much denser and more meaningful network diagram.

Sample bias, itself, is another significant problem. As the reviewer notes, anybody can freely download my database; much of its content has also been incorporated into the larger Chinese Biographical Database (CBDB). I made my data available with some trepidation. Because I compiled the database myself, over a period of about ten years, I have a relatively good understanding of what is included and what is not included. Other historians using the same data will not have a similarly acute sense of the sample biases until they have spent considerable time exploring the database and its source materials. The problem is compounded in the case of CBDB because CBDB’s data was compiled by numerous historians using a wide variety of sources. Most simple CBDB database queries comparing the Tang to the Song will identify differences that are essentially nothing more than artifacts of the sources – perhaps reflecting the different nature of excavated epitaphs (the basis of most Tang data) from the broader variety of source material used to compile CBDB’s Song-period data. I encourage historians using any of this data to spend substantial time exploring the underlying data sources. I would, myself, be happy to discuss sample bias (by email or in person) with anyone using my data in their own research.