Where the Quoting Stops

Baptism — because 6:3–4 is the only place Romans uses the word group at all, and the article will tell you the range of first-century immersion practice the readers might have had in mind. Note that the article will also tell you what the passage does NOT say: there is no water in these verses.

Slavery in the Greco-Roman world — because 6:16–22 is an extended argument from a transfer of ownership, and because Paul interrupts himself at 6:19 to say ἀνθρώπινον λέγω — I am speaking in human terms — which is an author flagging his own analogy as an analogy.

Marriage and divorce law, Jewish and Roman — because 7:1–3 argues from a rule about widows, and because the application in 7:4 does not match the illustration: in the illustration the husband dies, in the application the reader dies. Any dictionary article worth reading will say so.

Romans 7 and the ‘I’ — the one article to read with the most suspicion, because the entry you happen to open will usually present one answer as settled. There are at least six live readings and the dictionary you have will reflect its tradition.

Wages — a small one, and rewarding. ὀψώνιον at 6:23 is a soldier’s pay packet, not an abstract recompense; the other three New Testament occurrences are Luke 3:14 (soldiers asking John the Baptist what to do), 1 Corinthians 9:7 and 2 Corinthians 11:8, and all three are about being paid to serve.

Whether 6:1–14 describes something that happened to the reader at a moment in the past or something true of the reader continuously; whether 7:14–25 describes a Christian; whether σάρξ in 7:18 means the body or means a condition. Those are the three questions readers come to this passage with, and none of them is decidable by looking up a word. This page does not decide them either — see Stage 5.

Romans 7:15–19 states a problem that Greek and Latin writers had been stating for four hundred years.

⚠️ What that table is worth, honestly. It is worth exactly this much: the form of Romans 7:15 was available. It does not show that Paul read Ovid or Epictetus — Epictetus is later than Romans anyway — and it does not show that he did not. None of these texts was measured against Romans, because none of them is on this disk, and a Latin poem could not be run against a Greek letter by this method in any case. Anyone who tells you Romans 7 is or is not borrowing a Stoic commonplace is telling you a judgement, and so would we be.

What is measurably different is where Paul puts the blame. The classical topos locates the failure in the agent — passion overpowering reason. Romans 7:17 and 7:20 locate it in an occupying power: ἡ οἰκοῦσα ἐν ἐμοὶ ἁμαρτία, the sin that dwells in me. That is a difference of content, and it is visible in the text rather than inferred from it.

Two ancient realities stand behind chapter 6 and neither is measurable here.

Presenting yourself as a slave. 6:16 uses παριστάνετε ἑαυτοὺς — you present yourselves — and self-sale into slavery was a real legal act in the Roman world, usually to clear a debt. The verb παρίστημι and its by-form run five times in 6:13–19. This is reported from the standard handbooks; no documentary papyrus was consulted.

Sacral manumission. The Delphic manumission inscriptions, in which a slave is notionally sold to a god and thereby freed, are the parallel usually offered for 6:18 and 6:22 — set free from sin… enslaved to God. It is a genuinely attractive parallel and this page cannot check a word of it. The inscriptions are not on this disk. Deissmann popularised the comparison; later work has questioned how close it is. We name the debate and take no side.

Wages. Here the measurement does reach: ὀψώνιον occurs 4 times in the Greek New Testament and every occurrence is about pay for service — soldiers at Luke 3:14, a soldier at 1 Corinthians 9:7, Paul’s support at 2 Corinthians 11:8, and this verse. And in the complete Greek Old Testament — all 64 CATSS files as of 30 August 2026 — it occurs 3 times, every one of them in the identical form ὀψώνια that Romans 6:23 uses: 1 Esdras 4:56, 1 Maccabees 3:28 and 1 Maccabees 14:32. In 1 Maccabees 3:28 Antiochus opens his treasury and gives ὀψώνια to his forces for a year; in 14:32 Simon arms the men of his nation and gives them ὀψώνια out of his own money; in 1 Esdras 4:56 the garrison of the city is granted allotments and ὀψώνια. Every occurrence of the word in both corpora held here, Greek Old Testament and Greek New Testament alike, is a payer paying people who serve him. ⚠️ This sentence used to end “2 Maccabees is NOT on this disk and was not searched,” and that stopped being true on 30 August 2026. All four Maccabean books are held and screened now — 2 Maccabees 9:9 is the deepest partner of Romans 7:24 in Stage 4 below, on this same page — and the count re-run over the whole screened corpus is still three. The stale scope line is printed rather than deleted. Nothing here says Romans is drawing on 1 Maccabees — 6:23 shares no more than TWO consecutive words with any verse of it. What is measured is the word’s company, not a source.

6:3–4 has been read against Jewish proselyte immersion, against the Qumran washings, and against the Hellenistic mystery cults — the last of these most aggressively in the older German scholarship. The mystery-cult reading depends on dating initiation rites that are attested later than Romans, and on texts none of which are held here. This page reports that the comparison exists, notes that it has been heavily criticised since the middle of the twentieth century, and does not evaluate it.

What is on the page instead is a count: the whole letter to the Romans uses the baptism word group three times and all three are in 6:3–4. Whatever the background, the letter does not build on the word.

Counted from the MorphGNT lemma column across all 27 books: seven lemmas occur exactly once in the Greek New Testament, and that once is inside Romans 6:1–7:25. Two further lemmas are confined to this passage without being hapax — καινότης (6:4, 7:6) and παράκειμαι (7:18, 7:21), twice each and both times here — so nine lemmas in all occur in these 48 verses and in no other verse of the Greek New Testament.

Three of the seven begin with συν- or συμ-, together-with. The passage needed compound words for joint action and reached for words it never uses again — grown together with Christ, agreeing together with the law, delighting together with the law. That is an observation about the wordstock, not a doctrine; the count is the claim.

The two deepest verbatim runs Romans 6:1–7:25 has with anything in the Greek New Testament are tied at six, and both are full titles of Christ closing a paragraph.

Jude is 25 verses long. 1 Corinthians is 437. They tie. A run of six against a 25-verse letter and a run of six against a 437-verse letter cannot both be evidence of contact, and neither of them is: both are the same closing formula, and it is a formula precisely because short books and long books alike reach it.

Control first. The whole of Romans, all 430 MorphGNT rows, against all 30,348 verse units of Greek Old Testament this shelf screens — the 64 CATSS files less the Odes.

Across all 430 rows: 290 reach three, 91 reach four, 54 reach five, 33 reach six, 22 reach seven, 16 reach eight, 13 reach nine, 10 reach ten, and 2 reach seventeen. The method is not blind and this letter is not shy.

⚠️ TWO OF THOSE EIGHT NUMBERS WENT DOWN ON 31 AUGUST 2026, WHICH IS A DIRECTION THIS SHELF HAD NEVER HAD TO PRINT BEFORE. THE OLD FIGURES ARE KEPT HERE. They read 92 reach four and 34 reach six. What changed is the corpus, and it changed by getting smaller. The Odes — the fourteen liturgical canticles bound after the Greek Psalter, 289 verse units, which came in with the last twenty CATSS files — were taken out of everything this shelf calls “the Greek Old Testament” for screening. They are still held on the disk. They are no longer screened. HELD is 30,637 verse units in 64 files; SCREENED is 30,348, and the difference is exactly the 289 Odes units. Two reasons, either sufficient on its own: Ode 9 is Luke 1 and Ode 13 is Luke 2 — the Magnificat, the Benedictus and the Nunc Dimittis — so a New Testament verse screened against them is being measured against a copy of itself; and ten more canticles are second copies of passages the corpus already holds (Ode 1 = Exodus 15, 2 = Deuteronomy 32, 3 = 1 Samuel 2, 4 = Habakkuk 3, 5 = Isaiah 26, 6 = Jonah 2, 7 and 8 = Daniel (Th) 3, 10 = Isaiah 5, 11 = Isaiah 38), so any tie band touching a canticle was doubled.

⚠️ And exactly two rows of Romans lost depth, both outside this study, and it is worth naming what they lost it to. Romans 11:36 — to him be glory forever, amen — measured 6 in a band of one against Odes 12:15, and now measures 5 against 1 Esdras 5:58 and 4 Maccabees 18:24. Ode 12 is the Prayer of Manasseh, whose closing doxology is the same liturgical formula; the run was a hymn-ending matching a hymn-ending inside a Psalter appendix. Romans 9:5 went from 4 against Odes 12:15 and 14:35 to 3 against 1 Esdras 5:58, 2 Chronicles 6:2, 4 Maccabees 18:24 and Daniel 2:44 — Ode 14 being the morning hymn, the Gloria in excelsis. Neither loss was in Ode 9 or Ode 13, so neither was Luke measured against Luke; they are the other half of the problem, an appendix doubling doxologies the corpus already contained. Nothing inside Romans 6:1–7:25 moved at all, which the next table shows.

⚠️ THIS CONTROL WAS RE-RUN ON 30 AUGUST 2026 OVER A CORPUS ONE THIRD LARGER, AND THE OLD FIGURES ARE PRINTED RATHER THAN QUIETLY REPLACED. When this page was written the shelf held 44 CATSS files, 23,061 verse units, and the line above read: 273 reach three, 80 reach four, 48 reach five, 30 reach six, 20 reach seven, 14 reach eight, 11 reach nine, 9 reach ten, and 2 reach seventeen. The remaining twenty files — Greek Exodus, Leviticus, Numbers, both Joshua recensions, and the deuterocanonical books, which this shelf treats as witnesses, not Scripture and labels so every time — were fetched on 30 August 2026, taking the corpus to 64 files and 30,637 verse units with no gaps in it, of which 30,348 are screened. Every figure in the distribution rose, because a screen against a bigger target can only ever find more.

⚠️ And two rows of the control table are new, because Greek Exodus was not previously on the disk at all. Romans 9:17 — for this very purpose I raised you up, that I might show my power in you — now measures 11 against Greek Exodus 9:16 at a tie band of one. On the 44-file corpus that row measured 4, and its best partners were Jeremiah 16:21 and Ezekiel 43:7: the screen was answering with the wrong books because the right one was not there. Romans 9:15 — I will have mercy on whom I have mercy — moved from 2 to 9 against Greek Exodus 33:19. Both are outside the 48 verses this page studies, and both are printed here anyway, because a control is worthless if it is quietly re-baselined. A third row moved for the same reason and is worth naming: Romans 13:9 quotes you shall love your neighbour as yourself, and it went from 4 to 6 against Greek Leviticus 19:18 — a verse this shelf could not reach an hour before the fetch.

Now the same screen, unchanged, over the 48 verses of 6:1–7:25.

⚠️ This table survived the Odes removal without moving a digit. Re-run over 30,348 units it is still 18 / 27 / 3 / 0, the three four-word rows are still 6:4, 7:18 and 7:24 with the same partners and the same bands, and 6:6 is still 3 in a band of one at Greek Leviticus 9:15. No hit anywhere in these 48 verses was ever in an Ode. Fifteen tie bands narrowed by between one and thirteen rows apiece — 7:7’s 11 to 9 is the sharpest, 6:12’s 2,265 to 2,252 the largest — and not one of the fifteen is a figure this page prints. Three of the fifteen lost an Ode 9 row, which is to say a verse of Luke 1: 6:10 shed Odes 9:47, 6:12 shed Odes 9:69 and 7:6 shed Odes 9:48. All three were two-word runs inside bands of hundreds, so nothing here was ever reported as Luke matching Luke — but the Gospel was inside these numbers, and saying so is the point of printing the correction rather than the new figure alone.

⚠️ The ceiling was tested and it held. On the 44-file corpus this table read 21 / 25 / 2 / 0. Stage 5 predicted, in terms, that the four-word ceiling “could only rise if the missing books were added and could never fall.” The missing books were added on 30 August 2026 — all twenty of them, 7,576 further verse units — and the ceiling did not rise. Nothing in these 48 verses reaches five consecutive words of the complete Greek Old Testament. Nor did it fall when the Odes came back out on 31 August 2026. Three rows now reach four instead of two, and the third of them is printed below with the other two.

And here are the only three rows that reach four, with what they actually are.

⚠️ One more row moved, and it moved because Greek Leviticus arrived. Romans 6:6 — that the body of sin might be brought to nothing — measured 2 at a tie band of 94 on the 44-file corpus, which is the signature of a row that has found nothing at all. It now measures 3 at a tie band of one, against Greek Leviticus 9:15 — τῆς ἁμαρτίας τοῦ, from the sin offering. A tie band that collapses from 94 to 1 looks like a discovery and is not one: τῆς ἁμαρτίας stands in 37 verse units of Greek Leviticus alone, and the three-word string only goes unique because of the τοῦ that happens to follow it at 9:15 — exactly one verse unit in all 30,348 carries τῆς ἁμαρτίας τοῦ. The band moved because the corpus moved, not because the text did, and a tie band of one is not by itself evidence of anything.

Romans 7:7 is the only verse in the passage that attributes words to Scripture — if the law had not said — and the words it attributes are two.

⚠️ THE EXODUS ROW READ NOT VERIFIABLE FROM LOCAL CORPORA — GREEK EXODUS IS NOT ON THIS DISK UNTIL 30 AUGUST 2026. IT IS VERIFIABLE NOW, AND THE OLD ROW IS RECORDED HERE RATHER THAN QUIETLY OVERWRITTEN. Greek Exodus arrived with the last twenty CATSS files on 30 August 2026. Read off the file, CATSS Exodus 20:17 and CATSS Deuteronomy 5:21 are the same forty-four words in the same order — identical before normalisation, not merely after it. Which is worth saying plainly: the question which Decalogue is Paul quoting? cannot be settled from the Greek at all, because in this edition the two Decalogues do not differ by a single character at the tenth commandment. The Hebrew of the two does differ elsewhere in the commandment — Exodus 20:17 puts the house first and Deuteronomy 5:21 the wife — but the Greek column as CATSS prints it puts the wife first in both. The page reports that and does not resolve it.

⚠️ And one more text now reachable says the same thing Paul says, in almost the same breath. 4 Maccabees 2:5 — a witness, not Scripture — reads λέγει γοῦν ὁ νόμος οὐκ ἐπιθυμήσεις τὴν γυναῖκα τοῦ πλησίον σου — the law says, you shall not covet your neighbour’s wife. Romans 7:7 reads ὁ νόμος ἔλεγεν οὐκ ἐπιθυμήσεις — the law said, you shall not covet. Two Greek-speaking Jewish writers reach for the tenth commandment, introduce it with the law says, and cite the identical opening — and, like Paul, 4 Maccabees stops after the first clause instead of running the list out. ⚠️ What that is not. It is not evidence that Paul read 4 Maccabees, and this page does not say or imply it. Both are doing the obvious thing with the obvious verse. What it does show is that the citation-shape Romans 7:7 uses was available in Greek-speaking Judaism, which is exactly the kind of thing the shelf could not check before the deuterocanon landed — Stage 5 named 4 Maccabees as one of the two absences a reader of Romans 7 should feel most.

⚠️ Why the two figures differ, disclosed rather than patched. MorphGNT’s normalised column writes οὐ where the text has οὐκ; the CATSS column keeps οὐκ. So the shelf’s standard screen cuts every cross-corpus run that crosses a negative. The same defect was found and published on the Galatians page. The normaliser is frozen — every previously published number depends on it — so it was not changed; the two was obtained by re-running the identical screen against the text column, and both numbers are printed. Either way the point stands and is not close: two words is the whole quotation. The tenth commandment runs on for another forty-two words in Greek Deuteronomy and, word for word, in the Greek Exodus that reached this disk on 30 August 2026 — wife, house, field, servant, maidservant, ox, donkey — and Paul takes none of them.

Nor does the passage's most-discussed echo run any deeper. Romans 7:11 says sin ἐξηπάτησέν με — deceived me — which every commentary sets beside Genesis 3:13, ὁ ὄφις ἠπάτησέν με. Measured, the two verses share two consecutive words, με καὶ — and the very verb that makes the echo does NOT match, because Genesis has the simple form and Romans the compound. Screen the whole of Romans 7 against all 49 verses of Genesis 2–3 in Greek and nothing reaches three. Fourteen rows reach two. That is not an argument that the echo is not there — a compound of the same verb in a chapter about a commandment, a deception and a death is a reasonable thing for a reader to hear. It is a statement that the echo is not verbal contact, and the difference matters.

All 48 verses against every other book of the Greek New Testament, one book at a time. Ceiling per book — thirteen of the twenty-six shown. Titus, Colossians, 1 Thessalonians and 1 Peter also cap at four and are not printed.

Size does not predict the ceiling and neither does subject matter. The three longest books in the New Testament — Luke, Matthew, Acts — all cap at three. Jude, at twenty-five verses, reaches six. What predicts a deep run here is not shared argument but shared liturgical furniture: the four rows at the top of that table are all the same two closing formulas.

The 48 verses against the other 382 verses of the letter. All 48 reach two, 22 reach three, 5 reach four, and 2 reach six. The two sixes are 6:23 with 8:39 and 7:25 with 5:21 — the same two formulas again. The deepest non-formulaic internal link is four: 6:12 with 5:21, ἡ ἁμαρτία ἐν τῷ — sin… in the — band of one, and 5:21 is the verse that says sin reigned. 7:16 and 7:20 each reach four with 8:25.

The link a reader most expects is 7:25 to 8:1, and the two verses are consecutive so the screen has nothing to say about them. What it can say is that ἐν Χριστῷ Ἰησοῦ — three words — stands at 6:11 and at 8:1 and 8:2, in a band of six inside Romans alone. The phrase that carries the whole argument across the chapter break is a phrase the letter uses constantly. That is not a diminishment; it is what a governing phrase looks like when you count it.

This is the largest limitation on the page and it belongs at the top of the section. The six standing readings of the first-person block — Adam; Israel under the law at Sinai; Paul before his conversion; Paul as a believer; humanity under law in general; a rhetorical persona (prosopopoeia) — are arguments about theology, biography, and Greco-Roman rhetorical convention. A run-screen touches none of those.

What the measurement contributes, and its exact weight:

• Romans 7, screened against Genesis 2–3 in Greek, does not share three consecutive words with any verse of it. That does not refute the Adam reading — that reading has never rested on verbal reproduction, and 7:9–11 could allude to Genesis without sharing a syllable.
• The passage contains 0 occurrences of πίστις and 1 of πνεῦμα. That is a fact, and it is not an argument that the speaker is or is not a Christian. Silence about faith across 48 verses of a letter is compatible with several readings.
• The pronoun density is high. Density is not identity. A persona speaks emphatically; so does a man describing himself.

Anyone wanting this settled should read Kummel, Cranfield, Dunn, Moo, Stowers and Wasserman against each other; the list is below. They disagree, and the disagreement is not caused by a missing measurement.

The passage quotes the Decalogue, and both Greek Decalogues are now held and measured. The tenth commandment stands at Exodus 20:17 and Deuteronomy 5:21. Measured the same way against each — one verse against one verse — Romans 7:7 returns 1 on MorphGNT’s normalised column and 2, οὐκ ἐπιθυμήσεις, on the text column, against Exodus 20:17 exactly as against Deuteronomy 5:21; the Deuteronomy figures were reproduced first, on the same code, before the Exodus figures were taken. The two-word run ties across four verse units of the 30,348 screened: Exodus 20:17, Deuteronomy 5:21, Deuteronomy 7:25 and 4 Maccabees 2:5 (a witness, not Scripture). So the run cannot tell the two Decalogues apart, and the reason is measured, not assumed: CATSS Exodus 20:17 and Deuteronomy 5:21 are the same forty-four words. Stage 4 prints the detail.

Until 2026-09-24 this paragraph read: Greek Exodus is not on this disk, and the passage quotes the Decalogue. The tenth commandment stands at Exodus 20:17 and Deuteronomy 5:21. Deuteronomy is held; Exodus is not. So the run reported at 7:7 is a run against ONE of the two possible sources, and no comparison with Exodus 20:17 was made or could be made. The two words are identical in the printed editions of both, so the figure would very likely be the same — but ‘very likely’ is not a measurement and this page will not print it as one. The measurement has now been made, and the figure is the same.

⚠️ TWENTY CATSS FILES WERE ABSENT WHEN THIS PAGE WENT UP. THEY ARRIVED ON 30 AUGUST 2026, AND THIS SECTION KEEPS WHAT IT SAID AND ADDS WHAT HAPPENED NEXT.

The paragraph that stood here read: “Twenty CATSS files are absent — Exodus, Leviticus, Numbers, both Joshua recensions, Judith, both Tobit recensions, 2, 3 and 4 Maccabees, Odes, Wisdom, Psalms of Solomon, Baruch, the Epistle of Jeremiah, and the Bel and Susanna texts. Every count on this page against ‘the Greek Old Testament’ is a count over 39 books, not over the Septuagint — including the three occurrences of ὀψώνια reported in Stage 2, and the four-word ceiling reported in Stage 4, which could only rise if the missing books were added and could never fall. Wisdom and 4 Maccabees are the two absences a reader of Romans 7 should feel most, because both handle exactly this material: the passions, the law, and reason’s failure to govern the flesh. Neither could be checked.”

All twenty are now held, and every one of those three predictions has been tested.

That last result is the one worth sitting with. Wisdom and 4 Maccabees are witnesses, not Scripture on this shelf, and they were named as the absences that mattered because they argue about the same thing Romans 7 argues about — the passions, the law, and reason’s failure to govern the flesh. They are now on the disk, they were screened, and they produce no verbatim contact with Romans 6–7 at all. Two writers can work the identical material in the identical language and share no sentence. Which is the whole thesis of this page’s Stage 4, arriving from a direction the page did not choose — and it is worth more as evidence for having been flagged by the page against itself before it could be filled. ⚠️ What it does not show: that Paul was unaware of such writing, or that no relationship exists. A verbatim-run screen sees quotation. It does not see a shared intellectual climate, and the absence of a run is not the absence of a debt.

No critical apparatus was read for any verse on this page. The Greek printed here is one edition — the SBLGNT as digitised by MorphGNT — and 14 of its 834 word rows in these two chapters carry an editorial siglum. Where they fall: 6:1, 6:11 (three), 6:12, 6:13, 6:15, 7:13, 7:14, 7:17, 7:18, 7:20, 7:23, 7:25.

Two of those matter to sentences on this page. At 7:25 the first word is marked: the edition reads χάρις τῷ θεῷ, thanks be to God, where part of the tradition reads εὐχαριστῶ τῷ θεῷ, I thank God. The six-word run reported for 7:25 is downstream of that word, so the run itself is unaffected — but a page that reported the run without mentioning the variant would be presenting one edition as the text. At 6:1 the verb of the objection is marked. Neither variant changes any number on this page; both are disclosed because the reader cannot know that unless we say it.

No Qumran — and 1QS 11 and the Hodayot are where the closest Jewish parallels to the flesh-language of 7:14–25 are usually located. No rabbinic material, so nothing on the yetzer hara. No Philo, whose treatment of ἐπιθυμία and the tenth commandment is the closest Jewish-Greek analogue to 7:7. No Epictetus, no Ovid, no Euripides in machine-readable form — Stage 2 reports them from editions and measures nothing. No Delphic manumission inscriptions and no documentary papyri, which is where the slavery language of chapter 6 would actually be tested. No Apostolic Fathers.

And no patristic reading, which is the most consequential omission on this page. Augustine changed his mind about Romans 7 between the Propositions on Romans and the anti-Pelagian works, and Western reading of this chapter has run in the channel he cut ever since. That history is not a measurement and it is not on this page, and its absence is not neutrality — it is a gap.

• C. E. B. Cranfield, Romans (ICC, 1975–79) — the fullest defence of the Christian reading of 7:14–25.
• James D. G. Dunn, Romans 1–8 (WBC 38A, 1988).
• Douglas J. Moo, The Letter to the Romans (NICNT, 2nd ed. 2018).
• Richard N. Longenecker, The Epistle to the Romans (NIGTC, 2016).
• W. G. Kummel, Romer 7 und die Bekehrung des Paulus (1929) — the study that put the non-autobiographical reading on the map.
• Stanley K. Stowers, A Rereading of Romans (1994) — the prosopopoeia case, argued from Greco-Roman rhetorical practice.
• Emma Wasserman, The Death of the Soul in Romans 7 (WUNT 2/256, 2008) — Romans 7 read against Platonic moral psychology.
• A. J. M. Wedderburn, Baptism and Resurrection (WUNT 44, 1987) — the careful assessment of the mystery-religion comparison referred to in Stage 2.
• Dale B. Martin, Slavery as Salvation (1990) — on the slavery metaphor of chapter 6 in its Greco-Roman setting.
• None of these were consulted for the measurements on this page, which were run on the text files named in the footer.

The number that built this page is a number that is missing. Romans reproduces seventeen consecutive words of a psalm in chapter 10. In chapters 6 and 7 it reaches four, twice, and both fours are coincidences. The one place these chapters say the law said, the quotation is two words long and stops before the object: You shall not covet — and nothing after it.

It would be easy to make that mean more than it does, so here is what it does not mean. It does not mean Paul is not thinking about the Old Testament here; chapter 7 is about the law from beginning to end and uses the word twenty-three times. It does not mean these chapters are un-Jewish. It means that when Paul argues about sin and the law, he argues in his own sentences.

And what he does instead of quoting is answer. Four times somebody asks a question — shall we continue in sin? shall we sin because we are under grace? is the law sin? did the good thing kill me? — and four times the answer is the same two words. Then each half closes the same way, with the full name of the Lord: eternal life in Christ Jesus our Lord, and thanks be to God through Jesus Christ our Lord. Those two closings are, measurably, the deepest verbatim contact these chapters have with anything in the New Testament — six consecutive words each, tie band of one each. The passage’s strongest verbal ties are not to its arguments. They are to the way the whole church already said the Lord’s name.

New Testament shelf: Romans 6:1–7:25, 48 verses, Greek New Testament and ESV. The other Romans studies on this shelf are 1:1–2:29, 3:21–26, 5:1–11, 8:31–39 and 14:1–16:27.