This book counts. On almost every page there is a number, and behind every number lies a search through a text file. Whoever writes like that owes the reader two things: to say exactly what has been counted, and to say what has not been checked. This chapter does both.
It is also the place to which chapter 1 and chapter 10 refer, both for the same question: is the closing line of the prayer really there? Whoever arrives here from there will find it under §3.
Behind this chapter there are two more pages. The bibliography gives everything this book owes to others, divided into three according to what has and has not been consulted; the index gives the Scripture references, the measured words, the Names, and the people who walk through the chapters.
1. How this book measures
The three marks
Behind every claim in this book stands, written out or in a short note, one of three marks. They do not say how certain something is; they say what kind of claim it is.
mark
meaning
CALCULATED
Counted by machine over a text file. Can be recounted by anyone who has the same file. "ἁγιασθήτω occurs three times in the New Testament" is CALCULATED.
DOCUMENTED
It is there, and it can be pointed to. A verse, a form, a sequence. "In Genesis 22 here I am occurs three times" is DOCUMENTED.
CONSTRUED
My reading. The text does not say it itself. That Babel and Bethel mirror each other is CONSTRUED — Genesis puts no cross-reference between them.
Nowhere does this book claim a proof. What there is, is convergence: different counts pointing the same way. Convergence is not proof, and it does not need to be in order to be worth something.
Five rules I have imposed on myself
- One. A count in the running text is a mistake, even when it is right. Numbers belong in a box or in a note. The main text is for people. (The prologue is the exception: it has no boxes, and the two numbers in it therefore carry the clause that says what was counted.) - Two. Every number gets its denominator. "Eight times" without "in 7,000 verses" is not a measurement but an impression. Where the population is small, that is stated. - Three. With a striking number: work it through, do not hush it up. A hit that sounds suspicious is neither left out nor celebrated — it is tested under every other, equally defensible way of counting. If it holds, it may be mentioned with caution. If it does not hold, it comes here, to §4. - Four. The reading time is counted too. The minutes above each chapter are not an estimate but the number of words in that chapter divided by 240 — and that is why they range from 10 to 24. In a book that counts everything, that number ought to be right as well. - Five. Every dossier ends with what I have not seen. Not as a formula of modesty but because it is the only way a reader can weigh what he reads. The standing register for this whole book is in §5.
And the measure that applies on both sides
Every measure I apply to my own findings, I apply in the same breath to the prevailing view. Where this book says "the majority of textual critics read it differently", that is a statement about scholarship — not a null hypothesis that is right by default, and not an opponent that is wrong by default either. Where I have not consulted their work myself, that is stated. This happens in three places in this book, and they are listed in §6.
2. The textual basis
text
extent
Hebrew and Aramaic
Masoretic text, with morphological coding and Strong's numbers
306,785 words
Greek, New Testament
Textus Receptus, Stephanus 1550, with parsing
140,987 words
Greek, Old Testament
Septuagint, including the deuterocanonical books and the double text forms
623,693 words
In this file the Masoretic text has 929 chapters and 23,213 verses; the New Testament 7,957 verses. Of the 306,785 Hebrew words, 4,948 are Aramaic, in 269 verses: Daniel 2:4b-7:28, Ezra 4:8-6:18 and 7:12-26, Jeremiah 10:11 and two words in Genesis 31:47. (The four Hebrew words with which Daniel 2:4 opens and the seven Hebrew words of Genesis 31:47 are not counted; whoever includes the whole verse arrives at 4,959.)
What that means for the counts
Tokens versus verses. Where this book writes "occurs eight times", it counts words, not verses. Where the difference matters — because a verse has a word twice — it is stated: "nine tokens in seven verses". Whoever counts verses where this book counts words arrives at a different number, and both are correct.
Versification. The Masoretic text, the Septuagint and the English tradition diverge. The Psalms usually differ by one number; Isaiah 64 begins a verse earlier; Joel has three chapters where the Masoretic text counts four; Daniel 3:31-33 is English 4:1-3; Nehemiah appears in the Septuagint as 2 Esdras 11-23. This edition follows the English numbering and adds the other where it differs, with `[MT]` or `[LXX]`.
What is not there. There is no textual-critical apparatus on the disk of this project. No manuscript variants, no lists of witnesses, no collation between editions. Every statement in this book about what the manuscripts read is therefore taken over and not recalculated — and it is marked as such. See §3 and §6.
3. The closing line of Matthew 6:13
Chapter 1 (note 2) and chapter 10 refer here.
In this book the prayer ends with: "For thine is the kingdom, and the power, and the glory, for ever. Amen." Many modern translations do not have that line, or put it in brackets. This is the account of that choice — and of its limit.
What could be measured on the book's own disk, and has been measured
- One. The text used has the line. In the Textus Receptus Matthew 6:13 stands in full: ὅτι σοῦ ἐστιν ἡ βασιλεία καὶ ἡ δύναμις καὶ ἡ δόξα εἰς τοὺς αἰῶνας. ἀμήν.CALCULATED. - Two. Luke does not have it. Luke 11:4 ends at ἀπὸ τοῦ πονηροῦ, "from the evil one". CALCULATED — and this is the only datum in the whole question that is independent of which family of manuscripts one follows. Whoever reads the shorter Matthew and whoever reads the longer one read the same in Luke. - Three. The closing formula is a singleton in Matthew.εἰς τοὺς αἰῶνας occurs 28 times in the text used, and in Matthew exactly once: here. CALCULATED. - Four. The closing amen is a singleton in Matthew. Of the 106 occurrences of ἀμήν in the four Gospels, one hundred and one open a saying — amen, I say to you. Five close something, and four of those five are the final verses of the four Gospels. The fifth is Matthew 6:13.CALCULATED.
What could not be measured, and so has not been measured
The manuscripts. The line is missing in part of the oldest Greek manuscripts and present in the majority of the later ones; the majority of textual critics regard it as a later, liturgical addition. That is the state of scholarship and it may simply be stated here — but I have not recalculated it. Neither NA28 nor Tischendorf nor the ECM has been consulted; no textual-critical apparatus was consulted for this book. Every sentence in this book of the form "the oldest manuscripts read [...]" therefore rests on what others have reported. That is why it appears nowhere in this book as a finding, and here as a limitation.
Where the line came from. On that, nothing has been measured here and nothing is claimed here.
Why the book keeps the longer reading, and what that is and is not
It is an edition choice, not a textual-critical judgment. This book reads the Textus Receptus — the same text on which the counts of all 19 chapters rest. A book that counts on one text and quotes from another no longer counts.
And there is a methodological ground that I can account for. The prevailing explanatory model — that the text of the prayer grew along the way through liturgical use — explains both the short reading and the long reading: the short one because it would be original, the long one because the church would have supplemented it in worship. Chase himself calls that model "the theory" in 1891, and applies it to the closing line and to the bread petition — the two most difficult places in the prayer. An explanatory scheme that can absorb every outcome can be refuted by no outcome. This book demands a falsifier on both sides or on neither, and here it applies that measure to itself just as well: "the line belongs there" is not a finding in this book either. It is the edition on which the counting is done.
And what hangs on it in substance: nothing
The closing line says nothing that Scripture does not say elsewhere. Paul closes his last letter with the seventh petition and the doxology in one verse — "the Lord will deliver me from every evil work [...] to Him be the glory for ever and ever, amen" (2 Timothy 4:18; so also Romans 11:36). And the three words themselves are already found in David, in the prayer with which chapter 10 opens: "Yours, YHWH, is the greatness and the power [...] Yours, YHWH, is the kingdom" (1 Chronicles 29:11). Whoever leaves the line out loses not a single thought of this book. Whoever reads it prays after a king who gives back his crown.
One second textual question, and it is here because it touches a chapter
Luke 23:42 — the request of the man on the cross. The text used reads ἐν τῇ βασιλείᾳ σου, "in your kingdom"; other witnesses read εἰς τὴν βασιλείαν σου, "into your kingdom". Chapter 13 hangs on that verse and keeps the first reading, because this book counts on that text. The difference does not affect the finding of that chapter: what matters there are the two words your kingdom, and they are present in both readings.
4. Four counts that were tested and fell
The rule of §1 says: a number that does not hold comes here. These are they. They are not here out of modesty but because a book that shows only its successful counts tells you nothing about how strictly it measures.
4a. 666
The Hebrew word for "evil", רַע, enters Scripture as half of the name of a tree: the tree of the knowledge of good and evil (Genesis 2:9). When that word was recounted over the whole Masoretic text, a number came out that demanded attention.
The raw result is correct. Strong's number H7451 counts 666 words.
And it does not hold, and the reason can be pinpointed exactly:
way of counting
result
H7451, all words, as Strong delimits the lemma
666
of which: the adjective רַע
356
of which: the feminine noun רָעָה
306
of which: the masculine noun
3
of which: one form coded as a verb (Isaiah 3:11)
1
the adjective plus the masculine noun
359
H7451 together with H7455 (רֹעַ, same root, different lemma)
685
unique verses instead of words
623
The 666 is the sum of two different words under one number. Strong merges רַע and רָעָה; other lexicons separate them. Whoever asks "how often does רַע occur?" and means the word רַע gets 356.
And the number is breakable in three other ways besides: removing that one disputed coding gives 665; adding the third form of the same root gives 685; verses instead of words gives 623.
Conclusion: it is not a finding but a property of how a dictionary delimits its lemmas. Four other, equally defensible counts give a different number, and the most obvious one gives 356. That is why it appears nowhere in this book — not cautiously, not in a note. Not because numbers are not allowed; this book counts on every page. But because this count does not stand up under another, equally defensible division. An instrument that finds something everywhere measures nothing.
(Worked through on the explicit instruction of the author, who rightly rejected hushing it up: "on points like this I would rather have a double-check than silence". All the counts were run twice, in two independent searches.)
4b. "Scripture is permeated by this prayer" — measured, and it does not hold
This is the sentence that rises to the surface by itself in a book like this one, and it has been knocked down by the book's own instrument.
The test. For each of the seven petitions one Hebrew key word was taken — hallow (קדשׁ) · kingdom (מַלְכוּת and the Aramaic מַלְכוּ) · will (רָצוֹן, חֵפֶץ) · bread (לֶחֶם) · forgive (סלח) · test (נסה) · deliver (נצל). Then for each chapter of the Masoretic text it was counted how many of the seven occur there.
What comes out:
how many of the seven petitions
chapters
none
471
one
347
two
99
three
11
four
1(read strictly, zero — see the remark on Nehemiah 9 below)
five, six or seven
0
And then the comparison with chance. The positions of the same numbers of words were redistributed at random over the text 2,000 times, with the chapter lengths intact. What pure chance would produce: on average 70.6 chapters with three or more, and 14.6 with four or more. Observed: 12 with three or more, and one with four.
The probability that chance produces at least as much co-occurrence as Scripture itself is 1.0000. The seven petitions do not come together more often than this randomness — they come together significantly less often.
⚠ And then the objection to the book's own test, because it is justified and it is weighty. This null hypothesis is blind to what a word is. Words do not lie at random in a text; every word keeps to its own literature — hallow in Leviticus and Ezekiel, kingdom in Esther, Daniel and Chronicles, bread in the sacrificial laws and the narratives. Under such a null hypothesis virtually any set of seven Hebrew words clusters less than chance. So the 1.0000 says something about the null hypothesis and nothing about this prayer.
That is why the test was redone with a fairer comparison: 200 random sets of seven real Hebrew words, drawn from the 863 lemmas that occur as often as the seven key words. That changes the result, and it changes it in the other direction.
chapters with three or more
with four or more
the seven petitions
12
1
200 random sets of seven, average
7.3
0.64
same, median
6
0
Compared with real words, the seven petitions cluster not less but somewhat more than average: 84 of every 100 random sets of seven come together less often than they do. And that is still not a finding. 84 percent is the upper half, not the exception; one in six random sets of seven does just as well or better. The conclusion stands, but it stands on a different ground from the one it first stood on.
So what may not be said: that the petitions "cluster", that Scripture is "permeated" by this prayer in the sense of concentrated, that there are chapters where the prayer "lies ready". Every sentence of that form has been refuted by the book's own measurement — not because the seven repel one another, but because they do nothing that random words do not also do.
⚠ And one more limit, which does not reverse the result but does restrict it. For the fifth petition the test takes סלח and for the seventh נצל: the general words. But chapter 7 does not turn on סלח and chapter 9 does not turn on נצל — they turn on דְּרוֹר, the jubileejubileeThe fiftieth year in which debt is released and inheritance returns — a rhythm within the master clock., and on the Aramaic שֵׁיזִב. With eight and nine occurrences these two are too rare to mean anything in a cluster test, and an exhaustion of nine cannot be measured by density anyway. What has been refuted here is the density claim about the general vocabulary of the prayer. About the rare words on which those chapters actually stand, this test says nothing.
What does stand, and it is what this book stands on: not densities but exhaustions. Not "this word occurs often here" but "this word occurs n times in the whole of Scripture, and this is the only place" — n = 1, 2, 3, 4. Bread and testing stand in one single verse of the whole Hebrew Bible. That is not a density. That is a count that has run out.
And one thing from the test is too beautiful to keep quiet. The only chapter in the whole Masoretic text where four of the seven stand together is Nehemiah 9 — kingdom, will, bread and deliverance. Nehemiah 9 is itself a prayer: the long prayer of the Levites at the return, spoken aloud, in which the whole history is retraced — "bread from heaven You gave them" (9:15), "You delivered them" (9:28).
And then the honesty that goes with it, because it belongs with it. The word for "will" occurs there twice (9:24 and 9:37), and both times it is about the will of people: what the conquerors did with the land as it pleased them, and what the foreign kings do with the people as it pleases them. It is not the third petition. Whoever reads the test strictly is left with three petitions in this one chapter, and then there is not a single chapter in the whole Masoretic text with four.
(The count is CALCULATED; that it is a prayer of all things is CONSTRUED — and it is one chapter in 929, which is exactly the weight it may carry.)
4c. Two smaller ones, and why they appear nowhere
The Our Father three times a day, and Psalm 145 three times a day. The Didache prescribes the prayer three times a day; Jewish tradition prays Psalm 145 three times a day. Striking — and it is liturgical reception, not textual data, and the datings are not fixed. It carries nothing until it is dated, and it is not dated.
The Almighty in 1 Chronicles 29:12. There the Septuagint has παντοκράτωρ, "Almighty", in David's prayer — and the Hebrew does not have it. It is a Greek surplus. Usable only if it is explicitly called that, and therefore not used in chapter 10.
5. What has not been checked in this book
The standing register. Whoever relies on what stands above must read this too.
1. No manuscript has been consulted. No facsimile, no apparatus, no collation. Everything this book says about manuscripts has been taken over.
2. The coding of the text files has not been independently verified. The parsings and the Strong's numbers come from the files used. Where they make a mistake, their mistake counts as mine. That is not theoretical: the file misses Revelation 2:7 in the lemma list of paradise — chapter 13 recounts there by hand — and it codes Isaiah 3:11 as a verb where one can also read it as an adjective; that one word shifts a count from 666 to 665 (§4a). And there is a third kind of trap, which lies not in the file but in Hebrew: אֵל ("God") and אֶל ("to") are the same word without vowel points. Whoever searches on consonants for אֵל שַׁדַּי gets 11 verses and has three too many. Chapter 2 sets those three apart by name.
3. For the Septuagint there is no second file. For the New Testament a count can be held against a second text; for the Septuagint it cannot. That is the weakest side of this book's tools, and the Greek lemmas of the Old Testament are demonstrably less reliable there than the Hebrew ones.
4. The translations are mine. 138 quoted passages, and over 500 references: all from the original text, concordant and modern, no existing translation taken over. That makes them refutable, and that is intended: the original text stands alongside each time. A second pair of eyes has not yet been on them.
5. All counts were run twice, and the second time found 14 wrong. After the 19 chapters were written, every claim with a number was lifted out of the running text once more and measured again — 162 claims. 144 reproduced. 14 did not, and they have been corrected.12 of those 14 were in chapters 1 through 9; the other two in ch. 14 and ch. 18. Four of the 14 were not a calculation error but a missed occurrence, and those four have made the book better than it was. In addition, four counts are right only under a criterion the reader could not have guessed; there the criterion is now stated. 144 + 14 + 4 = 162.
⚠ And the limit of that audit belongs with it in the same breath, because it is real. From the running text 493 sentences containing a number and a counting unit were extracted by machine. Of those, 162 could be checked against the corpus; the remaining 331 count events, persons or distances within a narrative and have not been systematically retested. And the sieve itself is not perfect: a count hidden in words without a counting unit does not get caught. "Every count run again" therefore means: every count that could be found by machine and tested by machine. A second review, three days later, found four more errors in this way that the sieve had missed — among them a hitpael that occurs four times where the book said twice.
the audit
number
sentences from the running text with a number and a counting unit
493
of which checkable against the corpus
162
· reproduced
144
· did not, and corrected
14
· right only under a criterion the reader could not have guessed
4
of which not systematically retested (events, persons, distances)
331
second review, three days later: errors the sieve had missed
4
6. The structural question has been parked. Whether Matthew 6:9-13 has seven petitions or six, and where the hinge lies, is a question this book does not address. It keeps seven, counts them, and builds no argument on it.
7. What is said about the Kaddish dates nothing. Chapter 3 establishes that the Kaddish points to Ezekiel 38:23 and Daniel 2:20, and no more. Which is older, the prayer or the Kaddish, is not stated here and cannot be stated here.
6. Sources
What this book owes to others is stated in the place where it is used. This is the list, in three kinds — and the distinction is not politeness but the core of the method.
Primary reading — the text itself consulted
Raymond E. Brown, The Pater Noster as an Eschatological Prayer (1961)
The eschatological reading, petition by petition, is his. Seven of the connections this book uses are already in his work, and that is stated each time. He sets his own fence on p. 175: what the prayer meant when Jesus spoke it before his death "lies beyond the scope of our investigation". This book begins where that sentence stops.
Werner Grimm, Die Verkündigung Jesu und Deuterojesaja (1976)
His Tübingen dissertation, read in full. ⚠ This is not the book in which he lays the Our Father beside Isaiah 63-64 — that is another work of his, Die Motive Jesu: Das Vaterunser (1992), and that has not been consulted; it is listed below under "reported". Searched by machine, this 1976 book gave zero hits on "Vaterunser", "Herrengebet" and on Isaiah 63:15-16 and 64:7-8 [MT]. What it did yield is in two notes of this book: the aorist of Luke 12:32 as the grammar of a salvation oracle (p. 156, used in ch. 13) and 11QMelchizedek11QMelchizedek— Qumran text (±100 BC) that reads Daniel’s 490 as 10 jubilees, culminating in an end-time release by a heavenly Redeemer on the Great Day of Atonement. Strong as a witness to an ancient reading, weak (fragmentary, reconstructed) as a bearer of a precise number — no date on a hook (Appendix D §4). as a fusion of Leviticus 25:10 and Isaiah 61:1b (p. ~110, used in ch. 7).
Siddur Ashkenaz, Kaddish, Aramaic text
For chapter 3.
Reported — not consulted personally, and marked as such
The textual-critical apparatuses (NA28, Tischendorf, ECM)
Not consulted. Every statement about the distribution of manuscripts in this book comes from elsewhere. §3 and §5.1.
Ernst Lohmeyer, Das Vater-unser (1946)
The eschatological reading goes back to him; this book knows it through Brown.
Werner Grimm, Die Motive Jesu: Das Vaterunser (1992)
Not consulted. That in it he lays the Our Father beside Isaiah 63:15-64:11 [MT] — the comparison on which chapter 1 rests — is known from John Yieh's bibliography and not from the book itself. Chapter 1 therefore attributes the comparison to him and marks that attribution as [secondary]. This is the open item that would affect the manuscript most if it stays open: if 1992 says something different or something more on this point, the share this book claims for itself changes — not the measurement beneath it.
Jeffrey B. Gibson, "Matthew 6:9-13 // Luke 11:2-4: An Eschatological Prayer?" (Biblical Theology Bulletin 31, 2001) and The Disciples' Prayer (2015)
The direct contradiction of the reading in which this book stands, and chapter 8 gives it an Excursus of its own. His own summary of the article has been read word for word; the article itself and the book have not. What the Excursus says about his argument is therefore taken over and not weighed — and that is why the book refutes him nowhere, it only says why it reads the other way.
Frederic Henry Chase, The Lord's Prayer in the Early Church (1891)
The liturgical explanatory model and his own word "the theory" (§3) come from the working notes of this project; in the last round I did not open Chase again myself.
Samuel Sandmel, Parallelomania (JBL 81, 1962)
The charge has been looked up and applied to this book itself; the article has not been read in full. It has yielded one rule that applies everywhere: no line without an exhaustive count and a base rate.
Not opened, and on the wish list
Betz (1995) and Schnurr (1985) for early Christian interpretation. They are known through others and therefore not used. (Gibson was also here until 26 August 2026; he has moved to "reported" above, and chapter 8 now has an Excursus on him.)
The author's own work
Where this book takes up something from one of the 16 earlier books, the location is in the note with the word "self-quotation". That is not modesty but precision: a finding made 16 years ago is not a finding of this book, and whoever keeps quiet about that makes his new counts unreliable. In six places it turned out on checking afterwards that a line I took to be new was already in my own work. All six are now attributed.
Whoever finds something here that is not right: it is not right, and I would like to hear it. The counts in this book are written down in such a way that you can recalculate them. That is the intention.
This book counts. On almost every page there is a number, and behind every number lies a search through a text file. Whoever writes like that owes the reader two things: to say exactly what has been counted, and to say what has not been checked. This chapter does both.
It is also the place to which chapter 1 and chapter 10 refer, both for the same question: is the closing line of the prayer really there? Whoever arrives here from there will find it under §3.
Behind this chapter there are two more pages. The bibliography gives everything this book owes to others, divided into three according to what has and has not been consulted; the index gives the Scripture references, the measured words, the Names, and the people who walk through the chapters.
1. How this book measures
The three marks
Behind every claim in this book stands, written out or in a short note, one of three marks. They do not say how certain something is; they say what kind of claim it is.
Nowhere does this book claim a proof. What there is, is convergence: different counts pointing the same way. Convergence is not proof, and it does not need to be in order to be worth something.
Five rules I have imposed on myself
- One. A count in the running text is a mistake, even when it is right. Numbers belong in a box or in a note. The main text is for people. (The prologue is the exception: it has no boxes, and the two numbers in it therefore carry the clause that says what was counted.) - Two. Every number gets its denominator. "Eight times" without "in 7,000 verses" is not a measurement but an impression. Where the population is small, that is stated. - Three. With a striking number: work it through, do not hush it up. A hit that sounds suspicious is neither left out nor celebrated — it is tested under every other, equally defensible way of counting. If it holds, it may be mentioned with caution. If it does not hold, it comes here, to §4. - Four. The reading time is counted too. The minutes above each chapter are not an estimate but the number of words in that chapter divided by 240 — and that is why they range from 10 to 24. In a book that counts everything, that number ought to be right as well. - Five. Every dossier ends with what I have not seen. Not as a formula of modesty but because it is the only way a reader can weigh what he reads. The standing register for this whole book is in §5.
And the measure that applies on both sides
Every measure I apply to my own findings, I apply in the same breath to the prevailing view. Where this book says "the majority of textual critics read it differently", that is a statement about scholarship — not a null hypothesis that is right by default, and not an opponent that is wrong by default either. Where I have not consulted their work myself, that is stated. This happens in three places in this book, and they are listed in §6.
2. The textual basis
In this file the Masoretic text has 929 chapters and 23,213 verses; the New Testament 7,957 verses. Of the 306,785 Hebrew words, 4,948 are Aramaic, in 269 verses: Daniel 2:4b-7:28, Ezra 4:8-6:18 and 7:12-26, Jeremiah 10:11 and two words in Genesis 31:47. (The four Hebrew words with which Daniel 2:4 opens and the seven Hebrew words of Genesis 31:47 are not counted; whoever includes the whole verse arrives at 4,959.)
What that means for the counts
Tokens versus verses. Where this book writes "occurs eight times", it counts words, not verses. Where the difference matters — because a verse has a word twice — it is stated: "nine tokens in seven verses". Whoever counts verses where this book counts words arrives at a different number, and both are correct.
Versification. The Masoretic text, the Septuagint and the English tradition diverge. The Psalms usually differ by one number; Isaiah 64 begins a verse earlier; Joel has three chapters where the Masoretic text counts four; Daniel 3:31-33 is English 4:1-3; Nehemiah appears in the Septuagint as 2 Esdras 11-23. This edition follows the English numbering and adds the other where it differs, with `[MT]` or `[LXX]`.
What is not there. There is no textual-critical apparatus on the disk of this project. No manuscript variants, no lists of witnesses, no collation between editions. Every statement in this book about what the manuscripts read is therefore taken over and not recalculated — and it is marked as such. See §3 and §6.
3. The closing line of Matthew 6:13
Chapter 1 (note 2) and chapter 10 refer here.
In this book the prayer ends with: "For thine is the kingdom, and the power, and the glory, for ever. Amen." Many modern translations do not have that line, or put it in brackets. This is the account of that choice — and of its limit.
What could be measured on the book's own disk, and has been measured
- One. The text used has the line. In the Textus Receptus Matthew 6:13 stands in full: ὅτι σοῦ ἐστιν ἡ βασιλεία καὶ ἡ δύναμις καὶ ἡ δόξα εἰς τοὺς αἰῶνας. ἀμήν. CALCULATED. - Two. Luke does not have it. Luke 11:4 ends at ἀπὸ τοῦ πονηροῦ, "from the evil one". CALCULATED — and this is the only datum in the whole question that is independent of which family of manuscripts one follows. Whoever reads the shorter Matthew and whoever reads the longer one read the same in Luke. - Three. The closing formula is a singleton in Matthew. εἰς τοὺς αἰῶνας occurs 28 times in the text used, and in Matthew exactly once: here. CALCULATED. - Four. The closing amen is a singleton in Matthew. Of the 106 occurrences of ἀμήν in the four Gospels, one hundred and one open a saying — amen, I say to you. Five close something, and four of those five are the final verses of the four Gospels. The fifth is Matthew 6:13. CALCULATED.
What could not be measured, and so has not been measured
The manuscripts. The line is missing in part of the oldest Greek manuscripts and present in the majority of the later ones; the majority of textual critics regard it as a later, liturgical addition. That is the state of scholarship and it may simply be stated here — but I have not recalculated it. Neither NA28 nor Tischendorf nor the ECM has been consulted; no textual-critical apparatus was consulted for this book. Every sentence in this book of the form "the oldest manuscripts read [...]" therefore rests on what others have reported. That is why it appears nowhere in this book as a finding, and here as a limitation.
Where the line came from. On that, nothing has been measured here and nothing is claimed here.
Why the book keeps the longer reading, and what that is and is not
It is an edition choice, not a textual-critical judgment. This book reads the Textus Receptus — the same text on which the counts of all 19 chapters rest. A book that counts on one text and quotes from another no longer counts.
And there is a methodological ground that I can account for. The prevailing explanatory model — that the text of the prayer grew along the way through liturgical use — explains both the short reading and the long reading: the short one because it would be original, the long one because the church would have supplemented it in worship. Chase himself calls that model "the theory" in 1891, and applies it to the closing line and to the bread petition — the two most difficult places in the prayer. An explanatory scheme that can absorb every outcome can be refuted by no outcome. This book demands a falsifier on both sides or on neither, and here it applies that measure to itself just as well: "the line belongs there" is not a finding in this book either. It is the edition on which the counting is done.
And what hangs on it in substance: nothing
The closing line says nothing that Scripture does not say elsewhere. Paul closes his last letter with the seventh petition and the doxology in one verse — "the Lord will deliver me from every evil work [...] to Him be the glory for ever and ever, amen" (2 Timothy 4:18; so also Romans 11:36). And the three words themselves are already found in David, in the prayer with which chapter 10 opens: "Yours, YHWH, is the greatness and the power [...] Yours, YHWH, is the kingdom" (1 Chronicles 29:11). Whoever leaves the line out loses not a single thought of this book. Whoever reads it prays after a king who gives back his crown.
One second textual question, and it is here because it touches a chapter
Luke 23:42 — the request of the man on the cross. The text used reads ἐν τῇ βασιλείᾳ σου, "in your kingdom"; other witnesses read εἰς τὴν βασιλείαν σου, "into your kingdom". Chapter 13 hangs on that verse and keeps the first reading, because this book counts on that text. The difference does not affect the finding of that chapter: what matters there are the two words your kingdom, and they are present in both readings.
4. Four counts that were tested and fell
The rule of §1 says: a number that does not hold comes here. These are they. They are not here out of modesty but because a book that shows only its successful counts tells you nothing about how strictly it measures.
4a. 666
The Hebrew word for "evil", רַע, enters Scripture as half of the name of a tree: the tree of the knowledge of good and evil (Genesis 2:9). When that word was recounted over the whole Masoretic text, a number came out that demanded attention.
The raw result is correct. Strong's number H7451 counts 666 words.
And it does not hold, and the reason can be pinpointed exactly:
Conclusion: it is not a finding but a property of how a dictionary delimits its lemmas. Four other, equally defensible counts give a different number, and the most obvious one gives 356. That is why it appears nowhere in this book — not cautiously, not in a note. Not because numbers are not allowed; this book counts on every page. But because this count does not stand up under another, equally defensible division. An instrument that finds something everywhere measures nothing.
(Worked through on the explicit instruction of the author, who rightly rejected hushing it up: "on points like this I would rather have a double-check than silence". All the counts were run twice, in two independent searches.)
4b. "Scripture is permeated by this prayer" — measured, and it does not hold
This is the sentence that rises to the surface by itself in a book like this one, and it has been knocked down by the book's own instrument.
The test. For each of the seven petitions one Hebrew key word was taken — hallow (קדשׁ) · kingdom (מַלְכוּת and the Aramaic מַלְכוּ) · will (רָצוֹן, חֵפֶץ) · bread (לֶחֶם) · forgive (סלח) · test (נסה) · deliver (נצל). Then for each chapter of the Masoretic text it was counted how many of the seven occur there.
What comes out:
And then the comparison with chance. The positions of the same numbers of words were redistributed at random over the text 2,000 times, with the chapter lengths intact. What pure chance would produce: on average 70.6 chapters with three or more, and 14.6 with four or more. Observed: 12 with three or more, and one with four.
⚠ And then the objection to the book's own test, because it is justified and it is weighty. This null hypothesis is blind to what a word is. Words do not lie at random in a text; every word keeps to its own literature — hallow in Leviticus and Ezekiel, kingdom in Esther, Daniel and Chronicles, bread in the sacrificial laws and the narratives. Under such a null hypothesis virtually any set of seven Hebrew words clusters less than chance. So the 1.0000 says something about the null hypothesis and nothing about this prayer.
That is why the test was redone with a fairer comparison: 200 random sets of seven real Hebrew words, drawn from the 863 lemmas that occur as often as the seven key words. That changes the result, and it changes it in the other direction.
So what may not be said: that the petitions "cluster", that Scripture is "permeated" by this prayer in the sense of concentrated, that there are chapters where the prayer "lies ready". Every sentence of that form has been refuted by the book's own measurement — not because the seven repel one another, but because they do nothing that random words do not also do.
⚠ And one more limit, which does not reverse the result but does restrict it. For the fifth petition the test takes סלח and for the seventh נצל: the general words. But chapter 7 does not turn on סלח and chapter 9 does not turn on נצל — they turn on דְּרוֹר, the jubileejubileeThe fiftieth year in which debt is released and inheritance returns — a rhythm within the master clock., and on the Aramaic שֵׁיזִב. With eight and nine occurrences these two are too rare to mean anything in a cluster test, and an exhaustion of nine cannot be measured by density anyway. What has been refuted here is the density claim about the general vocabulary of the prayer. About the rare words on which those chapters actually stand, this test says nothing.
What does stand, and it is what this book stands on: not densities but exhaustions. Not "this word occurs often here" but "this word occurs n times in the whole of Scripture, and this is the only place" — n = 1, 2, 3, 4. Bread and testing stand in one single verse of the whole Hebrew Bible. That is not a density. That is a count that has run out.
And one thing from the test is too beautiful to keep quiet. The only chapter in the whole Masoretic text where four of the seven stand together is Nehemiah 9 — kingdom, will, bread and deliverance. Nehemiah 9 is itself a prayer: the long prayer of the Levites at the return, spoken aloud, in which the whole history is retraced — "bread from heaven You gave them" (9:15), "You delivered them" (9:28).
And then the honesty that goes with it, because it belongs with it. The word for "will" occurs there twice (9:24 and 9:37), and both times it is about the will of people: what the conquerors did with the land as it pleased them, and what the foreign kings do with the people as it pleases them. It is not the third petition. Whoever reads the test strictly is left with three petitions in this one chapter, and then there is not a single chapter in the whole Masoretic text with four.
(The count is CALCULATED; that it is a prayer of all things is CONSTRUED — and it is one chapter in 929, which is exactly the weight it may carry.)
4c. Two smaller ones, and why they appear nowhere
The Our Father three times a day, and Psalm 145 three times a day. The Didache prescribes the prayer three times a day; Jewish tradition prays Psalm 145 three times a day. Striking — and it is liturgical reception, not textual data, and the datings are not fixed. It carries nothing until it is dated, and it is not dated.
The Almighty in 1 Chronicles 29:12. There the Septuagint has παντοκράτωρ, "Almighty", in David's prayer — and the Hebrew does not have it. It is a Greek surplus. Usable only if it is explicitly called that, and therefore not used in chapter 10.
5. What has not been checked in this book
The standing register. Whoever relies on what stands above must read this too.
1. No manuscript has been consulted. No facsimile, no apparatus, no collation. Everything this book says about manuscripts has been taken over.
2. The coding of the text files has not been independently verified. The parsings and the Strong's numbers come from the files used. Where they make a mistake, their mistake counts as mine. That is not theoretical: the file misses Revelation 2:7 in the lemma list of paradise — chapter 13 recounts there by hand — and it codes Isaiah 3:11 as a verb where one can also read it as an adjective; that one word shifts a count from 666 to 665 (§4a). And there is a third kind of trap, which lies not in the file but in Hebrew: אֵל ("God") and אֶל ("to") are the same word without vowel points. Whoever searches on consonants for אֵל שַׁדַּי gets 11 verses and has three too many. Chapter 2 sets those three apart by name.
3. For the Septuagint there is no second file. For the New Testament a count can be held against a second text; for the Septuagint it cannot. That is the weakest side of this book's tools, and the Greek lemmas of the Old Testament are demonstrably less reliable there than the Hebrew ones.
4. The translations are mine. 138 quoted passages, and over 500 references: all from the original text, concordant and modern, no existing translation taken over. That makes them refutable, and that is intended: the original text stands alongside each time. A second pair of eyes has not yet been on them.
5. All counts were run twice, and the second time found 14 wrong. After the 19 chapters were written, every claim with a number was lifted out of the running text once more and measured again — 162 claims. 144 reproduced. 14 did not, and they have been corrected. 12 of those 14 were in chapters 1 through 9; the other two in ch. 14 and ch. 18. Four of the 14 were not a calculation error but a missed occurrence, and those four have made the book better than it was. In addition, four counts are right only under a criterion the reader could not have guessed; there the criterion is now stated. 144 + 14 + 4 = 162.
⚠ And the limit of that audit belongs with it in the same breath, because it is real. From the running text 493 sentences containing a number and a counting unit were extracted by machine. Of those, 162 could be checked against the corpus; the remaining 331 count events, persons or distances within a narrative and have not been systematically retested. And the sieve itself is not perfect: a count hidden in words without a counting unit does not get caught. "Every count run again" therefore means: every count that could be found by machine and tested by machine. A second review, three days later, found four more errors in this way that the sieve had missed — among them a hitpael that occurs four times where the book said twice.
6. The structural question has been parked. Whether Matthew 6:9-13 has seven petitions or six, and where the hinge lies, is a question this book does not address. It keeps seven, counts them, and builds no argument on it.
7. What is said about the Kaddish dates nothing. Chapter 3 establishes that the Kaddish points to Ezekiel 38:23 and Daniel 2:20, and no more. Which is older, the prayer or the Kaddish, is not stated here and cannot be stated here.
6. Sources
What this book owes to others is stated in the place where it is used. This is the list, in three kinds — and the distinction is not politeness but the core of the method.
Primary reading — the text itself consulted
Reported — not consulted personally, and marked as such
Not opened, and on the wish list
Betz (1995) and Schnurr (1985) for early Christian interpretation. They are known through others and therefore not used. (Gibson was also here until 26 August 2026; he has moved to "reported" above, and chapter 8 now has an Excursus on him.)
The author's own work
Where this book takes up something from one of the 16 earlier books, the location is in the note with the word "self-quotation". That is not modesty but precision: a finding made 16 years ago is not a finding of this book, and whoever keeps quiet about that makes his new counts unreliable. In six places it turned out on checking afterwards that a line I took to be new was already in my own work. All six are now attributed.
Whoever finds something here that is not right: it is not right, and I would like to hear it. The counts in this book are written down in such a way that you can recalculate them. That is the intention.