Week 2 Session: Reading the source text

Where our characters come from. Candidates come from repetition, they have to survive three tests, and the first one the Classic of Mountains and Seas offers is incomplete in more than half the records we cite.

1 March 2027 · Phase 1, Build a system that works

What this week does#

Week 1’s first move was choose characters, and it skipped the hard part: you cannot choose from nothing. Linnaeus had it easy, because every flower has stamens, so “how many stamens” is available for every case before he starts. Material nobody wrote for you rarely hands you anything that clean.

So this week is the step underneath step one — where candidates come from, how to tell a usable one from a plausible one, and what to do when the material does not say. Five sections:

1. Candidates come from repetition, which makes the first pass over any material a count rather than a close reading.
2. Three tests a candidate has to survive, the third of which Borges’ encyclopedist fails fourteen times running.
3. A record ruled into the columns — 鹿蜀, and then 蛊雕, where the name arrives first and the consequence is a behaviour rather than an omen.
4. The count that tests our own candidate — 21 of 45 cited lines carry the comparison and 11 carry the omen.
5. Three ways a cell can be blank — 夫诸, which has no use, and 軨軨, which has every slot and a flood at another scale.

Only sections 3 and 5 work a record. The other three are method and arithmetic.

Candidates come from repetition#

You do not invent characters. You notice the material keeps saying the same kind of thing, and you make a column out of it.

The first pass over any material is therefore counting, not close reading. Which phrases recur? A shape that repeats is a candidate. A shape that appears twice is an anecdote.

The 山海经 (Shānhǎijīng, the Classic of Mountains and Seas) repeats heavily. Its records are built on:

其状如 X，Y 首，见则 Z

shaped like X, with Y for a head, and when it appears, Z

A comparison, a distinguishing part, a consequence. Three candidates, handed over.

Three tests a candidate has to survive#

Comparable. You can check it on every case and get an answer of the same kind. “Number of legs” is comparable. “Impressive” is not.

Checkable. Another reader, same material, same answer. If two readers disagree and neither is wrong, it is not a character yet.

About the thing. It describes what the thing is. Not what it is worth, not what it does for you, not where you found it. This is the test Borges’ encyclopedist fails fourteen times running.

Ruling the columns#

This is the card you are filling in, for 鹿蜀 (lùshǔ) from the first route:

Now the working out. The line reads:

有獸焉，其狀如馬而白首，其文如虎而赤尾，其音如謠，其名曰鹿蜀，佩之宜子孫。

There is a beast there, shaped like a horse with a white head, marked like a tiger, with a red tail, and its cry is like singing. It is called 鹿蜀. Wear it and your descendants will prosper.

Ruled into the slots the formula offers:

Now 蛊雕 (gǔdiāo):

水有獸焉，名曰蠱雕，其狀如雕而有角，其音如嬰兒之音，是食人。

In the water there is a beast called 蠱雕, shaped like an eagle but with horns, its cry like an infant’s. It eats people.

Same slots, but the name arrives first and the consequence is not an omen. It is a behaviour: it eats people.

Count before you trust it#

Week 1’s lesson was that a classification can be wrong and you can say why. Apply that to our own candidate before building on it.

This course cites 45 base-text lines. Of those:

The formula is not a law the text obeys. It is a template the text reaches for and completes fewer than half the time.

Missing is not zero#

夫诸 (fūzhū) has a comparison, a name and an omen, and no use:

有獸焉，其狀如白鹿而四角，名曰夫諸，見則其邑大水。

There is a beast there, shaped like a white deer with four horns. It is called 夫諸. Where it appears, that town floods.

軨軨 (línglíng) has all three and nothing else:

有獸焉，其狀如牛而虎文，其音如欽。其名曰軨軨，其鳴自訆，見則天下大水。

There is a beast there, shaped like an ox with tiger markings, its cry like a groan. It is called 軨軨, it calls its own name, and where it appears, there are floods.

Two floods, two scopes: 其邑 is one town, 天下 is everywhere. Same slot, different magnitude, and the formula has nowhere to record the difference.

夫诸’s missing use is not “no benefit”. It is no information. Three different things can put a blank in a cell, and a system that stores them identically has destroyed information:

• the material says the value is nothing
• the material does not say
• nobody has checked yet

Only the first is a value. The other two are states of your own knowledge.

What you leave with#

Column names, agreed in the room and kept all semester. Name them now, while they are still arguable: an unnamed column gets quietly widened every time something does not fit, and nobody notices when it starts holding four unrelated kinds of thing.

Your first specimen card is due Sunday, for any creature in 南山經 other than 鹿蜀. It cites its line, fills the slots that line supplies, and leaves the rest blank. Marked on one thing: whether the blanks are honest.