Are there Really Codes in the Bible?

Having witnessed the worldwide sensation and the continuing debate over the Bible Codes, it can be difficult to know who to believe and what to think. On the one hand, the research is clearly sufficiently serious to merit debate and a slew of web sites, books and computer programs. On the other, the proposition that the literal text of the five books of Moses contains predictions about the future seems utterly absurd.

According to Orthodox doctrine, the text of the Torah was revealed by God to Moses while the Israelites were wandering in the desert, before they arrived in Israel. The Rabbis have been careful copying the text from generation to generation, so that the version Jews use today (the "Masoretic" text) is nearly identical to God's message for mankind, despite a few mistakes having crept in (minor differences have been found between Torahs in different parts of the world, and the Talmud itself alludes to there being more than one version).

From a modern academic viewpoint, a very different picture emerges. By analysing the text several different narrative strands have been discovered, each with a particular theological and political viewpoint. Many of the tales in the Torah are very similar to myths of the ancient Near East, of which records have been found predating the existence of any Jews. The text was composed and finally redacted during the Second Temple period, so that no matter how accurate the transmission was, what was being transmitted was essentially the result of human efforts.

Testing Hypotheses

The power of the Bible Codes is in arbitrating between two alternative hypotheses which are related, but not identical, to those mentioned above. Essentially, we set up two comprehensive and mutually exclusive propositions. It will be important for further discussion to note the precise form of these :

H0 : There is no supernatural phenomenon associated with the Masoretic Torah text.
H1 : There is some supernatural phenomenon associated with the Masoretic Torah text.

As with any statistical experiment, we begin by assuming the null hypothesis (H0). We then examine the evidence in the face of this assumption, and calculate a p-value, where p is the probability that such evidence would be obtained if the null hypothesis is true. The lower the value of p, the less likely it is that the evidence we have would have been obtained should H0 be true. If the value of p gets low enough (just how low is a difficult question), we begin to suspect that H0 in fact is not true, and therefore H1 must be (since one or the other must be). So a low value of p suggests the falsehood of H0.

However, p is not the probability that the null hypothesis is true, based on the evidence. For example, the probability of a girl having eight fingers on her right hand is extremely low. That does not mean that, if we find someone with eight fingers on their right hand, they must be a boy!

Focussing on the Question

Let us return to the Bible Codes. Let us imagine that, by simply reading a passage backwards, we obtain, in order, an exact location and date of every earthquake that has taken place in the last 200 years. If we assume H0, and examine this finding in that light, the p-value we obtain would be very small. If a text has no supernatural phenomenon associated with it, it would not be able to predict future geological events with such precision. (If you're thinking about various notions of transcendent prophecy / back-propagating quantum waves / time travel, etc..., all these possibilities fall under H1).

What about if we assume H1? Well, it is difficult to assign any sort of probability to the evidence under this hypothesis. If a text is somehow supernatural, we don't really quite know what to expect. However, it seems clear that it would be more likely for there to be such hidden prophetic information; exactly how much is a matter for philosophical and theological speculation. All we need realise is that, if a code with a low p-value under H0 were to be discovered, it should tilt the relative likelihood of the two hypotheses in our evaluation towards H1.

There are therefore two main questions to be answered with regards to the Bible Codes :

The rest of this article will focus on the first question, since if the answer to that is not particularly interesting, the second question is unnecessary.

Equidistant Letter Sequences

The basic medium for Bible Codes is through equidistant letter sequences. Codes are discovered by taking, say, every 42nd letter from the text starting in a certain place for a certain number of letters, and forming a new word as a result. The easiest way to do this is to write the text out in rows, with a certain number of letters in each row, and look for horizontal, vertical or diagonal sequences in this table. For example, take the first section of the Torah, from the book of Bereshit (Genesis) :

Bereshit Text

  1. When God began to create heaven and earth --
  2. the earth being unformed and void, with darkness over the surface of the deep and a wind from God sweeping over the water --
  3. God said, "Let there be light"; and there was light.
  4. God saw that the light was good, and God separated the light from the darkness.
  5. God called the light Day, and the darkness He called Night. And there was evening and there was morning, a first day.

Now, we remove spaces and punctuation, and write the same text out in rows of 50 letters :

Torah Table

Between the horizontal bars, you will see the Hebrew word Torah spelt out, at a skip distance of +50, starting at the 6th letter. Neat, eh? It is this method of decoding the plain text which all Bible Code research relies on, first discovered by Rabbi Weissmandl before there were any computers. However, such an example is not particularly relevant to the question in hand. It was not found in a formal statistical experiment and therefore cannot give us any p-value with which to support or reject a proposition.

If you think that the example above is amazing enough and you're already convined, consider this : Bill Gates the Third, the owner of Microsoft, is the richest man in the world. Convert BILLGATES to ASCII (computer code), add 3 (for the third), and you get 666. Does this mean the richest man in the world is also the most dangerous?

The Great Sages Experiment

All of the above is an introduction to the main experiment we will discuss, carried out in 1988 on the text of Bereshit (Genesis). According to the 3 individuals who performed it, Doron Witztum, Eliyahu Rips and Yoav Rosenberg, they selected a list of Rabbinic personalities from the book "Encyclopaedia of Great Men in Israel" whose dates of birth or death were specified and who had between 1.5 and 3 columns of text written about them. The original set they selected, of Rabbis with over 3 columns written about them, yielded a significant result - they decided to change this in order to demonstrate that the experiment had not been designed around the date in order to produce a strong result.

For each rabbi, they took several different traditional spellings or appelations of the name. For each rabbi, they also took the date of birth or death, expressed in a few different forms. Having prepared this list, they then proceeded to search for the names and dates, defining a measure of proximity between the two. The details of the experiment can be found in the article referenced below. What is important is that, based on a measure of proximity between the names and dates, they obtained a highly-significant p-value of 0.000016 by comparing the search on the correct text of Bereshit with that on 999,999 jumbled versions of Bereshit. This means that, based on the experiment's methodology, the chance of the Rabbis' names being so well encoded in Genesis under the original H0 (that the text of the Torah has no supernatural phenomenon associated with it) is 1 in 62,500.

This was an exceptionally strong result, more than enough to justify sensible rejection of the H0 hypothesis. However, research was conducted by Maya Bar-Hillel, Dror Bar-Natan and Brendan McKay into the actual names and dates used, and it soon became clear that several things were slightly amiss. The date forms used were not entirely consistent and the appelations chosen were not exhaustive. Nonetheless, this need not necessarily invalidate the experiment; if the list was selected a priori, before the experiment was carried out, then the result remains significant.

The Selection Process

So the question on the original experiment becomes : how was the list of names and dates selected? The answer lies with Professor Havlin, who supplied the original data to the experimenters. In a 13-page article written in October 1996, he describes the process by which he selected the names and dates, where by his own admission a major factor was "personal judgement". In other words, the method he applied cannot easily be specified rigorously, and if we were to select the list again ourselves, we might come up with something very different.

In the absence of rigorous selection criteria, we still need not necessarily reject the validity of the experiment. If the list was chosen a priori (before the experiment was performed), the result can still stand. Unfortunately, when it comes to asking whether or not the list was chosen independently of the experimenters, we have to rely on peoples' word. Witztum, Rips and Rosenberg claim that Havlin prepared the list for them, and then they independently performed the search based on what he gave them.

Can we trust them? I don't know. But when it comes to such a momentous discovery, one has to be sceptical. According to Alex Lubotsky, Member of Knesset (Israeli Parliament) and Professor of Mathematics, "Psychology literature is full of cases where a researcher's expectations have influenced his judgement and/or consideration." Bar-Hillel, Bar-Natan and McKay decided to test whether it was possible that the list had been 'cooked' (modified to get the best results), and set about producing a new list, openly admitting they were cooking this too. They then demonstrated that the new list, when used in a search on the Hebrew translation of War and Peace, obtained results even more significant than Havlin's list did on Genesis. So it is definitely possible to cook a list in order to obtain the results the experimenters did, but the question remains : did they?

One additional piece of information came to light recently. A cassette was discovered that recorded a lecture given in 1985 in Russian regarding the Torah Codes (three years before the official experiment was carried out). This exposes the fact that Rips had been experimenting with rabbis' names in various appelations years before the official experiment came out. However, it was claimed that the official experiment was performed entirely a priori, without any previous investigations of proximity between rabbis' names and dates. The uncovering of this irreconcilable inconsistency between the espoused and actual chronology of the experiment directly affects our trust in the claim that the list was prepared a priori, before any experiment was carried out. If they had been playing with the data for years, surely they had time enough to cook it?

The Cities Experiment

In at attempt to find out once and for all, another experiment was subsequently carried out by Harold Gans, a senior cryptologic mathematician at the National Security Agency, who got involved with the codes via Aish HaTorah (a Jewish outreach organisation who use the codes in their Discovery seminars). Gans took all 66 rabbis from the two lists Rips et al had prepared. Then, instead of using dates of birth and/or death, he searched for these names against their cities of birth or death from the Encyclopaedia of Great Men in Israel and Encyclopaedia Hebraica.

The p-value he obtained was less than 0.000005, i.e. the chance of the rabbis' names and cities being so well encoded in Genesis under the original H0 (that the text of the Torah has no supernatural phenomenon associated with it) is 1 in more than 200,000. This would seem to both relieve the original experiment from accusations of cooking, and strengthen the evidence against hypothesis H0. A victory for H1?

Not so fast. It was subsequently alleged that the data Gans had used for his cities experiment was also full of spelling errors and other flaws. Around the end of April 1999, Gans made the following statement :

As is known to many, I am in the process of redoing and confirming the cities experiment that I had done several years ago. Several individuals informed me that the list of cities, as provided to me by Zvi Inbal, had many mistakes in them that were designed to make the results appear artificially significant. I take such claims very seriously and decided to conduct a thorough investigation of every aspect of the city selections and spellings. That investigation is extremely detailed and is not yet complete. As a result, I will not make any public statement of the outcome of that investigation until my study is completed. This unwillingness to speculate on an outcome of an investigation while it is still ongoing has prompted some people to interpret that as evidence that I am no longer convinced that the Torah codes phenomenon, as detailed in WRR, is a real phenomenon or that I no longer believe that the conclusions drawn from my original cities experiment are correct.

Let me then state in absolute terms that this is not true. To date, I have not uncovered a single fact or even a hint that the list of cities that I was provided was manipulated in an attempt to make the results of the experiment appear significant when, in fact, they are not significant. I have not uncovered a single fact that causes me to doubt that the conclusions drawn from the original cities experiment were accurate. I have not seen any argument advanced by anybody which convinces me that either the WRR experiment or the cities experiment is not valid and truly statistically extremely significant.

Nevertheless, so as to be completely thorough and honest, I will not now claim that I have verified the cities experiment : there are still a few things left to check. When my investigation will be completed, the results and all the details will be made public in an appropriate way, regardless of whether the results confirm the conclusions of the original cities experiment or not.

No further developments have been made public since this statement. However, it seems that the data for the cities experiment was actually supplied by Zvi Inbal, who lectures on the codes for an Orthodox outreach organisation called Arachim (which is connected to Aish HaTorah). He clearly might have a vested interest in cooking data.

The Periodic Table Experiment

Having seen and learnt all of the above, I decided to conduct an experiment of my own, and did so in collaboration with Brendan McKay, who has been a major character in debunking the Bible Code. To avoid being dogged by difficulties over questionable data, we agreed on a simple experiment based on the Periodic Table of Chemical Elements. All details of the experiment were discussed before it was carried out, and anyone is welcome to an archive of all the email correspondence as evidence for this.

We considered the possibility that there would be evidence in the text of Genesis of some supernatural phenomenon, which is 'meant' to be discovered, and made three inferences from this :

  1. the evidence contained would be relatively simple (c.f. Ockham's Razor).
  2. the evidence would pertain to some universal knowledge.
  3. the evidence would be in the same language as the text.
Therefore, if some phenomenon were to be discovered, we would be able to present it to the world in a form which they could appreciate and understand, without resorting to obscure historical and religious sources and methods. Although evidence need not necessarily be of this form, a more obscure phenomenon might be subject to the criticism that has been levelled at other parties for their previous work. A list of the Chemical Elements, in Hebrew, was obtained from the General Hebrew Encyclopaedia, 1989 Bulletin, page 462 (in Hebrew : enciklopedia ha'ivrit klalit, yedion 1989, daf 462).

This contained a table of every element with atomic numbers 1 to 107, and 109 (I do not know why 108 was omitted). For each element, a Hebrew spelling was listed, plus the atomic symbol, atomic number, atomic mass (of most common form), and year of discovery (where known).

A cursory analysis of the Hebrew element names revealed that, for the majority, the name was a straight transliteration from the known English name, and most of these were at least 6 letters in length. We agreed to perform our experiment on those elements whose names were specifically Hebrew in origin, because these were generally shorter (and therefore had more chance of being discovered as an equidistant letter sequence), and this is more in keeping with principle 3 above.

The experiment was designed to test whether the text of Genesis might indicate knowledge of the atomic number of those elements for which there is an original Hebrew name. We suggested that such knowledge might be indicated by inserting these elements' names into the Genesis text at skip distances equal to their atomic number. We would compare the number of occurrences at this distance with the number of occurrences at all distances 1 to 82 (82 is the highest atomic number of all the elements tested).

As a result of the test, no evidence was found to support the existence of knowledge of the atomic numbers of the chemical elements for which there were Hebrew names. Furthermore, we found no evidence that these element names had been encoded at any short skip distance within the Genesis text.


The issue by no means closed, but for now we can reach the following conclusion :

Based on the experiments performed so far and in consideration of their flaws, there is no significant statistical evidence for the existence of codes in the Masoretic Torah text which would indicate some supernatural phenomenon associated with it.


Many other minor references to email messages, letters and articles were used.
© Mayim, 2005.