I compared it to a text in which I had distributed the Voynish glyphs at random.
I measure exactly what is highlighted in the diagrams: grid columns, atoms, left-aligned, without spaces; a column is considered conspicuous if a glyph occupies ≥40 per cent of ≥4 cells.
The mixed text has the same layout — the same number of pages, the same number of lines, the same line lengths, the same glyph frequencies. Only the order has been disrupted. It nevertheless produces 175 conspicuous columns; this is the baseline level produced by the frequency distribution alone. But the Voynich text has 497!
So: the effect seen in the graphs is genuine. It is not what a randomly shuffled alphabet would produce on its own. At the start of a line, it is eleven times as strong as chance; in the rest of the line, it is a good twice as strong.
Note: the figures were calculated using Claude Faible and I have cross-checked it using Chatgpt Sol
(A Bavarian text adapted to the specifications yields 176 such columns (in a first quick test)).
One reason is likely to be the highly structured slot structure of the VMS, which does not exist in any language in this form – but I cannot yet say whether that is the only reason.
I measure exactly what is highlighted in the diagrams: grid columns, atoms, left-aligned, without spaces; a column is considered conspicuous if a glyph occupies ≥40 per cent of ≥4 cells.
The mixed text has the same layout — the same number of pages, the same number of lines, the same line lengths, the same glyph frequencies. Only the order has been disrupted. It nevertheless produces 175 conspicuous columns; this is the baseline level produced by the frequency distribution alone. But the Voynich text has 497!
So: the effect seen in the graphs is genuine. It is not what a randomly shuffled alphabet would produce on its own. At the start of a line, it is eleven times as strong as chance; in the rest of the line, it is a good twice as strong.
Note: the figures were calculated using Claude Faible and I have cross-checked it using Chatgpt Sol
(A Bavarian text adapted to the specifications yields 176 such columns (in a first quick test)).
One reason is likely to be the highly structured slot structure of the VMS, which does not exist in any language in this form – but I cannot yet say whether that is the only reason.