How hand copying fails, and why a shared mistake carries more information than a frequent one.
22 March 2027
Three weeks ago you controlled what changed and measured the effect. Last week you had one output and no history. This hour starts a different kind of problem: many copies, no original, and a question about how each copy diverged rather than whether it did.
Three ways a hand copy fails#
A copyist’s eye and hand fail in a small number of recognisable ways, and naming which one happened is more useful than counting how many differences there are.
Omission. Something present in the exemplar is simply missing from the copy — a word dropped, a negation lost. An omission that reverses a rule’s sense, rather than merely shortening it, announces itself: the copy stops making sense on its own terms.
Repetition and its opposite. A copyist can write a word or phrase twice where the exemplar has it once — a dittography — or can skip from one occurrence of a repeated word or phrase to a later occurrence of the same thing, silently dropping everything between — haplography, or, when the skipped span ends on matching word-endings rather than a matching whole phrase, homoeoteleuton. The direction of these two is usually not in doubt: an eye skipping ahead over a repeated anchor is a well-attested failure, and a copyist spontaneously inserting a repeated clause that was never there is not.
Substitution. One reading stands in for another — a synonym, a misreading, or a deliberate correction. This class is the one that most needs a second question asked of it, and it is where the week’s check comes from.
The reading that lies to you#
Not every substitution is scribal error in the sense the previous three are. The commonest substitution is quiet normalisation — regularising an archaic idiom or an odd spelling into its more familiar form — and it is a change scribes make unconsciously and independently. Two witnesses that agree only on a normalisation may share nothing more than the same ordinary instinct toward tidiness. That agreement looks exactly like shared inheritance and is not: it is what a polygenetic change looks like, one that entered more than once because nothing about producing it depends on having a particular ancestor.
Contrast that with a haplography, or an omission that breaks the sense of the line it is in. Neither has an ordinary mechanism for happening twice, by different copyists, in exactly the same place, in exactly the same way. A reading like that is the kind of evidence that actually locates a shared ancestor.
The rule the week hands you. A shared error is strong evidence for shared ancestry in proportion to how implausible it is that two copyists made it independently. A shared normalisation is weak evidence for exactly that reason: it is entirely plausible that two copyists made it independently.
Anonymised, and withheld on purpose#
Wednesday’s five witnesses disagree with each other in several places, and none of you can see the document they all come from — the Ordinance stays in the print collection’s withheld half until week 6’s reveal. That is deliberate: reconstructing a text you already know how to read is a memory exercise, and this course is not testing memory.
Your job this week stops at classification: what kind of mistake happened at each disagreement, and how confident that classification is. Week 6 is where the table you produce gets turned into a claim about who copied whom.
Before Wednesday#
Read the branching instructions linked from the workshop
page. Do not open reveal/ — it holds the answer to a puzzle you have not
been given yet.
Teaching team
- Dr Wren Halloway — convenor