[Index page]   [Part One]

The Chanukah experiment of Doron Witztum

This is the second of three parts of the long reply to the article "Did they really find codes in War and Peace" by Doron Witztum.

Text which is indented and italicised is quoted from that article, except that the Hebrew letters have been transliterated according to the Michigan-Clairmont scheme.


The "Case Study" of Doron Witztum concerning the Holiday of Chanukah is demonstrated to have been constructed by fraudulent means.

Selecting the central word.

Witztum's "experiment" consists of one central word surrounded by some number of related expressions.

To understand what is going on here, realise that Witztum has looked at this data before. There is a whole chapter on it in his book [Wit]. All of the words he is going to mention are there, even all the distances. The only thing that is not there is this "experiment" he is now cooking up. What appears in his book instead is a much larger "experiment" that includes these words amongst many others.

Apparently Witztum the Scientist thinks it is permissible to make an "a priori experiment" using known data.

Of course, Witztum might claim that the experiment he is presenting now was done by him over 7 years ago, before the larger experiment in his 1989 book. That possibility is easily dismissed by just reading the book. There is no concept of a preliminary experiment to select one title word, as he is about to do here. Mishna Brurah is absent, as is the 1/700,000 claim. In fact, his present description refers to a 1996 paper for both the general form of the experiment and the method of analysis. In conclusion, it is positively established that he is simply putting together a few things he previously found to work during a much more extensive search years ago.

From Witztum's explanation so far we can infer that he is going to use the definite article, as otherwise he wouldn't have brought it up. That gives him four possibilities to choose from. From the data in his book he already knows which one works, namely HXNWKH, so somehow he has to eliminate the others.

So, what can he do? Sometimes (examples in [Wit] and [WRR3]) he just says "it can be written X or Y, we will write it X". Other times he applies some retrospectively-designed "a priori rules". He can't really do that here, because he knows that the spelling rules given in all three of his papers that are mentioned above select the short spellings. He wants the long spelling, so he has to do something else. He is usually pretty good at these games, but this time his effort is very disappointing:

Oh dear, a Halachic source used as a spelling authority. Most people use a dictionary. Of course, he is not really using the Mishna Brurah for the spelling. Almost any book written without nikud (vowel marks) would have done as well. The edition of Mishna Brurah printed with nikud uses the short spelling XNKH, but this time he has to suppress his usual preference for vowel-reduced spelling and hope that nobody notices.

It is illuminating to note that in a recent document [Wit2] (Assertion 4, Point C), Witztum hypocritically argues against using Halachic works as sources of spelling.

In any case, he is not interested in correctness here. Rather, he is searching for an excuse to use HXNWKH that he can try to fool us into believing is objective. So he looks around for a place with a few good words that appear in the same sentence. There are thousands of sentences to choose from, so perseverance is the only requirement. In the end he settles for a few words that are not especially good at all, but compensates by inventing a trick to make them seem good. Let's watch.

Incredible it is, but he really wrote that. XNW does not appear very close to those words by the usual measures. Witztum has to invent a new method to make it work. The tiny word XNW has 20 minimal ELSs, and he marks just one of them. Note what he is asking us to believe, that in advance of doing any of the calculations or looking at the data he knew that this particular technique was the correct one at this point. If I insulted the intelligence of my students the way he is insulting ours, they would walk from the classroom and leave me lecturing to empty chairs.

I cannot reproduce the above calculations without more details, but since it is so obviously concocted it hardly matters. Witztum's excuse for using the article H is that HXNWKH does better than XNWKH with respect to a single one of the 20 minimal ELSs of a single 3-letter word! Pathetic.

The above excuse for selecting the spelling HXNWKH is extremely weak even by Witztum's standards. In fact, he could do lots better if he wanted to. It seems that his disdain for the reader is so great that he just can't be bothered. Or maybe he spends too much time with disciples who believe every word he says, no matter how ridiculous it is.

Selecting the related expressions.

Now watch what Witztum the Probability Expert is going to do. After one experiment with HXNWKH he is going to do another. It is a mathematical fact that the result of one experiment with HXNWKH biases the next experiment with HXNWKH. In this case the nature of the bias is very complicated, but Doron Witztum, true to form, doesn't even indicate there is an issue to consider. Does he not know, or not care? Such carelessness abounds in his work and disqualifies him from being taken seriously.

So the cooking of the experiment continues. The subject of Judah the Maccabee "came first to mind" in the following way: from the numbers in his book he already knows the closeness to HXNWKH of a considerable number of Chanukah-related words. Almost 30 are presented, and it is obvious he tested many more. Amongst the words doing the best are "Maccabee", "The Maccabee" and "Hashmonai". There are others, but those are a good choice for his present purposes because he can put them all under the umbrella "Appellations of Judah the Maccabee".

Armed with those three star words, he now has to make it look objective. Adding long appellations will look good and be safe - they won't appear as ELS's. (Witztum believes in a deity who is only capable of encoding short words). There is a serious problem though: the appellation YHWDH is a bad one. It gets a really awful distance of 108/125 and threatens to wreck his result. He has to do something about it, but what? In his book he invented a trick for turning 108/125 into 6/125, but it will look bad to present such an obvious fiddle here. Better to get rid of the word altogether. He could forget to mention it, but that would be a bit too obvious, so he has to put it in and then take it out again. Somehow. Let's watch:

What a let-down. Surely he could have thought of a better excuse than that! To see how genuine it is, look at his book to see YHWDH marked in Chanukah pictures despite it really referring to a different YHWDH. Moreover, he uses far less specific words plenty of times, for example XTWM (sealed).

Incidentally, the removal of HX$MWN)Y is mathematically illegal. Since permutations of its letters may have ELSs, the original spelling has now been artificially distinguished from them. Errors like this are abundant in Doron Witztum's work.

There is another published experiment of Witztum that involves Judah, namely the "Sons of Mattityahu" experiment in [WRR3]. The header phrase is BN LMTTYH (son to Matitya). Not satisfied with the names of Judah and his brothers, Witztum added three descriptions of Mattityahu himself (as though he was his own son!). Those descriptions are: BN YWXNN (son of Yochanan), KHN GDWL and KHN GDL (high priest, in two different spellings). Appearing immediately after those descriptions in the source they are taken from, the 'Al Hanisim' prayer, appears the much better known and specific description "Hashmonai". Why didn't he use that? Let's see:

Experiment:          HXNWKH           BN LMTTYH

YHWDH   distance:    0.8640            0.0165
       word used:      no                yes

X$MWN)Y distance:    0.0769            0.6154
       word used:      yes               no

In at least three of the four cases, the word could have been put in or left out. Somehow, without peeking, Doron Witztum made all three choices in his favour. As always.

More criticism of the "Sons of Matitya" experiment can be found in our paper on Witztum's Auschwitz experiment.

And all without looking at his own book!

It is interesting that he bothered to present an argument for using HXNWKH at all. In his book he presented these arguments:

He checks both, all right. He also tried both including the W and excluding it, as the pictures in his book attest.

The Results

Note that all three of the above articles explicitly state that the method of distance calculation does not work for words of less than five letters. All three of them reject four-letter words as too short. How many letters does "MKBY" have?

                         StatSci experiment    This experiment

Four-letter words good?         no                   yes
Four-letter words used?         no                   yes
So, we see three things:
  1. Witztum lied about calculating the 1/700,000 by the method given in the sources he quoted.
  2. Witztum lied about his method not working for four-letter words.
  3. Witztum has an uncanny knack of making "a priori" decisions in his own favour.

This is hilariously funny. In another document we show a War and Peace experiment that produces a "probability" of 1/4,000,000 using one central word and 7 related words. If Mr Witztum challenges any of our 7 related words, we can reply "The critical reader is at liberty to double the probability". We can even answer in that way 7 times, as doubling 1/4,000,000 even 7 times leaves us with a perfectly healthy result of 1/31,000. By Doron Witztum's logic, we have a 1/31,000 probability without the need to justify anything!


And so ends the exceedingly instructive Chanukah experiment of Doron Witztum. May his credibility rest in peace.


Astounding Discoveries in War and Peace, http://cs.anu.edu.au/~bdm/dilugim/chanukah.html (1997).
D. Witztum, HaMeimad HaNosaf (The Additional Dimension), self-published, 1989.
D. Witztum, A Retutation Refuted, part 2. Available as a PDF file.
D. Witztum, E. Rips and Y. Rosenberg, Equidistant Letter Sequences in the Book of Genesis, Statistical Science, 9 (1994) 429-438.
ibid, Equidistant Letter Sequences in the Book of Genesis, II. The relation to the text, preprint ca 1995.
ibid, Hidden Codes in Equidistant Letter Sequences in the Book of Genesis, The Statistical Significance of the Phenomenon, Hebrew preprint accompanying a lecture given by E. Rips to the Israeli Academy of Sciences, 1996.

This page was written by Brendan McKay, who wishes to gratefully acknowledge the help and support he received from a considerable number of his friends.

On to the third part of this article
Back to the title page
Back to the Torah Codes page