The Silversmith

The talking leaves were not magic. They were a technique — and a technique can be reinvented by anyone who knows only that it exists.

The institute's summary of the stimulus-diffusion argument, §5

§1 of 5The Brief

Sequoyah set himself the problem. There was no king, no council and no commission: a Cherokee silversmith, born around 1770 and known in English records as George Guess or Gist, decided some time around 1809 that what white settlers did with marked paper could be done with Cherokee.

The brief was unusually pure, because it contained no specification. Sequoyah could not read English. He could not read Cherokee, because Cherokee was not written. He had seen writing — documents, letters, printed pages, the "talking leaves" of the phrase that follows him through the literature — without ever having been taught what any individual mark meant. What he possessed, and it turns out to be the only indispensable ingredient, was the knowledge that speech could be fixed on a surface and recovered by someone who had not heard it.

That absence of specification is what makes this chapter different from every other in the series. Sejong had scholars and a phonological tradition. Mashtots had Greek. Wulfila had a Bible to translate and two alphabets to raid. Sequoyah had a fact and no method.

§2 of 5The Constraints

The linguistic constraint was favourable and he found it by experiment. Cherokee is an Iroquoian language with a modest syllable inventory and a strong preference for open syllables — consonant plus vowel. A language shaped like that can be written with a syllabary of manageable size, where a language full of consonant clusters cannot.

The social constraints were hostile. Sequoyah worked for roughly twelve years on something his neighbours could not evaluate and many regarded as at best a waste and at worst witchcraft. The tradition — consistent across accounts, and the sort of detail that does not get invented — is that his work was burned, either by his wife or by neighbours, and that he began again.

He was not translating a system. He was deducing that a system was possible, from the single fact that other people had one.

The technical constraint was that he had no analyst to tell him what unit to write. This is the hard part of inventing writing, and the reason it has happened independently so few times in human history: the units of speech are not audible as units. A speaker of any language hears words, not phonemes, and not syllables.

§3 of 5The Decisions

The first decision was wrong, and abandoning it was the second. Sequoyah began with a sign for each word, and the accounts agree he accumulated thousands before concluding that the inventory would never close. That failure is the single most informative event in the chapter: it is an empirical discovery, made alone, that writing cannot be a picture-book of meanings and must decompose speech into recurring parts.

The decision that followed was to write the syllable. Cherokee's phonology rewarded it: the finished system has 85 characters, each standing for a whole syllable, and it fits the language closely enough that a fluent speaker has very little left to guess.

The third decision is the one that unsettles people, and it is a deliberate separation of two things usually welded together. Sequoyah took letterforms from an English spelling book — Roman capitals, some numerals, shapes resembling Greek and Cyrillic letters — and assigned them values with no relationship whatever to their originals. Ꭰ is not D. Ꮃ is not W. He could not read the shapes, so their sounds were not available to be borrowed, and he took the one thing he could use: an existing stock of distinguishable marks.

This is form without phonetics, and it is close to unique. Every other borrowing in this building — Phoenician to Greek, Greek to Gothic, Greek to Glagolitic — carries sound values across with the shapes, adapting them. Sequoyah carried across the shapes alone.

§4 of 5The Adoption Curve

The curve is the steepest in this series, and the reason is that the design was demonstrable. In 1821 Sequoyah presented the syllabary to the Cherokee National Council using his young daughter Ahyokah: the two were separated, a message was dictated to one and read back by the other. The test settles the question in minutes and requires no one to trust the inventor.

The Cherokee Nation adopted the syllabary officially in 1825. In 1828 the Cherokee Phoenix began publishing at New Echota with type cut for the script, printing in Cherokee and English. Literacy spread through households rather than schools, because the system could be taught by anyone who had it to anyone who had not, in days.

Then the curve meets history. The forced removals of 1838–39 took the Cherokee from their homeland along the route remembered as the Trail of Tears; the press was seized, and the community that had built the highest literacy rate in the region was dispersed. The script survived the removal — carried west, used in Indian Territory, printed again — but the chapter cannot claim a clean ascent, and does not.

§5 of 5The Verdict

Separate two claims that are often run together. Sequoyah did not invent writing without knowing writing existed; that has been documented only three or four times in human history and this is not one of them. What he did is what anthropologists call stimulus diffusion: he took the idea that speech can be recorded, and rebuilt the entire technique from it, deriving the syllable as the working unit without help.

Judged as design, the system is very good. It is closely fitted to the language, learnable in days by a fluent speaker, and it has now been in continuous use for two centuries. Judged as evidence, it is the best case in this series for the argument the institute keeps making: writing is a technology, not a possession. It can be analysed, and it can be rebuilt from scratch by someone with no education in it at all — which also means no community's illiteracy is ever evidence about that community's capacity.

The Unicode record adds a quiet postscript. Cherokee was encoded in 1999; the lowercase letters, developed later in the script's life, were not added until 2015. A script does not stop being designed when its inventor stops. See the registry plaque for the current inventory, and Floor 3 for where the language now stands.