The Bible Code: A Scientific Experiment

Someone created a science thread on this topic and it got deleted for T&C problems, along with my reply, so I've just edited my reply somewhat to make it a thread instead of a reply. This isn't the kind of thread I normally make but I was a little surprised to see some people still take this seriously and think it's scientific.

I searched and found numerous bible code threads talking about predictions and prophecies, etc. (see link at the end of this post for one). That is not the topic of this thread which is why it's posted in the science forum instead of predictions and prophecies.

Eliyahu Rips

became known to the general public following his coauthoring a paper on what is popularly known as Bible code, the supposed coded messaging in the Hebrew text of the Torah....

In 1994, Rips, together with Doron Witztum and Yoav Rosenberg, published a seminal article in the journal Statistical Science, "Equidistant Letter Sequences in the Book of Genesis", which claimed the discovery of encoded messages in the Hebrew text of Genesis. This, in turn, was the inspiration for the 1997 book The Bible Code by journalist Michael Drosnin. While Rips originally claimed that he agreed with Drosnin's findings, he later distanced himself from his interpretations.[3] Since Drosnin's book, Bible codes have been a subject of controversy, with the claims being criticized by Brendan McKay and others.

Reading the wikipedia article might give one this is a scientifically controversial topic, because that article fails to mention the one scientific experiment that was performed with participation from both sides of the "debate".

The purpose of this thread is primarily to make people aware of that experiment, to put the bible code issue in a scientific perspective. The source here is a BBC documentary for which we have the transcript and the youtube video:

The Bible Code

DR LYNDA WHITE: Rips original experiment gave us odds of 62,500 to 1. When we did our experiment we had two lists, we had the sceptics list and we had the Rips list. The odds for the sceptics list turned out to be 2 to 1 against, much smaller than Rips’ original odds. In very simple terms this represents a chance of about 1 in 3 that the clustering is accidental. So no evidence at all for hidden codes. When it came to the Rips’ list we get odds of 2 to 3 which again putting it simple represents a chance of just over 50% that the clustering is accidental. So when we look at these two together there is absolutely no evidence at all, absolutely none that there are hidden codes here.

NARRATOR: One experiment is never accepted as proof in science, however the only Bible code experiment involving both sides of the debate had failed to show any evidence of their existence.

So unless someone can post additional scientific experiments which have been peer-reviewed, this scientific attempt to validate the bible code showed that there was nothing to it. In fairness I should add that Rips who agreed to the experiment beforehand objected to it afterward when it didn't support his claims, citing errors, however his critics responded that even after accounting for these error there was still no statistical validity to the claim.

As people are fond of saying, you can't prove a negative. so this doesn't conclusively demonstrate there is no bible code. But what it does demonstrate is that using a rigorous statistical approach, the claims of Rips failed to materialize in the scientific experiment to see if there is really a bible code.

His critics claim Rips fails to appreciate all the statistical nuances that change the odds, like all the different spellings a name can have for example, that may have artificially inflated his results when he didn't accurately account for that statistically. One of his critics, McKay managed to find "codes" in Moby Dick:

NARRATOR: Brendan McKay is an expert in advance probability theory. He’s been investigating mathematical mysteries for almost 30 years. He decided to take up Drosnin’s challenge. He bought a copy of the 150 year old American novel Moby Dick and using a search programme similar to Rips began to comb it for hidden messages. The results were at first glance remarkable.

But he explains they aren't really codes at all and the reason we can find coincidences in large texts:

BRENDAN MCKAY: What’s really going on of course is that the computer can search in so many ways, millions or hundreds of millions of ways that even though any particular pattern is unlikely, it’s going to find something sooner or later.

If anyone prefers to see the video instead of the transcript, here is the documentary:


I'd like to ask that any replies be limited to the science involved, and not any specific predictions or prophecies which are already addressed in other threads such as this (which failed to materialize in 2006 by the way):

Bible Code predictions

Cool, but actually....when related topics are superimposed in the same location, like 4 or 6 related topics.....that's interesting.
also, moby dick and other books than the Bible may have supernatural help....don't discount supernatural design....it's a miracle world that we have here for sure, babe.......then........there's the Rabin assasination (he was forewarned.....actually forewarned from Bible codes.)

It's only statistically controversial among non-statisticians. There are lots of people who claim to be statisticians who really aren't educated in the field. Even though these individuals may be (but generally aren't) adequate in their usual application, as soon as they get outside of those specific applications where they follow well-defined procedures they tend to make huge errors in assumptions and methodology choices.

If you have a sufficiently short piece to find and millions of combinations to search through, you will get a match eventually. A 5 character key to find, with a 22 letter alphabet (like Hebrew), will occur at random on average with about 5,153,632 independent searches if the letters are random generated.

That sounds like a lot until you realize there are skips of 1, 2, 3, 4 and 5, etc characters and checking is done in at least 8 different directions (+ and X). If you divide 5153632 by 40 (up to skips of 5 letters), you will find the key word about 1 in 130,000 checks if the letters were distributed at random. This is a little off because one letter is in common in all 8 directional checks. The frequency of finding matches of the key word can also be increased by sliding the word back and forth along these 8 directions.

Letters however are not distributed randomly in texts, so this will vary depending on the frequency of the letters in the language and how the text is written. Numerically calculating frequencies given actual letter frequencies in the text is possible but would be much more tedious that an experiment. I'd go for the experiment first!

I'm not sure what McKay did, but I would use two lists: one with key words chosen to be meaningful in a prophetic context and the other not meaningful. Examples might be: "world war" (meaningful) and "ice cream" (not meaningful). The letter combinations should be roughly equivalently likely given the frequency of the letters in the language. Then search for all matches in both lists and compare the hit and miss rates.

We can use the Fisher's exact test to determine if there is a higher (or lower) chance to find matches with and without prophetic word selections.

I'd probably want to work with a linguist to ensure that I didn't overlook something that I could include in the structure of the experiment which would induce a bias towards one or the other outcome. Similar outcomes (aka not significant) would, of course, imply there was no evidence that special messages existed in the text unless you want to assume the text was written to provide that outcome because it was somehow known that this test would occur when the text was written and a lack of significance was desired.

If you tested many thousands of key words in this manner from each group, you might find significance just due to te sheer volume ot patterns examined. To prevent this, the length of the two lists should be short -- maybe a few hundred entries at most in each list.

More than not being able to prove non-existence of a Bible Code, statistical methods also cannot prove existence. All the statistical methods do is to provide a measures of certainties of the hypothesis, essentially just an indication one should or should not believe the two lists were equivalent.

McKay probably did something similar to the above. I assumed, the first time I heard about the Bible Code and how the searches are performed, that it was unlikely to be real. The key wrods were too short and the Bible has a huge number of letters. Given a reasonable statistical test, an outcome of not supporting a difference is only statistically controversial among non-statisticians.

Originally posted by GBP/JPY
Cool, but actually....when related topics are superimposed in the same location, like 4 or 6 related topics.....that's interesting.
also, moby dick and other books than the Bible may have supernatural help

Using the scientific method, you could make the hypothesis that the "codes" found in Moby Dick exceed expected statistical probability, and test your hypothesis. McKay says they don't, and he's an expert on statistical probability.

....don't discount supernatural design....it's a miracle world that we have here for sure

As I said this is a thread about the science. Science can't be used to prove or disprove things which are beyond the scope of science, like the supernatural.

What science has demonstrated in this case is that the "bible codes" were not so miraculous. Finding the correlations was about as likely as getting a "heads" on a coin toss, which hopefully most would agree is no miracle.

Originally posted by BayesLike
It's only statistically controversial among non-statisticians. There are lots of people who claim to be statisticians who really aren't educated in the field. Even though these individuals may be (but generally aren't) adequate in their usual application, as soon as they get outside of those specific applications where they follow well-defined procedures they tend to make huge errors in assumptions and methodology choices.

Good comments, thanks.

That's apparently what happened in the case of Rips. This is from the BBC link in the OP:

NARRATOR: But as McKay examined Drosnin’s findings, he thought he saw a flaw. Rips is a brilliant mathematician, but he is not an expert in statistics. It’s a very different science. To get a figure recognised by statisticians you need to follow a rigorous statistical procedure.

BRENDAN McKAY: In order to actually give a statistical significance to a Bible code type finding it’s essential that you specify in advance what it is that you have to find in order to be successful and you specify the way in which this experiment’s results will be analysed and without such a formal experiment the numbers are simply meaningless, they’ve got no statistical value.

One big difference between the experiment McKay was involved in and Rips earlier work was carefully agreeing on the specific parameters in advance.

My understanding is that Rips found things and tried to assign statistical probabilities after finding them, and this resulted in a big error because he didn't account for all the other variants.

This can get a little complicated, even for a statistician, so there are lots of ways to get the statistics wrong. Even the reviewers of Rip's earlier paper missed the problems at first. I think his subsequent paper on this topic remains unpublished.

This is the first I've heard of codes within the bible but it wouldn't surprise me if there is just by the strange way its structured and written. You also have to remember the current bible has been revised numerous times, and the code may not be present in an original edition. Then you have to wonder what is the original edition? much of the bible contains passages from The Book of Enoch which was written 300 BC and some of it dating back to the Egyptians.

Originally posted by Eonnn
Then you have to wonder what is the original edition?

Good point. ATS member Sigismundus wrote a detailed reply to the thread that got deleted about just that. I think he said there were at least 6 versions.

We'd have better luck finding the original version of Moby Dick.

But I can't tell if you understand that the claimed code was not confirmed, in either Moby Dick or the Bible. In fact Moby Dick was used to show that it's NOT a Bible code.