This seems to be the right place to continue the discussion on the auto-copy hypothesis.
(This post has a general introduction first, and then (further down) some new statistics that argue against the auto-copy hypothesis.)
My thoughts are completely in line with those of Emma, stated near the start of this thread namely that there is a fundamental difference between:
- observing that similar words tend to appear near each other
- saying that the Voynich MS was created by arbitrarily writing words similar to previous words
The result does not provide evidence of the intent of the 'actor'. We cannot judge just from the statistics what was the process that generated these statistics.
It is understood that the evidence presented by Timm (and Schinner) lies with the computer algorithm for generating Voynich look-alike text, but the details of this algorithm have not been shown here, so there remain two important questions:
- how complicated is this algorithm?
- does it include any anachronistic elements?
One way to look at this is to observe the appearance of the most frequent word daiin (daiin)
Obviously, one word has to be the most frequent, so it makes no difference which particular word it is.
However, now there are two conflicting hypotheses:
1) This word is the most frequent because it is a representation of some very frequent element (not further specified) of an encoded source text
2) This word is the most frequent because it most naturally appears by the auto-copy hypothesis, i.e. by making small modifications of earlier words.
This is something that can be tested.
I have looked at a transliteration of each individual page of the MS, and compared the first occurrence of 'daiin' with the preceding text. (This is something I did several months ago, but I had not yet reported on it). The complete list is included
in a spoiler tag below.
The page is indicated by a file name starting with a two-character code, that some people may recognise, many not, but that is not too relevant.
Then there is a number, which indicates the appearance of the word on this page. If it says 20, it means that the first occurrence of daiin is the 20-th word on this page.
The last two columns show the Levenshtein distance to the most similar previous word on this page, and what this word was. Note that all distances were computed based on Eva.
Let me first include a graph of the frequency by Levenshtein distance.
daiin is a 5-character word (in Eva) and in about 1/3 of the cases, the Levenshtein distance is 3 or higher, meaning that the word is mostly different from the one it was supposed to be copied from.
I can only invite people to look through the detailed list.
When I do this, I can only conclude that in a large number of cases, it is not reasonable that the word daiin, the most frequent word in the MS, was copied from a very different word earlier on the page, but it simply appeared because it is a frequent word for another reason.
(This post has a general introduction first, and then (further down) some new statistics that argue against the auto-copy hypothesis.)
My thoughts are completely in line with those of Emma, stated near the start of this thread namely that there is a fundamental difference between:
- observing that similar words tend to appear near each other
- saying that the Voynich MS was created by arbitrarily writing words similar to previous words
The result does not provide evidence of the intent of the 'actor'. We cannot judge just from the statistics what was the process that generated these statistics.
It is understood that the evidence presented by Timm (and Schinner) lies with the computer algorithm for generating Voynich look-alike text, but the details of this algorithm have not been shown here, so there remain two important questions:
- how complicated is this algorithm?
- does it include any anachronistic elements?
One way to look at this is to observe the appearance of the most frequent word daiin (daiin)
Obviously, one word has to be the most frequent, so it makes no difference which particular word it is.
However, now there are two conflicting hypotheses:
1) This word is the most frequent because it is a representation of some very frequent element (not further specified) of an encoded source text
2) This word is the most frequent because it most naturally appears by the auto-copy hypothesis, i.e. by making small modifications of earlier words.
This is something that can be tested.
I have looked at a transliteration of each individual page of the MS, and compared the first occurrence of 'daiin' with the preceding text. (This is something I did several months ago, but I had not yet reported on it). The complete list is included
in a spoiler tag below.
The page is indicated by a file name starting with a two-character code, that some people may recognise, many not, but that is not too relevant.
Then there is a number, which indicates the appearance of the word on this page. If it says 20, it means that the first occurrence of daiin is the 20-th word on this page.
The last two columns show the Levenshtein distance to the most similar previous word on this page, and what this word was. Note that all distances were computed based on Eva.
Let me first include a graph of the frequency by Levenshtein distance.
daiin is a 5-character word (in Eva) and in about 1/3 of the cases, the Levenshtein distance is 3 or higher, meaning that the word is mostly different from the one it was supposed to be copied from.
I can only invite people to look through the detailed list.
When I do this, I can only conclude that in a large number of cases, it is not reasonable that the word daiin, the most frequent word in the MS, was copied from a very different word earlier on the page, but it simply appeared because it is a frequent word for another reason.
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)