The 'Chinese' Theory: For and Against

[Jorge_Stolfi]

Some of the "bumpiness" features listed by Nick seem to affect a relatively small fraction of the text, and (like transcription errors, or missing fragments on clay tablets) should not be a big obstacle to decipherment.  

I don't see what on Nick's list affects a small fraction of the text, except maybe the right justified titles on a minority of folios.  The rest of the issues - repetitions and word types varying across different positions of the text - are chronic problems, so chronic it seems something has been done to the plaintext, if there is one. 

What I mean is this.  The computer transcriptions we have all contain some number of explicit 'unreadable' glyph codes ('*') and must have an unknown number of unmarked transcription errors.  The VMS itself surely has an unknown number of errors that the Author committed while writing the draft (where he meant to write one glyph but skipped, doubled, or wrote the wrong glyph) and more when the draft was copied to vellum (by him or by a distinct Scribe). 

But as long as those gaps and errors are less than 90% of all glyphs, they will only make the decipherment harder, not impossible.  

It seems pretty certain that the one-leg gallows are just alternatives to the two-leg ones, possibly in combination with other letters, that are used almost only on the first line of each paragraph.   Until we figure out exactly what the rules are, we can replace all one-leg gallows, anywhere, by '*'.  

Ditto for the m and g glyphs that seem to be extra common at the ends of lines and/or paragraphs.  Maybe they are features of the language, like "-す" and "-た" in Japanese. Maybe they are abbreviations that the Scribe was allowed to use when he got close to the right margin and there was only one or two words remaining in the paragraph.  Whatever.  Just replace every 'm' and 'g' by '*'.

Then replace all weirdos and very rare characters by '*'.

Then discard all labels, and isolated glyph tables and sequences.

Those changes will add '*' to a relatively small percentage of the text, say 10% of all words.  What remains still has hundreds of stretches of several dozen consecutive words without any '*'.   Even if the text uses some fancy encryption and/or is in some "exotic" language, those fragments should be enough to decipher it.  I cannot imagine an encryption scheme that would be viable for a text of that size and could have been conceived in that epoch, but cannot be deciphered without knowing every character of the text. 

And in fact we know that the Zipf plot and word entropy of the text are similar to those of natural languages.   Even if the text is encrypted, these results indicate that the encryption is one-to-one on the lexicon (the set of all words).  That would exclude Vigenère-type ciphers, unless the key is synchronized at each word boundary.

So, are those fragments still "bumpy"? 

Line patterns at different positions of the text are serious problems for any idea that we are only seeing a natural language.

Considering the above conjectures for one-leg gallows and m/g, I don't think they are a significant problem.

I find it interesting in that it could imply that if we have copying scribes following a layout by an "Author", they have agency in terms of how they lay out the text:  they are not obliged to match their line starts with the Author's line starts, nor their line ends with the Author's line ends by cramming in text or widening spaces between words to make the ends match.

This is almost certainly the case, even if the text was encrypted.  The parag breaks would (usually) be clear in the draft, and the Scribe would have to respect them.  However, within each parag, the line breaks in the draft would be determined by the width of the paper and the size of the Author's handwriting, and would not be significant.  The line breaks in the vellum version would be expected to be different, and neither would have any significance (except that they would count as word spaces). 

Said another way: if line breaks were significant (as in a poem, song, list, etc.) we would expect the lines to have variable lengths.  The fact that they are all of about the same length, except at the ends of paragraphs, is strong evidence that the line breaks were chosen "on the fly" wherever the writing reached the right margin.

Given how word types seem to undergo complex mutations at these positions

I question whether there are "bumps" at line breaks.  If we look only at line breaks within a parag, excluding the parag breaks, do we still see such anomalies?  Are they statistically significant?

By Top Row, I mean the first line of each paragraph but with its first word and last word omitted so as to isolate a top row effect from separate paragraph/line start effects and line end effects.

This is not sufficient, since the second word on the first line of a parag may have one-leg gallows, or may be special in some other way.  To look for line break anomalies, one should exclude the first line of the parag entirely, and look only at internal breaks.