Regaining the lost order

[geoffreycaveney]

Here are a few examples of what the original text could have looked like, before its encryption by shifting the last 2 characters or units (using Cuva, not EVA) of each vord forward and attaching them to the beginning of the following vord to create the actual Voynich ms text. In other words, I created the text below by shifting the first 2 characters or units of each vord backward and attaching them to the end of the preceding vord. Importantly, I also deleted the first 2 characters or units of each line, which could have been null filler characters according to this idea. 

f1v, 1st para:
[sycha daiinol ol tcheycha rcfha ram..]
[eeaycha ror och ydch olk odyok odarcho dy..]
[ckhy ckho ckhy shy dk sheeycthy ko tchodyda l..]
[lcho keodair da mso cheycho kody..]

f20r, 1st para:
[chodycho pyche eyqo tcholqo toeeydch or choiin..]
[deycthe ycho tolod aiirqo tchyctho dycho dchy..]
[teeycho cho daiinsho qo chyche ytch eodalda ral..]
[olol te eyot olchey..]

f77r, last para:
[lshe dyqo eedyqo kaiinchcph eyqo llt aiinshe dyqo l..]
[aiinche eyshe ckheylsh eeyqy kaiinshe edylaiin ..]
[ol ee edyok aiinshe eolqo taiinsho dyqo ty..]
[sh qo lcheyqo tedyqo talrain chl rar ol..]
[keedyqo tedyqo keedyqo keedyqo keeyraiin al ..]
[eeol che edyqo taiinqo teedyqo tedyraiin ..]
[eyqe pchedyqo lche edyqo kearche eylo lydy ..]
[kaiinche slch earda lchr sl s ain ol raiin lo d..]
[chl chpsheeyta lche olda mar oteydaiin ..]
[eeyda lsaiir olsa lda lotain da ryda lo..]
[darol oky..]

I chose the last passage because it contains the famously repetitive [qokeedy qotedy qokeedy qokeedy qokeey ...] line. It is still repetitive, but now as [keedyqo tedyqo keedyqo keedyqo ...]. 
Note: I treated [ee] as 2 separate units for this exercise. Perhaps they should be treated as a single unit.
I deliberately avoided imposing my own ideas (well, that I took from Koen and ran with) about verbose cipher units such as [od], [ok], [ot] on this analysis. I stuck to the traditional interpretation of [o] as its own single unit everywhere.

Observations:

We find a much greater variety of vord-initial and vord-final characters and sequences in the text above! This looks much more like a European natural language pattern, at least in this respect.

However, this transformation now produces a different, less blatantly obvious but nevertheless still present, curious restriction on positional character/unit occurrence: The restriction has been shifted to the antepenultimate (third to last) character of each vord! If a vord indeed still has 3 or more characters/units after this transformation, the antepenultimate character is now in the great majority of vords either [y], [n], [l], or [r]. I have never heard of anyone doing a study of the letter distribution of the antepenultimate letters of words in any natural language, but I doubt it could be so restricted as this. For example, German "bleiben" would be unlikely to be allowed to be represented with such a restriction on antepenultimate letters to a small set of very frequent letters.

Still, a big advantage of this approach is that [aiin], for example, can now occur at any position in the vord: initial, second, medial, or final. In this respect it now behaves much more like an actual unit in an actual natural language. Likewise, [e], [o], and [a] can now occur at any position in the vord: initial or final as well as second or other medial positions. Again this seems much more like actual natural language.