Early in my graduate school career, I was affected deeply by a comment from textual scholar Peter L. Shillingsburg. He asked, about some work that I was doing–and I paraphrase to the best of my memory–“If your quotation of the work absolutely had to be correct, would you cite this edition?” He referred to an edition that had been prepared on the basis of microfilm copies. The procedure in that edition was that the text was keyboarded and carefully proofed. When microfilm was not clear, original was checked.
On thinking long and hard, I decided that the answer was no. I would not trust the edition. Although the edition was prepared with typical standards, I silently swore to myself that from here on out I would always attempt nothing less–in editorial work–than accuracy of transcription with excruciating attention to detail. I would here remark that it requires considerable self-discipline and knowledge about tools. One can never, ever take an easy way out. And one has to understand how to use electronic collation software.
Therefore, editorial work requires at least two independent keyboardings. When I worked on Era UTC, I performed one keyboarding myself, I compared every keystroke to a text prepared by a keyboarding prepared for another electronic text transcription project (Accessible Archives). Every variant character between the two keyboardings was individually checked against an original printing and corrected. Then I and a partner orally proofed the corrected text. To test the accuracy of oral proofreading, I had someone insert errors in the text. During oral proofreading, we caught approximately 80 percent of the inserted errors. And thus based on the perhaps shaky assumption that the rate of catching planted errors matched errors in two keyboardings, I was able to estimate the number of errors that remained in the text. To discover a missed error in the oral proofreading was horribly disappointing–every time–and to have an estimate of the errors that remained was also horrible. I also compared text to a third keyboarding (of the Jewett text). Every time I corrected an set of errors, I always corrected them in two places, to confirm that the correction made properly. For full details of procedure, see chapter 1 of my dissertation.
What I did not do, when all was said and done, was to go back and silently proof the entire text. I was stymied by the two versions of the text on my site, and to proof the public text would require two silent proofings. As the public version of my site was only available for about three weeks before my defense, I went on vacation and put off another proofing until after dissertation was done. Being a human, the impulse to go back and re-proof the text twice (but also thinking philosophically about whether I have a responsibility to keep text in same form in which it was published), when silent proofreading is so mind-numbing and the work that is more likely to provide professional advancement is writing, has been postponed until I need to update the text.
I’ve proofread two installments since and found one error. The individual error is a case where I made a mistake in transcription, Accessible Archives made the same one, I missed it in oral proofreading, and I did not catch it when comparing the Jewett text, probably because I had another error in the same phrase, a reversed quote mark. Mistakes are not unlikely in a text of 700,000 characters. The error is this. When Eliza says
[…] give your consent—to—to”
the last “to” should be followed by an em dash. But it is annoying because one error per two installments would be a higher rate of error than I estimated in my edition.
I am currently reviewing all of the variants in between the Jewett text and the Era. Based on initial work, I begin to think I’m going to find 30 or more variants that by some reasonable criteria would count as errors in the Jewett text. So, onward.