Claims of Scientific Accuracy in Scriptures

by Gabe Czobel

There are claims made by some that the religious scriptures to which they adhere contain many
accurate statements, when viewed in the light of modern science, that the human authors of those
scriptures could not have known at the time they were written. It is then claimed that this could only
happen if the authors of the scriptures were guided by an omniscient “divine” power already possessing
these truths of science, especially since scientific truths would have been a consequence of the creative
act of this divine power. There are a number of problems with such claims, when soberly examined in
greater detail, that seriously call into question the thrust of this line of argument .

• Each particular claim of this nature relies heavily on a very liberal interpretation of particular
scriptural elements since the scriptures often derive from some archaic language with broad
possibilities of translation, they often contain archaic terminology, and are often saturated with
myth and metaphor. They are anything but precise, unambiguous scientific statements. [1]
• Even where a more specific fact related assertion is made, the comparison with science may be
in the nature of comparing apples to oranges in a very lax manner to spin the comparison in the
most positive light. On closer detailed examination, such assertions appear far less impressive
than on first sight.[2]
• Often, only a handful of scriptural assertions are presented in detail in the typical argument of
this type but many more are simply asserted to be equally “predictive” and “accurate” without
giving the actual instances. The arguments also studiously ignore the blatant factual errors and
other erroneous assertions that are equally present in the same scriptures in great quantity, some
of which may be the same statements that are purported to represent truths of science, but
interpreted more literally.
• The arguments take as given the assumption that these “facts” could not possibly have been
known, or even reasonably surmised, in ancient times when the scriptures originated (i.e.
knowledge that the earth is spherical instead of flat). But this vastly underestimates the known
sophistication of ancient philosophy, science, and mathematics such as possessed by the
Greeks, the Chinese, the Babylonians, the Mayans, and the Egyptians, to name a few. It also
ignores the likelihood of scriptural revision over time to better reflect the evolution of
knowledge.
• Even generously granting that a good number of these claims are interpreted correctly as
possessing some modicum of scientific validity, that in no way implies that the only possible
alternate source of such scientific knowledge, aside from the human authors of the scriptures,
was the divine creator of the universe as envisaged by the adherents of the particular scriptures,
who was the purported guiding hand in their origin. One could, with equal plausibility, posit
alternatives such as a hidden or lost advanced ancient civilization (Atlantis for example),
visitations by advanced off-world aliens, visitations by time travelers from the future, clever but
local spirits or minor deities, or even the deity of some other incompatible religion, and so on.
Only the imagination places limits on the number of plausible, non-human, alternate
possibilities that are no less plausible than some particular divine source. Thus, such arguments
rely on a blatantly false dichotomy to buttress their favoured thesis.

In this article, I will not delve into more details concerning these problems. I want to address a more
technical issue regarding some claims of probability related to such arguments.
Some of these arguments are couched in terms of probability. For example, if the scriptural source says
that the Moon reflects light rather than emits it, then there are two possibilities, and the probability of a
correct guess (the former) at random is ½. If there are, say, 10 likely plausible choices (i.e. flat Earth,
spherical, cubic, various polyhedra ...), the probability of a correct guess at random is 1/10. If a
number of independent “scientific truths” of this type are presented to buttress the argument, the
probability of randomly getting them all correct is the product of their individual probabilities. This
product quickly becomes a very small number even for a handful of assertions and the claim is then
made, “what are the chances of that happening if the authors of the scriptures were merely guessing?”
It is such probabilistically framed arguments for a divine scriptural source that I will examine in detail
because, to the uninitiated in the basics of probability, arguments framed in such a manner appear very
impressive and compelling. But are they really? Let's take a closer look.

What is usually not considered in these probabilistically framed arguments is that a few “apparently”
correct guesses on their own cannot be taken in isolation. In assessing probability, all factual scriptural
assertions regarding domains of knowledge, scientific or not, that are not likely available to ancient
authors, need to be considered. To make this clear, take the case of 10 tosses of a coin. Let's say that
one is given a chance to predict the outcome of each toss before that toss is made and a tally is kept of
the correct and false guesses. The chance of correctly guessing a single toss, the choice between a
“head” or “tail”, is ½, or 1 out of 2, assuming we have a “fair” coin. Let's call a correct guess as true
and an incorrect guess as false. If one can predict every single outcome in these 10 tosses, that is get ten
true guesses out of 10 and no errors, then the chances of that happening are indeed ½ x ½ x ... x ½ (ten
products of ½), or (1/2)10, or 1 in 210 or 1 in 1024. That is an eye opener – not a very likely event! But
if one simply gets 10 true guesses anywhere in, say, 100 tosses of a coin, that is much more likely to
occur because possibly the other 90 guesses could be wrong. We'll see below just what is the exact
probability of correctly guessing any 10 of all the tosses of a coin out of 100.

A general formula for calculating the probability of obtaining a specific number of correct predictions
among a series of random guesses may be derived. There would be three variables involved:

• Let X stand for the total number of random guesses in the situation where X assertions are
being made by guessing from a number of possible choices for each assertion. In relation to the
arguments for true scientific assertions in scriptures, X would be the superset of all factual
assertions made by these scriptures, that may be reasonably construed to have been beyond the
knowledge of the authors.
• Let Y stand for the number of possible ways to make each guess, only one way being correct or
true (T), and the other ways being false in Y-1 different ways (F1 , F2 , F3 , F4 , ... FY-1 ) . This
would vary from assertion to assertion in real situations but we will simplify the model by
keeping this constant over all assertions.
• Let Z stand for the number of guesses found to be correct among the X guesses. In the case of
arguments purporting to show true scientific assertions in scriptures, Z would be the number of
such assertions actually presented to be in accord with scientific truths. Thus, if we make X
guesses randomly, we want to calculate the probability of being able to find Z correct guesses
among the X. For this to happen, we need to find at least Z true guesses, but more may be also
be true by chance.

In summary, we want to calculate the probability of finding Z true guesses by systematically searching
among all X guesses, if all X were guessed at randomly from Y possible choices for each. If more than
Z are randomly true in any one attempt at guessing the X assertions, that does not impact the
probability of finding the Z since the search for Z true guesses is systematic; once Z are found, the
search is complete. To view it another way, an argument will present only Z true assertions in isolation.
We do not know if other assertions among the X are also true, even by chance, since the argument does
not examine them. So either Z true guesses may be found or they may not. If the X assertions have less
than Z true ones, then Z true ones will certainly not be found. If there are Z or more true assertions in
the X, then Z true ones will certainly be found.

For X assertions, there are a total of YX possible ways that they may be randomly stated since each
guess has a choice of Y possibilities and for each one of the possibilities the next assertion has Y
possibilities and so on up to X guesses. Thus, on any single attempt at randomly guessing the value for
X assertions, we will see one of these YX possible ways, which may be viewed as a list of X results. For
instance, a typical list of X results, represented in terms of truth value, may look something like this,

F4 F1 T T F2 T . . . . . . . . . . F7 T F3
1 2 3 4 5 6 (X-2) (X-1) X

where T indicates that that assertion was guessed correctly and Fn indicates one of the Y-1 possible
incorrect (false) guesses. Now we need to find Z true assertions distributed among any particular
instance of such a list, although that list may have more than Z true ones. If there are more, we can still
systematically pick out Z true assertions with room to spare. We need to lay out all the possible lists
among the total YX possible ones, where, by chance, there are enough true assertions to find Z true
ones. Therefore we need to ignore any lists which have less than Z true assertions. Here are the possible
ways to have Z or more than Z true assertions in a random list of X:

• There are exactly Z true assertions among the X, each one in any position of the list. Among a
total of X assertions, this may happen C(X,Z) ways or combinations. [3] But with each such
way of finding Z true assertions, the other (X-Z) assertions which are false, may each be false
in (Y-1) different ways. Thus, the total number of unique different ways to have Z true
assertions in the list is C(X,Z) x (Y-1)(X-Z) which is equal to C(X,X-Z) x (Y-1)(X-Z) since C(n,r)
= C(n,n-r) . [4]
• There are exactly Z+1 true assertions among the X. The total number of ways to have Z+1 true
assertions in the list is C(X,Z+1) x (Y-1)(X-Z-1) = C(X,X-Z-1) x (Y-1)(X-Z-1) since there is room
only for X-Z-1 false assertions.
• Similarly for Z+2 true assertions, we can have
C(X,Z+2) x (Y-1)(X-Z-2) = C(X,X-Z-2) x (Y-1)(X-Z-2) ways.
•
•
•
• This pattern may be repeated for increasing number of true assertions, and the last two ways
will be when we have X-1 true assertions in C(X,X-1) x (Y-1) = C(X,1) x (Y-1) ways and
• when all X assertions are true in only one possible way, that is,
C(X,X) x (Y-1)0 = C(X,0) x (Y-1)0 = 1 .
Thus, the total number of ways to find Z true assertions among the X, along with none, some, or all the
others being in some way false, is the sum of the above distinct cases. Adding in reverse order we have:
(X-Z) X

∑ C(X,r) (Y-1)r then adding and subtracting the same term ∑ C(X,r) (Y-1)r gives
r=0 r=X-Z+1

(X-Z) X X

= ∑ C(X,r) (Y-1)r + ∑ C(X,r) (Y-1)r - ∑ C(X,r) (Y-1)r then combining the first two terms
r=0 r=X-Z+1 r=X-Z+1

X X

= ∑ C(X,r) (Y-1)r - ∑ C(X,r) (Y-1)r and using the binomial expansion [5] for the first term
r=0 r=X-Z+1

X X

= (1+[Y-1]) - ∑ C(X,r) (Y-1) = Y - ∑ C(X,r) (Y-1)r

X r X

r=X-Z+1 r=X-Z+1

In general, the probability of some type of event occurring at random is calculated by dividing the total
number of possible ways that particular event may occur, by the overall total possible ways all such
events may occur. For instance, the probability of rolling a “5” by tossing a die, is 1 / 6 since there is
only one face on the die with “5”, but there are six faces altogether, each equally likely to come up
using a fair die.

Hence, the probability, denoted here as p(Z/X/Y), of finding Z true assertions distributed among X
assertions guessed randomly from Y choices is equal to the total number of ways to find at least Z true
assertions distributed among X, divided by the total number of ways to state X assertions, or
X

p(Z/X.Y) = { Y - ∑ C(X,r) (Y-1)r }

X
/ YX
r=X-Z+1

p(Z/X/Y) = 1 - {∑ C(X,r) (Y-1)r / YX}

r=X-Z+1

Now lets see how this works for some typical situations. In the following calculations, I will round off
the values at 2, 3 or 4 significant digits to keep the calculations compact. We're only interested in
trends here, not exact values.

1. Taking the original example of ten tosses of a coin where each toss is predicted true, X=Z=10
and Y=2 (heads or tails). Applying the formula above,
10

p(10/10/2) = 1 - { ∑ C(10,r) (2-1)r / 210 } then subtracting and adding 1 in the numerator
r=1

10

p(10/10/2) = 1 - { (-1 + 1 + ∑ C(10,r) (1)r ) / 210 } bringing the +1 into the sum gives
r=1

10

p(10/10/2) = 1 - { (-1 + ∑ C(10,r) (1)r ) / 210 } note the r starts at 0 now and
r=0

applying the binomial expansion

 p(10/10/2) = 1 - { (-1 +(1+1)10 }/ 210 = 1 - { ( 210 - 1) / 210 } = 1 - { 1 - 1 / 210 } = 1 / 210

as earlier calculated by simply multiplying the 10 probabilities of each toss. Thus it is only 1 in
1024 that by random guessing one can predict all ten tosses of the coin.

2. Looking at the same case but expanding the total tosses of the coin to X=100, the formula for
the probability now gives,
100

p(10/100/2) = 1 - { ∑ C(100,r) (2-1)r / 2100 }

r=91

= 1 - { (C(100,91) + C(100,92) + ... + C(100,100) ) / 2100 }

= 1 - {(1.9E+12 +1.86E+11 +1.6E+10 +1.2E+9 +7.5E+7 +3.9E+6 +1.6E+5 +4.9E+3 +100+1 ) /1.27E+30}

and since the denominator in the bracketed term is 18 orders of magnitude greater than the
numerator, the bracketed term is near 0, so the probability is near 1. That is, it is very near
certain that in 100 tosses of a coin, one can correctly predict at least 10 of the tosses.

3. Let's take a look at assertions where the probability of guessing right is much lower than a coin
toss, say 1 in 10. Then Y=10. Then let's say we claim to get Z=5 assertions right, but this is part
of a group of X=20 assertions altogether. What are the chances of that occurring by simple
guessing? Taken in isolation, getting 5 such assertions right by guessing would occur 1 out of
105 or 1 out of 100,000 ! That sounds very impressive and seems to very strongly discount that
guessing was behind the true assertions! But what is the probability when taken in the context
of 20 assertions? In that case,

20

p(5/20/10) = 1 - { ∑ C(20,r) (10-1)r / 1020 }

r=16

= 1 - { (C(20,16)(916) + C(20,17)(917) + ... + C(20,20)(920) )/ 1020 }

= 1 - {(4845(1.85E+15 )+1140(1.67E+16 ) +190(1.50E+17 )+20(1.35E+18 )+1(1.22E+19 ) ) / E+20 }

= 1 - {(8.96E+18 +1.90E+19 +2.85E+19 +2.7E+19 +1.22E+19 ) / E+20 }

= 1 - 0.957E+20 / E+20 = 0.043

This is 1 in about 23 which is way more likely by orders of magnitude than what it would
appear to be if taken out of context, that is 1 in 100,000. And this is in only 20 assertions with
quite a low chance of guessing each one right ( 1 in 10).
 4. If we change the total assertions in case 3 above to X=50, that should make it even more
probable to randomly predict 5 correctly.
50

p(5/50/10) = 1 - { ∑ C(50,r) (10-1)r / 1050 }

r=46

= 1 - { (C(50,46)(946) + C(50,47)(947) + ... + C(50,50)(950) )/ 1050 }

= 1-{(2.3E+5(7.86E+43 )+1.96E+4(7.07E+44 )+1225(6.36E+45)+50(5.73E+46)+1(5.15E+47 ) ) / E+50 }

= 1 - {(1.81E+49 +1.39E+49 +7.79E+48 +2.86E+48 +5.15E+47 ) / E+50 }

= 1 - 0.4317E+50 / E+50 = 0.5683

This is better than 1 out of 2 that 5 assertions will be found to be guessed correctly out of 50
assertions even if there is only a very low 1 out of 10 chance of getting any single assertion
correct by random guess. This is better than a flip of a coin!

These examples should make it clear that looking in isolation at merely a small set of assertions in
scriptures that are picked out as scientifically true, according to current knowledge, is highly
misleading. We can see that the likelihood of finding a handful of true assertions among a larger
number of assertions is not at all low and may even be quite likely, even if all these assertions were
guessed randomly – vastly more likely than merely considering the correct assertions in isolation. Such
purported true assertions must be viewed in the total context of all assertions of which the authors of
the scriptures would not likely have had direct knowledge, including assertions of which we may even
now not know the correct answer.

Of course we must also not overlook the troubling questions regarding the amount of laxity in
interpretation needed to present any particular scriptural assertion as indeed corresponding to a
scientific fact currently taken as true.

Notes:

[1] There is a claim that the Old Testament or the Torah “knew” about dinosaurs. These are the
references given:
• (Genesis 1:21) - And God created great whales ... or the Torah translation is given as “great sea
monsters” (“taninim” in Hebrew)
• (Isaiah 27:1) - In that day the LORD with his sore and great and strong sword shall punish
leviathan the piercing serpent, even leviathan that crooked serpent; and he shall slay the dragon
that is in the sea. (the Torah also used “tanin” as the word for dragon)
• (Exodus 7:9) - When Pharaoh shall speak unto you, saying, Shew a miracle for you: then thou
 shalt say unto Aaron, Take thy rod, and cast it before Pharaoh, and it shall become a serpent.
( the Torah also used “tanin” as the word for serpent, which in this context would be deemed to
be a snake)

Thus the interpretation centers around the single Hebrew word “tanin”, variously interpreted in
different contexts as whales, sea monsters, serpents, snakes and dragons, possibly something
reptilian. Of course such creatures, either of the real or mythological variety, would hardly be
unknown to ancient authors. But the claim that these referred specifically and unequivocally to
dinosaurs hinges merely on the loose interpretation that dragons and great sea serpents are taken
as reptilian and mythologically large at that, just as dinosaurs are commonly viewed to have
been. But dinosaurs came in a large variety of sizes, nor were they all sea creatures as the
scriptural references appear to restrict. Regarding any other characteristics of dinosaurs, such as
the great variety of species, variety of habitat, herbivorous or carnivorous nature, span of time
during which they were dominant, the fact that they were extinct millions of years before the
arrival of hominids on the scene, and so on - that is, all the things which evoke the genuine
scientific concept of dinosaurs - are entirely absent in the scriptures. Only with the aid of an
overactive imagination or perhaps unbridled confirmation bias can one claim that these
scriptural references could only be about dinosaurs.

[2] In one argument to divine guidance, the Hebrew Talmud has Rabban Gamliel, who flourished
around the mid first century, stating, “I have received tradition from my father Abba, that the
moon will renew no sooner than 29 days, and a half, two thirds of an hour, and 73 parts of an
hour.” Then the argument claims that Rambam, more commonly known as Moses Maimonides
(1135-1204) commented that a full hour has 1080 parts. From this, it is deduced that the phases
of the moon, by Gamliel's reckoning, have a period of 29 + ½ + (2/3) (1/24) + (73/1080) (1/24)
= 29.53059 days.
Then it is brought up that modern science has determined the cycle of the phases of the moon
(called the synodic month) to average 29.530588 mean solar days. Finally the question is
raised, how Rabban Gamliel, not possessing modern telescopes, could have known this unless
guided by God.
This claim does indeed look impressive and convincing on the face of it considering that the
numbers match up to 5 significant decimal places! But let's look closer at the details. The
numerical value for the phrase “parts of an hour” came from a source one millennium after
Rabban Gamliel. Did they both divide the hour into the same number of parts? We don't really
know and no other evidence is given in this argument that they did.
But even granting that the parts of the hour were the same and that the derived value, 29.53059
days, is what Gamliel had in mind, we are really comparing apples and oranges here. Gamliel
claimed that the cycle of which he spoke is the smallest renewal period (“... will renew no
sooner than ... “). But the modern scientific value for the synodic month given in this argument
is the average one. The synodic month actually varies between 29.27 days to 29.83 on account
of orbital perturbations. So Gamliel's number should really be compared to 29.27 which is no
longer so impressive. Another important point to remember is that these numbers are dealing in
units of days! A day has 24x60 = 1,440 minutes. Converting the difference between Gamliel's
number and the modern value into minutes we have 1440 (29.53059 – 29.27) = 375 minutes or
6 ¼ hours! This is time enough to take a good nap and is well out of the bounds of the range of
error expected even of ancient astronomy!
Finally, regarding the claim in the argument that a modern telescope is needed for accurate
astronomical observations, especially of an object as large to the naked eye as the moon, we
need only mention the name of Tycho Brahe, the 16th century Danish astronomer who
 catalogued a great number of highly accurate observations even before practical telescopes
were first introduced by Galileo.

[3] C(n,r) is one notation that stands for the number n! / r!(n-r)! . This number is equal to the total
number of ways (combinations) one can pick out, in any order, r items from a total of n items.

[4] Since C(n,r) = n! / r!(n-r)! = n! / (n-r)!r! = n! / (n-r)!(n – [n-r])! = C(n,n-r)

[5] From the binomial theorem we have the expansion for (1 + x) n as

n

(1 + x) n = ∑ C(n,r) xr
r=0

© 2010 by Gabe Czobel