How to decode?
Luckily, this hypothesis can be put to the test, and the Mark II should be considerably more easy to decipher than Mark I — namely, because the main advantage of the Mark I, the blurring of plaintext letter boundaries, gets lost.
If every ciphertext word represents exactly two plaintext characters, there should only be some 500 different words at all. (Of course, this is not really the case. But it’s conceivable that transcription errors were introduced and, more importantly, that rare characters in the plaintext like numbers would also require rare strokes, and thus rare Voynich letters.) This should be manageable by a simple frequency analysis. The pure number crunching can of course be supplemented by educated guesses about the meanings of Robert’s odd and even groups. For example, it’s highly probable that “I” or “l” would only be encoded by single-letter groups, because they require only one pen stroke in most typefaces. “g” or “k” are probably more complex and need longer groups.
And here is where I get my suspicions. Obviously, Robert was also on the same track as me. And he seems apt enough that he could have done the final steps himself, even without the stroke theory in the back: The simple assumption that two different code groups alternate for an expanded substitution cipher should be sufficient to break this code, even if you don’t know that you have to interpret the letters as pen strokes.
By the simple fact that we never heard from Robert again I must assume that he tried and didn’t achieve at readable results. Which would mean that I am on the wrong track.
Read on here about the latest experimental results regarding the Stroke theory!