I think it was Don who once described an apologetic method of debate as something along the lines of coming into a room, dropping a huge pile of feces on the floor, and then leaving the skeptics to sort out the mess.
We were recently hit by such an apologist on our AE TV list who wrote to give us “60 pages” of evidence for god’s existence. Not that I’m not interested in giving evidence a fair review—but please have mercy and some sense. We get hit with loads of requests from theists and atheists to “please look at this and tell me what you think about it.” Yes, we have a small team of people—but we are all volunteers with real lives outside of AE. And while most people understand and respect that, some seem to think that we’re obligated to read (actually in most cases reread) anything they want to dump on us for their god. Fair enough that we should consider claims before dismissing them, but how about doing what Matt sometimes asks: Give me your very best argument or evidence.
This way we can start with what you think is most compelling and examine that first. Then, if I don’t find it compelling, there is no reason for me to have to wade through the other 59 pages of crap that you admit isn’t quite as compelling. Fair enough? At any rate you’d have to admit it would be a big time-saver that benefits both the apologist and the counter-apologist.
Well, this apologist wrote to some others on the list. Some things I read, some I didn’t. But with me, she was very interested in the Bible Code. It had been a few years since I had encountered anyone serious about the Bible Code. In fact, it’s so infrequently used by apologists who contact us (and for good reason), that I thought other counter apologists may or may not have ever had any reason to investigate this “code” for themselves. I had read a bit back when I was a theist, but I investigated it just enough to find it utterly uncompelling, and that was that. Since I already believed in god then, it really didn’t matter.
So, I was a bit rusty in my responses to the latest claims, and had to do some refresher reading, which I did. And to be fair, I learned some things I didn’t recall from my prior reading, which is always nice. To be honest, though, what I learned made me even less inclined to accept this code as anything but hogwash.
My first complaint about the code is the actually name: Bible Code.
At least in the West, when we see the word “Bible,” we think of books that contain basically what is contained in a standard King James anthology or, perhaps, a Catholic version. In fact, while I was dialoguing, I mentioned the New Testament more than once. She never corrected me to let me know that, in fact, the “Bible” Code has nothing to do with the Christian content of the “Bible”—it only applies to restricted portions of the Hebrew holy texts—the Torah (Genesis through Deuteronomy). This means that if there is any reason to think that a book containing codes is the handiwork of any god, nothing in the Christian “scriptures” would be demonstrated as text from god, due to this code. In fact, another name for the Bible Code is “Torah Code”—which I hold to be more honest. When a Christian calls it “Bible Code,” that’s misleading, unless they also clearly note that none of what has been considered compelling in any of the research—the codes in the main debate over what is generally considered the context of the “Bible Code” issue—applies to what they would usually mean when they use the word “Bible.”
Where It All Began and the Koran Code
Although the idea of finding Bible codes goes back a ways, the real hot button came when, Doron Witztum, Eliyahu Rips and Yoav Rosenberg (WRR) published a paper titled: “Equidistant Letter Sequences in the Book of Genesis” in the journal Statistical Science. The main criticism from me is that we have no original manuscripts from which to work. So, we begin any such examination for divine codes with a copy, and with little means of demonstrating that copy’s firm adherence to any original version(s).
Ironically, Muslims have noticed this same problem. There is also, not surprisingly, a Koran Code, and one of the reasons given for the superiority of the Koran Code by Muslim proponents is that the Koran text is closer to “original” than any Bible text could ever hope to be. While the Koran does have a history that leaves room for translation “adjustments,” the claim that it is closer to “original” is not without a bit of merit. I should note that I do not claim the Koran Code is ELS-type code. It's somewhat variant. But as long as the Torah guys can make up their code rules, why not the Muslims? At any rate, it seems anyone can have some sort of superhuman, magical code in their holy book from god. But here is a link to a Muslim making his case for why the Koran Code beats the pants off any Bible Code:
And here is a load of info if you’re interested in what impresses some Muslims about the Koran Code:
And here is an article in the Egypt Daily News, talking about the miracle of the Koran Code, in which Meer Hamza, who has a Ph.D. in software engineering from University of Paisley in Scotland, says it “will be one of the main reasons to make non-Muslims turn to Islam.” In fact, the team who “cracked” the code claimed, “no person on earth, not even a computer software is capable of writing even one word abiding with the Quranic mathematical code”:
I include this, not because I find it any more compelling that the Bible Code, but because I was told, by the apologist with whom I was dialoging, that they’d never heard of any such “Koran Code.” Not only is it there, but it’s hailed as an undeniable “miracle” by the Muslims who subscribe to it. And if you’ve ever argued with a Muslim, you know they have what can only be called “a thing,” for number-play in the Koran.
Forgive me for one sideline. I know I go long. But I’ve heard people claim the Koran doesn’t contain prophecy. (It’s funny how many Christians make claims about the Koran, that are hotly disputed by Muslims I encounter.) As I was looking up the code material last night, I found a Koran prophecy for the Apollo Moon mission:
What makes me laugh is how much like Christian prophecy this Koran prophecy works. It predicts so much—after it has already happened. It’s very rare to find someone hollering for extra security at some political event because Isaiah or Moby Dick or the Koran or Nostradamus predicted an assassination attempt. How many disasters have been averted by someone recognizing an embedded prophecy before such a prophecy took place? And as I toss this out as rhetorical, watch someone find me an example! It seems there’s always someone.
Bible Codes, the Later Years
Believe it or not, there is a guy who came after WRR, Michael Drosnin, who says he found a prophecy in the Torah Code before it happened—the assassination of Yitzhak Rabin. Since Rabin was successfully assassinated, I have to note that this prophecy must not have been very clear, since it was of zero use in averting a tragedy it supposedly foretold. Drosnin claims he tried to report the threat. If so, that would be to his credit. It's too bad there wasn't sufficiently specific information in the prophecy to make a warning more useful. Armed beforehand with a name of the assassin, a specific date and a few more details, and someone in authority could have perhaps helped Rabin avoid being killed. How was this “prophecy” even helpful? And how many future events will be "found" in the text--like the name of Rabin's assassin, said to have been found after he was already identified? That’s painting a target around my arrow after I’ve randomly hit some tree.
But Drosnin was so impressed with himself that he went back looking for other, unhelpful and useless, historic assassination prophecies, and—surprise, surprise—found them. Of course, they were all “encoded” in different ways—it appears there is no set method in this code of expressing that someone will be assassinated. You sort of have to know how to read it in various disguises. But, it’s there, says Drosnin—if you know how to look. This sort of sloppy work is a trademark of Drosnin, though. It is examined in sufficient detail in Marvin V. Zelkowitz’s research paper, “The Bible Codes,” found at the University of Maryland website:
It’s slightly painful to read Zelkowitz’s responses, because it’s hard to accept anyone could be as careless as Drosnin, and still be hailed as a hero by so many gullible followers. Drosnin, to be fair, doesn’t claim to be either a scientist or a researcher. He’s just a writer—selling books about the Torah Code. Nothing wrong with being a writer, unless your writing about something that rests upon research data, and you don’t understand good research methods.
The first thing to understand about research with something like the Torah Code is how data can be bias. Here’s a simple example of something you have to understand about numbers and how they impact interpretations of data:
I am going to do a study. In my study I have a subject who will flip a silver dollar. I won’t tell him why (I want to keep it “blind,” so my subject won’t skew my results). I also want the coin to land on the floor—I don’t want the subject to touch it after he’s tossed it—in order to make my results as random as I can. After all, I want to be fair. My hypothesis, known only to me, is that I believe the coin toss experiment will reveal that I get “heads” as a result far more often than “tails.” My subject tosses the coin into the air—and it lands, exactly as I predicted! Heads up! I thank my subject and write my paper demonstrating that “not only did I get heads more often than tails, but I actually achieved a much greater result than anticipated: I achieved heads 100% of the time!”
I did nothing to manipulate the data. I blinded the study and removed myself as a potential factor that could bias the data. And I inserted something in my method to make the result random. And I was very honest about the data I gleaned from my research.
I’m guessing you can see the problem with my conclusion that coins, when tossed, will most often (maybe even 100% of the time) land on heads.
Sample size impacts results. How many coin flips did I test? One. And that sample size invalidates my findings. Obviously I need more coin tosses. So, next time I do my research, I do two tosses. And I get heads both times. Have I fixed the problem? No, because we all know that it’s possible to get what is called a “run” with things like this. Even though over an extended period of tossing the coin, I should get heads something close to 50% of the time, I could, theoretically, get heads for 100 tosses in a row. It would be funny, but not impossible.
What was the sample size in the original Torah Code experiment? How many books other than the Torah did they test? None. They ran multiple codes, but only on the Torah. They didn’t run those codes on other similar samples. And, as we’ll discuss, that’s more of a problem than just “sample size.” But to be fair, there is some debate from what I read about whether the results held up as “statistically significant,” after other researchers ran the codes on other, similar books as samples. We’ll address that later.
Repeatability, Part I
Beyond increasing sample size, repeatability is also good thing. Instead of doing one experiment where I toss the coin 100 times and calculate the rates, I now decide we’ll do ten 100-toss experiments. As I increase my sample size to something more reasonable—100 tosses per experiment, rather than two—I will more and more start getting results closer to 50% heads. But since I recognize it’s still realistically possible to get an anomaly like a 100-head-result run in one experiment, it’s a good idea to do the experiment more than once. With the Torah Code, if there is demonstrated statistical significance, we still have to understand, this could be our 100-head run. So, I see high results found in one book, as potentially not even relevant. If I run the test on 100 books, I expect to see some hit high, some low, and some closer to center. If we can’t explain how one book generated higher rates—the fact that it did so could be nothing more than our 100-head run. The simple fact is: Nobody knows.
The way to demonstrate that the high hit book is significant, and not just a possible anomaly, would begin by offering a plausible explanation for how it hits so much higher than the other books, something for which we could test. “God” could not be offered as a cause in a universe where no gods have yet been demonstrated to exist, since things that do not exist cannot be the cause of other things. To posit god as an explanation, would first require a study to demonstrate there are gods and that those gods would be inclined to produce book codes. A daunting task. But short of any actual plausible explanation available to us, we would be left only with a high hitter and no means demonstrate how it hits so high.
But I will share with you a statement from Robert Aumann. Aumann is a Game Theorist. He was impressed by the codes at the time he said, “for many years I thought that an ironclad case had been made for the codes; I did not see how ‘cheating’ could have been possible.” However, when the research was critiqued by other researchers, Aumann had to admit that “Though this work [in reference to published criticisms of the code methods] did not convince me that the data had been manipulated, it did convince me that it could have been; that manipulation was technically possible.”
What Is a Code?
Aumann’s statement leads us into our next issue, which requires an understanding of how the codes are generated. A method known as ELS is used—but you could apply loads of different patterns. Any pattern could be a “code.” But in ELS, the researcher gets to pick a starting point in the manuscript—which may or may not be the first letter. Then he gets to choose another number (so far the researcher has manipulated two variants) that he then “skips” until the next letter in the manuscript. So, let’s say every tenth letter is selected after your chosen start number. The fact you can configure it however you like is a big part of the problem. There is no “prescribed” matrix that we know will work to find secret messages from god in books, so it’s up to the researcher to start picking numbers. With no known “correct” matrix for finding secret messages from god in texts, we all have free reign over how we build out favorite matrices.
So, we end up with a string of random letters, which we then go over with a fine-toothed comb looking for words that we get to label as “meaningful” (to us) or not. Obviously we can expect to find loads of “words” or “strands” of words—but only the ones we decide “count” will be selected. So we get to keep and toss whatever we think fits.
As you might imagine, this is a big no-no. In relation to my coin toss experiment, let’s say that 25 times out of 100, the coin bounces off a wall before it lands on the floor. Would it be a problem if I said, “anytime the coin bounces off the wall and lands on tails, I am going to say it doesn’t count toward the final 100; but anytime it bounces off a wall and lands on heads, we’ll count that”? The answer is “yes, it would be a problem.” It would skew my results. Each time I hit the wall and don’t get my desired heads, I get to do it over again, which increases the chance I’ll get heads more often over tails. In these codes, whatever words the researchers find that don’t mean anything to them, they don’t have to report as a “miss.” But whatever they find that they have predetermined will be “meaningful” to them—literally whatever they call meaningful—they get to report as “hits.” So you ignore the garbage, and just report the positive findings.
If meaningful phrases are evidence of the existence of a code, then why isn’t “noise”—random letters with no discernable meaning found in the code (to a large degree)— counted as evidence against the existence of a code in the text? In other words, if a bit of “not crap” = “coded,” then why doesn’t “crap” found all over in these codes demonstrate “not coded”? Why does this magical code include any crap at all? What’s the crap for? Why doesn’t it tell a cohesive story using all the code letters with no noise? I suggest something that starts with “Congratulations, you’ve cracked my secret god code! I have so much to reveal to you…” and on from there.
In the original study, though, the words that the researchers decided would be counted were words connected with biographical data about famous rabbis. The findings were interesting and the researchers claimed success. In subsequent criticisms, it was demonstrated that by manipulating the matrix settings, you can increase or decrease the level of significance. It would be along the lines of discovering that if you start your coin flip with the coin on tails, you are more likely to get heads as a result. As Aumann says, you may not have intended to start more of your flips on tails, but if you did, however unintentionally, it could still potentially skew your results. Even without meaning to, the moment you select which letters and numbers to use, you have already influenced the results. And this is only one way to subjectively, and problematically, “tune” your results. More later.
Repeatability, Part II and Statistical Significance
For now, let’s go back to my Coin Toss experiment and the idea of “repeatability.” Most likely I will get some variance if I increase my sample size to 100 tosses within my single experiment. For example, I may get heads 30% and tails 70% in one experiment. So, it’s good to “repeat” the experiment. The second time, I get heads 48% and tails 52%. So, I repeat and repeat, and eventually I get 52% heads after 10 experiments of 100 tosses each, plus or minus a few points for error rate—which we would all recognize as a more realistic expectation for a Coin Toss result.
This “plus or minus” consideration is what determines something called “statistical significance.” In other words, if we get 52%, that’s still in line with results we would expect from random chance for a two-sided coin flip. The two percent is not considered “significant” to us. It is not be impressively outside the range of our expectations. If we had a good sample size, and we did the experiment repeatedly a good number of times, and somehow we kept getting heads “significantly” less than tails, like only 2% of the time, we’d consider something was not in line with “chance.” Either our experiment was somehow biased, or there’s something else influencing the flips that we must identify. But what is “significant”? Is there a way to determine whether the variance we’re seeing is “chance” or something else at 53%? 55%? 60%? In fact, clever researchers have worked out methods of figuring this out. It’s not a guess. You don’t get to say, “Well I just can’t believe we could get these results by chance—so it must be significant!” Your level (or my level, or their level) of credulity is not how statistical significance is measured:
What you find personally significant or impressive, as far as assessing results, is meaningless, because human beings are biased. What can be demonstrated as significant, in research, is something else entirely.
Sometimes you can figure out if something is beyond the norm with what is called a “control.” “Controls” are handy. A control would be helpful, for example, in a drug trial. Let’s say I make a drug to cure disease X. X kills 5,000 people each year in the U.S. within one year of infection. Nobody who dies from X lives past the first year. And anyone who survives it simply exhibits natural immunity and survives with no further detectible infection. But there is no known effective medical treatment currently for curing X once you contract it.
I take 100 subjects infected with X and give them my drug protocol within the first few days of diagnosed infection. At the end of one year, 50 of my subjects are still alive and show no signs of infection. I announced that of 100 people infected with X, I successfully cured 50 with my treatment.
Should we celebrate that we’ve made a dent in medical treatment of X? If you think “yes,” slap yourself—very hard.
Before we pop the champagne, there’s something we forgot to consider: Each year in the U.S. 10,000 people become infected with X. And if I would have used a “control” group—a group of people who weren’t treated with my drug—I would have discovered 50 out of 100 subjects alive in that group as well at the end of the year—all with no signs of infection. And the study would have failed to demonstrate my drug helped anyone in that case. It’s simply a fact that half the people who become infected with X are able to fight it off successfully through their own natural immunity. The other half die within a year. Unless we get better results in future research, my drug appears to be wholly ineffective.
A “control” in the case of the Torah Code would be using other books—the more the better—to run patterns to see what I get. And in fact, the lack of controls was not ignored by critics. In 1999, in a paper published in Statistical Science (Brendan McKay, Dror Bar-Natan, Gil Kalai, Maya Bar-Hillel [MBKB]) there were a list of criticisms against work done by Code proponents (WRR), that included a control test claim. The claim was that a Hebrew copy of War and Peace had been tested and achieved high levels of statistical significance. I won’t lie to you. In this debate—as in all religious debate—there are claims and counter claims and counter-counter-claims. For every person I find claiming “statistically significant” data, I find someone else claiming they have demonstrated that same level somewhere else, or have demonstrated that the first level was achieved using faulty methods. I will let the “experts” hash that one out. All I can say is that the original data did not include controls, which other researchers had to add later. And the initial lack of control in the research should be counted as a demonstration of sloppy method.
How Do We Interpret the Data?
I said at the outset that I had learned this is about the “Torah” Code, and not the “Bible” Code. I was dismissive when the apologist on the list used an Old Testament verse in her examples, because there is something interesting about Old Testament manuscripts that I was already aware of, that, in my mind, makes them extremely suspect—if not entirely useless—in a setting like the code studies. The writing contains no vowels.
Why is this important? Consider this: Let’s say we run our code on a regular book and we get a strand that includes these letters: T H E R A P I S T. The codes use no punctuation and you get to decide not only if this set of letters is “meaningful” in your greater context, but how to interpret it. Is it “the rapist” near a set of letters that look like “bundy”? Or is it “therapist” next to a set of letters that look like “freud”? Or if it is in close proximity to “bundy,” do you only consider it as “therapist” (and fail to see the “rapist” possibility) and find no link to “bundy” (who was a rapist, but not a therapist) nearby—so you toss it out as a “miss”—instead of a “hit” or even an identified error in your code? If the code says “bundy” was a “therapist”—shouldn’t that be reported as an error in the code? Does the code say erroneously that “bundy” was a “therapist”—but you manipulated it and made “bundy” a “rapist” and wrongly attributed a “hit”? What if in 10 years, a famous therapist arises named “Bundy”? Which “bundy” does the code mean? Maybe neither; maybe the “bundy” reference is just noise? How do we tell? Even with benefit of vowels, it’s a subjective mess.
Speaking of undesired outcomes of the code—meaningful misses—it’s hard to overlook the humorous work of Dr. James D. Price, professor of Hebrew and OT at Temple Baptist Seminary, Chattanooga, TN, who found repetitive “self-contradicting” codes, and “negative codes” with messages like “there is no god” and “Satan is Jehovah”:
But back to the problem of “no vowels.” Here’s what happens when we don’t have to deal with vowels: Using English as an example, let’s say we find an “R” in the code. Just an “R” in a strand of letters. Since I get to add the vowels at will, here is a sample of what I can do with just an “R” and my choice of vowels:
And if I add an “S” after the “R” (R and S are, after all, common letters in English):
And so on.
I think you get my point. It’s incredibly subjective and easy to manipulate. In fact, it isn’t just the case that I can manipulate this data. Since there are no vowels, and I am trying to make words from this mess, I must manipulate this data. This is bad, bad, bad. I’m not just interpreting the results now—I’m actually creating the results I want. And this is called “bias,” and it’s every research paper’s worst nightmare. If I can demonstrate reasonable room for bias in your methods, you have seriously compromised your right to label your results “valid.”
Believers as Critics
As I wind this down, something interesting I also found was that some of the most damning critics of the Torah Code are religious people or, at least, people who are sympathetic toward it. Remember Zelkowitz, I mentioned earlier? He actually thinks there is something to the work done by WRR. He says in his Bible Code criticism, in fact, that WRR’s results had yet to be “satisfactorily explained.” But he still went on to shred Drosnin’s claims from just about every angle imaginable. Again, I make no claims about WRR—you can read the debate on your own if you find any of this potentially compelling. Zelkowitz published the same year as MBKB’s criticism of WRR—so I have no idea if he had a chance to read their criticisms by the time he published or not.
Also, remember Dr. Price, at the Baptist University? Here’s a comment posted by him at an online list:
“In any segment of Scripture literally thousands of such codes can be found on thousands of words. One may pick and choose among them to imagine any message he desires. The same is true for secular Hebrew literature. Hundreds of false and self contradictory statements have been found. The alleged ‘statistical’ proof has been seriously challenged by expert mathematicians. In my opinion, the topic is not worthy of serious thought. It is a waste of one's time.”
Some other religious folk who take the codes to task include the Web site “bibleonly.org,” with their posting of “Does the Bible Code Bear the Signature of God?” by Ed Christian Ph.D., Department of English, Kutztown University, Kutztown, PA. Christian’s answer is pretty clearly “no.”
And at the scp-inc.org site (Spiritual Counterfeits Project), actually a religious site dedicated to the truth of god’s word, they have a multipart shredding of the codes called “Bible Codes, or Matrix of Deception?” Again, after reading a bit of it, I’d say the conclusion I’m supposed to come to is “matrix of deception.” Of course, religious people have their own bones to pick with why these types of Bible parlor tricks are blasphemous. I’ll leave them to sort that one out as well.
Meanwhile one really uplifting tale I found on the Internet was the story of a current theist, Lori Eldridge. Lori is dedicated to the lord and runs a Bible study site. Lori used to be extremely dedicated to the defense and support of the truth of the Torah (Bible) Codes. At her site today, however, she has only this to say about her past as a defender:
“I used to be the owner of the Tcode mailing list where some of the top notch Torah Code experts in the world discussed the Bible Codes. The majority of us finally came to the conclusion that the codes cannot be of God because they were not statistically relevant and you can even fine ‘bible codes’ in the daily newspaper. And, most important, even Jesus' name (in Hebrew) was not found in any book of the Bible.”
Well, if it doesn’t say “Jesus,” then it can’t be from god, right? In jest, I have to ask, is it possible it doesn't say Jesus as a message from god? I guess what I find hopeful about Lori’s statement is that she was finally able to see past a falsehood that would have supported her beliefs about god. She had every biased reason, any other believer would have, to hang on tightly no matter how ridiculous the claim or how sorry the “evidence,” but ultimately somebody, somehow got Lori to understand how research works and how the Torah Code fails. While she is still a believer, she changed her mind about at least one piece of evidence when honestly confronted by other compelling evidence to the contrary. And I respect that and applaud her for it. That’s a heck of a fine character trait in any human being—theist or not.