Reader Comments

Post a new comment on this article

Misunderstandings in the Montemurro & Zanette paper

Posted by gordonrugg on 22 Jun 2013 at 10:29 GMT

There are some unfortunate misunderstandings in this paper which have implications for its conclusions.

First, its finding that the Voynich Manuscript’s text is non-random is already well known. In addition to the Landini paper which they cite, there is also work by Stolfi, Perakh and others reporting similar findings, via a range of methods.

Its claim that this finding is inconsistent with the hoax hypothesis, because hoaxes would produce random text, is based on a serious misunderstanding. The whole point of the hoaxing mechanism that I described in 2004 is that it produces non-random text. This issue was discussed in detail by Schinner (2007), cited by the authors, in my 2012 paper for the Voynich centenary event in Italy, in one of my blog posts earlier this year, and in my book “Blind Spot” which was published a couple of months ago.

The authors conclude that they’ve found evidence for “genuine linguistic structure” but they do not mention the very substantial and well documented arguments against the Voynich Manuscript’s text being in an unidentified real language. Nor do they mention the equally substantial and well documented discussions about the constraints that such structures would impose on possible codes. These constraints have serious implications for the authors' conclusions, and should have been discussed in their article.

I’ll post more detail about these issues on the Search Visualizer blog site and the Hyde & Rugg blog sites.

Gordon Rugg



Competing interests declared: I'm the author of various articles describing a mechanism for producing meaningless gibberish text similar to Voynichese, one of which is cited in the Montemurro & Zanette article.

RE: Misunderstandings in the Montemurro & Zanette paper

mmontemu replied to gordonrugg on 22 Jun 2013 at 12:48 GMT

Dear Dr Rugg,

In our work we do no claim that a hoax would necessarily produce uniformly distributed random text. To suggest that would be a misunderstanding and misrepresentation of our work. What we show is that there are more layers of statistical structure in the text, which any hoax hypothesis needs to explain to be considered a likely solution. Furthermore, our study is not focused on demonstrating that the text is non-random, but on giving quantitative evidence of consistency between the distribution of words in the Voynich text and real linguistic sequences. We put forward, for the first time, putative sematic structures in the text, and show that there is consistency between the pictorial and linguistic levels.

We cite previous work and acknowledge the controversy on the issue. However, the scientific method works by contrasting any model or explanation against the empirical evidence. In that regard, and as we explicitly point out in the paper, our work together with previous evidence constrains further the scope of any hoax hypothesis. In particular, any hoax hypothesis needs to show how the proposed hoaxing mechanisms lead naturally to Zipf’s law and the further statistical features we show in our study. If the hoax hypothesis cannot do that, then its credibility can hardly be sustained.

Marcelo Montemurro

Competing interests declared: I'm one of the authors of the study referred by Dr Rugg.

RE: RE: Misunderstandings in the Montemurro & Zanette paper

gordonrugg replied to mmontemu on 23 Jun 2013 at 11:07 GMT

Dear Dr Montemurro,

With regard to your comment above, the article also makes the following mistaken claims:

Claim: These statistical features could only be produced intentionally. (“It is also worthwhile remarking that the description of such features which the hoax’s fabricator should have been familiar with in advance were unquestionably out of reach of sixteenth-century mathematics.”)

This is mistaken. Andreas Schinner pointed out in 2007 that the binomial word length distribution in Voynichese could easily be explained as an unintended side-effect of a hoaxing mechanism. I showed in detail how this could be done in my book “Blind Spot”. I also showed other statistical effects that could be produced unintentionally, in my 2012 paper for the Voynich centenary conference.
I’ve posted an article summarising these demonstrations here:

Claim: The hoax hypothesis does not explain textual features across substantial sections of text ( e.g. “the accumulated evidence about organization at different levels, limits severely the scope of the hoax hypothesis”).

This is also mistaken. My April blog article about textual structures in the Voynich Manuscript dealt with precisely this point, and shows how these structures are a likely side-effect of the table and grille hoaxing method.
This is also discussed in some detail in Blind Spot, where I discuss both accidental and deliberate simple ways of producing such regularities within a hoax.

Claim: The statistical features of Voynichese are consistent with the presence of genuine linguistic structure.

This is well recognised; however, the same statistical features are also consistent with a hoax produced using the mechanism I described. There are numerous well-documented problems with the natural language hypothesis, which are not mentioned in this article.

I’ve posted a more detailed analysis of these and other issues on the Hyde & Rugg blog site:

Gordon Rugg

Competing interests declared: As declared above - author of articles about the hoax hypothesis