[split] Does Voynichese evolve from A to B?

[npcompl33t]

The rules and preferences evolved during centuries, they can be dated, but not precisely because the copyist may have preserved some of the source's abbreviation style and some copies used old-fashioned ways to abbreviate words, almost extinct in their own century. However, in most cases, the scribes re-abbreviated the text entirely using their own preferences and choices: each copy of the same source material differs even when the copies are not several decades apart, which is very helpful when trying to figure out difficult abbreviations. So a composite manuscript can have several different abbreviation styles, causing a drift in statistics.

This is a great point, and also to be clear things like inconsistently applying abbreviations could indeed result in some random drift. Since none of these incorporate time, they aren't perfect comparisons, but I thought it was interesting to compare / see how random errors manifest vs abbreviations.

To answer the questions you raised about drift due to different abbreviation styles, I looked at Psalms 101. I have over 40 variants of Psalms 101, all of which use slightly different abbreviation styles. There is some randomness, but the offsets still all fall within a 90 degree angle window from the original.  Note the % of tokens abbreviated appears to have no affect on the angle of displacement, only on the magnitude or distance of the displacement. 

   

The above chart is combining different abbreviation preferences with different wording choices, so to make things clearer, i split the text into identical words before abbreviating, and words that are not identical.

The chart below shows:

1. comparisons of only identical words (before abbreviating)
   
2. comparisons of only identical words (before abbreviating) -- removing "and"
   
3. comparison of all words after expanding abbrevations
   

   

The identical abbreviation comparison above shows 2 dominant angles of offset, which appears to be a result of abbreviating the word 'and'. The middle plot removes 'and', we still see 2 dominant vectors, related to different choices around abbreviating certain words, specifically the upper group appears to prefer to keep`in` (10), `quia` (5) and `mei` (4) unshortened. 

The final plot, showing the expanded abbreviation comparison, shows how expanding abbreviations can create ambiguity. The Left and Top-most plots specifically differ primarily with changes like compeditorum versus conpeditorum.

The specific ambiguity in these expansions You are not allowed to view links. Register or Login to view.. The sign ꝯ can represent either com- or con-,so the decision to expand into compeditorum vs conpeditorum is actually an editorial choice, that can make spelling differences appear larger than they actually are.  This also explains why the clusters are split by the total % of abbreviated words. More heavily abbreviated versions offer more opportunities for an editorial convention to affect the comparison. 

The bottom cluster uses generacion- versus generation. This is a known medieval spelling convention, interchange of ti and ci before a vowel.


So yes, different abbreviation conventions can alter the direction of the offset. But not as much as the random copying errors. They are still mostly contained within a 90 degree zone, with many of the variants falling within 1 of 2 much tighter bands.