Gripla - 2020, Page 27

GRIPLA26
figures. It is our hope that this will make the investigation both accessible
and transparent. It is also our hope that this iterative process will assist the
reader in assessing whether, indeed, we are successful in removing bias
from the test environments (in keeping with point 2). Between each itera-
tion, we attempt to determine whether the results can be explained by style
rather than something else (in keeping with point 3).
Preparing the documents
The basic precondition for stylometry is to have the texts in machine-read-
able format and split into different documents. Given the reconstructive
editorial history of Ljósvetninga saga discussed above, we could not rely on
a particular edition at the outset. We opted instead to produce new texts of
the A- and C-redactions according to a best-text approach. As mentioned
above, the texts of the redactions from the Svart á hvítu Íslendingasögur
edition were used as a beginning text56 and were updated to be closer to
the manuscript readings of 561 for A and 162 and 485 for C,57 applying
consistent normalization practices between the two versions.
Once this was achieved, the texts of each version were split into the
following documents:
Document 1: Ljósvetninga saga A chapters 1–4, 19–21 (hereafter
“A-parallel”)
Document 2: Ljósvetninga saga C chapters 1–4, 19–21 (hereafter
“C-parallel”)
Document 3: Ljósvetninga saga A chapters 13–18 (“A-divergent”).
Document 4: Ljósvetninga saga C chapters 13–18. (“C-divergent”)58
56 Íslendingasögur og þættir, II, ed. Bragi Halldórsson et al.
57 On the selection of 485, see n. 45.
58 A- and C-parallel also include the last part of chapter 18 in A and C, which is where the
versions converge. The A-divergent document omits the seventeenth-century summary on
37v of 561 mentioned in n. 5.