Transcriptions produce the raw data from which you calculate the text of the edition. The purpose is not really to provide alternative readings of this or that passage. The variants you find will, at times, do this. Most, however, are errors that have intruded at some point in the transmission of the text, and it is as errors that they are extremely valuable. Agreements of certain kinds of errors allow the reconstruction of families of texts descended, in a closed recension, from a common hyparchetype. The errors will also allow you to identify any witnesses that descend directly from the archetype. These independent witnesses and hyparchetypes together chart the transmission of the text and are the fundamental evidence for the reconstruction of the archetype.
In order to serve a purpose, however, they must be recorded with accuracy. Errors in the recording of errors will cloud relationships that exist, falsely suggest relationships that do not, and undermine the reconstruction of the text. Of course, we all make mistakes and will probably make some in recording the readings of each witness. However, there are at least two safeguards that should be put in place, both common sense steps to reduce our own errors to, hopefully, nil. First, when we see an error shared between otherwise unrelated witnesses, we go back to the manuscripts to check our readings. These mistakes are easy to spot when the manuscript families are distinct and patterns of error consequently clear. This check will help correct any recording error on our part. Second, once the text has been reconstructed and the apparatus put in, it is a good idea to read the edition back against each of the witnesses, moving back and forth between text and apparatus as you move through the manuscript copy (more on this later).
The greatest weakness of Liebermann’s edition of the Instituta was created when he recorded the readings of his witnesses. His rate of error with this text was spectacular (see O’Brien, ‘The Instituta Cnuti and the Translation of English Law’, Anglo-Norman Studies 25 , 179-80). This might seem surprising since his transcriptions of the Old English texts in his Gesetze were so accurate. The errors, however, are there, and their presence undermines the persuasiveness of his editorial decisions.
For the edition of the Instituta, I have transcribed 14 witnesses (nos. 1-4, 6, 8, 10, 12, 13, 15, 16, 21, 23, 25, as identified in the previous blog). There are a few ways to do this. One way is to use a copy of the standard edition as a kind of template on which to record the errors of each witness. I could have printed out Liebermann’s edition and then noted on it the readings of each witness. This is how I worked when I edited the Leges Edwardi. For the Instituta, I decided not to do this partly because Liebermann’s edition nowhere occurs in one piece. Instead, it is scattered throughout the Gesetze, wherever its Old English sources appeared, and then, at the end of volume one, the sequence of the chapters as a whole is presented, albeit in abridged form. The other reason I decided not to do this was that there is a very early and relatively uncorrupt witness available, whose transcription could serve the purpose. This is the copy in Textus Roffensis, written around 1123 and originally standing at the head of the entire collection. It is the full text and not one of the witnesses that ends by some accident midway through what Liebermann called book 3.
So my first task was to make an accurate copy of the text in the Textus. Because I was interested also in the structure of the whole, I incorporated into my transcription all large colored initials, capitals or any kind, and all punctuation. This transcription then became my base text for recording the errors or variants in all the witnesses. To keep the process as clean as possible, I recorded the readings of only one witness on each copy of the transcription. With some texts, you might be perfectly fine to record (with different colored pencils) the variants of a couple of witnesses. The danger with this is that you might find yourself pulled to resolve illegible bits or suspensions found in the second witness in the same way you did with the first witness. In my case, the witnesses to the Instituta were sufficiently messy—especially with their spellings of vernacular terms embedded in the Latin—that combining witnesses was impracticable.
When there are too many witnesses, and the text is very long, you would want to take a series of samples as a first stage rather than record all errors in all witnesses for all of the text. This exploration of the patterns of error might allow you to identify derivative copies which can then be ignored for the edition. The Instituta, while popular, was not that popular, and the whole was less than 40 pages long (A4 or Letter).
I did eliminate some of my 25 witnesses and brought the total down to 14. In some cases, early modern copies stated explicitly that they were copies of manuscripts which are still extant. In some, comparisons of sources and their copies as identified by Diana Greenway confirmed her judgments about which copies were derivative. All of these, along with the now lost manuscripts, reduced the pile significantly.