State of the Art

Network analysis of the manuscript context of Old Icelandic literature

Katarzyna Anna Kapitan

kak@hum.ku.dk

University of Copenhagen, Denmark

Aim

This paper explores the possibilities of applying computer-assisted methods to
the field of Nordic manuscript studies, with a special emphasis on a network
analysis—in a broad sense—of manuscript context. A case study of one Icelandic
legendary saga's manuscript tradition is used to test the hypothesis that the
manuscript context can carry information about ethnic genre associations of the
text (on ethnic genre in Old Norse literature see: Harris, 1975; on
legendary sagas as a genre see: Quinn, 2006).

Research Questions

Hromundar saga Gripssonar traditionally belongs in the corpus of legendary
sagas (fornaldarsogur); it was included in the second volume of Rafn's (1829) 
Fornaldarsogur Nordrlanda, and in Bjorner's (1737) Nordiska kampa dater. The
saga as it is known today, however, is a post-medieval re-working of a
metrical version of the story known as rlmur (Brown, 1946), and is probably not
much older than seventeenth century. Therefore it does not necessarily fit well
with the other texts included the corpus of legendary sagas, as they usually
date from the fourteenth and fifteenth centuries (Driscoll, 2005:207). This
makes Hromundar saga Gripssonar an interesting case study for investigations of
the text's genre affiliation in the extant manuscripts preserving the saga.
Does it appear frequently in manuscripts with the older legendary sagas or with
younger rimur-based narratives? To answer this question, I first examine the
manuscript context of legendary sagas as a corpus, based on collaborative
research with Rowbotham and Wills (Kapitan et al., 2017); second, I examine the
position of Hromundar saga Gripssonar within the corpus and its relationships
with other texts.

Much discussion in the field of Old Norse studies centers on whether the
legendary sagas deserve to be considered a separate literary genre, or should
instead be analyzed as chivalric literature (Quinn, 2006). One of the main
reasons for these considerations seems to be the fact that the term
fornaldarsogur is not attested in the medieval texts; it was introduced in the
early nineteenth century by C.C. Rafn, who published a collection of texts
under the title Fornaldarsogur Nordrlanda (Rafn, 1829). Rafn's selection of
texts and the definition of fornaldarsogur as a corpus of texts dealing with
events taking place in Scandinavia before the settlement of Iceland, however,
was not detached from pervious scholarship of early eighteenth century 
(Lavender, 2015). The current discussion on the legendary sagas as a corpus (or
a genre) is polarized around contradicting opinions. Some scholars suggest that
the legendary sagas had to be considered a separate category in pre-modern
period, because they are frequently bound together in the manuscripts 
(Gudmundsdottir, 2001:cxlvii; Mitchell, 1991:21) while others, using the same
argument, emphasize strong connections between the legendary sagas and the
chivalric sagas (Driscoll, 2005:193). Additional problems arise when
classifying the generic hybrids (Rowe 1993; 2004) appearing within the corpus,
or distinguishing between the legendary sagas (fornaldarsogur) and the late
legendary sagas (fornaldarsogur sldari tlma; Driscoll 2005). Even though
scholars eagerly turn towards the manuscript context to support their claims
regarding the genre classification, no comprehensive overview of the legendary
sagas' codicological context has yet been presented. This gave rise to the
project Stories for all times conducted at the University of Copenhagen, which
created a complete catalogue of manuscripts preserving legendary sagas. The
catalogue contains 818 TEI-conformant XML-based manuscript descriptions with
over 8000 items, 1764 of which are classified as legendary sagas and 920 as
chivalric sagas. This amount of data is much too large to be analyzed manually,
therefore it is necessary to apply computer-assisted analysis in order to draw
some general conclusions regarding this corpus and the relationships between
these texts.

Methods

The first part of my paper, which aims to establish the position of Hromundar
saga Gripssonar within the wider context of the manuscript, draws on
the network analysis of the corpus, conducted in collaboration with Rowbotham
and Wills (Kapitan et al., 2017). There, the codicological context of a text
was considered as a system of relationships between texts, and following Hall's
(2013) approach in his network of chivalric sagas, texts were represented as
nodes, manuscripts as edges, both visualized with the free visualization
software Gephi. The second part is based on database queries aimed at obtaining
detailed information about particular manuscripts and their contents. The main
focus of the analysis was to examine the manuscripts preserving the
complete texts of Hromundar saga Gripssonar in Icelandic, therefore the
manuscripts containing excerpts and translations were ignored. The distribution
of texts appearing frequently alongside Hromundar saga Gripssonar by century
has been obtained using XPath queries of the online catalogue Stories for all
times. Main Findings

As a result of this research, the hypothesis can be confirmed: generally, texts
belonging to one genre appear most frequently in manuscripts with other texts
belonging to the same genre. However, an interesting transmission history of 
Hromundar saga Gripssonar suggests a close association of this saga with the
late legendary sagas, and in particular Bragda-Olvis saga. Both Bragda-Olvis
saga and Hromundar saga Gripssonar are post-medieval reworkings of older metric
versions of the stories (rimur), and for both texts the manuscript AM 601 b 4to
(Arni Magnuisson Institute, Reykjavik) was suggested as the witness carrying
the best text of the saga (Andrews, 1911; Brown, 1946; Hooper, 1934; Hooper,
1932). Even though Hromundar saga Gripssonar appears most frequently with
texts classified as late legendary sagas in pre-1800 manuscripts, after 1800
the texts classified as (traditional) fornaldarsogur start to dominate.
The late Bragda-Olvis saga dominates the pre-1800 setting, but the distribution
changes in the nineteenth century when Porsteins saga Vikingssonar,
Starkadar saga gamla, Fridpjofs saga ins fr&kna, and Halfs saga Halfsreka
appear more frequently (as presented on figure below). Starkadar saga gamla is
a late-eighteenth century saga written by Snorri Bjornson (1710-1803),
utilizing traditional legendary motifs of Saxo's Gesta Danorum and legendary
sagas (Driscoll, 2009:209; Simek and Hermann Palsson, 1987:331), so its
co-occurrence with other legendary sagas starting from the eighteenth century
onwards is not surprising. The three remaining texts that started to appear
more frequently alongside Hromundar saga Gripssonar in nineteenth-century
manuscripts were all published in the same volume of Rafn's Fornaldarsogur
Nordrlanda, in which Hromundar saga Gripssonar was published (volume II);
likewise Fridtjofs saga ins fr^kna, and Halfs saga Halfsreka appeared in
Bjorner's edition from 1737 together with Hromundar saga Gripssonar. This shows
how printed editions influenced the saga's transmission in the manuscript form.
A text, which once showed strong connections to another rimurbased narrative,
became detached from its previous setting and gained new, print-influenced
context after becoming part of printed editions.

[342-1]

Figure 1. Texts appearing frequently with Hromundar saga Gripssonar in
manuscripts by century

Relevance

The topic of this paper fits in the advertised panel "Quantitative stylistics
and philology, including big data and text mining studies," as it employs
database quarrying and network analysis of significant amount of data.

Acknowledgements

The author of this paper owes a great debt to Tarrin Wills for merging the data
from various databases (fasnl.ku.dk, skaldic.abdn.ac.uk,
handrit.is, onpweb.nfi.sc.ku.dk) to obtain the input data for the network
analysis, and to Tim Rowbotham, who worked on standardizing the uniform titles
and genre classifications in the XML-files in the Stories for all times
catalogue.