A Map of Meaning

The map

Each dot is one chapter: 1,189 of them, from Genesis 1 to Revelation 22. The layout aims to place chapters with similar numerical representations near one another. Distances are approximate, so use the chapter links and similarity scores to explore a connection. The legend under the map says what the colors mean in the view you are looking at.

Color

The map can be colored six ways, and any view can be switched to any of them. By testament: two colors for the Old and New Testaments, using Christian terminology. By section: seven colors for the seven sections, cool colors for the Old Testament and warm ones for the New, or three colors for the Jewish order. By the groups a view calls out, so that Samuel and Kings and Chronicles, or the four Gospels, each take their own color. By book, one of seven colors repeating in order so that neighboring books differ. By groups calculated without using testament or section labels in the clustering step, which is a way to ask whether it sees the same shape readers do. And by the surprises, the chapters that sit with another section, colored by the section that draws them.

Inside one book

Choose a book under the map and its chapters appear numbered, pale at the first chapter and deep at the last, with the rest of the map faded. The map turns to the angle that spreads the book widest and zooms in to fit it. Under the figure: the book's two most alike chapters and its two least alike, the book nearest to it and the book farthest from it, and every chapter listed with the chapter outside the book that sits nearest to it and the chapter inside the book it is least like. Some books form compact groups; others spread across the map. These patterns can provide starting points for comparing their chapters.

Headings

The map can carry two sets of labels. Seven sections is the familiar Protestant arrangement: Law, History, Poetry and Wisdom, Prophets, Gospels and Acts, Letters, Revelation. The Jewish grouping labels the Hebrew Bible as Torah, Prophets and Writings, and shows the New Testament in gray. This changes the grouping labels, not the underlying Christian translation or its book and chapter divisions. It is not a separate analysis of a Jewish edition.

Browse all twelve map views

Each view is an angle, a coloring and a caption that says what to look for. Open one on the map above, or study every view here as a still picture.

• Law
• History
• Poetry and Wisdom
• Prophets
• Gospels and Acts
• Letters
• Revelation

• Group 1: Prophets and Poetry and Wisdom (342 chapters)
• Group 2: mostly Law (130 chapters)
• Group 3: mostly History (212 chapters)
• Group 4: mostly Law (67 chapters)
• Group 5: History and Law (39 chapters)
• Group 6: mostly Poetry and Wisdom (170 chapters)
• Group 7: Gospels and Acts and Letters (229 chapters)

• Matthew
• Mark
• Luke
• John

• Old Testament
• New Testament

• Law
• History

• Psalms
• Job
• Proverbs and Ecclesiastes
• Song of Solomon
• Songs from other books

• Revelation
• Daniel 1 to 6
• Daniel 7 to 12

• Samuel and Kings
• 2 Chronicles
• 1 Chronicles

• Chapter order, first chapter pale to last chapter deep, in the book's section color

• Law
• History
• Poetry and Wisdom
• Prophets
• Gospels and Acts
• Letters
• Revelation

• Law
• History
• Poetry and Wisdom
• Prophets
• Gospels and Acts
• Letters
• Revelation

• Torah
• Prophets (Nevi'im)
• Writings (Ketuvim)

For young readers

Imagine every chapter of the Bible is a person at a very big party. People who like talking about the same things stand together. The songs stand with the songs. The letters stand with the letters. The stories about kings stand with the other stories about kings.

This picture is that party. Every dot is one chapter. Choose "Testament" under the map and the blue dots represent the Old Testament and the orange dots represent the New Testament. Choose "Section" and each kind of book gets its own color.

Things to find:

• The songs. The book of Psalms is a book of songs, and its dots make a big crowd near the top.
• The stories about Jesus. Matthew, Mark, Luke and John make a crowd near the bottom. The Christmas story is in there: Luke 2 and Matthew 2.
• A chapter to compare. Choose "Testament". Almost every orange dot is on one side and every blue dot on the other. But one orange dot is far away from its friends, over with the blue ones. It is Matthew 1, which begins with Jesus' genealogy and also tells of his birth. The model places it near chapters with other family lists. Read it to see what they share and how they differ. Can you find it?

Now turn the map. Find two crowds that are touching. Turn it again. Are they still touching? Sometimes two crowds only look like they are touching because of where you are standing. That is a good thing to remember about maps.

For students

A computer read every chapter of the Bible and gave each one a place on this map. It did not read the way you do. It has read a very large amount of English, and from that it learned patterns used to represent and compare text. When it reads a chapter, it turns it into a long list of numbers that describe what the chapter is about. Two chapters with similar lists land near each other.

That is enough to see some real things:

• Matthew, Mark and Luke overlap. They share many accounts of Jesus' life and teaching, while each Gospel has its own selection and presentation. Twenty-three of Matthew's twenty-eight chapters have a chapter of Mark or Luke as their nearest match outside Matthew.
• Chronicles sits on top of Samuel and Kings. Samuel and Kings tell the story of Israel's kings; Chronicles tells it again. Many chapters in these books appear near one another in this layout. Thirty-one of the sixty-five chapters of Chronicles have a chapter of Samuel or Kings as their nearest match outside their own book.
• The Law and the Histories share a shelf. Genesis to Deuteronomy and Joshua to Esther sit together on the left, with similarities that may reflect shared narratives, vocabulary and other features.
• Revelation sits among the prophets. It is the last book of the New Testament, but its pictures and its language come from Daniel, Ezekiel and Isaiah, and the map keeps its chapters beside Daniel, Ezekiel, Jeremiah and Nahum.

Some chapters appear apart from the rest of their book. Matthew 1 begins with a genealogy, and it sits with the family trees of Genesis and Chronicles, far from the rest of Matthew. James 3 is about the tongue, and it sits with the Proverbs. 2 Samuel 22 is a song David sang, and it sits with the Psalms; it closely parallels Psalm 18, with some differences in wording.

Try this: choose one book under the map. Find the chapter that sits furthest from the others. Open it and work out why. Then find a book that sits somewhere you did not expect. The map is a good way to ask questions; the text is where the answers are.

What the computer did, and what it did not

The model used for this map also powers the site's search by meaning. Give it a passage of English and it answers with a list of 768 numbers. That list is not a summary you could read: an individual number is not a labeled topic or an independently readable measure of meaning. The numbers work together, like coordinates, to represent patterns learned from a large training dataset, and two passages that share ideas end up with lists that point in nearly the same direction, even when they share few words.

This page keeps the first 256 of the 768. That is not a shortcut invented here. The model was trained so that its list can be cut short from the end and still work: the first numbers carry most of what the whole list carries, the way a rough sketch keeps the outline while the later numbers add detail. Keeping 256 makes every comparison cheaper, at a small cost in precision that the model's makers publish, and the site's search keeps the same 256, so the map and the search see the same numbers.

For this map, every chapter was cut into overlapping windows of four verses, each window was turned into its numbers, and the chapter took the average. That gives 1,189 points in a space with 256 dimensions. You cannot draw that. So a method called UMAP was used to squash the 256 dimensions down to three, keeping each chapter's nearest neighbors as close as it can, and this page draws those three.

That last step is the one to understand, because it is where the map can mislead. Think of a globe pressed flat into a map of the world: Greenland looks larger than it is, and Alaska and Russia, which face each other across the Bering Strait, land at opposite edges. Every flat map of something round stretches some distances and hides others. This map is a shadow of a shape you cannot see whole, and turning it shows you a different shadow each time. Two groups that touch from one angle pull apart from another. That is why the map lets you turn it: to distinguish apparent overlaps from separation in the three-dimensional layout. Rotation cannot recover information lost when the original 256 dimensions were reduced to three.

There is one more thing to try before reading on. Choose "Computed groups" under the map. These seven groups were calculated from the chapter representations without supplying testament or section labels to the clustering step. Seven was chosen in advance, not discovered by the algorithm. One group contains most New Testament chapters. Then the Psalms as a group of their own, the histories as one, the Law in two, and the prophets together with Job, Proverbs and Ecclesiastes. Some groups overlap with familiar sections; others cross those boundaries. For example, Acts 7 and Matthew 1 group with chapters from Genesis and Exodus. Shared vocabulary and literary form may help explain these associations, but the grouping alone cannot establish their cause.

What does a cluster show? It identifies chapters grouped together by this model and these settings. Shared wording, topics or literary forms may help explain a grouping, but it is not proof of shared meaning. It does not prove that they were written together, or that they teach the same thing, or that one quotes another, though the map finds many chapters that do. The two sections below show how well the map's sense of "similar" matches what readers have known for centuries, and where it produces something new to look at.

Checking familiar parallels

Several biblical books contain parallel passages. Kings and Chronicles recount overlapping events, while Matthew, Mark and Luke share many accounts of Jesus' life and teaching. Psalm 18 closely parallels 2 Samuel 22. Ezra and Nehemiah carry the same list of returning families. These familiar parallels offer useful examples against which to check the model's similarity scores.

• Law
• History
• Poetry and Wisdom
• Prophets
• Gospels and Acts
• Letters
• Revelation
• A pair inside the Old Testament
• A pair inside the New Testament
• A pair across the testaments (dashed)

1. Obadiah and Zephaniah, 0.868
2. Obadiah and Micah, 0.864
3. Micah and Zephaniah, 0.858
4. Joel and Zephaniah, 0.851
5. Amos and Obadiah, 0.845
6. Joel and Obadiah, 0.844
7. 2 Peter and Jude, 0.840
8. Amos and Zephaniah, 0.839
9. Mark and Luke, 0.838
10. Ephesians and 1 Peter, 0.836

The strongest pair is Obadiah and Zephaniah at 0.868, and the minor prophets fill the top of the list: the six strongest pairs are all among Joel, Amos, Obadiah, Micah and Zephaniah. The first New Testament pair is 2 Peter and Jude at 0.840, seventh on the list. Of the forty, 23 are pairs of letters, 13 pairs of prophets and four pairs of Gospels; every pair sits inside one section of one testament. The strongest pair across the testaments, Proverbs and James at 0.787, ranks 100th of 2,145. Development comparisons with an earlier model retained several familiar parallels but changed other relationships, both within and between the testaments, and this was not a comprehensive test of stability across models; the cross-testament link shown here is one reading with one model.

All forty pairs

#	First book	Second book	Average score	Chapter pairs
1	Obadiah	Zephaniah	0.868	3
2	Obadiah	Micah	0.864	7
3	Micah	Zephaniah	0.858	21
4	Joel	Zephaniah	0.851	9
5	Amos	Obadiah	0.845	9
6	Joel	Obadiah	0.844	3
7	2 Peter	Jude	0.840	3
8	Amos	Zephaniah	0.839	27
9	Mark	Luke	0.838	384
10	Ephesians	1 Peter	0.836	30
11	2 Thessalonians	1 Peter	0.834	15
12	2 John	3 John	0.834	1
13	Philippians	2 Thessalonians	0.833	12
14	Ephesians	Colossians	0.831	24
15	Ephesians	Philippians	0.831	24
16	Ephesians	2 Thessalonians	0.830	18
17	Joel	Micah	0.829	21
18	Amos	Micah	0.827	63
19	Philippians	2 Timothy	0.826	16
20	Philippians	Philemon	0.826	4
21	Philippians	1 Peter	0.826	20
22	1 Thessalonians	2 Thessalonians	0.826	15
23	1 John	2 John	0.823	5
24	James	Jude	0.822	5
25	Philippians	Colossians	0.821	16
26	Ephesians	1 John	0.821	30
27	Isaiah	Obadiah	0.820	66
28	Joel	Amos	0.819	27
29	Colossians	1 Peter	0.819	20
30	Nahum	Zephaniah	0.817	9
31	Colossians	2 Thessalonians	0.816	12
32	Isaiah	Zephaniah	0.815	198
33	Matthew	Luke	0.815	672
34	2 Thessalonians	Jude	0.813	3
35	1 Peter	Jude	0.813	5
36	Matthew	Mark	0.812	448
37	Mark	John	0.812	336
38	2 Timothy	1 Peter	0.811	20
39	Titus	1 Peter	0.811	15
40	James	1 Peter	0.810	25
100	Proverbs	James	0.787	155

These are the closest pairs of chapters from different books, measured in the full 256 numbers rather than on the drawn map. The score is cosine similarity, which ranges from -1 to 1; a chapter compared with itself scores 1. The scores shown here are positive. They are not percentages of shared meaning.

All forty closest pairs

#	First chapter	Second chapter	Score
1	2 Kings 19	Isaiah 37	0.993
2	2 Samuel 22	Psalms 18	0.992
3	Ezra 2	Nehemiah 7	0.989
4	Matthew 26	Mark 14	0.987
5	2 Samuel 10	1 Chronicles 19	0.984
6	2 Samuel 7	1 Chronicles 17	0.983
7	1 Kings 8	2 Chronicles 6	0.979
8	2 Samuel 8	1 Chronicles 18	0.977
9	1 Kings 10	2 Chronicles 9	0.976
10	Matthew 24	Mark 13	0.976
11	2 Kings 25	Jeremiah 52	0.970
12	Mark 12	Luke 20	0.970
13	Matthew 27	Mark 15	0.969
14	1 Kings 22	2 Chronicles 18	0.967
15	Matthew 19	Mark 10	0.966
16	Mark 14	Luke 22	0.964
17	Matthew 27	Luke 23	0.964
18	2 Samuel 23	1 Chronicles 11	0.963
19	Matthew 26	Luke 22	0.963
20	1 Samuel 31	1 Chronicles 10	0.961
21	Mark 15	Luke 23	0.959
22	Matthew 14	Mark 6	0.957
23	Mark 5	Luke 8	0.957
24	2 Kings 18	Isaiah 36	0.955
25	Exodus 20	Deuteronomy 5	0.955
26	1 Kings 12	2 Chronicles 10	0.953
27	2 Kings 11	2 Chronicles 23	0.950
28	Ephesians 1	Colossians 1	0.949
29	Matthew 12	Luke 11	0.949
30	Psalms 69	Lamentations 3	0.948
31	Matthew 27	John 19	0.947
32	Mark 8	Luke 9	0.945
33	Mark 2	Luke 5	0.944
34	Matthew 13	Mark 4	0.943
35	Matthew 22	Mark 12	0.943
36	Matthew 17	Mark 9	0.943
37	2 Samuel 5	1 Chronicles 14	0.941
38	Mark 13	Luke 21	0.941
39	Mark 9	Luke 9	0.941
40	Mark 10	Luke 18	0.941

Other useful comparisons include Psalm 14 and Psalm 53, closely related versions of a psalm with a score of 0.981. As a within-book pair, they are not in the cross-book table. Exodus 20 and Deuteronomy 5, two accounts of the Ten Commandments, score 0.955 and appear in the full table. Jude and 2 Peter 2 share closely related warnings; they are each other's nearest chapter in this dataset, at 0.890, below the top forty.

These parallel passages were not supplied as labels during this analysis. However, we cannot rule out the pretrained model's exposure to biblical texts or discussions of them. Recovering familiar parallels is a useful check, but each additional connection still needs to be examined.

And the least alike

The other end of the scale is as telling. The lowest-scoring chapter pair in this dataset is Esther 10, three verses on the king's tribute and Mordecai's greatness, and Psalm 150, six verses calling everything that breathes to praise: a brief closing narrative and a song of praise. Nearly every pair at this end is the same contrast, Esther against a psalm. Esther 10 is the least alike chapter for 543 of the Bible's 1,189 chapters; Psalm 150 is for 108; Leviticus 13, the laws on skin disease, is for 98. The Bible's language runs from records to praise, and a few very short chapters sit at the far ends of it.

The minor prophets are each other's nearest books: Obadiah, Zephaniah, Micah and Joel share a form and a vocabulary. Esther, whose Hebrew text does not explicitly name God (unlike its Greek additions), has low similarity to the Letters in this analysis⁵ and it has the lowest average chapter similarity to more than half the books in the dataset: 35 of the 66 books have Esther as the book least like them.

Two more measures, both from the same numbers. The chapter with the highest average similarity to the other chapters in this dataset is Deuteronomy 32, the Song of Moses, which includes themes of creation, rescue, rebellion, judgment and mercy. After it come Psalm 78, Jeremiah 17, Psalm 105 and Nehemiah 9, and three of those four are themselves retellings of Israel's story. The chapter with the lowest similarity to its own nearest match is Psalm 133, three verses on brothers living in unity, followed by Esther 10 and Nehemiah 3, the list of who built which stretch of the wall.

The twenty least alike pairs, the most and least typical chapters, and the book pairs

#	Chapter	Least alike chapter	Score
1	Esther 10	Psalms 150	0.147
2	Esther 2	Psalms 150	0.191
3	Esther 10	Psalms 142	0.195
4	Esther 10	Psalms 6	0.196
5	Esther 10	Psalms 43	0.214
6	Esther 10	Psalms 143	0.215
7	Esther 1	Psalms 150	0.217
8	Esther 10	Psalms 88	0.218
9	Esther 10	Psalms 70	0.219
10	Exodus 26	Esther 10	0.220
11	Genesis 5	Acts 25	0.224
12	Esther 10	Psalms 38	0.227
13	2 Chronicles 27	Psalms 142	0.229
14	Psalms 150	Jeremiah 41	0.232
15	Esther 10	Psalms 3	0.233
16	Esther 10	Psalms 23	0.233
17	Esther 10	Psalms 54	0.234
18	Esther 10	Psalms 61	0.235
19	Jeremiah 41	1 Corinthians 13	0.238
20	Esther 10	Psalms 86	0.239

Highest average similarity to other chapters

#	Chapter	Mean similarity
1	Deuteronomy 32	0.746
2	Psalms 78	0.726
3	Jeremiah 17	0.726
4	Psalms 105	0.725
5	Nehemiah 9	0.724
6	Isaiah 41	0.723
7	Amos 5	0.722
8	Isaiah 1	0.721
9	Isaiah 8	0.721
10	Ezekiel 24	0.720

Lowest similarity to a chapter's nearest match

#	Chapter	Its nearest	Score
1	Psalms 133	Proverbs 27	0.653
2	Esther 10	Esther 8	0.698
3	Nehemiah 3	Nehemiah 11	0.731
4	2 Chronicles 26	2 Kings 15	0.734
5	Job 25	Job 15	0.737
6	Jonah 3	Jonah 4	0.750
7	Ruth 1	Ruth 2	0.759
8	Genesis 38	Genesis 30	0.761
9	Revelation 12	Revelation 17	0.761
10	Judges 12	Judges 11	0.761

Books, most alike

Books	Score
Obadiah and Zephaniah	0.868
Obadiah and Micah	0.864
Micah and Zephaniah	0.858
Joel and Zephaniah	0.851
Amos and Obadiah	0.845
Joel and Obadiah	0.844
2 Peter and Jude	0.840
Amos and Zephaniah	0.839

Books, least alike

Books	Score
Esther and Philemon	0.373
Esther and 2 John	0.386
Esther and 1 John	0.388
Esther and 1 Thessalonians	0.400
Esther and 1 Timothy	0.401
Esther and Ephesians	0.401
Esther and Philippians	0.405
Esther and 2 Thessalonians	0.410

Scores are cosine similarities in the 256-dimensional representations, not percentages of shared meaning. Book scores are the average similarity of every chapter pair across the two books. Every figure in this section is in the data file and checked against it; the extremes are recomputed over every pair of chapters.

The surprises

Fifty-eight chapters sit among another section's chapters: six or more of their ten nearest neighbors, in the full measure, belong to a section other than their own. These associations may reflect shared wording, subject matter or literary form. They do not show that a chapter belongs in a different canonical section.

• Revelation 18 sits among the prophets, beside Jeremiah 51, Ezekiel 7 and Isaiah 13. It is the fall of Babylon, written in the manner of the oracles against the nations.
• James 3 sits with the Proverbs and Ecclesiastes. A chapter about the tongue, in short sayings.
• Lamentations 3 sits with the Psalms. It is a lament in the first person, and its nearest chapters are Psalm 69, Job 19 and Psalm 35.
• Deuteronomy 32, the Song of Moses, sits among the prophets, beside Psalm 78, Isaiah 51 and Isaiah 5.
• Genesis 10 and 36, and Matthew 1, all genealogies, sit with the genealogies that open 1 Chronicles.
• Ezekiel 43 to 46 sit with Leviticus and Numbers: the law of the temple reads like the Law. Ezekiel 48, the division of the land, sits with Joshua's.
• Daniel 1 to 6 sit toward the histories and Daniel 7 to 12 beside Revelation. The book is stories in its first half and visions in its second, and the map sees the seam.
• Psalms 2, 87 and 126, and half a dozen more, sit among the prophets: psalms of the king, of Zion and of the nation's return, in the prophets' language.
• Joshua leans on Deuteronomy throughout: chapters 1, 2, 14, 22, 23 and 24 all sit among the Law. Several of these chapters explicitly recall Moses' instructions.
• Leviticus 26 and Deuteronomy 28, the blessings and the curses, sit together among prophetic chapters. Similarity alone does not establish a direct quotation.
• Numbers 23 and 24, Balaam's oracles, sit among the prophets, while the story that carries them stays with the stories.
• 2 Kings 19 and 25 sit among the prophets, beside Isaiah 37 and Jeremiah 52, the chapters of the prophets' books that carry the same events; Jeremiah 41, the murder of Gedaliah, goes the other way and sits with the histories.
• Acts 7, Stephen's recital of Israel's story, sits with the Law: its closest chapters anywhere are Exodus 3 and Exodus 32.
• Hebrews 9 argues from the sacrifices of Leviticus, and this model puts it there: six of its ten nearest chapters are in the Law, and Leviticus 16, the Day of Atonement, is its second-closest chapter in the whole Bible. Hebrews 1, a chain of quotations from the Psalms, sits with the poetry.
• John 17, Jesus' prayer, sits with the Letters, beside Colossians 1; 2 Peter 2 sits with Jude and with Job.

All fifty-eight, with their three nearest chapters

Chapter	Its section	Drawn toward	Neighbors there	Three nearest
Lamentations 3	Prophets	Poetry and Wisdom	10 of 10	Psalms 69, Job 19, Psalms 35
Revelation 18	Revelation	Prophets	10 of 10	Jeremiah 51, Ezekiel 7, Isaiah 13
Genesis 36	Law	History	9 of 10	1 Chronicles 1, 1 Chronicles 4, 1 Chronicles 8
Joshua 23	History	Law	9 of 10	Deuteronomy 7, Deuteronomy 4, Deuteronomy 29
2 Samuel 22	History	Poetry and Wisdom	9 of 10	Psalms 18, Psalms 118, Psalms 22
Psalms 2	Poetry and Wisdom	Prophets	9 of 10	Isaiah 14, Isaiah 41, Jeremiah 25
Psalms 87	Poetry and Wisdom	Prophets	9 of 10	Isaiah 62, Isaiah 66, Micah 4
Jeremiah 41	Prophets	History	9 of 10	Jeremiah 40, Judges 8, 1 Kings 22
Ezekiel 44	Prophets	Law	9 of 10	Numbers 18, Leviticus 21, Leviticus 22
James 3	Letters	Poetry and Wisdom	9 of 10	Proverbs 24, Proverbs 16, Ecclesiastes 7
Numbers 24	Law	Prophets	8 of 10	Numbers 23, Numbers 22, Jeremiah 51
2 Kings 19	History	Prophets	8 of 10	Isaiah 37, Isaiah 36, Isaiah 10
Ezra 9	History	Prophets	8 of 10	Jeremiah 2, Daniel 9, Jeremiah 14
Psalms 126	Poetry and Wisdom	Prophets	8 of 10	Isaiah 49, Jeremiah 31, Isaiah 61
Ezekiel 45	Prophets	Law	8 of 10	Ezekiel 46, Ezekiel 43, Numbers 18
Ezekiel 46	Prophets	Law	8 of 10	Ezekiel 45, Ezekiel 43, Numbers 28
John 17	Gospels and Acts	Letters	8 of 10	John 15, John 14, Colossians 1
Genesis 10	Law	History	7 of 10	1 Chronicles 1, 1 Chronicles 4, 1 Chronicles 6
Genesis 14	Law	History	7 of 10	Deuteronomy 2, Numbers 21, 2 Chronicles 25
Leviticus 26	Law	Prophets	7 of 10	Deuteronomy 28, Ezekiel 5, Ezekiel 20
Numbers 26	Law	History	7 of 10	Numbers 1, Nehemiah 7, 1 Chronicles 6
Deuteronomy 28	Law	Prophets	7 of 10	Leviticus 26, Deuteronomy 7, Jeremiah 15
Deuteronomy 32	Law	Prophets	7 of 10	Psalms 78, Isaiah 51, Isaiah 5
Joshua 1	History	Law	7 of 10	Deuteronomy 31, Joshua 24, Deuteronomy 1
Joshua 22	History	Law	7 of 10	Numbers 32, Deuteronomy 1, Numbers 14
Judges 13	History	Law	7 of 10	Exodus 4, Exodus 34, Numbers 11
Ruth 4	History	Law	7 of 10	Ruth 3, Ruth 2, Genesis 24
2 Kings 22	History	Prophets	7 of 10	2 Chronicles 34, Jeremiah 26, Isaiah 8
1 Chronicles 16	History	Poetry and Wisdom	7 of 10	Psalms 132, 1 Kings 8, Psalms 135
Psalms 50	Poetry and Wisdom	Prophets	7 of 10	Deuteronomy 32, Amos 5, Isaiah 43
Isaiah 64	Prophets	Poetry and Wisdom	7 of 10	Psalms 102, Psalms 74, Psalms 80
Ezekiel 1	Prophets	Revelation	7 of 10	Ezekiel 10, Revelation 4, Daniel 7
Ezekiel 43	Prophets	Law	7 of 10	Ezekiel 46, Ezekiel 45, Ezekiel 44
Matthew 1	Gospels and Acts	History	7 of 10	1 Chronicles 3, 1 Chronicles 1, Genesis 46
Acts 7	Gospels and Acts	Law	7 of 10	Exodus 3, Exodus 32, Numbers 14
Hebrews 1	Letters	Poetry and Wisdom	7 of 10	Hebrews 2, Isaiah 40, Job 36
Genesis 3	Law	Prophets	6 of 10	Genesis 2, Ezekiel 3, Deuteronomy 32
Numbers 23	Law	Prophets	6 of 10	Numbers 24, Numbers 22, Jeremiah 11
Deuteronomy 3	Law	History	6 of 10	Deuteronomy 2, Deuteronomy 1, Joshua 13
Joshua 2	History	Law	6 of 10	Deuteronomy 1, Genesis 42, Numbers 14
Joshua 14	History	Law	6 of 10	Numbers 13, Deuteronomy 1, Joshua 24
Joshua 24	History	Law	6 of 10	Judges 2, Deuteronomy 1, Joshua 7
Judges 1	History	Law	6 of 10	Deuteronomy 2, Numbers 21, Joshua 11
1 Samuel 2	History	Prophets	6 of 10	Deuteronomy 32, 1 Samuel 12, Ezekiel 24
2 Kings 25	History	Prophets	6 of 10	Jeremiah 52, Jeremiah 39, 2 Kings 24
2 Chronicles 4	History	Law	6 of 10	1 Kings 7, Exodus 37, Exodus 38
Psalms 14	Poetry and Wisdom	Prophets	6 of 10	Psalms 53, Psalms 94, Isaiah 1
Psalms 24	Poetry and Wisdom	Prophets	6 of 10	Isaiah 42, Isaiah 60, Isaiah 52
Psalms 53	Poetry and Wisdom	Prophets	6 of 10	Psalms 14, Psalms 94, Psalms 73
Psalms 83	Poetry and Wisdom	Prophets	6 of 10	Psalms 69, Isaiah 1, Isaiah 10
Psalms 110	Poetry and Wisdom	Prophets	6 of 10	Psalms 89, Psalms 29, Jeremiah 17
Ezekiel 48	Prophets	History	6 of 10	Joshua 18, Joshua 19, Numbers 34
Daniel 1	Prophets	History	6 of 10	Daniel 5, Daniel 2, Daniel 6
Daniel 7	Prophets	Revelation	6 of 10	Daniel 8, Revelation 13, Revelation 17
Habakkuk 1	Prophets	Poetry and Wisdom	6 of 10	Isaiah 59, Jeremiah 5, Psalms 73
Zechariah 3	Prophets	History	6 of 10	Joshua 24, Joshua 7, Exodus 33
Hebrews 9	Letters	Law	6 of 10	Hebrews 10, Leviticus 16, Hebrews 8
2 Peter 2	Letters	Poetry and Wisdom	6 of 10	Jude 1, Job 15, Job 34

Do short chapters sit at the extremes?

On the first score, short chapters do not sit at the ends. Among the 119 least typical chapters, six have fewer than ten verses, or 5%, about the share of such chapters overall (71 of 1,189, or 6%). The lowest of all is Esther 10 (three verses) at 0.391, but next comes Genesis 5, a genealogy of 32 verses, at 0.434, and seven of the lowest ten are lists, measurements and detailed laws: that genealogy, the camp in Numbers 2, the laws on skin disease in Leviticus 13, the altar and courtyard in Exodus 27, the temple in Ezekiel 40, 41 and 42. Psalm 117, the shortest chapter at two verses, sits at 0.518, the 50th lowest of 1,189. At the top, Deuteronomy 32, 52 verses, is the most typical chapter at 0.746.

On the second score, they do. Psalm 133 (three verses, 0.653) and Esther 10 (three verses, 0.698) have the two weakest nearest matches in this measure, and chapters of fewer than ten verses have a median nearest match of 0.861 against 0.914 for chapters of 40 verses or more; the best-matched chapter in this measure is Isaiah 37 (0.993, to 2 Kings 19). A chapter of a few verses is a smaller sample of language, and that may contribute to its weaker nearest match; the scores alone do not establish why. The same short chapters turn up at the ends of the pairwise lists (the least alike pairs, the loners) rather than at the ends of the typicality list.

The ten shortest chapters, the ten most typical and the ten loners

The ten shortest chapters

Chapter	Verses	Average similarity	Nearest match	Nearest chapter
Psalm 117	2	0.518	0.819	Psalm 96
Esther 10	3	0.391	0.698	Esther 8
Psalm 133	3	0.515	0.653	Proverbs 27
Psalm 131	3	0.575	0.788	Psalm 102
Psalm 134	3	0.606	0.806	Psalm 68
Psalm 123	4	0.588	0.822	Psalm 86
Jeremiah 45	5	0.554	0.782	Jeremiah 18
Psalm 93	5	0.559	0.863	Psalm 29
Psalm 70	5	0.579	0.899	Psalm 40
Psalm 43	5	0.579	0.864	Psalm 71

The ten most typical chapters

Chapter	Verses	Average similarity	Nearest match	Nearest chapter
Deuteronomy 32	52	0.746	0.920	Psalm 78
Psalm 78	72	0.726	0.929	Psalm 105
Jeremiah 17	27	0.726	0.891	Isaiah 65
Psalm 105	45	0.725	0.929	Psalm 78
Nehemiah 9	38	0.724	0.923	Psalm 106
Isaiah 41	29	0.723	0.922	Isaiah 40
Amos 5	27	0.722	0.918	Isaiah 1
Isaiah 1	31	0.721	0.927	Jeremiah 6
Isaiah 8	22	0.721	0.884	Isaiah 30
Ezekiel 24	27	0.720	0.899	Jeremiah 9

The ten loners: the weakest nearest matches

Chapter	Verses	Average similarity	Nearest match	Nearest chapter
Psalm 133	3	0.515	0.653	Proverbs 27
Esther 10	3	0.391	0.698	Esther 8
Nehemiah 3	32	0.482	0.731	Nehemiah 11
2 Chronicles 26	23	0.494	0.734	2 Kings 15
Job 25	6	0.565	0.737	Job 15
Jonah 3	10	0.565	0.750	Jonah 4
Ruth 1	22	0.556	0.759	Ruth 2
Genesis 38	30	0.545	0.761	Genesis 30
Revelation 12	17	0.569	0.761	Revelation 17
Judges 12	15	0.519	0.761	Judges 11

Where a book changes

• Law
• History
• Poetry and Wisdom
• Prophets
• Gospels and Acts
• Letters
• Revelation

1. Esther 10 to Job 1, 0.403 (book boundary)
2. Ezekiel 39 to 40, 0.478
3. Ezekiel 48 to Daniel 1, 0.489 (book boundary)
4. Numbers 29 to 30, 0.498
5. Ezra 2 to 3, 0.508
6. Psalm 87 to 88, 0.518
7. Malachi 4 to Matthew 1, 0.521 (book boundary)
8. Judges 16 to 17, 0.530
9. Judges 12 to 13, 0.532
10. Psalm 150 to Proverbs 1, 0.533 (book boundary)
11. A Daniel 6 to 7, 0.629 (seam, 47th deepest)
12. B Isaiah 39 to 40, 0.644 (seam, 61st deepest)
13. C Revelation 3 to 4, 0.669 (seam, 84th deepest)
14. D Job 2 to 3, 0.680 (seam, 100th deepest)

Read straight through, the similarity from one chapter to the next drops furthest at a handful of book boundaries and at a few places inside books. Of the 1,188 steps, 65 cross from one book to the next, and four of them are among the ten deepest: Esther 10 to Job 1 (0.403), Ezekiel 48 to Daniel 1 (0.489), Malachi 4 to Matthew 1 (0.521), Psalm 150 to Proverbs 1 (0.533). The step from Malachi 4 to Matthew 1, where the testaments meet, is the seventh deepest at 0.521. The deepest step inside a book is Ezekiel 39 to 40 at 0.478, where the oracles end and the vision of the temple begins.

The seams listed below are mostly not single-step drops. Five of the twelve seams are among the 100 deepest steps: Ezekiel (second), Daniel (47th), Isaiah (61st), Revelation (84th), Job (100th). The other seven are shallower; the shallowest, Zechariah 8 to 9, at 0.860, is above the average step and only the 893rd deepest of 1,188. A seam compares the chapters before it, as a group, with the chapters after it; a step compares one chapter with the next. A book can shift gradually over several chapters and still show a seam on the map.

The twenty deepest steps

#	From	To	Score	Where
1	Esther 10	Job 1	0.403	book boundary
2	Ezekiel 39	Ezekiel 40	0.478	a seam of section 07
3	Ezekiel 48	Daniel 1	0.489	book boundary
4	Numbers 29	Numbers 30	0.498	inside the book
5	Ezra 2	Ezra 3	0.508	inside the book
6	Psalm 87	Psalm 88	0.518	inside the book
7	Malachi 4	Matthew 1	0.521	book boundary
8	Judges 16	Judges 17	0.530	inside the book
9	Judges 12	Judges 13	0.532	inside the book
10	Psalm 150	Proverbs 1	0.533	book boundary
11	Ezra 1	Ezra 2	0.538	inside the book
12	Nehemiah 6	Nehemiah 7	0.540	inside the book
13	Lamentations 5	Ezekiel 1	0.548	book boundary
14	Psalm 133	Psalm 134	0.549	inside the book
15	Ezekiel 1	Ezekiel 2	0.553	inside the book
16	Ezra 3	Ezra 4	0.559	inside the book
17	1 Chronicles 9	1 Chronicles 10	0.559	inside the book
18	Philemon 1	Hebrews 1	0.561	book boundary
19	Jeremiah 41	Jeremiah 42	0.569	inside the book
20	2 Samuel 21	2 Samuel 22	0.571	inside the book

The following selected divisions between chapters correspond to separation in this layout. They were supplied for comparison, not discovered automatically, and do not establish when or by whom a book was written. For each of these, the chapters before the seam and the chapters after it sit as two groups on the map, further apart than either group is spread:

• Isaiah, between 39 and 40: judgment on Jerusalem and the nations, then comfort for a people in exile. Readers have noticed the change of voice for a very long time.
• Exodus, between 18 and 19: the story of the escape from Egypt, then the law at Sinai.
• Leviticus, between 16 and 17: the sacrifices and the Day of Atonement, then the laws of holiness.
• Ezekiel, between 39 and 40: the oracles, then the vision of the temple, whose measurements and land divisions sit with Numbers and Joshua.
• Daniel, between 6 and 7: the stories, then the visions.
• Zechariah, between 8 and 9: visions and messages about restoration, followed by further prophetic oracles.
• Job, between 2 and 3: the opening narrative about Job's suffering, then the poetry.
• Jeremiah, between 25 and 26.
• Proverbs, between 9 and 10: the long instructions, then the short sayings.
• Acts, between 12 and 13: Peter and Jerusalem, then Paul and the journeys.
• Romans, between 11 and 12: Paul's theological argument, followed by an extended section of practical exhortation.
• Revelation, between 3 and 4: the letters to the churches, then the visions.

Genesis does not split so much as scatter: its first eleven chapters, the creation, the flood and the genealogies, spread across the map, and the rest, the family stories, sit together. Choose any of these books under the map to see its seam for yourself.

What it shows, taken together

These studies explore similarities in the language of selected English Bible editions. Together, they offer several starting points for reading and comparison.

The groupings depend on the method. In this model's seven-group clustering, one group contains about 88% of the New Testament's chapters. Other groups bring together passages from different Old Testament sections. These are patterns in the model's representations, not a measure of theological difference.

Similar forms of writing can connect different books. Fifty-eight chapters meet the study's definition of an unexpected neighborhood. Examples include Matthew 1, which begins with a genealogy, and Exodus 15, which includes songs. The linked passages let you examine those connections in context.

Books can contain several forms of writing. The twelve examples above compare changes in the layout with changes in the text. A visual division can suggest a passage to examine; it does not establish a literary structure or a book's composition history.

Some comparisons are sensitive to chapter length. Esther 10 and Psalm 150 are the least similar pair under this measure. Deuteronomy 32 has the highest average similarity to the other chapters. Neither result makes a chapter more representative of biblical teaching or more spiritually important.

Many close parallels persist across translations. The forty closest cross-book pairs on the main map remain within each other's ten nearest chapters in the four other editions tested. Other neighboring matches change. This comparison tests chapter similarities, not whether separately drawn maps would have the same shape.

A weak search result does not establish absence. In the topic-search experiment, 921 of 1,001 titles score below the search threshold. That includes 200 of the 271 titles with literal word matches in the text. These results describe this model and these queries, not what the Bible does or does not discuss.

The sections are a human overlay. Changing them does not move a dot, and the model's own groups follow them in places and cross them in others.

Results also depend on the model. An earlier version used BGE; the current version uses EmbeddingGemma. Development comparisons retained several familiar parallels but changed other relationships, both within and between the testaments. This was not a comprehensive test of stability across models.

The companion studies explain the translation comparisons, explore books beyond the main map's Protestant canon, examine the limits of topic searches, show twelve topic searches chapter by chapter, and list the words most characteristic of the chapters the model grouped together. They do not rank traditions, canons or translations by religious value.

What the map cannot tell you

• It reflects one translation. The map was measured from the Berean Standard Bible. A companion page, Does the translation change the map?, compares the King James Version and three other editions with the same model. Many close chapter parallels persist, while other neighboring matches change.
• It reflects one model. EmbeddingGemma was trained on multilingual web, technical and task-specific data. It measures similarity of language as it understands it, which is not the same as similarity of meaning as a careful reader understands it.
• Distance on the drawn map is approximate. The underlying similarity scores are calculated from the 256-dimensional representations, and the lists above use that. The map is a shadow, and some chapters that are close in the full measure sit apart on the map. Turn it before you conclude anything from where two dots fall.
• Chapters blur short parallels. A chapter's representation is the normalized average of its overlapping passage representations. Micah 4 and Isaiah 2 share three verses almost word for word, but at chapter level Isaiah 2 ranks twenty-third among Micah 4's matches, while Micah 4 ranks twentieth among Isaiah 2's matches, because the rest of both chapters pulls the averages apart. A verse-level comparison could help examine shorter parallels, though it would have its own limitations.
• Form can influence similarity. Genealogies sit with genealogies, lists with lists, and a song in the middle of a story sits with the songs. That is often illuminating and occasionally misleading; the surprises list is where to see both.
• Least alike is a lopsided measure. A handful of extreme chapters, mostly very short ones, are the least alike chapter for hundreds of others. That is why the book explorer compares a chapter with its own book for least alike, and why the extremes are shown once, as the ends of the scale, rather than for every chapter.
• Nearness is not agreement. The layout uses similarities between model-generated representations. Its rankings do not measure spiritual importance or settle interpretation.
• Placing your own text is an estimate. The Your own text control scores a passage against the 1,189 chapter representations and marks the similarity-weighted center of its ten nearest chapters on the drawn map. The scores are similarities between the model's representations, not a judgment of the text or of its faithfulness to Scripture, and the marker's position comes from those ten chapters, not from a new layout. Read the ten chapters before drawing a conclusion from where the diamond falls.

How this map was made

To cite a figure elsewhere, say which measure it is (a score in the full 256 numbers, or a position on the drawn map) and carry the translation and the model with it. Corrections are welcome through the contact page.

Sources

1. The Berean Standard Bible, released into the public domain, read on this site at /bible/bsb/.
2. Leland McInnes, John Healy and James Melville, UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction, 2018.
3. Google, EmbeddingGemma (embeddinggemma-300m), the sentence embedding model, its model card, and the Gemma Terms of Use it is released under; run locally for the corpus and on Workers AI for search queries.
4. USCCB, Introduction to Esther, on the Hebrew text and the Greek additions.
5. The chapter count of 1,189 follows the standard division of the Protestant canon; see Bible Stats for how this site's editions are counted.

Questions people ask

Is this a statement about what the Bible means?

No. This is an exploratory comparison of language using one translation and one model. Its similarity rankings do not measure spiritual importance, establish theological agreement or replace interpretation in context.

Why does one chapter sit next to another?

The layout places chapters using similarities between their model-generated representations. Nearby dots suggest passages to compare, but the drawing is approximate. Read both chapters and consult their similarity scores rather than relying only on their positions.

Would another translation give the same map?

The companion study tests chapter similarities, not new layouts. The forty closest cross-book pairs remain within each other's ten nearest chapters in all four other editions tested, but other neighboring matches change. This does not establish that separately drawn maps would have the same shape.

Why chapters rather than verses?

Chapters keep the overview manageable: this dataset has 1,189 chapter points. However, averaging passages across a chapter can obscure shorter parallels. The chapter links let you read those passages in context.

Can I use the coordinates?

Yes. The method section links to the downloadable coordinates, computed results and settings. Please identify the translation and model when reusing them. The source Bible text is public domain; a link back to this page is welcome.

Can I see where my own text lands on the map?

Yes. In the map's controls, the Your own text control takes up to 600 characters. The site's search model turns the text into the same kind of numbers as the chapters, finds the ten chapters whose representations are most similar, and marks the similarity-weighted center of those ten on the drawn map. The map itself is not recomputed, and the scores measure similarity between representations, not the worth or accuracy of the text.