This is a working paper at a relatively early stage. All feedback is welcome, to The URLs of files linked to from this page may not be stable, so don't rely on them. The URL of this page will, however, contain or point to this article long-term and can, therefore, be cited.

References will be found at Where possible, references link to online versions of the texts cited.

A stemma of Sigurgarðs saga frækna

Alaric Hall, University of Leeds


Since 2009, I've been chipping away at making a stemma of Sigurgarðs saga frækna, a fifteenth-century romance saga which I have also been involved in translating (Hall, Richardson, and Haukur Þorgeirsson forthcoming). This involved a major detour as I tried to establish a sound, transparent, and efficient method for making the stemma by testing approaches on the relatively well understood romance saga Konráðs saga keisarasonar (Hall and Parsons 2013). But I'm finally (in 2013) back to sorting out the stemma of Sigrgarðs saga. This will produce a pretty reliable stemma of the saga, but also some further methodological advances on Hall and Parsons 2013. So this page should be developing fairly rapidly through August and September 2013.

The saga saw a popular edition in 1884 (Einar Þorðarson 1884), which is included in the stemma, and a scholarly edition in Loth’s milestone edition of Icelandic romance-sagas (1962--65, V 41--107). Loth used the oldest manuscript, AM 556a 4to (Eggertsbók), where available, and otherwise the near-identical AM 588m 4to, occasionally recording or using alternative readings from AM 123 8vo and AM 167 fol. The edition which I was involved in simply normalised Loth's text.

Why Sigurgarðs saga?

In reality there are various tangential reasons why I wanted to work out the stemma of this saga. But the short answer is that I'm interested in studying the kinds of social networks between scribes and other literary practitioners during the long history of Icelandic manuscript production from the Middle Ages right down to the early twentieth century. This is a bit like the kind of work cool people like Davíð Ólafsson (2009) have been up to, but from more of a big-data perspective. Romance sagas were widely copied, widely adapted into verse, and so clearly widely read during this period, but have received less attention from editors and literary critics than the Íslendingasögur or fornaldarsögur, so they seemed a fun place to start (for a survey of work so far, with links to stemmas, see Hall and Parsons 2013, §1.3). With a sprightly story and some great characters, as well as nearly twice as many manuscripts as the average for an Icelandic romance-saga (56), Sigurgarðs saga seems a pretty reasonable case study. Working out the stemma of Sigurgarðs saga is intended as part of a raft of investigations into romance-saga stemmas which will cast considerable new light on how Icelandic scribes worked: who copied off whom? How did they alter the texts they copied? How did they choose which texts to copy? And, ultimately, why did they bother at all?

For a sense of the diversity and complexity of the tradition, check out this diagram, which combines the romance-saga stemmas established by previous scholarship (for Konráðs saga, Dínus saga drambláta, Gibbons saga, Mírmanns saga, and Viktors saga og Blávus) into one monstrous über-stemma:

Mostly all you'll see is a mess, but one striking thing is how seldom two of these sagas ever appear in the same manuscript. Yet at the same time, the romance sagas composed in medieval Iceland, of which we have over thirty, are pretty cohesive in their manuscript groupings---romance-sagas tend to be grouped with other romance-sagas (Hall and Parsons 2013, §1.3). What this tells us is that we need a much large number of accurate stemmas before we reach a critical mass for understanding how scribes chose (or, as the case may be, were constrained in choosing) what sagas to copy. Alongside Sigurgarðs saga, I and/or Sheryl McDonald Werronen are also working on Nítíða saga (cf. McDonald Werronen 2013), Nikulás saga leikara (working paper, drawing on Wick 1996), Jarlmanns saga og Hermanns, and further down the track perhaps Bærings saga fagra, Sigurðar saga turnara (cf. Spaulding 1982), Ambrosius saga og Rósamundu, Flores saga og blankiflur, and Halfdanar saga Eysteinssonar.


Hall and Parsons 2013 shows that most Icelandic saga manuscripts introduce so many unique variants that you can make a pretty reliable stemma on the basis merely of short transcriptions from the beginning and end of the saga (c. 300 words in total), except that the stemma is liable to be unreliable right at the top, where data is patchiest. The present paper uses much the same methods. However, it develops them in two ways, giving a three-stage research process:

  1. On the model of my previous study, I first establish a stemma using two short samples (this time, in order to better capture data from incomplete early manuscripts, one from the beginning and one from near the end but not the end itself), putting them through the Pars program of the phylogenetic software Phylip, and then building on that with human analysis. (I've done most of this part, but there are a few MSS which I missed out on the first pass, so I'll have to put those in to complete this stage.)
  2. While I established that my Konráðs saga stemma based on small samples was, apart from problems right at the top, otherwise about as good as the previous efforts, I did not explore whether fuller sampling would produce substantially different results. Therefore I took five samples, spread through the saga (otherwise chosen for being moments of literary interest whose manifestation I wanted to trace throughout the textual tradition), totalling 616 words. I'm manually testing the stemma from (1) against this larger dataset, correcting some errors and finding reasons to rejig the stemma slightly (but seldom profoundly).
  3. Even with this extra data, the stemma is likely to be unreliable at the top, where the data is most sparse. Therefore I'm also taking readings in manuscripts which look like they might be primary witnesses to the top of the stemma from the 151 occasions where Loth recorded or used an alternative reading, providing a cross-section of readings already known to be sites of variation throughout the saga. I'm gathering this data but haven't yet analysed it systematically.
Hopefully when I've done all this I'll not only have a nice verdant stemma, but I'll also have developed a stronger sense of how worthwhile all that extra effort was, and whether I could have done without some of it. I'd also like to explore the value of minor variants (such as function words) a bit more and---not least because Lethbridge has been discussing their literary interest (2012, 361--63)---titles. It would be interesting to have a sense of how far scribes intervened in titles and rubrics. And it would be very handy if you could work out a stemma from the title of a text, as these are normally given in library catalogues. Albeit that probably you can't. Worth checking though.

All the samples are available in their raw format, keyed to Loth's edition, in this spreadsheet. Or they will be when I've finished doing them---mostly it's done.

Stage 1: the stemma based on samples 1 and 4

My spreadsheet of aligned readings from samples 1 and 4 is here; the Pars infile derived from it is here. This was run through Pars not using the factory settings (as in Hall and Parsons 2013) but multiple times using randomised input orders; this process produced consistent results (and results consistent with running using the factory settings). The unrooted stemma derived by running the output in Drawtree is here. It's pretty similar to what I eventually decided was the right tree, actually---probably more successful than the attempt with Konráðs saga. In due course, when I've got all the data in place, I'll check its similarity to my stage 1 and stage 3 final stemmas using Teemu Roos's program Randkistance. And it shows some cool clustering: one of the handy things about those Pars stemmas is that they give you a good sense both of clustering and of distance between MSS.

I then used the unrooted stemma as the starting point for filiating the manuscripts manually. Eventually, once I've actually finished this stage, I'll do a nice .dot file to show this but for now here's the one I made earlier:

Stages 2 and 3

I'm working on them. The spreadsheet of extra readings for the top of the stemma is here. The final stemma in .svg format is taking shape thus:

And in .html format (the creation of which is integral to the process of checking readings---albeit perhaps an unduly laborious way to do it) thus. Text in bold shows alterations from the parent manuscript. Bold ellipses in curly brackets ({...}) indicate omissions from the parent. $ indicates an illegible letter; uncertain readings are marked with [?]. Line- and page-breaks in the MS are not replicated, but NB that hyphens marking word-breaks across lines, catchwords, etc., are still included.

The main surprise so far in stages 2 and 3 has been the conclusion that the saga isn't rooted in AM 556a 4to. Rooting in AM 556a 4to seemed pretty sensible at first. It works for the pars stemma, is indeed a parsimonious interpretation and so is methodologically attractive, and would be similar to Hast's positioning of Harðar saga, from the same MS, as the root of all surviving MSS of that saga (1960a, 1960b), along with the view that this MS is the root of all MSS of the shorter version of Gísla saga (Þórður Ingi Guðjónsson 2010, 108). On the other hand, Loth obviously assumed a lost archetype of which AM 556a 4to and AM 123 4to were independent witnesses. This would make AM 556a 4to more comparable to the vellum Vestfirðir MSS of Viktors saga og Blávus, which are short on direct descendants, most of the textual tradition deriving instead from one or more lost medieval manuscripts. Certainly this interpretation would explain why it's only AM 556a 4to (and sometimes AM 588m 4to) that call Ingigerður's kingdom Taricia instead of Tartaria---otherwise you have to assume that at least two if not more scribes changed that to Tartaria independently. But that's far from unthinkable---lectio difficilior potior etc. I became convinced that 556a can't be the ancestor of all MSS because of some archaic spellings in (particularly) Lbs 423 fol (which I had initially written off as being the product of an archaising scribe), which seem likely to go back to a medieval exemplar independent of AM 556a 4to: Lbs 423 fol sometimes has -o for -u in unstressed syllables; v̈vine in §1; at in §§2, 3; an instance of lute for hlute in §3, and liöp for hljóp in §4 where AM 556a has hliop, and it's hard to imagine a later intermediary scribe dropping the h- there; Tartariam in §3, rendered as Tartaria in catchword. Those geofumm (for gjöfum) type spellings are really weird---what's going on with them? And a few marginal corrections in a hand that looks as old as the scribal one (is it the scribal one?), though never in the transcribed passages, showing that someone's taking care in their copying. Some of these archaic forms may echo in Rask 32 and even (at least úvini) LBS 3891 4to or ÍB 165 4to, but that might be wild fantasy on my part. I'll have to have a proper look. Given all this, it might even make us think twice about previous rootings of Harðar saga and Gísla saga.

As with Konráðs saga, this stemma also emphasises how a text that looks so radically different from the archetype that it must be a different recension can actually be the product of gradual alteration.

Cool stuff about scribal culture

I plan to do with this saga what I did with Konráðs saga (Hall and Parsons 2013, §5.3) and look at scribal cultures. Similar patterns appear to what I found there, though I haven't refined this data yet: around C17, the kind of long-distance links between episcopal seats and the magnates in the Vestfirðir that Springborg sketched in 1977 (the yellow and pale blue plots on the rough and ready map below). But around C18--19, more localised patterns of distribution appear, with one tight cluster around the Dalir (red plots), and another around Reykjavík (purple). Those Dalir manuscripts! Strikingly conservative, and tightly distributed. If my current stemma is right, Magnús Jónsson í Tjaldanesi, writing in the early twentieth century, gives us an independent witness to a lost medieval manuscript. Pretty amazing.

Cool stuff in different versions


mechanical vs determined variation. Ólafur Halldórsson in postola saga edn 1994 on errors.
genetic and genealogical variants (tilblivelsesvarianter og overleveringsvarianter)
Bengel: variants should be weighed, not counted
Jón Helgason: 'storeyed apparatus'

Codex WOrmianus, mid C14, disses shift of ǿ > ę'

lexical variation? Openness to loanwords? Anything composed by these scribes when not copying?
Einarr Hafliðason as case study?

Þordís Edda Jóhannesdóttir thesis on Sigrdrífumál

Haraldur Bernharðsson, Málblöndum í sautjándu aldar uppskrftum íslenskra...


This project is thoroughly indebted to the kindness and hospitality of the National Library of Iceland, the Stofnun Árnamagnússonar in Reykjavík and the Arnamagnæanske Hꜹndskriftsamling in Copenhagen. It also received funding from the British Academy, for which I am grateful.