Layer upon Layer. "Computational Archaeology" in 15th Century Middle Dutch Historiography

Layer upon Layer. “Computational Archaeology” in 15th Century Middle Dutch Historiography.
Stapel, Rombert, Fryske Akademy (KNAW) / Leiden University, Netherlands, rstapel@fryske-akademy.nl
A scholar or student who wishes to engage in ‘non-traditional’ authorship attribution would be wise to choose a test corpus that is as free as possible of ‘interfering’ features such as genre, external editing, or a corpus that is stretched over a large number of years. The higher the consistency throughout the corpus, the larger the chance of a successful investigation of authorship and/or stylistic features.

Medieval manuscripts are characterized by a much greater amount of, what we could call, ‘interfering’ features. Scribes manually copied texts again and again, not seldom altering the content and often altering the orthography. Most original works have been lost, just as much of the copied material. To add even more difficulties, what we nowadays will easily refer to as ‘original work’ is much less clear-cut in the Middle Ages. Our modern notion of ‘copyright’ is virtually unknown to medieval men and women and for a long time the concept of auctoritas (author) was primarily used in referral to classic writers such as Aristotle and Augustine. Many of the medieval texts are thus written anonymously.

The situation could be characterized as chaotic by scholars used to relatively straightforward text corpora. Before you can begin your quest for a medieval author, you first have to find out what content is scribal related and what can be attributed to the author. And just when you think you are making some progress, you find out that your ‘author’ has been merely compiling source texts, who he (or she) is copying word for word. A scholar addressing these texts should therefore meticulously peal of the different layers of the text. When confronted with these scenarios, it might not sound that surprising that the number of studies involving the use of computational techniques and medieval texts is not that great. In recent years though, some progress has been made (e.g. Van Dalen-Oskam and Van Zundert 2007; Van Dalen-Oskam, Thaisen and Kestemont 2010). Most of all, these studies show that it is possible – using computational techniques such as Burrows’ Delta – to overcome some of the difficulties in distinguishing for instance scribal and authorial layers within a single text.

The Case
In this paper I will contribute to this evolving field of research and discuss some possible methods that can be used to differentiate these different layers within medieval texts. Although the text corpus that I will be using originated from fully tagged TEI/XML files, for this purpose I will be using plain text files. This could be of great interest to scholars who are not able or not willing to spend large amounts of time, energy and skill in preparing their texts. The texts are transcribed without changing spelling to modern use (thus euen instead of even and Iherusalem instead of Jherusalem), but abbreviated words are expanded.

One of the more interesting aspects of this particular text corpus is that it has been written by a single person, whose name is recorded in one of his charters: Hendrik Gerardsz. van Vianen, most likely secretary of the Utrecht Land Commander of the Teutonic Order. The Teutonic Order was one of the three major military orders – beside the Knights Templars and Hospitallers – that defended the Holy Church in the Holy Land, the Iberian Peninsula and the Baltic region and received goods and land all over Europe, including the Low Countries. Hendrik van Vianen wrote at least 25 Middle Dutch charters containing land contracts for the Teutonic Order between 1479 and 1491. He also wrote a few Latin charters between 1489 and 1509 that are not included in this study. Furthermore, he manually copied a manuscript containing a Dutch version of the popular Sachsenspiegel that belonged to two Land Commanders of the Teutonic Order in Utrecht. Last but not least, he was responsible for a chronicle on the history of the Teutonic Order, known as the Croniken van der Duytscher Oirden or Jüngere Hochmeisterchronik (Stapel and Vollmann-Profe 2010). Codicological evidence suggests that the manuscript of his hand, now in Vienna, is an autograph, although it is not sure whether parts of the text existed before. Autographs are relatively rare in a medieval context, a recent survey of Middle Dutch manuscripts mentions barely more than a hundred examples (Houthuys 2009).

As a result, we have the unique opportunity to study a medieval text corpus – perhaps not as large as the ones modern literary scholars work with, but still substantial in medieval terms – of around 131.000 words that includes the work of a single person working as manuscript scribe; writer of land charters; and possible author of a history of his order.

Ever since the Croniken van der Duytscher Oirden has been studied in a scholarly context, there have been questions about the original composition. The original Middle Dutch version of the chronicle consists of a long prologue, a part that contains the deeds and lives of the Grand Masters of the Order, alternated by privileges granted by popes and emperors. It ends with a history of the regional bailiwick of Utrecht and its Land Commanders. Especially that last part is often put aside as a later addition to the chronicle. It is an interesting question, the more so since it has historical significance in the sense that it has an influence on the original function and aspiration of the text.

Methods
To examine the relationship between the different parts of the text corpus we have collected, I employed several methods, but for now, I would like to focus on the use of the Delta Spreadsheets, freely made available by David Hoover (Hoover 2009). Some primary samples were selected from of the text corpus using the Intelligent Archive (Craig, Whipp and Ralston 2010) to create word frequency lists: the combined charters; the Sachsenspiegel; the table of contents of the Croniken; its prologue; some of the privileges; three parts of the Grand Masters’ part; and the bailiwick chronicle. These were computed against 2000 word pieces, overlapping and 500 words advancing. The results are shown in Figures 1, 2 and 3.

Figure 1: Moving Delta, Charters

Full Size Image

Figure 2: Moving Delta, Saxons’ Mirror

Full Size Image

Figure 3: Moving Delta, Croniken

Full Size Image

What stands out is that in all three instances, the primary samples of the table of contents and the bailiwick chronicle (in light and dark blue) outperform the other samples, especially if one leaves out the samples in Figure 3 which perform off course very well in their own consecutive areas.

Conclusion
What can be concluded from this observation? I think the table of content and the bailiwick chronicle best describe the personal writing style and orthography of Hendrik van Vianen. In both the charters, the copied Sachsenspiegel and the Croniken this layer is present, and it raises the question if he indeed can be held responsible for the bailiwick chronicle and the table of contents. The table of contents is clearly written at a later stage, added at the end of project, as is shown by watermark evidence and the distribution of some abbreviations and specific letter forms, quantified in Figure 4 and 5 below.

Figure 4: Different forms of letter “w”. The TOC in front corresponds with the latter part of the chronicle.

Full Size Image

Figure 5: Use or absence of abbreviations in “ende” (Dutch for “and”).

Full Size Image

Moreover, to what extent can Hendrik van Vianen still be held responsible for the rest of the chronicle? Was there an older exemplar of a chronicle available to him? It is hard to say for certain. There is a good possibility that he was not so much the – what we would now call – author, but more a compiler for large parts of the chronicle. The old source texts would then form another layer within the text. Again, in the Middle Ages it is not always possible (and even helpful) to make a strict distinction between the two. An old-fashioned approach to research the sources of the chronicles and how these source-texts were implemented could be a logical step further and will be one of the issues addressed in the remainder of this project.

References:
Craig, H. Whipp, R. Ralston, M. 2010 Intelligent Archive. Budgerigar Version., (link)

Dalen-Oskam, K. van Thaisen, J. Kestemont, M. 2010 “Computational approaches to textual variation in medieval literature, ” Digital Humanties 2010 Conference Abstracts, 37-44

Dalen-Oskam, K. van Zundert, J. van 2007 “Delta for Middle Dutch – Author and Copyist Distinction in Walewein, ” Literary and Linguistic Computing, (link) 22 345-362

Hoover, D. 2009 The Excel Text-Analysis Pages, (link)

Houthuys, A. 2009 De autograaf van de Brabantsche yeesten, boek VI (vijftiende eeuw). Hilversum 2009., Middeleeuws kladwerk,

Kestemont, M. Dalen-Oskam, K. van 2009 “Predicting the Past: Memory-Based Copyist and Author Discrimination in Medieval Epics, ” Proceedings of the Twenty-first Benelux Conference on Artificial Intelligence, 121-128

Stapel, R.J. Vollmann-Profe, G. 2010 “Cronike van der Duytscher Oirden, ” Encyclopedia of the Medieval Chronicle, Dunphy, G. 328-329

Thaisen, J. 2010 “A Probabilistic Analysis of a Medieval English Text, ” Digitizing Medieval and Early Modern Material Culture, Terras, M. Nelson, B. 328-329