The Workshop

The structural data, the counts and the statistics behind "Sit at My Right Hand" — and, just as fully, the tests that did not clear their threshold.

Why this appendix exists. Every count in this book was carried out mechanically over the complete Leningrad Codex (WLC/OSHB, morphologically tagged) and exhaustively at the lemma level. Whoever wishes to check a claim will find the basis here. And whoever wishes to know what did not come out will find that here too: §IV and §V consist exclusively of negative results. A book that shows only its hits does not measure — it sells.

Counting convention. The maqqef splits words; letters are counted consonantally. Where a count depends on that convention, it is noted.

Status: 8 August 2026.

I. THE METHOD

Pre-registration. For every structural hypothesis the following was fixed before the run: the hypothesis, the test to be run, and the decision rule. After that the test was run once and reported, whatever the outcome. The registrations lie in the Structure project.

The instrument. The analysis pipeline ("the Runner") determines colometry, articulations, anchors and compositional centres mechanically, and lays the structure found alongside the content. Agreement between structure and content counts as the strongest signal; structure alone counts as a by-catch.

The null hypothesis is given the same freedom. For every test the null is granted the same search space as the hypothesis. Where that could not be done, the test was rejected — see §IV.

Two independent counts. Every count that goes into the book was carried out mechanically twice, independently. Two errors that passed this check were later caught after all and corrected; they stand in the verification report.

What statistics does and does not do here. A test can show that something is very probably signal. It can never falsify signal. A failed test therefore means: this instrument found nothing here — not: here is nothing. That distinction is consistently maintained in §IV.

II. THE FOUR PSALMS — BASIC DATA AND STRUCTURE

A2. Psalm 2 — two grids laid over each other

Grid A (the mechanical interlock).

``` vv 1-4 28 words 123 letters vv 5-8 32 words 125 letters vv 9-12 32 words 123 letters ```

The words bind the two last units (32 = 32), the letters the two outer ones (123 = 123). Double-layered. p = 0.011 (n_perm = 999); with the formula layer weighted in, p = 0.019. The interlock holds in both variants. Pre-registered, reproduced, and robust against the version bump.

Grid B (the four panels, exhaustive).

``` I 2:1-3 the rebellion of the nations 22 words 99 letters II 2:4-7a the heavenly response through "I proclaim" 22 86 III 2:7b-9 the quoted decree of YHWH 22 88 IV 2:10-12b the warning to the kings 22 86 2:12c coda 4 12 92 371 ```

4 × 22 + 4 = 92. No remainder word. Blocks II and IV are equal in both layers (22 words and 86 letters).

Every seam is independently marked. After verse 3 a discourse anchor · in verse 7 the ole we-yored (ole on אֶל, yored on חֹק — the heaviest disjunctive of the poetic system, which also rules out the construct reading; see Appendix B2) · after verse 9 the וְעַתָּה of verse 10 · in verse 12 the atnach on אַפּוֹ, directly before the coda.

The verse rhythm. Words per verse: 6 · 10 · 6 · 6 · 6 · 7 · 11 · 8 · 6 · 6 · 6 · 14. Seven of the twelve verses count exactly six words. Tested against a null that draws verse lengths from the actual distribution of all 2,527 psalm verses — in which 6 is the mode, 591 times — this gives p = 0.019. Among the 126 psalms of eight verses or more, Psalm 2 stands fourth in metrical uniformity. As a control: Psalm 8 gives p = 0.87 and Psalm 110 p = 0.98.

The marked centre. At 92 words the middle pair lies at #46|47 = אָמַר | אֵלַי (verse 7); the middle letter (#186 of 371) falls in the word before it, יהוה. The arithmetical heart is thus יְהוָה אָמַר אֵלַי. This is by-catch. The corpus test for this category was rejected (§IV.4); the datum is noted as a convergence and used nowhere as an argument.

The two grids do not coincide, and that is itself the datum. The question is not which grid is "the" structure, but that a song of twelve verses evidently carries two at once. Note: the counts well documented; that both grids are design — weighed.

A3. Psalms 8, 45 and 110

Psalm 8 has no exhaustive counting architecture. An active search was made for a structure of the 4 × 22 + 4 type such as Psalm 2 carries. It does not exist. The qualification that belongs with this: that does not mean Psalm 8 "does nothing numerically" — it means this instrument finds no exhaustive division there.

III. THE WORD FIELDS — ALL COUNTS IN ONE PLACE

Exhaustive, at the lemma level, over all thirty-nine books. The column "status" gives the mark of the claim, not of the count; the counts themselves are all well documented.

A4. The four chains that carry "the coronation descends"

The most important limitation of this book stands here. שִׁית does not occur in Psalm 2. The chain runs through three of the four royal psalms, and the missing link — the crown on the head — comes from Psalm 21, which is nowhere cited in the New Testament. That is a datum, not a defect: it makes the four a selection and not a grab.

A5. Psalm 2

The wisdom register. Of the five Proverbs passages that touch this field, three lie in the speech of Wisdom (1:26 · 8:15 · 8:23) and two in the instruction of the queen mother (31:2 · 31:4). To these are added the wisdom terms of the close: הַשְׂכִּילוּ · הִוָּסְרוּ · אַשְׁרֵי. Note: the counts well documented; that the poet writes in this register — weighed; that he had Proverbs before him — not asserted.

The hook to Psalm 1. אַשְׁרֵי (1:1 ↔ 2:12c) · אָבַד with דֶּרֶךְ (1:6 ↔ 2:12b) · הָגָה (1:2, the righteous one who meditates on the Torah ↔ 2:1, the nations who meditate on emptiness). The linkage is chiastic: the coda of Psalm 2 hooks onto the beginning of Psalm 1, the warning onto its close. That is a text-internal argument for the coda status of 2:12c.

A6. Psalms 8, 45 and 110

A7. The address structure of the four

Second-person forms, counted mechanically on the morphological tagging (suffixes · verb forms · independent pronouns):

This count was made for this book and was not in the Phase 0 groundwork; it was carried out twice independently.

IV. THE TESTS THAT DID NOT CLEAR THEIR BAR

What stands below has not been demonstrated with this instrument. It stands here because it was run.

IV.0 What a p-value does and does not say here

This paragraph stands first because it qualifies everything below it.

A p-value measures one thing: how unusual the observation is under the assumption that nothing is going on. It says nothing about the size of the effect, nothing about practical importance, and it proves no hypothesis. A low p-value makes chance improbable; it does not make the explanation true.

And the threshold of 0.05 is a convention, not a law of nature. It was fixed in advance in this project to make after-the-fact shifting impossible — that is its entire use. But an outcome of 0.089 is not thereby "nothing": it does not clear the agreed bar, and that is something other than being refuted.

Therefore the word "failed" has been replaced in this appendix. What stands there is: does not clear the bar set in advance. Where a test is genuinely refuted — because the instrument demonstrably could not measure — that is noted separately.

The discipline remains unchanged. What does not clear the bar does not go into the book as an argument. Not because it is untrue, but because this book has imposed on itself to carry only what it can carry.

IV.1 The formula layer — run twice, and both times unusable

Hypothesis (H-FORM): weighting in divine-oracle formulas increases the coincidence of articulations.

First round, corpus delimited by place in the canon. On the Psalter p = 0.089 — below the bar set in advance; on the prophetic corpus p = 0.0096 — accepted.

That delimitation is unsound, and it is an error of this project. The Psalter contains prophetic psalms — Psalm 110 carries both נְאֻם יְהוָה and the oath formula — and a large part of the prophetic literature is poetry with the same colometry. That formulas are "scarce in the Psalter" (31 hits) says nothing about the Psalter, but everything about the fact that most psalms are not divine oracle. It is not place in the canon that is decisive, but genre.

Second round, corpus delimited by genre. The test was pre-registered anew and run once, with the detection rule untouched; only the corpus changed. Unit: the biblical chapter; corpus: every unit in the whole codex with at least one formula hit (N = 438 of 929 scanned chapters).

Outcome: p = 0.0020 (z = 2.96; 5,000 permutations). Formally accepted.

And yet this outcome too is not usable. Three independent checks point the same way:

(a) The genre criterion turned out too broad. Of the 1,073 hits, 510 are of the type אמר + divine name — a pattern that also catches the narrative וַיֹּאמֶר יְהוָה. As a result Genesis and Numbers stream massively into the "genre corpus." What was measured is thus not the divine oracle but a much broader speech formula.

(b) The lower p does not come from a larger effect. The effect size is nearly equal to that of the first round; what changed is the width of the null, which at N = 438 is three times narrower. The whole outcome rests moreover on ten points of contact against roughly five expected — a narrow basis for an accepted hypothesis.

(c) And the layer measures two things at once. About forty percent of the hits are genre-neutral connectives (לָכֵן, וְעַתָּה). The layer is moreover direction-blind: it counts also where a human speaks to God. The formula anchor of Psalm 2 is the וְעַתָּה of verse 10 — a word of the poet, no divine oracle.

Verdict. The corpus criticism is confirmed, and a second defect has been added that a corpus switch does not repair: the layer must be split — divine oracle (direction-sensitive) apart from connectives — with its own outcome measure per genre. Moreover the two corpora probably measure different phenomena: in the prophets a compilation seam between oracles, in the psalms a shift of voice within one poem.

As long as that split has not been made, the book does not use the formula layer as an argument — in neither direction. (Open, Appendix D.)

What the repetition did yield, and it is philology and not statistics. In re-counting נְאֻם, something sharper emerged than the groundwork had. The word stands 376 times in the Hebrew Bible, but is almost never verse-opening: four times, against 276 times within the verse and 96 times at the close. It is a signature, not an opening — which in itself is already an objection to using this lemma as an articulation signal.

And the header construction — נְאֻם followed by the name of the speaker, at the head of a poem — carries in the whole Bible only six names: Balaam (Numbers 24:3, 4, 15, 16), David (2 Samuel 23:1), Agur (Proverbs 30:1), "transgression" (Psalm 36:1) — and once YHWH himself: Psalm 110:1.

That is sharper than "the only place in the Psalter." Everywhere else נְאֻם־יְהוָה is an attached hallmark within or after an oracle. At the head of a poem, as superscription, that formula stands with a human seer — except in one place. Verified mechanically, exhaustively. Note: well documented.

IV.2 The accent layer — tested twice, and the first time demonstrably unmeasurable

H-ACC: accent-bound delimitation as a basis for symmetry detection. Outcome: 2 of 148; binomial 0.9956.

This outcome is not interpretable as a finding, for there was a meter error in it: isometry is multiset-dependent, so that the permutation null saturates and p runs toward 1.0. The test could not succeed, regardless of the reality. That was reported as it was — this is the only case in the series where the instrument demonstrably did not measure.

H-ACC2: the same hypothesis with repaired meter — only order-dependent templates. Outcome: 2 of 148 on words, 6 of 148 on letters. Does not clear the bar.

What this means for the book. The Masoretic accents are used in this book where they decide a question of interpretation (the ole we-yored of Psalm 2:7, the atnach before the coda) — that is philology and stands firm. They are not used as a basis for structure detection, for as such they were tested twice and twice yielded no signal.

IV.3 Psalm 8 — no exhaustive counting architecture

Actively searched for a division of the 4 × 22 + 4 type. Not found. See §II.A3.

IV.4 The numerical by-catch — corpus test negative

The pre-registered corpus test for the marked centre is negative: counted primarily (without the superscription) the centre shifts to וְלֹא — no divine name.

Consequence for the book: the marked centre of Psalm 2 (§II.A2) is noted as a convergence and used nowhere as an argument. The whole by-catch layer has been kept out of the main text.

IV.5 Psalm 110 — the anchor detector gives a false negative

On Psalm 110 the anchor detection demonstrably fails. That is diagnostic and not accidental: it is precisely the case type for which the instrument has received a proposed improvement. As long as that improvement has not been carried out, the figure for Psalm 110 is unusable — in both directions. The book therefore does not use it.

V. THE NEGATIVE CONTROL ON THE LEXICAL AXIS

This is the most important negative finding of the whole project, and it forbids a sentence that would otherwise have been obvious.

The question. Do the four royal psalms share a strikingly large vocabulary?

The test. The number of shared lemmata among the four, set against randomly drawn quartets of psalms of comparable length.

The outcome.

The bar was set in advance at 95. The collocation test reaches 94. That is a miss, and it is reported here as a miss.

What follows from this, and it is binding for the book:

The sentence "these four psalms share a strikingly large vocabulary" may not stand in this book. The aggregate is utterly ordinary.

What does stand is something else and far more specific: four separate chains, each exhaustively countable, that all four describe the same movement (§III.A4). That is a claim about a few words, not about the vocabulary as a whole — and that is how it is presented in the book.

V-bis. THE GENRE — THE FOUR PSALMS AND THE PROPHETIC REPERTOIRE

This is not a test but a finding, and it arises from the criticism of §IV.1. The question is not whether these psalms carry one formula, but whether they behave as prophetic literature — a genre that draws on a whole repertoire of forms, including those in which a human speaks to God.

The opposition "psalm versus prophecy" does not hold at the level of form

The canon supplies its own refutation. Habakkuk 3 is a psalm, complete with liturgical apparatus, in the middle of a prophetic book: a superscription in psalm form (תְּפִלָּה לַחֲבַקּוּק הַנָּבִיא עַל שִׁגְיֹנוֹת), three times סֶלָה, and as a closing line the choirmaster's rubric לַמְנַצֵּחַ בִּנְגִינוֹתָי. Outside the Psalter that apparatus occurs nowhere else. Jonah 2 is likewise a psalm within a prophetic book.

Conversely the prophets draw amply on the forms one calls "psalmic": lament, prayer, hymn — Habakkuk 1 and 3, Jonah 2, the confessions of Jeremiah, Isaiah 12, 38 and 63-64. Isaiah 42:10 and Psalm 96:1 open word for word alike.

What does separate the two is not the content but the frame: the messenger apparatus.

Counted mechanically. The rîb form is shared property; the woe-cry and the messenger formula are not.

The four, weighed

"Not measured" and not "weak": the instrument sees almost nothing in Psalm 45, and a scale on which a text lies that you barely measure gives not a degree but a gap. See the void line below.

What this does and does not bound

The messianic weight of Psalm 8 in this book rests on theme and reception — on what Hebrews 2 does with it, not on what the psalm formally is.

That is no relapse but the starting point. Scripture is its own interpreter; when Hebrews 2 relates Psalm 8 to the Son, that is the decisive reading, and not a second choice because form criticism found nothing. The form-critical finding bounds only what may be claimed on formal grounds: this book does not say of Psalm 8 that it is by its form a divine oracle, for it is not. It says that the canon reads it messianically, and that is a stronger ground, not a weaker one.

The distinction that counts here is thus not "truly prophetic versus merely thematic," but: what the text carries in its form versus what the canon reads into it. Both are revelation. Only the second carries the thesis in this book.

Two routes, and what they carry together

The mechanical test of §IV.1, blind to this form-critical work, did take Psalm 2 and Psalm 110 into the corpus and Psalm 8 and Psalm 45 not; the form-critical weighing arrived at the same order. But the calibrated follow-up test below shows how far that convergence carries and how far not: both routes confirm that Psalm 110 stands apart, and both place Psalm 8 at the other extreme. About the relation between Psalm 2 and Psalm 45 they together say nothing — there the ranking is undecided. Convergence at the extremes, no proof, and no ranking of four.

The honest limit of this finding

The human-to-God forms — lament, prayer, hymn — are shared property of both corpora. They can demonstrate genre membership in no direction. That bounds the thesis of this book and it is noted here as such.

And the proper follow-up test is not a membership test. A form profile per text, direction-sensitive, with a distance measure — and obligatorily calibrated on the hinge cases that the canon itself supplies. That test has been run; it stands below.

The follow-up test: the calibrated form profile

In full in `REPORT — FORM PROFILE (opus-lens).md`; pre-registration written out before the first outcome, one run, seed 42. All figures below have been independently recomputed by me from the result file.

The setup. Per unit — the psalm for the Psalter (150), the chapter for the prophets (233), plus one partial passage — a density vector over the five form families, with direction per hit (God→human, human→God, about-a-third). Two centroids, and a position t on the axis between them: t = 0 is the psalm centroid, t = 1 the prophet centroid. The five calibration cases and the four royal psalms were kept out of both centroids — without that exclusion the whole exercise would have been circular. 2,532 hits over 384 units.

The direction rule repairs the error of §IV.1. Psalm 16:2 (אָמַרְתְּ לַיהוָה, "you said to YHWH") counted there as divine oracle; now it falls correctly into the human-to-God family. All six known traps yield zero transmission hits.

The calibration — the gate, with the threshold fixed in advance at five out of five.

Passed, five out of five. The three psalms-in-a-prophetic-book land on the psalm side despite their canonical place, the oracles against the nations on the prophet side despite their poetry. The instrument thus measures form and not the canon — precisely what it had to prove before it was allowed to say anything. Isaiah 38:10-20 cleared it even without its superscription, on lament and address forms alone.

What the axis turns out to be. The two corpora differ in transmission density by a factor of 394 and in prayer density by a factor of 2.7 — the other way round. The axis is thus carried by the messenger apparatus (weight +1.26) against the prayer direction (−0.88). That is, mechanically, over 383 units, what the form-critical work established by hand: it is not the repertoire that separates, but the frame.

The four, positioned — not tested.

Psalm 110 is the only one of the four psalms that crosses the axis, and it does so on the messenger apparatus. Of all 150 psalms two reach the prophet side: Psalm 49 (+0.825) and Psalm 110 (+0.810). Within the Psalter it lies at the 99th percentile, within the prophetic corpus at the 42nd — an exceptional psalm and an average prophetic chapter at once.

Psalm 8 lies below every one of the 233 prophetic chapters. It is not "weakly prophetic"; it is more psalm than the average psalm — its prayer density lies a third above the psalm average. For the third time, and now along a third route: the messianic weight of Psalm 8 rests on theme and reception, not on genre.

The expectation was 110 > 2 ≫ 45 > 8, with the great gap between 2 and 45. The machine puts the gap somewhere else: between 110 and the rest. The nominal order comes out, but it comes out in a segment where the instrument cannot distinguish — and that makes it no finding. What the measurement carries are the two extremes: Psalm 110 on the one side, Psalm 8 on the other. The middle is undecided; see below.

The limit of this instrument — three things it does not carry

1. Psalm 2 and Psalm 45 cannot be distinguished with it. Their margins overlap widely. Whoever says on this measurement that Psalm 2 is form-critically more prophetic than Psalm 45 goes further than it carries. What does stand is philology and not geometry: Psalm 2 carries the only יְהוָה אָמַר אֵלַי of the whole Psalter, Psalm 45 carries no transmission, vision or announcement feature at all.

2. The void line. A text in which the instrument detects nothing lands not on zero but on t = +0.320 — recomputed, and 24 units lie exactly on it, among them Psalm 1 and Psalm 23. Geometrically unavoidable: an empty text resembles a prophetic chapter more than a psalm because it prays nothing. Psalm 45 lies 0.013 above that line. Its position is thereby no finding but an artefact of the thinness of its profile. Psalm 2 lies 0.093 above it, and that difference is one hit. Only Psalm 110 and Psalm 8 stand where they stand on content.

3. The direction determination has a measured error fraction of 20 % [10.5 % · 34.8 %], from a sample of forty checked by hand. It lies almost entirely in one marker and works systematically conservatively: it shortens the axis and thus underestimates the distance of the psalms from the psalm centroid. One registered marker did not fire at all owing to a lemma error in the pre-registration (הִנְנִי carries in the OSHB tagging Strong 2005, not 2009 — checked by me). In accordance with the one-run rule, not repaired; that error too works against the outcome, not for it.

No p-value, and there is none to compute either. Four units carry no test. t is a position, not a membership: that Psalm 110 lies above the decision boundary says that its form profile lies closer to the prophetic centroid — not that it "is prophecy."

VI. WHAT REMAINED UNDECIDED

The frequency 65 = 31 + 34. That this division falls on a defensible seam in two of the four psalms is an indication and no pattern. Two cases carry no argument. Undecided, not negative.

The border witnesses. Three psalms came into view without belonging to the four, each for its own reason:

The verification debt. What remains open in primary literature and edition control is gathered in Appendix D.

VII. ACCOUNTING IN ONE PARAGRAPH

Of the tests in this appendix, three did not clear the bar set in advance (the formula layer on the Psalter, the repaired accent layer, the corpus test on the marked centre), one was demonstrably unmeasurable because the instrument carried a meter error (the first accent layer), one was not executable owing to a false negative of the anchor detector (Psalm 110), one was actively searched for and not found (the counting architecture of Psalm 8), and the negative control on shared vocabulary came out as negative controls ought to come out: ordinary.

In one of these the error lies not with the finding but with the framing of the question. The formula layer was assessed on a corpus delimited by canonical place instead of by genre — while prophetic psalms exist and most of the prophetic literature is poetry. That test must be redone, and until then it does not count, in neither direction.

And over everything stands §IV.0. None of these outcomes refutes anything. They say only that this instrument, on this data, with this prior agreement, did not find enough to build on. What remains — the double interlock of Psalm 2, the interlock of Psalm 45, the verse rhythm, and the four word chains — is what this book uses, and no more than that.