Do patterns in Scripture prove it is God's word?

We believe the Bible is the word of God. That is where we start, and it is why we care about getting the evidence right.

For as long as people have read the Bible closely, some have pointed to patterns in it as proof that it came from God. A prophecy that names the year. A hidden word spelled out at equal spacing. A story folded like a mirror around its centre. If even one of these is beyond any human author, the argument goes, then the sceptic has to listen.

Peter tells believers to be ready to give a reason for the hope that is in them, with meekness and fear (1 Peter 3:15). An honest reason has to survive a fair test. So our research team took the most famous patterns and tested each one against the Hebrew and Greek texts, following one rule throughout: the test and its pass mark were written down before we looked at the answer.¹1 Inspiration itself is a matter of faith that no count decides, so the plan tested only the evidence offered for it.

What would compel a sceptic#

Many impressive patterns are things a gifted, careful writer could also produce. So the first job is to separate two kinds of evidence.

Evidence that no human could produce comes in two forms:

• a specific prediction that came true, in a copy of the text that can be dated, by some independent means, to before the event;

• a hidden pattern that keeps turning up in places nobody chose to look, far more often than chance allows.

Everything else, such as careful design, consistency across many writers, accuracy about the ancient world and faithful copying, is what God's word should show. It encourages faith. But a sceptic can always reply that skilled human writers do some of it too.

What held up#

A great deal. The flood story in Genesis really is built as a mirror around "God remembered Noah": its matching halves echo each other better than 1,000 other Hebrew stories cut the same way. The Bible's alphabet poems, such as Psalm 119 and Lamentations, were all found by an automatic search, and nothing else in the text looks like them.

The Bible names 3,130 people with family lines, across books written centuries apart, and not one of them ends up with two fathers or as their own ancestor. Of 17 journeys whose length in days is given, 14 fit the pace ancient travellers kept.

The text knows its world. People named in ancient inscriptions cluster in the books about the kings, where history expects them. In books whose dates are not disputed, Persian loanwords appear almost only in those written after Israel met Persia: 154 of 155 times.

And it was kept with care. The Samaritan copies of Genesis, handed down apart from the Jewish ones for over two thousand years, carry the flood's mirror almost unchanged.

A sceptic can point out that mirrored stories are also found in Homer, and alphabet poems in Babylon. That is true, and it is no surprise to Christians. God has always spoken through human writers, in their own languages and with their own crafts. The care is theirs and the word is his.

The near misses#

Now for the patterns that would point beyond any human author. Each came closer than you might expect, and each fell short of the bar we set.

Daniel's seventy weeks. Daniel 9 speaks of seventy weeks until an anointed one is cut off, and a famous Christian reading counts 69 weeks of years from a Persian decree to the death of Jesus. Read one of the traditional ways, the count does land in the years when Pontius Pilate governed Judea, where Tacitus and Josephus place the crucifixion. Only about 4 in 100 random windows of time do as well. But there are a dozen ways to read the decree and the count, and only 2 land there. Once we allowed for that, as we had planned to, the result was no longer strong enough to count as proof.²2 2 of 12 readings end in AD 26 to 36; 3.7% of random 11-year windows do as well; Holm-adjusted p = 0.11 against a bar of 0.05. Anderson's and Hoehner's famous counts of exactly 173,880 days come out at 173,882 and 173,884 days between their own dates.

Scholars who date Daniel late read the same verses another way, with the first seven weeks running from the fall of Jerusalem to the Persian king Cyrus, and that reading fits at least as well. None of the ancient copies of Daniel that can be dated preserves these verses, so there is no copy we can prove is older than the events.³3 The rival reading: Holm p = 0.045. Track 64 checked 15 prophecies against 171 dated manuscripts from the Judaean Desert; none has a copy older than every fulfilment on record.

Daniel 11. This chapter traces a long run of wars between Greek kings in Syria and Egypt in remarkable detail. Two independent readers scored every checkable claim against the ancient historians. Up to verse 39, 96% of the claims match what happened.

Then something changes. Verses 40 to 45 describe the end of a persecuting king. If he is Antiochus IV, who died in 164 BC, just 1 of their 6 checkable claims matches: the history has him campaigning in the east, not invading Egypt.⁴4 Track 95r. Against the control passage (90% correct, 19 of 21), the drop is significant: Fisher p = 0.0014. It rests on 1 Maccabees and Josephus; with the Greek and Roman historians alone it is undecided.

From Jerome onwards, many Christian interpreters have read those last verses as looking past Antiochus to a final enemy of God's people. On that reading, a break at verse 40 is exactly what you would expect, because the prophecy moves on to events still to come. Scholars who date the book to Antiochus's own reign read the same break as the point where the writer's own history ran out. The test finds the break. It cannot tell us which reading is right.

Could a copy settle it? One Dead Sea Scroll of Daniel 11 has been radiocarbon dated. The chance that its animal skin is older than Antiochus's reign is 91%. Our pass mark, set in advance, was 95%. The handwriting suggests a later date. It is the closest any test came to a prophecy provably older than its fulfilment, and it is not quite there.⁵5 4QDan^c, recalibrated on IntCal20: probability 0.912 against a bar of 0.95. Palaeographic date 125 to 100 BC (track 64c).

Hidden codes and numbers. In Genesis and Exodus, every 50th letter from the first "t" really does spell "Torah", and that is rare. But it never happens again, with 22 other words or in any other book. John 1:1 adds up to a number times its own reverse, like Genesis 1:1, but only with a spelling that medieval scribes used, and the device turns up at chapter openings no more often than chance.

God's name in Esther. Esther is the one book of the Bible that never mentions God, yet his care for his people runs through every chapter. Tradition points to four places where the first or last letters of four words in a row spell God's name. Those four are really there, and Esther has more of them than any other book. But about one shuffled copy of Esther in 27 does as well, which is not rare enough to rule out chance.⁶6 4 against 1.21 expected, p = 0.037 against a bar of 0.01. Across the whole Bible the device runs at chance: 125 against 134.9 expected (track 73).

Sevens. Claims that Genesis 1 and Matthew 1 are built on hidden sevens passed in one edition of the text but failed in half of the others. A pattern that depends on which edition you print is a pattern in the edition.⁷7 6 of 12 edition settings (track 12).

Where our confidence rests#

No pattern we tested is a proof that would compel a sceptic. Several came close. None of the hidden codes recurs, and no fulfilled prophecy yet has a copy we can prove is older than its event.

That should not unsettle anyone. Christians have never rested their confidence in Scripture on counting letters. The Reformers taught that our full assurance of God's word comes from the Holy Spirit bearing witness through the word itself, and above all from Christ, who received the Scriptures as his Father's word and fulfilled them.

What the tests do show is a book made with great care and kept with great care: shaped on purpose, consistent across many hands and centuries, rooted in the real world, and copied faithfully for thousands of years. We read those as the fingerprints of a faithful God. A sceptic reading the same numbers may not, and we would rather offer them honest evidence and the plain gospel than a clever proof that fails.

There is more to test. Prophecies matched against newly dated manuscripts could yet move the needle, and any new hidden-pattern claim can be tested the same way: written down first, then checked.

Every test here was written down, with its pass mark, before it was run. The full reports, with their methods and every figure in this post, are to be published with Carestead Research's open data. Scripture quotations are from the King James Version.