Week 6 Workshop: Reading the damage backwards

Which family tree does the evidence support, and what remains unknown?

31 March 2027

Which family tree does the evidence support, and what remains unknown?

Learning target. Justify proposed edges and compare at least one competing history.

Before the workshop#

Bring the anonymised witnesses and error table another group handed you at the end of week 5 — or, if your own exchange did not happen, use the instructor fallback pack: five witnesses of the Ordinance of the Copy-House, witness-a.txt through -e.txt, with collation.csv listing every point where they disagree. Do not open reveal/ yet.

Check — 5 minutes#

Two candidate trees are shown below, built from the same four-line error table.

Tree 1: {A, C} share an ancestor because they share point 3. Tree 2: {B, C} share an ancestor (point 2), and A’s point 3 is independent.

Which tree does the table support, and which single point is doing the misleading work?

Try it on any pair from this same three-witness table: pick two witnesses and see which points they share and how strong each one is, using week 5’s strong/weak rule. This is practice on the toy table above, not your own workshop evidence — it has nothing to say about the five real witnesses (A through E) you work with from Investigate onward.

Receive — 15 minutes#

Confirm every disagreement point in your table is classified by type (from week 5) before you start building anything. A tree built on an unclassified error table is guessing at which agreements are worth trusting.

If you want to watch error accumulate down a chain of copies before you reconstruct one from a table, try the chain simulator: each generation there corrupts only the previous generation’s output, the same “copy of a copy, not of the original” structure your five witnesses share. It is a supplement to this workshop, not a substitute for the tree you build from the actual evidence table.

Investigate — 40 minutes#

Construct a candidate graph (20 minutes). For each shared error, ask: is this the kind of mistake two copyists could plausibly make independently, or does it need a common immediate source? Group witnesses that share an implausible-to-repeat error under one hypothesised ancestor. Draw the nesting — which groups sit inside which — as a tree, or as an edge list if drawing is not available to you.

Apply the same strong/weak rule the Check-stage tool above practised, now to your own five-witness evidence table — nothing here reports the answer for you, since that stays locked behind reveal/.

Justify and challenge selected edges (10 minutes). For every edge you drew, write the one shared feature it rests on. Then pick your least certain edge and argue against it: what would have to be true for two witnesses to share that feature without the ancestor you proposed?

State the strongest alternative and one unresolved relationship (10 minutes). Name one pair or group your evidence cannot order — say what would resolve it, not just that it is unresolved. Then state the strongest competing tree the same table would also support.

Evaluate — 30 minutes#

Write your proposed tree, its evidence table, your named alternative and your unresolved relationship — before opening reveal/.

Deposit — 30 minutes#

Open reveal/production-log.md (ask your tutor for the link once your tree is drawn) and compare it against your proposed tree. Write a short correction note: which edges matched, which did not, and — if your tree differs from the log — whether your reasoning was still defensible given only the evidence you had. Deposit your tree, your evidence table, your alternative and this correction note into the print collection.

What you should find#

The pack supports one tree with one deliberately unresolved edge, not a fully resolved one. A lacks the shared haplography and sits outside it entirely. B shares the haplography but not the following homoeoteleuton, placing it above the split that produced C from one side and the identical pair D and E from the other. D and E agree at every disagreement point in the table, so the evidence cannot choose between “D copied from E”, “E copied from D”, and “both copied from a lost sibling” — that is the unresolved edge, and naming it as unresolved is the correct answer, not a shortfall.

The trap is point 4 in the full table (the log’s numbering, not this page’s check): A and C share a normalisation that suggests a direct kinship the rest of the evidence rules out. A tree that trusted every shared reading equally would draw an edge between A and C that the haplography and homoeoteleuton evidence directly contradicts.

Accessible route#

Textual edge list and evidence table alongside every diagram; no drawing tool is required, and the deck and this page both give the tree as a table of parent/child rows with the evidence for each row.

Assessment#

The Stemma’s curated pack releases now, at the end of this workshop. Open the assessment brief for the submission format and deadline. The method is what you practised today — evidence table, candidate tree, one named alternative, one honestly unresolved edge — applied to a separate, larger set of witnesses with no reveal until after the assessment closes. Marks reward that method, not agreement with a hidden answer: “the evidence does not resolve this edge” earns full credit where it is argued.