A script born old

Claim (verbatim)

The Tangut script was promulgated at a stroke by imperial decree in 1036 under Weiming Yuanhao — tradition credits the scholar Yeli Renrong with its design — as roughly six thousand logographs built in deliberate visual differentiation from Chinese. Natural logographies are eroded by use: in Chinese, frequent characters tend to be simpler, because centuries of writing economy wear the common signs smooth. A script issued complete, used for under three centuries by a chancery-and-clergy class, and never passed through mass scribal attrition should lack this wear pattern: its complexity should be flat across the frequency spectrum. Prediction: joining per-character frequencies from digitized Khara-Khoto texts to stroke counts, the rank correlation between frequency and stroke count in Tangut will be no stronger than -0.1, while the matched computation on a pre-modern Chinese corpus yields -0.25 or stronger; secondarily, Tangut characters will average at least 25% more strokes than the Chinese characters glossing them in Li Fanwen's dictionary (primary clause: the correlation contrast at the stated thresholds; the verdict follows it). Kill (not yet built): a Tangut frequency-by-stroke table, buildable now from the digitized Khara-Khoto facsimiles in Ecang Heishuicheng wenxian (the Kozlov collection of the IOM, St Petersburg; Shanghai Guji facsimile series) joined to the character data of the BabelStone Tangut database (the dataset behind Unicode Tangut) and Li Fanwen's Xia-Han zidian (1997).

Prediction clause (verbatim)

Prediction: joining per-character frequencies from digitized Khara-Khoto texts to stroke counts, the rank correlation between frequency and stroke count in Tangut will be no stronger than -0.1, while the matched computation on a pre-modern Chinese corpus yields -0.25 or stronger; secondarily, Tangut characters will average at least 25% more strokes than the Chinese characters glossing them in Li Fanwen's dictionary (primary clause: the correlation contrast at the stated thresholds; the verdict follows it).

Kill-dataset (verbatim)

Kill (not yet built): a Tangut frequency-by-stroke table, buildable now from the digitized Khara-Khoto facsimiles in Ecang Heishuicheng wenxian (the Kozlov collection of the IOM, St Petersburg; Shanghai Guji facsimile series) joined to the character data of the BabelStone Tangut database (the dataset behind Unicode Tangut) and Li Fanwen's Xia-Han zidian (1997).

Novelty / leakage triage

already answered in the literature

Andrew West (author of the BabelStone Tangut database the conjecture names as its kill-source) has published exactly this analysis: there is no relationship between frequency and stroke count for Tangut characters — normal Tangut text is uniformly composed of characters of about 12 plus-or-minus 6 strokes — whereas high-frequency Chinese characters are disproportionately simple. The primary contrast is his stated result.

• A. West, 'How Complex is Tangut?', BabelStone Blog, August 2009 (babelstone.co.uk/Blog/2009/08/how-complex-is-tangut.html)
• Li Fanwen, Xia-Han zidian (Zhongguo shehui kexue chubanshe, 1997)

Predictions

No prediction registered yet.