Medieval Historical, Hagiographical and Biographical Networks

Abstract

In recent years, a new method to study narrative texts was introduced, using network analysis. The approach is original and adventurous; instead of focusing on the literary or narrative bases of the texts, it involves extracting data for a formalised analysis of network structures. These are determined from descriptions of events in the texts and their statistical properties are studied using standard network-analysis tools. In this way comparisons between chronologically and geographically different texts are possible. Furthermore, we can compare these textual networks with real social networks, studied by modern sociologists or, indeed, fictional ones. These studies have clearly shown that social-network analysis forms an effective bridge between very different disciplines. It can connect scientists and humanists in joint research; it can depict old research questions in a new light and connect different phenomena belonging to the worlds of nature and culture. The key to this bridge is the understanding of complex systems and their emergent properties. But we are still in the very beginning of exploring these issues and in developing an adequate methodology as we seek to incorporate a number of tools recently developed. Here we attempt to cross the bridge from the humanities side. To this end, we present the results of two studies of medieval sources. Our focus is on visualisation and interpretation of local network properties, an approach which is complementary to complexity analyses. We show that the method can offer powerful augmentation to traditional approaches to the humanities and we outline ways in which these can be developed for the future.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

This chapter is aimed at demonstrating the outcomes of network studies strongly driven by humanities perspectives. Each section involves a comparison of medieval sources. In contrast to the previous chapter by P. MacCarron and R. Kenna, however, here the focus is rather on small networks and the roles of individuals within them. We partition the chapter into two sections. The first is by Máirín MacCarron, a medieval historian at the University of Sheffield, and focuses on texts from late-seventh- and early-eighth-century Britain. The supporting material, which aids the visualisation, was generated by Pádraig MacCarron at Oxford. The second section was authored by German medievalist Robert Gramsch from the University of Jena, and concerns descriptions of the events surrounding the succession to the German throne in 1002. That work was supported by Joseph Yose, a complexity scientist at Coventry’s Applied mathematics Research Centre. Besides the comparisons and contrasts discussed within each section, we believe comparisons between these two contributions as well as between this chapter and the previous one are worthwhile. The latter comparison gives insight into the differences between the research questions and modes of thinking emanating from a humanities background and those with a more science-based orientation.

Scientists, for example, are primarily interested in large texts so that they can bring the full power of their statistical tools to bear on the problem. From experience in the physical world, they are acutely aware that sample size determines the amount of information available and, consequently, the level of confidence we have in our calculated estimates. Their interests are in the macro-world: average properties and trends in entire populations or their textual equivalents. Humanities scholars, on the other hand, are more used to operating with small data sets. Indeed, the networks examined in this chapter contain between 21 and 84 network nodes. Such small sample sizes are not well suited to statistical analysis, and any such statistics are presented under that caveat. By way of comparison, a commonly used social network study is that of Zachary’s karate club (Zachary 1977) which, although having only 34 nodes, for a time provided a benchmark for community detection algorithms (Girvan and Newman 2002). Animal social networks also tend to have small samples; for example Kasper and Voelkl (2009) studied 70 networks which range in size from 6 to 35 primates. Transport networks also can have small samples, and Derrible (2012) analysed 28 metro networks, half of which have 20 or less nodes. We can glean some information from such small social networks through visualisation and by focussing on the properties of the central characters. Indeed, we hope this chapter demonstrates that humanities scholars with precisely the right expertise can extract information from network visualisations that would probably not be accessible to more road-brush scientists.

Some elements of modern research on dramatic and epic textual corpora already include comparative network-based modelling and interpretation (e.g., Trilcke 2013; Fischer et al. 2015; Jannidis et al. 2015). Network analyses are also increasing in prevalence in medieval research, such as using historical networks based on the links between clergy or political actors (Padgett and Ansell 1993; Malkin 2011; Gramsch 2013; for reviews, see Lemercier 2012; Jullien 2011). In such research, the texts form important sources of descriptions of political and personal constellations, of events and the people involved. They serve as a sort of “quarry” for the mining of data. It therefore seems necessary to study these texts for their network properties in order to gain understanding of their realism and representational accuracy.

We open the chapter with MacCarron and MacCarron’s networks-inspired analysis of four texts from Anglo-Saxon England in the early middle ages. This is followed by the Gramsch-Yose contribution in which two tracts of text, covering the same events, from medieval Germany are compared.

Section 1: Networks in History and Hagiography: The Two Lives of St Cuthbert and Early Histories of Wearmouth-Jarrow

In this section, we present the findings of a pilot study which examined four Latin texts written in England in the Anglo-Saxon period of the late-seventh and early-eighth centuries. The texts are from the kingdom of Northumbria and involve people and events from that region. Specifically, they concern the monasteries of Lindisfarne and Wearmouth-Jarrow (religious houses were centres of literary production in northern Europe at that time). The texts that we assessed are¹ (see Table 1):

• (i) the Life of Cuthbert by an anonymous author from the monastery of Lindisfarne;

• (ii) the Life of Cuthbert by the Venerable Bede (AD 673–735), which was written in the monastery of Wearmouth-Jarrow;

• (iii) the Life of Ceolfrith, written by an anonymous author based in Wearmouth-Jarrow;

• (iv) Bede’s History of the Abbots of Wearmouth and Jarrow.

Table 1 The four texts analysed herein categorised according to their subject matter, genre, authorship and the monastery in which the author was based

Our primary purpose is to compare the networks of (i) and (ii) to each other as hagiographies concerned with the same central character but by different authors. We also compare (iii) and (iv) which cover the same monastery, in the same time period, but are examples of two different genres. Our analysis addresses three main issues: firstly, whether we can use networks to detect or visualise differences between hagiographies and history; secondly, our investigation illuminates the position of women in the first two texts; thirdly, we shed light on the roles of authors in such texts.

Before turning to the network analyses, a brief comment on the genres of hagiography and history may assist those unfamiliar with their meanings in a medieval context. Genre or type of text was very important for medieval writers, as each had its own rules and conventions. Of the four texts that we examined, three are hagiographies and the fourth is a history (Table 1). Hagiographies are concerned with the lives of those who are considered saints in the Christian tradition: they most often take biographical form and usually have the word Life in their title.² Hagiographies vary in style and substance but they are identifiable in their focus on the life of a saint and their intention to edify the reader. Their Christian character is clear from the outset as they occasionally refer to portentous signs at their protagonist’s birth or during their childhood which are presented as indicators of future sanctity. Where these texts differ most obviously from conventional biography is in the attention paid to the saint’s influence after their death, with several chapters usually devoted to post-mortem miracles intended to demonstrate their sanctity. Their primary purpose was to promote and commemorate the main protagonist, either to bolster the reputation of someone already regarded as a saint or to boost an embryonic cult.

The history genre follows different conventions, and such works usually contain the word History in the title. The medieval understanding of history was defined by Isidore of Seville (AD 560–636) in his famous encyclopaedia, the Etymologies (Barney et al. 2006: 67).³ He stated that a history is a narration of deeds accomplished (Etymologies Book i.41); and histories usually concern peoples, rather than individuals, and cover a longer period of time than the lifespan of one person (Book i.43). Medieval writers worked within these conventions and authors frequently produced works in different genres as is the case today. The Venerable Bede, who will play a prominent role below, is an example.⁴

Lives of Cuthbert: Texts (i) and (ii)

Cuthbert of Lindisfarne (AD 634–687) was one of the most important early Anglo-Saxon saints and was commemorated shortly after his death by his monastic community. The first account of Cuthbert (i) was written by an anonymous Lindisfarne monk sometime around AD 700 (Thacker 1989: 115). About 20 years later, the community at Lindisfarne asked Bede to produce another Life of Cuthbert (ii). Bede is the most famous and widely-read writer from the Anglo-Saxon Age and his best known work today is the Ecclesiastical History of the English People (Colgrave and Mynors 1999). That the Lindisfarne community asked a monk from another religious house to write an account of Cuthbert indicates the esteem within which Bede was held by his contemporaries. Bede also had a long-standing personal devotion to Cuthbert, as he had previously written a Life of Cuthbert in verse, and later included several chapters about Cuthbert in his Ecclesiastical History.⁵

These two Lives of Cuthbert are structured very differently: the anonymous Life divides its material into four books, while Bede presents his in a single book of 46 chapters. However, both are biographical in character as they chart Cuthbert’s life from childhood through his monastic career to his elevation to a bishopric in Northumbria, retirement to a hermitage on Farne Island where he died, and posthumous miracles (Colgrave 1940). Bede, who was writing 20 years later, at a time when Cuthbert’s fame was continuing to increase, also included additional stories and more post-mortem miracles (Thacker 1989).

In an approach which is new for these types of texts, we analysed the social networks of both. Our data collection focussed on connections between individuals, but we did not weight the links: therefore the network does not distinguish between strong relationships (such as spousal) and those between individuals who happened to meet only once. This approach has the benefit of producing data that is simple and we deemed it appropriate for a pilot study. In Section 2 of this chapter, an alternative approach is introduced.

Bede’s account (ii) contains more characters than the Lindisfarne Life (i): 59 compared to 49; and (i) and (ii) have 19 characters in common. There were 12 unnamed or unidentifiable characters in (i) and 18 in (ii). If these were discernible, one may have a higher number of shared characters between the two texts. There are 41 identifiable characters in Bede’s Life and 37 in the anonymous Life. The 19 shared characters include clerics and monks in the Northumbrian church, members of the royal family, two angels, as well as Hildmer and his wife. The angels interact only with Cuthbert in each Life and are without identifiable gender. There are 78 edges (connections between characters) in the anonymous Life (i) and 86 in Bede (ii). Both writers record interactions between Cuthbert and the population of Northumbria. However, for the most part these people are unnamed and unidentifiable, or, with the exception of Hildmer and his wife, when names are provided the sources do not agree. Despite these discrepancies, the graphical configurations appear quite similar, as shown in Figs. 1 and 2. The network properties are also similar, as we shall see.

Fig. 1

The social network of the 49 characters in (i) the anonymous Life of Cuthbert: Red nodes represent female characters, blue nodes are male, white nodes are angels. Cuthbert is indicated by the largest node and the anonymous author is in the bottom middle

Fig. 2

The social network of the 59 characters in (ii) Bede’s Life of Cuthbert: Red nodes represent female characters, blue nodes are male, white nodes are angels. Cuthbert is indicated by the largest node and Bede, the hagiographer, is the next largest in the bottom left

The Early History of Wearmouth and Jarrow: Texts (iii) and (iv)

The monastery of Wearmouth and Jarrow is familiar to many today primarily because of the Venerable Bede. The first part of this monastery was established at Monkwearmouth in AD 672/3 by Benedict Biscop (AD 628–690). A second house was founded at Jarrow in AD 681/2 at Benedict’s instigation, which was ruled by Ceolfrith, and linked to Wearmouth (Wood and Grocock 2013: xxv–xxxii). The establishment of a single monastery in two locations, approximately 9 miles apart, is unusual, but the anonymous writer of the Life of Ceolfrith (iii) and Bede in his History of the Abbots (iv) both present the community of the two houses living in unity. The texts are concerned with the monastery’s early years during the abbacies of Benedict Biscop, Eosterwine, Sicgfrith and Ceolfrith, who ruled both contemporaneously as co-abbots and successively. Both accounts conclude during the abbacy of Hwaetberht, soon after the death of Ceolfrith in AD 716. Much attention has been paid to the relationship between the two accounts (iii) and (iv) and some have speculated that Bede was the author of both texts (McClure 1984), though recent scholarship has rejected this assertion (Wood and Grocock 2013: lxi–xcv).⁶

The pair of texts (iii) and (iv) is roughly contemporary with the Lives of Cuthbert (i) and (ii), but, unlike them, the Wearmouth-Jarrow accounts represent two different genres of writing. Although each has sometimes been regarded as a hagiography (see Wood and Grocock 2013: xxii), Bede’s account (iv) is more properly a history. It follows the conventions of historical writing and Bede himself categorised it as such. He provided a list of his extensive writings in an autobiographical note at the end of his magisterial Ecclesiastical History (Book v. 24). Organising his output according to genre, he collected his hagiographies under one heading, and grouped his works of history under another. He described his account of the abbots of Wearmouth-Jarrow (iv) as: ‘A history of the abbots of the monastery in which it is my joy to serve God, namely Benedict, Ceolfrith, and Hwaetberht, in two books’ (Colgrave and Mynors 1999: 571).⁷ Bede did not place his History of the Abbots with his hagiographies, nor did he call it a Life or Lives of the abbots.

The anonymous account (iii) has 21 characters while Bede’s version (iv) has 23.⁸ The texts have 28 and 38 edges, respectively. Of the characters, 11 are shared between the texts: the five abbots of Wearmouth and Jarrow; two kings of Northumbria (Egfrid and Aldfrid); three popes (Agatho, Sergius and Gregory II); and John the Arch-Cantor, whom Benedict Biscop had escorted to England from Rome in AD 678.⁹ The difference in character sets is partly because the Life of Ceolfrith (iii), in line with hagiographical convention, has much additional information on the protagonist’s early years and the end of his life, whereas the History of the Abbots tells us more about wider developments in the Anglo-Saxon church and society.

Network Statistics and Genre Identification

A summary of some of the network statistics is given in Table 2. An obvious question is whether any of these can distinguish between hagiographical and historical genres. Notwithstanding the obvious caveat associated with small sample sizes, two measures suggest themselves. The first is the clustering coefficient (transitivity), which, in the history (iv), is larger than in the hagiographies (i)–(iii). This means there are proportionally more closed triads in the history network than in the other three. Indeed, one might expect that two acquaintances of a character are more likely to be connected to each other in a history than in a hagiography. Secondly, but less significantly, the distribution of distances is more varied in (iv) than in any of (i)–(iii). Despite (iv) being a smaller network than (i) and (ii) and having a relatively low mean path length, it has higher standard deviation. These differences are small however and further investigations are required. Indeed, a large-scale quantitative assessment of hagiography and history texts from different parts of Europe is required to test whether bigger data sets can deliver different results.

Table 2 Network statistics for all four texts

Fig. 3

The network of the anonymous Life of Ceolfrith (iii). Only Ceolfrith and Benedict Biscop have more than three links. Ceolfrith is represented by the largest node. The second largest is Benedict Biscop

In the three hagiographies, the titular character is the most connected by a considerable degree (see Figs. 1, 2 and 3). In the anonymous Life (iii), for example, Ceolfrith has 16 connections and Benedict has eight and no other character has a degree above three. In Bede’s History of the Abbots (iv), however, Benedict Biscop is the most connected with 14 links followed by Ceolfrith with eight (Fig. 4). But all five abbots of Wearmouth-Jarrow (Benedict, Ceolfrith, Eosterwine, Sicgfrith and Hwaetberht) are also well connected and located relatively centrally in the network diagram. In fact, and in contrast to the hagiographies, seven characters have a degree above three.

Fig. 4

The network of Bede’s History of the Abbots (iv). Benedict Biscop is represented by the largest node and Ceolfrith is the second largest. Eosterwine, Sicgfrith and Hwaetberht are also prominent in terms of their connectedness

Importance of Gender

In Bede’s hagiographical account of Cuthbert (ii) there are 44 men and 12 women. Between these we have 57 male-male interactions; 1 female-female interaction; and 26 male-female interactions. The anonymous Life (i), on the other hand, has 39 men and 8 women. There are: 58 male-male interactions; 0 female-female interactions; and 18 male-female interactions. These data are summarised in Table 3.

Table 3 Numbers (and proportions) of characters broken down by gender and their associated interactions for Texts (i) and (ii)

The conventional view in modern scholarship is that Bede pays less attention to women than the anonymous author. For example, Stephanie Hollis has argued that Bede deliberately curtailed the influence of women in his writings, and her hypothesis has received support from many quarters (Hollis 1992; Blanton 2007: 22–63; and Lees and Overing 2001: 15–39). In examining the Lives of Cuthbert, in particular, she suggested that, in contrast to the anonymous author, Bede deliberately suppressed the activity of Abbess Aelfflaed of Whitby (Hollis 1992: 179–207).

Our observation suggests that the women in Bede’s text are more integrated into their social network than women in the anonymous Life. As can be seen from Table 3, Bede included a female-female interaction (the connection is between Ecgfrith’s queen and her sister), while the Lindisfarne Life did not. In addition there is a higher proportion of female characters in Bede as well as a higher proportion of interactions involving females. That the attention paid to females is higher in Bede is further borne out by assessing the betweenness centralities for women in both Lives of Cuthbert. In a network, the betweenness centrality of a node gives a measure of the number of shortest paths that the node lies on in proportion to all shortest paths. For information to travel optimally through the network, it is more likely to pass through nodes with a higher betweenness centrality measure. A betweenness of 0 means that node is not on any shortest paths between any other pair of nodes. Here we see that half of the women in Bede’s Life (ii) have a betweenness centrality greater than zero, while only a quarter of the women in the anonymous Life (i) have non-zero betweenness centralities (see Table 4).

Table 4 Betweenness centralities for women in Texts (i) and (ii)

By way of contrast, in (i) only 44 % of male characters (17 nodes) have a non-zero betweenness, and only three of these have higher betweenness than Aelfflaed. Similarly, in Bede (ii), 41 % of male characters (18 nodes) have non-zero betweenness, with Aelfflaed again having the fourth highest centrality. The most central character in each network is Cuthbert whose betweenness centrality is 953 in (i) and 1278 in (ii). (Women do not feature in either of texts (iii) or (iv), so we do not have analogues of Tables 3 and 4 for these works.)

Our research suggests that women collectively play a greater part in the social network of Bede’s account (ii) than in the anonymous Life (i). As noted Aelfflaed is the fourth most connected person in both texts, which is unsurprising from the perspective of Anglo-Saxon society, as she was sister to two successive kings of Northumbria and abbess of the influential monastery of Whitby.¹⁰ However, Aelfflaed’s prominence, especially in Bede’s text (ii), has, to our knowledge, not been identified in previous scholarship and demonstrates the merits of using networks to aid the analysis of these texts. This also allows us to contribute to an important historiographical debate concerning the treatment of women in these sources, particularly the argument that Bede reduced the role of women in his writings. Our analysis shows that Bede was not suppressing the role of women relative to the writer from Lindisfarne, indeed as Bede included more women and a greater percentage of the women in his Life of Cuthbert (ii) have a betweenness centrality above zero, we may speculate that he presents a more realistic picture of Anglo-Saxon society—or, at least, of Anglo-Saxon social networks—than the writer from Lindisfarne. This has wider implications for the study of Bede, both for his presentation of women and for his depiction of society. A full analysis of the social networks of his Ecclesiastical History along with a similar assessment of other contemporary Anglo-Saxon hagiographies could therefore be very revealing.

Role of the Author

The role of the author within each text is also an important issue. In Texts (i) and (ii) each author asserts his credibility by citing their sources for various anecdotes that are presented in the works. This serves to insert the hagiographer into the social network of the hagiography. Although this is obvious to any reader of the texts, it is striking that the authors are the most connected people in these networks after the person about whom the text was written. In particular, Bede frequently referred to himself in his Life of Cuthbert (ii) to demonstrate the credibility of his witness to Cuthbert’s sanctity. Bede did not, however, insert himself into the social network of his History of the Abbots (iv). This is perhaps a little surprising as he lived in the monastery for much of the period concerned and was part of the community’s social network. He may not have seen the need to assert himself in the same way when writing for his own community in what is a work of history.

In contrast, the authorship of (iii) is an interesting question and we do not know if the anonymous author inserted himself into the Life of Ceolfrith (iii). As mentioned above, it has been suggested that Bede wrote this text (McClure 1984). Although recent scholarship overwhelmingly rejects that suggestion, (see, e.g., Wood and Grocock 2013), it is interesting to reflect on what evidence our network approach provides. In our analysis of texts (i) and (ii) we have shown that Bede paid more attention to women than the anonymous hagiographer in (i). In (iii), however, women do not appear despite it covering Ceolfrith’s early life when clearly women must have been present. Given the strength of our first observation, we might expect that, were Bede the author, women should appear in (iii) too. It could be argued that their absence indicates that Bede is not the author of (iii). Further evidence for this comes from the observation that Bede is the second most connected character in the hagiography (ii). But he is not the second or even third most connected in the hagiography (iii). [These considerations do not apply to (iv) because that is a history and not a hagiography.] Thus, the network-visualisation analysis delivers the same conclusion as traditional techniques (Wood and Grocock 2013).

Indeed it remains possible that the anonymous author does feature in the text. We suggest that he may be the otherwise unidentified ‘small boy’, who, we are told, was a student of Ceolfrith’s and a priest in the monastery at the time of writing, and was eager to promote the life of his former abbot (Life of Ceolfrith, chapter 14). An alternative theory is that this boy is Bede, indeed he has recently been described as ‘confidently identifiable as Bede’ (Holsinger 2007: 162). However, the suggestion that he could be the anonymous author, who was clearly devoted to Ceolfrith’s memory, cannot be ruled out. (If the small boy were identified as Bede it would reduce the number of nodes in Fig. 3.)

Section 2: Networks in High Medieval German Historiography: A Comparison Between Thietmar of Merseburg and Adalbold of Utrecht

Introduction

The purpose of this section is to introduce and evaluate methods which can be used for network analysis of historiographical texts. To demonstrate them, a clear and simple example is used. It is not the objective of this study to gain new insights into the historical events described or into the texts in general. Rather it is to show how the network analytical approach basically works and which cognitive possibilities it would offer within the frame of larger studies.

Concretely, we focus on two German historiographical works from the early eleventh century: the Chronicle of Bishop Thietmar of Merseburg and the Vita Heinrici II Imperatoris of Utrecht’s bishop Adalbold (Ottonian Germany, ed. Warner 2001; Adalbold von Utrecht, ed. Schütz 1999). We consider only excerpts from these works which refer to a specific chain of events, namely the struggle for the German throne in the year 1002.

Our first aim is to compare the networks of political actors according to the narrations in the two chronicles. We will examine to what extent, viewed through the lens of network analysis, the descriptions of Thietmar and Adalbold match. We have to take into account that Adalbold had Thietmar’s account to hand when drafting his Vita Heinrici. Therefore, one might expect that Adalbold’s account is at least as dense as that of Thietmar, and possibly even more accurate since he had additional information. Can network analysis confirm this? And to what extent are differences in the narrative strategies of the authors reproduced in the network models?

A second aim of our study is methodological. In gathering the information from Thietmar’s chronicle, we employed two different data-harvesting techniques; we used different rules for the determination of the protagonists (nodes) and their relations (links). One approach is the same as in MacCarron’s and Kenna’s original papers—disregarding any knowledge external to the texts and, as with Section~1, taking them on their own merits. These networks are unweighted, to use the standard parlance (MacCarron and Kenna 2013). The other approach is informed—or weighted—by expert knowledge of medieval Germany from the outset. We term this hermeneutical. It is important to note that the distinction between the two approaches refers to data harvesting only; once the data are gathered they can be statistically processed using either approach and interpreted in the light of, and in parallel to, expert humanities-based knowledge. Our second aim is therefore to investigate the extent to which different rules for the determination of the protagonists and their inter-connections affect the resulting network models. What types of data-harvesting should be applied in future studies?

Background to the Texts and the Events Described

We should start with a brief description of the German throne crisis of 1002. The crisis was triggered by the unexpected death of Emperor Otto III in January 1002 near Rome, aged only 21. His body was brought to Aachen where he was buried on 5 April. As Otto died childless, the problem of his successor arose and several princes proclaimed their ambitions to the throne. The first of them was Duke Henry of Bavaria, the future King Henry II. He was a great-grandson of the first Ottonian ruler, Henry I, and thus the next of kin of the deceased. In February 1002 Duke Henry received the funeral procession of Otto III in Polling near Augsburg and sought support for his candidacy. However, according to Thietmar’s account, he didn’t meet with the approval of the princes led by Archbishop Heribert of Cologne. These princes declared themselves, by majority, in favour of the Swabian Duke Hermann II. Simultaneously in Saxony, the Margrave of Meissen, Ekkehard I, staked his claims to the throne. He had been a close friend of Otto III. Although he enjoyed the support of some, such as the Saxon Duke Bernhard, powerful enemies in the Saxon nobility opposed him. These prevented his election and ensured that the Saxons agreed to Henry II as the new king. Ekkehard set out to the west of the kingdom in order to establish contact with Hermann II. However, on 30 April 1002 he was murdered by Count Henry and Udo from Katlenburg in Pöhlde in an unrelated case of private revenge.

Duke Henry of Bavaria then took the initiative. He rushed with an army of loyal followers to Mainz where he was elected king by some of the German nobility on 6 June 1002. His most important ally, Archbishop Willigis of Mainz, crowned him immediately. Thus, a fait accompli was established which Duke Hermann II couldn’t undo, although he still had strong and loyal followers, especially in the west of the Reich. A direct military confrontation between Heinrich and Hermann failed to manifest itself, although there were proxy engagements in southwest Germany which tended to result in favour of Hermann’s supporters. Nonetheless, the newly crowned King Henry II steadily gained the recognition of all leading princes by making a royal tour (“Krönungsumritt”) through Thuringia, Saxony and Lower Lorraine. On 8 September 1002 he was crowned at the “right” place, Aachen, by the “right” Archbishop, Heribert of Cologne. On 1 October Duke Hermann finally conceded. This was essentially the end of the crown struggle.

Thietmar’s description of these dramatic events leaves nothing to be desired in terms of detail and credibility. It is one of the classic reports of a German “Königserhebung” (the process of appointing a king) in the Middle Ages. Adalbold’s account of Henry’s “Königserhebung” is also relatively detailed. However, in comparison to Thietmar, his treatment is more liberal and there are a number of omissions which have a clear tendency to favour Henry II. For instance, Adalbold softens Thietmar’s report on the events in Polling to create the impression that the funeral procession of Otto III was more favourable to Henry II from the beginning. The episode of Ekkehard’s throne candidacy is almost completely ignored—just a brief note of Ekkehard’s assassination in Pöhlde remains. Hermann’s struggle for the throne is severely curtailed as well.

Even without using network analytical methods, the quite different character of each report is evident to the reader. While Henry II is the main focus and the hero for both authors, Thietmar painted a picture of an initially uncertain struggle in which Duke Henry asserted himself with luck, skilful manoeuvring and successful political networking. In contrast, Adalbold described Henry’s victory as an ab initio safe bet, with Henry’s enemies relegated to auxiliary players or nearly completely written out. Although he had basically the same information available, he obviously was interested in creating a panegyric instead of a historically accurate account. This corresponds to the typical characteristics of the literary genre of Adalbold’s work, the royal biography (as opposed to a history).

We considered network-analytical modelling of both texts as potentially suitable for accurately visualising this general finding. Due to the fact that this is a novel and unusual approach to medieval studies, the process should be described below in more detail.

The Methodologies of Gathering Data

The first step of data harvesting consists in identifying the actors or players (nodes) and the ties (edges) between them. On the surface that seems to be very easy and non-ambiguous, but in practise the process raises questions.

The simplest procedure is at the level of “scene structure” of a text. The Romanian mathematician Solomon Marcus already described this very elementary method in the 1970s as follows:

Let us imagine a spectator who is during the theatrical performance just able to observe the actors’ entries and exits and recognize each character, so to distinguish two different kinds of characters from each other. The amount of information, received by every viewer, we should call the scene structure of the play (Marcus 1973, p. 289).

Such a data set cannot, of course, represent the content of a play, novel or historical work, which is obviously much more complex. But it is a controlled process allowing mathematical comparison with other narrative texts. Already this very simple procedure involves a certain act of interpretation, at least in regard to the distinct identification of the actors. But the network-analytical researcher aims to model more information contained within the texts. This opens further scope for interpretation and we have to acknowledge that there are different ways to render a narrative text for an adequate network model. One may, for example, define positive (friendly) and negative (hostile) links between actors in the following way (see chapter “A Networks Approach to Mythological Epics”).

Friendly Links

are formed between two characters if they speak directly to one another; know each another; are in the same immediate family (siblings or parents); or are present together in a small congregation.

Hostile Links

are made between characters if they physically fight or are at war with one another. If two characters argue or are aggressive to one another, this is not sufficient to constitute a hostile edge.

The placing of a single link between characters following the above rather mechanistic rules may be called an unweighted approach. It is already considerably refined in comparison to Solomon Marcus’ approach in that the discrimination between positive and negative links makes the model more realistic.

There are various ways to weight the network. For example, one could weight links in proportion to the number of encounters between pairs of characters in the text. However these procedures, and the mechanistic classification of positivity and negativity, remain in some aspects simplistic. As the historiographical reports of the throne struggle show, hostile behaviour may not be simply the exercise of violence. Already one may interpret the verbal opposition of a throne candidacy as an act of hostility to the pretender. We refer to methods of data gathering which not only reproduce the formal appearance of communicative acts (like knowing each other, meeting each other, fighting each other using violence), but also reflect the intentions behind them (political support, impairment), as the hermeneutical approach. Such a procedure is more interpretative and admittedly cannot be described in a few sentences.

To demonstrate the differences between the unweighted and the hermeneutical approaches, we discuss one example from Thietmar’s chronicle in detail, namely the description of an assembly of princes in Frohse (book IV/52 = Ottonian Germany, p. 188s):

Meanwhile, having learned of their lord’s premature death, the leading men of the Saxons sadly convened at the royal estate of Frohse, held as a benefice from the emperor by Count Gunzelin. Giselher, archbishop of Magdeburg, along with his suffragan bishops, Duke Bernhard, the margraves Liuthar, Ekkehard, and Gero, and the great men of the region pondered the condition of the realm. When Margrave Liuthar realized that Ekkehard wanted to exalt himself over them, he called the archbishop and the worthier part of the magnates outside for a secret discussion. During this discussion, he proposed that they swear an oath to refrain from electing a lord and king, either as a group or individually, until a meeting could be held at Werla. This was agreed and consented to by all, with the exception of Ekkehard. Angry at being held back somewhat from the royal dignity, he shouted: ‘Oh Count Liuthar, what do you have against me?’ Liuthar responded, saying: ‘Indeed, have you not noticed that your cart lacks its fourth wheel?’ Thus the election was interrupted, and the saying of the ancients was confirmed: to delay for one night is to postpone for a year, and that means to defer for a lifetime.

Which personal interrelations can we deduce from this short report? The following two sociograms (Fig. 5) show, how the unweighted and the hermeneutical approaches lead to very different network models. The small boxes (nodes) represent the actors who are mentioned in the text, the links between them represent the friendly and hostile edges, which can be deduced from the narration.

Fig. 5

The meeting at Frohse in 1002 according to Thietmar IV/52. The unweighted and hermeneutical methods deliver the left and right sociograms, respectively (the red line marks a conflictual or hostile link)

The unweighted approach of MacCarron and Kenna pictures this passage simply as a positively associated cluster of six named princes (left panel of Fig. 5). A negative link between Ekkehard and Liuthar is not set because Liuthar’s intrigue is not a hostile act in the strict sense of violence or open threat. If we retranslate this unweighted data set back into a text, the narration of Thietmar may essentially be reduced to only one sentence: “The princes … meet each other in a congregation.”

In contrast, the hermeneutical approach seeks to model content oriented as precisely as possible with the text (right panel of Fig. 5). In addition to the persons mentioned by name, the bishops of Zeitz and Meissen may be introduced. Their participation at the meeting can be deduced from the text because they are Giselher’s of Magdeburg suffragan bishops who, according to Thietmar, were accompanying their metropolitan bishop. Between them and Giselher positive relations are set in the context of a hierarchical subordination due to which they follow him as an act of allegiance (in network analytical terms, a subordination can best be modelled by a star structure). In an analogous manner the secular magnates are assigned to their chief, the Duke Bernhard of Saxony. These attributions are without doubt acts of interpretation, which are however oriented with the wording of the source which is by no means arbitrary in its listing of names. A special position is occupied by Graf Gunzelin who hosted the meeting. He was Ekkehard’s brother so that it can be assumed that he wanted to promote primarily Ekkehard’s candidacy. Therefore, a positive link is set between him and Ekkehard. Another positive link is directed to the deceased Emperor Otto III, from whom Gunzelin had the royal court of Frohse as a benefice. However, the real key position in the narrative is held by Markgraf Liuthar. As Thietmar explicitly describes, he establishes at the meeting a connection to Archbishop Giseler, in order to prevent Ekkehard’s candidacy. Thus Liuthar is on the one hand a broker between the secular and the ecclesiastical greats; on the other hand he is Ekkehard’s opponent which is represented by a hostile link. Since its action is politically decisive—in the subsequent narrative his role as a preventer of Ekkehard’s candidacy is further specified—it seems justified in regard to the content that we speak here of an open hostility.

The network model derived using the hermeneutical approach requires specialist knowledge in the field of medieval studies as well as knowledge of the modelling options offered by the network approach. While one may claim it follows the text in a manner that the flat unweighted approach cannot, it is not an incontestable objective process. But one should not be quick to reject it because of this; its structure depends on the questions being asked. Different sociograms derived using different interpretations may each be meaningful in its own way.

The aggregation of all the small sub-networks which are displayed in the various chapters of the Thietmar chronicle leads to an overall network. This network constitutes a model of the entire personal constellation of the struggle for the German throne of 1002 as described by Thietmar. As expected, due to the different methods of data gathering, the models have quite different appearances and the unweighted and hermeneutical networks are shown in Figs. 6 and 7, respectively.

Fig. 6

The unweighted network of the German princes in 1002 according to the Chronicle of Thietmar of Merseburg

Fig. 7

The hermeneutical network of the German princes in 1002 according to the Chronicle of Thietmar

In the sociograms, the actors are located in a manner which shows important nodes and clustering as clearly as possible. The size of each node is determined by the normalised betweenness centrality (denoted by C) of the actors and, for simplicity, the edges do not distinguish between positive and negative links. The two sociograms differ even in their size. The unweighted one has 84 nodes and 284 edges while the hermeneutical one consists of 55 nodes and 122 edges. Even the ranking of the players varies: Henry II (C = 0.605) and Ekkehard (C = 0.317) clearly dominate in the hermeneutical record (Fig. 7). They have relatively less prominence in the unweighted approach (with C = 0.379 and C = 0.213, respectively, in Fig. 6). Moreover, in the unweighted approach, there appears other (relatively) important actors, namely Hermann II (C = 0.186) and Boleslaw Chrobry (C = 0.111). Particularly striking is the denser clustering in Fig. 6. A fully connected cluster of 17 actors (including Henry II) is manifest, which has no counterpart in Fig. 7. The structure is nothing more than a reflection of a scene, in which Thietmar enumerates a total of 17 participants in the by-election of Henry II in Merseburg (including Henry II). According to the unweighted approach, this meeting is considered in terms of the existence of positive relations between all participants, generating 136 dyads at one stroke! The hermeneutical approach interprets the message of Thietmar in a more guarded manner and the election of Henry II is modelled as the establishing of positive relations between Henry and his voters present in Merseburg, i.e. just 16 star-shaped dyads outgoing from Heinrich are set.

It should be stressed once more that there is no simple “right” or “wrong” model. If we consider an assembly primarily as a communication network, it makes sense to model it as a fully connected cluster. If we look, on the other hand, at the political decision arrived at (Henry’s election), it is more appropriate to represent the assembly as a star in which there is a clear hierarchy with the elected king in the centre and his princely supporters on the periphery. In principle, the researcher is quite free to choose an unweighted or hermeneutical approach; this decision mainly depends on the particular research question. Of course, the researcher must use the selected method consistently, especially to compare material between narrative texts or with “real” social networks. Here we continue our investigations using the hermeneutical approach because it promises a more representative depiction of the political constellations of the year 1002.

Comparison of the Networks of Thietmar and Adalbold Using the Hermeneutical Approach

The hermeneutical network for Adalbold’s texts is depicted in Fig. 8 and is to be compared to Thietmar’s network of Fig. 7. Thietmar’s is not only bigger but it is also more complex than Adalbold’s network. Adalbold’s report puts Henry II at the centre, with an overwhelmingly high betweenness centrality (C = 0.926). Although in Thietmar’s chronicle Henry II takes also a dominant position (C = 0.605), a second centre is also looming in the account: the person of Ekkehard (C = 0.317). The exceptionally high level of interest Thietmar has in the events in Saxony is very clearly reflected in the network analytical result. Duke Hermann II of Swabia, however, the other major opponent of Henry II, appears very marginal both from Thietmar’s perspective (C = 0.092) and Adalbold’s (C = 0.077).

These results may not be so surprising for an expert reader; all these findings can equally be made by a close reading. Nevertheless, in line with the main objective of the survey to discuss the network analytical methodology on a clear and simple example, this outcome is markedly instructive. It becomes apparent that historiographical (or more in general: literary) texts can be usefully “measured” mathematically: Which persons acting have central relevance, how dense is the net of the described interactions, how large the complexity of the plot consequently is? In the given case the graphical visualisations convincingly show how differently two authors with a congruent level of information can portray the same events. The distinction between the literary genres—one a chronicle and the other a royal biography—becomes visible at a glance. This is an advance at least in didactical regard. In addition to that we can suppose that the benefit of such a methodology will be much greater if we work with large text corpora. New research questions arise: Does a royal biography also sketch strong secondary characters? Which level of complexity can the historical networks reach? Can we uncover additional insights into the structural properties of the described personal networks, for instance with regard to the existence and the extent of different political factions?

Fig. 8

The hermeneutical networks of the German princes in 1002 according to Adalbold’s ‘Vita Heinrici’

Figure 9 shows the political balance of power at the climax of the succession crisis in the summer of 1002. In this case an algorithm was used (Gramsch 2013) which on the basis of the principle of structural balance divides the personal network into groups which are internally free of conflicts. For the sake of comparability, fixed places have been assigned to the actors in the sociograms. These places nearly match their topographical position. To improve the legibility of the graph, only the most important actors are labelled by names.

Fig. 9

The hermeneutical networks of the German princes 1002 according to the Chronicle of Thietmar of Merseburg (left) and Adalbold’s “Vita Heinrici” (right). Node positions nearly match their geographical locations. Node colours indicate political partisanship; link colours indicate friendly (blue), neutral (dotted blue) and hostile links (red)

The geographical focuses of the reports are now easy to recognise. Adalbold is more accurate in his account of the circumstances in Bavaria (bottom right) where he enumerated the followers of Henry II. However, the conditions in Saxony appear far less important to him which explains the relative emptiness of the upper right quadrant of the right panel of Fig. 9. Thietmar listed here a lot more actors but amongst them were indeed a number of politically insignificant persons. For example, Ekkehard’s household knights forming part of a “ring” of actors (in green) on the right side of the left panel of Fig. 9.

The colouring of the nodes corresponds to the division of the network into several groups, determined by the analytical software. They are separated by various conflicts which are represented by red lines. If we consider the distribution of colours in Fig. 9, the big difference between Thietmar’s and Adalbold’s reports is particularly clear. Adalbold’s narrative network is completely dominated by a blue group which is associated with the main actor, Henry II. Only a tiny group of followers (marked turquoise) of the rival candidate, Duke Hermann II, is opposed to the new king’s widespread network of supporters. On the other hand, Ekkehard is completely isolated (marked in grey in the right panel). Besides, the dyad of Margrave Henry of Schweinfurt and Duke Boleslaw of Poland stick out as a small group (in green in the right panel); Adalbold here refers already to the protagonists of the uprising of 1003. According to Thietmar, the blue group led by Henry II is significantly smaller. In earlier periods of the conflict (before Ekkehard’s assassination and Henry’s recognition by the Saxons on the so-called by-election of Merseburg the 24 July 1002) this group is even outnumbered. Ekkehard has his own followers (represented in green in the left panel) although they consist only of low ranking (knightly) actors. The turquoise group of Hermann’s (potential) followers is significantly greater than that one described by Adalbold and spreads over all parts of the empire. It is particularly striking that the deceased Emperor Otto III is assigned to this cluster. It points to a circumstance which is repeatedly emphasized by medieval studies: the power shift of 1002 represented a real change in policy, an exchange of the leading elites and a redefinition of policy guidelines, for instance in relation to Poland (Schneidmüller and Weinfurter 1997). This complex scenario is largely hidden in Adalbold’s report where even Archbishop Heribert appears as Henry’s II follower.

Summary and Conclusions

Here we presented the results of two pilot studies, each carried out from strongly humanities-based perspectives. The first examined the social networks underlying four short texts from early Anglo-Saxon England. Our investigation illuminates the position of women in medieval hagiographical social networks, and allows us to challenge the prevailing view that Bede’s Life of Cuthbert pays less attention to women than the anonymous Life. We also raised questions about the place of authors in hagiographies. The hagiographers in the Lives of Cuthbert inserted themselves into the social network of their texts and it is possible that the anonymous author of the Life of Ceolfrith did likewise.

The second pilot study compared two historiographical reports on Henry’s II coronation in Germany in the year 1002. Although both reports are based on the same information, they exhibit a number of significant structural differences. Adalbold’s report is completely focused on the successful candidate, Henry II, who reigns supreme in the centre of a large network of supporters while his competitors appear only marginally. Adalbold’s network has a unipolar and politically homogeneous structure. Thietmar, on the other hand, draws a very different picture which leads to doubt over the inevitability of Henry’s success. His network appears bipolar and highlights the fact that the crown struggle of 1002 was a quarrel between Henry II and the nobility who were leading under the reign of the deceased Emperor Otto III.

Many of these findings are not new from the point of view of medieval studies. But their reproducibility by means of network analysis confirms the practicability of this new methodological approach. In many respects, it can substantiate existing knowledge, as the acknowledged differences between texts may be reflected in their network properties. In addition, we can observe considerable additional benefits of this methodology: firstly, the network-analytical models can illustrate historical findings very clearly and convincingly and, secondly, they illuminate some hitherto underexposed aspects. They can inspire new questions of familiar material, when used alongside traditional methods of investigation.

Thus, despite the very small sample sizes, our investigations suggest that network-visualisation is a promising approach to illustrate narrative strategies of medieval historians through network analytical methods. Medieval historiography is frequently characterized by an emphasis on personal relationships, and Wolfgang Christian Schneider speaks in this context of a “personal-relational understanding” which formed the historical perception of historians at that time (Schneider 1988, p.~40ss.). The actors do not appear as “closed, individual person[s] but as included in a personal texture of relationships” (p. 30), making these sources an appropriate field for network-theory oriented research.

A second focus of our study was to demonstrate that different methods have different knowledge potentials. A formalistic unweighted approach like that in the first part of this chapter facilitates data gathering and could perhaps be automated in the future. However, when applied to the second data sets, it could be viewed as coarsening the information contained in the sources. It was argued that source-critical interpretive modelling methods (the hermeneutical approach) can map the text more accurately and can achieve improved results in terms of the historical interpretation. However, they have not yet been tested sufficiently and they make high demands on the processor who has to combine historic and network theoretical knowledge. It will be worthwhile in the future to use both methods of data gathering and to tailor them to the aims of the analysis.