4. Narratives and textual sources

Textual sources can give detailed and complex information about religion, but because of the dependence on literacy acquired through the church and Christianisation, they are problematic and need to be treated in a way that presents the evidence with sufficient apparatus to evaluate it.

An example text

Snorri Sturluson (d. 1241) gives an account of an episode involving Thor and the giant Hymir where Thor attempts to catch and kill Miðgarðsormr, the serpent that encircles the world. This episode and the various forms of evidence for it are treated at length by Preben Meulengracht Sørensen (1986) and other scholars (cf. Abram 2011). The poem Hymiskviða includes this story within a broader narrative, but the clearest presentation of this part of the story is in Snorra Edda. Part of the episode reads thus:

And when Thor had shipped his oars, he got out a line that was pretty strong, and the hook was no smaller or less mighty-looking. On to this hook Thor fastened the ox-head and threw it overboard, and the hook went to the bottom … The Midgard serpent stretched its mouth round the ox-head and the hook stuck into the roof of the serpent’s mouth. And when the serpent felt this it jerked away so hard that both Thor’s fists banged down on the gunwale. Then Thor got angry and summoned up his As-strength, pushed down so hard that he forced both feet through the boat and braced them against the sea-bed, and then hauled the serpent up to the gunwale. And one can claim that a person does not know what a horrible sight is who did not get to see how Thor fixed his eyes on the serpent, and the serpent stared back up at him spitting poison. And just at the moment when Thor was grasping his hammer and lifting it in the air, the giant fumbled at his bait-knife and cut Thor’s line from the gunwale, and the serpent sank into the sea. (Faulkes 1987, 47)

The translation gives an aid to interpretation for an audience not familiar with Old Norse. Ideally the text would be linked to the primary materials, which would ultimately include the manuscript(s), viewable as digital images of them. The basic structure that can represent this is a series of links representing a one-to-many relationship between images of the physical object and ultimately the repository which is responsible for it. This can be represented by a relational data model (Figure 1).

Figure 1: Links between manuscripts, collections and images

The joins represent one-to-many relationships (the small square is the many side). In other words, one repository may contain one or more collections; one manuscript page can have many images. The tables can include multiple items that are linked individually to an item in the ‘parent’ table. The relationships can be served via a web interface and exported in a machine-readable way using Semantic Web technologies. Users in this way can browse and search the various categories in order to locate manuscripts and view images of individual pages.

The same structure can be used for any physical object which has a visual or textual representation of a religious phenomenon. In most cases the relevant holding institution will have web-based materials that document the various phenomena here and can therefore be linked as URIs (web links). Depending on the collection, some of these structures may have to be incorporated into the electronic resource itself containing the necessary metadata to fit the materials into the structure. Eventually such data can be replaced by an authoritative external resource.

This structure provides a means of electronically identifying the physical objects which provide evidence of the religions, and a means of incorporating information about their provenance and date which can then be incorporated into the analysis by electronic links.

At this point we can introduce a specifically textual environment which is linked to the data associated with the physical object. Abstracting the text from the manuscript allows for incorporating and linking different versions of the text. It requires linguistic and textual divisions (words, sections, chapters) to be useful, rather than the physical divisions of the material item (codex, page). The most flexible way of creating the link between the object and text is to use the most specific physical phenomenon which contains text, namely, pages. There may be different images of a single page, but the text on it is the same for each image of the whole page. The page can form the basis of linking to the texts that are represented on it and which may occur in more than one manuscript. It would be technically feasible to link the text to individual lines in the manuscript, but this would require a very large amount of additional data entry and processing. The page is therefore the most specific practical link that can be made between the text and the physical object.

In order to effectively link pages to different parts of the text we need a relatively small but meaningful division of the text in its edited form. Most editors use chapter divisions which normally are found in one or more manuscripts. Chapters therefore provide a useful subdivision of the text for the purposes of linking information. The basic structure can be represented as a series of relationships that link the ‘page’ resource above to the textual environment (Figure 2).

Figure 2: Links between texts and manuscript pages

Because there is a many-to-many relationship between chapters and pages (one chapter may occur on more than one page, and one page may contain more than one chapter), a separate table (‘Text on pages’) is needed to record these links in a relational data model.

Different manuscripts, editions and translations will sometimes use slightly different divisions and may result in different numbering. The resource therefore needs to select a single set of subdivisions in order to accurately link information to the text. A table containing at least the metadata necessary for identifying the particular relevant piece of text will be necessary for linking the chapter to other phenomena. This will contain numbering that is unambiguous; ideally this will be based on the most authoritative edition of the text, but need not be as long as the reference arising from it is unambiguous: the text provides a way of identifying the correct subdivision of the text for linking authoritative versions and other phenomena. For internal reference, a representative text should also be incorporated into the resource. Digitised out-of-copyright texts of relevant sources are often available on the internet and can serve this purpose.

Figure 3: Fragment of tables linking texts and manuscripts

The links between rows in each table can be seen in the Figure 3, which represents a fragment of the overall tables. Within each row, columns store information associated with each item, such as dates and locations (for manuscripts and objects), biographies for authors, introductions to texts, editions and translations for chapters, and so on. The lines represent links between rows in each table, with one-to-many relationships shown by links between single rows in one table and multiple rows in another. A user interface can provide various means of searching and browsing the relevant texts and content, with links to the manuscripts and images of manuscript pages which contain the relevant text.

There is considerable potential for populating such a structure because of the large number of resources that have been digitised and are available via relatively stable URIs. For editions and translations of Gylfaginning ch. 48, for example we have publicly available Faulkes’ edition (p. 44), Finnur Jónsson’s edition (ch. 32, p. 62), Brodeur’s translation (p. 69) and Guðni Jónsson’s text (ch. 48). In some cases, such as particularly important and relatively short extracts, members of the PCRN project may make editions and translations of their own.

Some of these resources are bibliographic and further reference can be made to an internal bibliography or external resources such as openlibrary.org. The references can be incorporated into the relational model (Figure 4).

Figure 4: Links between texts and editions

In some cases the text will form part of another text, such as þættir and sagas in compilations (e.g. Gylfaginning forms the first part of Snorra Edda). These relationships can be encoded as a recursive reference to the texts table within the texts table.

The same basic structure can be applied to poetry. In the case of stanzaic poetry (the overwhelming majority of Norse poetry), the most useful division is the stanza itself. For poetry then, ‘chapter’ can be taken as the equivalent of ‘stanza’. For non-stanzaic poetry such as is found in other Germanic languages (e.g. the Old English poem Beowulf) another set of divisions may be needed in order to break longer poems into more manageable units for referencing. For Beowulf the 40-odd numbered sections in the manuscript are used.

Runic inscriptions normally have a one-to-one relationship between the text and object which makes the structures above redundant for representing the relationship between the text and object. However, this is not always the case: some objects contain more than one runic inscription, and some inscriptions are repeated on more than one object, or very occasionally extend over more than one object. In addition, the text may be recorded on more than one side of the same object. This paper proposes retaining this structure for runic inscriptions, which additionally allows for linking images to specific sides of the inscription.

Non-textual phenomena can also fit in this structure, albeit with some conceptual adjustments. Physical objects with pictorial representations of potentially religious phenomena can normally be divided into one or more separate pictures (such as the panels and sub-panels of the Franks Casket). These pictures can be treated as the equivalent of chapters, stanzas and inscriptions as linkable units for linking interpretative data.

References