Tyndall-John: Quality Engineering (STEP 4b)

Date: 2026-10-09. QE agent (Opus). Inputs read: _text/manifest.json, all 45 digest files (every page of every chunk), deep/ (all five, by section), facts.json, claims.md, sites.md, research/ notes, physicist.md, mathematician.md, contrarian.md (by section), arbiter.md and primer.md (in full), example.py (run). Spot checks were made against the extracted text in _text/ and against page images rendered from the source PDFs into a scratch directory. Nothing in the source folder was modified. Tags: [read] from the project documents; [read: page image p.N] checked on a rendered page; [calc] my own arithmetic (project venv); [memory] recalled, with confidence.

Verdict: MINOR

The central claims and the arbiter's verdicts stand. No verdict needs to be redone. The defects are: (1) one chapter of Heat that bears on claim 2B and claim 5 is still covered by the digest alone, and it holds a result nobody used; (2) the arbiter and primer describe the ten-foot extrapolation in misleading words; (3) a "Tested, not supported" verdict rests on unsourced [memory]; (4) a few items from the sources and from the Investigator dropped out of the arbiter or primer. All of these can be fixed by updating facts.json, claims.md, arbiter.md (a "QE revision" section) and primer.md.

1. Coverage

All 45 chunks have a digest (45 files in digest/). No document was skipped. Every manifest document has a facts.json entry, and the multi-essay New Fragments is split into its essays as STEP 1 asks. Coverage: PASS.

2. Shortcuts

• Heat Ch. XII (PDF pp.407-441) is still digest-only. claims.md itself asked for an Opus deep read and none was run. The arbiter says "no verdict below turns on them alone". That is true of the verdict labels, but the chapter holds the hydrogen-flame series (section 3, item 1). That series is Tyndall's own measurement of how strongly absorption depends on the source, which is the very point the arbiter and primer make about the ten-foot claim (2B) and about claim 5. It also holds the liquid and vapour tables that underlie the "thermal continuity" weak form, the breath table and the water-layer opacity figures the arbiter quotes. This is a defect: evidence for claim 2B and claim 5 rests on a digest. Remediation: deep read 1 below.
• Heat Ch. XIII (pp.442-485) is digest-only. It mostly repeats the 1866 paper, which was deep-read, so this is acceptable for claims 3 and 4. However, it adds three items that are not in the 1866 paper and that bear on claim 3 (section 3, items 7-8). Deep read 2 (Sonnet) covers them.
• Heat Ch. XIV, paragraphs 646-662 (pp.486-495: dew, Wells, Pouillet, Glaisher, lunar heat) are digest-only. These are the meteorological evidence for claim 2 that Tyndall drew from others. Deep read 3.
• Heat Ch. XI, paragraphs 435-466 (pp.362-378: perfumes, ozone, dynamic radiation, long and short columns) are digest-only. Claim 1's "radiating power equals absorbing power" draws partly on this. The arbiter cites ozone from the digest. Deep read 4.
• The other digest-only material (Heat Ch. I, IV, V, VII, VIII; the non-physics essays in New Fragments; Six Lectures I-IV and VI) is peripheral to the seven claims, so digest-only coverage is acceptable for it.

3. Missed material (read in the digests and checked where marked)

None of the following appears in facts.json key claims, claims.md, the three reviews, arbiter.md or primer.md, unless a note says otherwise.

1. Hydrogen-flame series (Heat paragraphs 530-531, PDF pp.428-429). In a polished tube 4 ft long, undried laboratory air absorbed 17.2 per cent of the radiation from a hydrogen flame, against 0 for dry air. On a humid day it absorbed 20.3 per cent. A platinum spiral glowing about as brightly lost only 5.8 per cent in the same air ("one-third"), and 8.6 per cent with the spiral plunged in the flame [read: page image p.428; text p.429]. Tyndall reads this as "accord" between the periods of hot and cold water vapour. The modern reading is that the emission bands of the hot H2O line up with the absorption bands of the cold H2O [memory: high]. This bears on claim 2B: the same 4 ft of the same air absorbs 6 to 20 per cent depending on the source. It also bears on claim 5 (selectivity) and on the primer's lesson that "the source matters". Primer Part C should cite it as Tyndall's own demonstration.
2. Breath table in Heat (paragraph 541, PDF p.433), three pairs. Chemical (Frankland) against physical (radiant heat): 4.311 against 4.00, 4.66 against 4.56, and 5.33 against 5.22 [read: page image p.433; the text layer misreads 5.22 as "622"]. On Radiation (1865) prints only the last two pairs. The contrarian's "agreement within 2 per cent" (contrarian.md, strongest case) holds for those two pairs only. The first pair in Heat differs by 7 per cent. Both versions are Tyndall's, and Tyndall himself calls the agreement "very fair". The full table should be on record.
3. Planetary habitability (Heat paragraphs 545-546, PDF pp.435-436). Tyndall answers an unnamed work that held the outer planets uninhabitable by the inverse-square law. He argues that an atmosphere which passes sunlight and absorbs planetary radiation can keep a distant planet warm: a 2-inch layer of air saturated with ether vapour absorbs about 35 per cent, and an olefiant-gas shell 1 inch thick absorbs "at least 26 per cent". This is the greenhouse argument extended to other planets, and it is the source of the "35 per cent ether shell" figure. The physicist kept the figure in the verdict table, but the arbiter and primer dropped it. The 26 per cent comes from a measured 1-inch value [read: digest p.435]. The 35 per cent has no experimental record in the folder.
4. Janssen's water-vapour bands. In summer 1866, on the Titlis, Tyndall "repeated M. Janssen's observations of water-vapour absorption bands" with a hand spectroscope [read: digest New Fragments p.472]. Janssen's telluric water-vapour lines are an independent, spectroscopic line of evidence that atmospheric vapour absorbs [memory: medium on Janssen's 1860s work]. They are relevant to claim 2 and to the Magnus dispute. The Investigator did not search for them. This is a lead for the next Step I run (Janssen, telluric lines, 1864-66). Heat p.483 also cites Janssen on absorption by the eye's humours.
5. Dew and atmospheric absorption of sunlight (Heat Ch. XIV, PDF pp.486-506). - Wells's dew theory: grass under a clear sky is 9-14 F colder than the air. - Tyndall's supplement that ice-making pits in Bengal need dry air as well as clear air, with a vapour screen over wet straw (paragraph 655). - Glaisher's grass-chill table. - Pouillet's estimate that the atmosphere absorbs 25 per cent of sunlight at the zenith and 40 per cent over the sunlit hemisphere, which Tyndall ascribes "mainly" to water vapour acting on the longer waves (paragraph 684).

These are the meteorological record for claim 2 [read: digest pp.487-506]. claims.md lists Wells only among the recommended reads. 6. Magnus beyond the vapour dispute. - Heat paragraphs 291-295: Magnus held that hydrogen conducts heat like a metal. Tyndall argues that convection was not excluded and calls gaseous conduction "still an open" question [read: digest pp.251-254]. - Heat p.533, closing note: Magnus had published a further paper controverting the vapour results, "by which I am not convinced" [read: digest p.533]. - New Fragments p.256: Magnus's radiant-heat experiments in gases "occupied him for the last years of his life in discussion with Tyndall".

These add to the record of how the dispute ran (arbiter, Absence-of-evidence item 4) and show that Magnus's side was still active in print in 1868. 7. Luminous fractions of other sources (Heat paragraphs 584-586, PDF pp.457-458). White-hot platinum spiral: 1 part luminous in 24. Gas flame: 1 in 25 (forty experiments). Electric light: 1 in 10. A worked check for the reader [calc]: an infrared-to-visible ratio of 23 corresponds to a blackbody near 2400 K, which is hotter than the melting point of platinum, about 2041 K [memory: high]. Two readings fit: the iodine method under-counts the visible light, or platinum's emissivity, which is higher in the visible than in the infrared [memory: medium], raises the visible share at about 2000 K. This is a cheap test of the filter method and is not in the primer. 8. Tyndall corrects Melloni's alum assumption (Heat paragraphs 641-643, and footnote p.319). Concentrated alum transmits 20 per cent of the obscure heat that passes the iodine filter. Melloni's luminous and obscure shares (oil flame 10/90, platinum 2/98, alcohol flame 1/99, quoted in Heat p.320 and in the 1866 paper p.3) rest on the assumption that alum stops all obscure heat, so they understate the obscure share. Tyndall's footnote calls this "an error, as we shall subsequently learn". This bears on how far claim 3's comparison figures can be trusted. The same chapter gives the ox-eye result ("one-fifth" of obscure rays reach the retina; On Radiation gives 18 per cent). 9. Tyndall's priority concession to Schaffgotsch (Heat p.293). On singing flames Tyndall wrote that "all the merit of priority rests with Count Schaffgotsch". This is relevant context when weighing Tyndall as an interested party in the Akin dispute: it shows a case where he conceded priority. It should be recorded beside, not instead of, the Akin material. 10. Minor: the French translator of Six Lectures (F. Moreno) disputed Tyndall's credit to Kirchhoff over Foucault, and Tyndall replied in a footnote (Six Lectures pp.288-290). Six Lectures p.229 records Tyndall's own 1861 attempt to reverse the sodium lines. These are peripheral priority material.

4. Accuracy spot checks (against _text/ and page images)

Accuracy defect (wording, not arithmetic): arbiter item 4 says "neither is what he claimed", and primer Part C Step 2 says "Neither matches Tyndall's printed claims". Tyndall's claim is "at least 10 per cent". Both of the reviewers' extrapolations (17.6 and 37.5 per cent) and the scaling of his own polished-tube figure (over 10 per cent in 4 ft gives about 23 per cent in 10 ft [calc]; 17.2 per cent for the hydrogen flame gives about 38 per cent) exceed 10 per cent. So on his own data the number is conservative, not contradicted. The real objections are those the primer already gives: saturation, and the change from a 373 K source to the 288 K Earth. Reword it as "both exceed his lower bound; the open question is whether the lower bound survives saturation and the cooler source". The verdict (Plausible, untested as stated) does not change.

example.py was run with the project venv. It exits 0, and every number in primer Part A, B and C matches its output (777.5; 769; 793.6, 749.8, 765.3; the 7.74 and 1.07 ratios; 3004 K and 2977 K; 37.5 and 17.6 per cent; exponent 0.19; band fractions; 255 K).

5. Retention check

The test is whether pertinent material from the sources or from the Investigator was left out of the arbiter or primer because a reviewer judged it negatively.

• Kept well: Foote and Franz priority; the Magnus, Buff and Lecher-Pernter negative results (with "more transparent than a vacuum"); Very's request for "further elucidation"; Röntgen; the Akin dispute with competing explanations for his silence; Stokes's and Croll's letters; Macdonald's viXra paper (stated fairly, with the reviewers' reply); the pink glow (Never properly tested, explicitly "ignored, not dismissed"); Tyndall's unsourced "most delicate experiments"; the RI tube; the 1859 journal entry. Arbiter and primer both carry the "interested party" caveat. The record of single-source evidence is described, not labelled.
• Dropped (defects, minor): 1. The 35 per cent ether shell ("proved by experiment", On Radiation p.44; Heat paragraph 546) and the Sahara night ice (Heat p.395, no source). The physicist listed the first in the claim 2 verdict as Plausible, untested. Both are Tyndall's single-source statements and were left out of the arbiter's claim 2 and the primer. Restore them as quoted statements, with a note that they have no experimental record in the folder. 2. Ferrel, as quoted by Very, accepted Magnus's film explanation (sites.md, "Evidence against"). This is evidence against Tyndall, and SOURCE HONESTY says evidence against is stated too. It appears in neither the arbiter nor the primer. 3. Babbitt 1878 (calorescence taken up in an alternative colour theory), Airy's "beginning Meteorology", Soret on the Geneva printing, and the RI archive references (JT/2/13/1158) are in the arbiter but not the primer. This is acceptable for a study document. Babbitt is worth one line in the primer's influence and priority paragraph, because the reader profile includes an interest in alternative science. 4. The breath table's third pair (item 3.2) was not dropped for negative reasons; it was never read. Add it.
• No pertinent Investigator finding was removed because it was judged false. Retention: minor defects only.

6. Process

• Source tags: present and generally honest throughout. The primer's [web] items are mostly "per Investigator", and it says so. Good.
• Web checks: physicist 3, mathematician about 6, contrarian 4, arbiter 2 (one failed, reported as failed). Defect: the arbiter's claim 5 strong-form verdict "Tested, not supported" rests on "[memory: high] spectroscopy alone; no source fetched". STEP 2 says "Never state that replications or tests happened (or failed) without a source". The scale term asserts that controlled tests exist. Fetch one source for the vapour/liquid water infrared difference (for example a published optical-constants table for liquid water, or a water-vapour window reference) and retag it. Otherwise relabel the strong form "Not supported by modern spectroscopy [memory: high], source not fetched", still on the scale. The same applies, at lower priority, to "superseded by orbital forcing" (claim 7), which is high-confidence memory and acceptable as tagged.
• Verdict scale: applied consistently, with split verdicts stated. Note that "Internally inconsistent" is used for the "red waves more powerful than blue" sentence, which the arbiter itself calls "an error of statement". That fits the scale only if the sentence contradicts Tyndall's own definitions. The mathematician's reading (energy per wave against his own wave theory) supports it, so this is acceptable, but the primer should say why the label applies.
• Primer against STEP 4: sections 1-9 are present and in order; there are 5 check-yourself questions with answers; example.py exists and runs; symbols are defined. Reader-profile slips: "HITRAN" (primer section 7) is used without being defined. Define it as the high-resolution transmission molecular absorption database, used for line-by-line calculations. "Line-by-line calculation" also deserves one sentence of explanation.
• Corrections recorded by the arbiter (items 1-11) and the primer's note on 777.9 against 777.5 are sound. I checked the arbiter's correction that d ln J / d ln gamma is 2.5, not 1.67: J is proportional to gamma/(gamma - 1), so d ln J / d ln gamma = -1/(gamma - 1) = -2.5 at gamma = 1.40 [calc]. The arbiter is right.

7. Remediation actions

Deep reads (most important first)

1. Heat-35e701, PDF pp.407-441 (Ch. XII), Opus. Transcribe from page images: - the liquid-absorption table (paragraph 507) and the corrected vapour table (paragraphs 512-515); - the source-temperature tables (paragraphs 519-525); - the hydrogen-flame and CO-flame series (paragraphs 530-537); - the breath table (paragraph 541); - the water-layer transmission table (paragraphs 542-543); - the olefiant and ether shell figures (paragraphs 545-546).

Questions: how strongly does Tyndall's own absorption by humid air depend on the source, and what does that imply for "at least 10 per cent within ten feet"? Does the liquid/vapour ordering support the weak form of claim 5 for water specifically? 2. Heat-35e701, PDF pp.442-485 (Ch. XIII), Sonnet. Check the luminous fractions (1 in 24, 1 in 25, 1 in 10) on images. Check the alum correction of Melloni (paragraphs 641-643) and the ox-eye "one-fifth" figure. Check the solar trial dates (15 Nov 1864) against the 1866 paper. Do these change the arc and solar figures of claim 3? 3. Heat-35e701, PDF pp.486-506 (Ch. XIV, paragraphs 646-684), Sonnet. Cover Wells, Tyndall's dry-air supplement (paragraph 655), Glaisher's table, Melloni's 2 C limit, Pouillet's 25 and 40 per cent and its attribution to vapour, and lunar heat. What meteorological evidence beyond Strachey does Tyndall bring for claim 2, and how much of it is his own? 4. Heat-35e701, PDF pp.362-378 (Ch. XI, paragraphs 435-466), Sonnet. Cover the perfume table (resolving the text-table conflicts), the ozone figures (20-136 times oxygen), the dynamic radiation and absorption tables, and the long and short column test (paragraphs 463-466). Does the reciprocity half of claim 1 rest on measured pairs, and are the ozone numbers as the arbiter states? 5. Heat-35e701, PDF pp.250-254 and p.533, Sonnet. Read Magnus on gaseous conduction (paragraphs 291-295) and Tyndall's closing note on Magnus's later vapour paper. What exactly did Tyndall say about Magnus's 1868 position, and does it change the arbiter's account of the dispute? 6. New-fragments-John-Tyndall-626498, PDF pp.471-473, Sonnet. Read the 1866 Titlis spectroscope passage repeating Janssen's water-vapour bands. What did Tyndall observe, and does he present it as evidence for atmospheric vapour absorption?

facts.json

• Heat entry: add key claims for the hydrogen-flame series (17.2 and 20.3 per cent against 5.8), the planetary-habitability argument, the luminous fractions (1/24, 1/25, 1/10), the alum correction of Melloni, and Pouillet's 25 and 40 per cent. Replace the OCR breath values with the image-checked 4.311/4.00, 4.66/4.56, 5.33/5.22.
• Update how_read once deep reads 1-5 are done.
• New Fragments, Old Alpine Jottings entry: add the 1866 Janssen-band observation.

claims.md

• "Specific testable statements": mark the CO2 pressure table as image-confirmed (Heat p.431). Add the hydrogen-flame series and the full breath table.
• Source note: remove Ch. XII from "digest only" once deep read 1 is done.

arbiter.md: add a "QE revision" section

• Item 4 and claim 2B: reword "neither is what he claimed" as described in section 4. Add Tyndall's hydrogen-flame series as his own evidence of source dependence. The verdict is unchanged (Plausible, untested as stated).
• Claim 2: restore the 35 per cent ether shell and the Sahara ice as Tyndall's quoted, unsourced statements. Add Ferrel (via Very) to the evidence against. Add Pouillet, Wells and the Janssen observation as leads after the deep reads.
• Claim 5 strong form: fetch one source, or relabel it as in section 6.
• Remove the "OCR, not image-checked" caveat on the CO2 table.
• Breath measurement: give all three pairs.
• Record that "150 of the lower degrees" is image-confirmed.

primer.md

• Part C Step 2: fix the "Neither matches" sentence. Add a Step 1b with the hydrogen-flame and platinum-spiral numbers as Tyndall's own proof that "the source matters".
• Source note and Part C Step 3: remove the OCR caveat on the CO2 table.
• Section 4, priority and influence: add one line each for Babbitt 1878 (influence) and Ferrel (evidence against, via Very).
• Section 7: define HITRAN and line-by-line calculation.
• Optional exercise (Part B or Check yourself): the white-hot platinum "1 part in 24" against Planck (about 2400 K, above platinum's melting point), with the two competing readings.

sites.md (next Investigator run)

• New keywords: Janssen raies telluriques vapeur d'eau 1864-1866; Tyndall Titlis spectroscope 1866; Magnus 1868 Wasserdampf Absorption reply.

Corrections to earlier documents

1. arbiter.md item 4 and primer.md Part C Step 2. The wording "neither is what he claimed" and "Neither matches Tyndall's printed claims" is misleading. Tyndall's claim is a lower bound ("at least 10 per cent"), and both extrapolations exceed it. Evidence: Heat p.387 ("no doubt at least 10 per cent"); [calc] 17.6 and 37.5 per cent, and about 23 per cent from his own 4-ft polished-tube figure.
2. claims.md, arbiter.md and primer.md give the CO2 CO-flame table as "OCR, not image-checked". It is now image-checked and correct as printed (Heat p.431).
3. digest Six-lectures part 06, p.214 ("150 may be 15"). The page image shows 150; the arbiter's quotation is right.
4. digest Heat part 13, p.433 reads the physical breath values as "4.00, 4.56, 6.22". The page image reads 4.00, 4.56, 5.22, and the first chemical value is 4.311. The contrarian's "within 2 per cent" applies to the 1865 two-pair table only.
5. facts.json Heat entry describes the CO2 table as "digest only". After this check its numbers are image-confirmed, though the rest of Ch. XII still needs deep read 1.

Recheck

Date: 2026-10-09. Checked arbiter.md, primer.md, claims.md and facts.json (valid JSON) against the defects above.

Remaining: arbiter.md line 19 (claim 2B background item 4) still carries the withdrawn wording without an inline marker. Suggested fix: append "(withdrawn, see QE revision item 3)" to that sentence, and "(superseded by QE revision item 11)" to line 75.

Final: OPEN (1 issues)

Orchestrator fix (2026-10-09)

• arbiter.md:19 (claim 2B background item 4): appended "(withdrawn, see QE revision item 3)"; item 3 in the QE revision is the Claim 2B wording correction, so the number matches.
• arbiter.md:75 (claim 5 strong form): appended "(superseded by QE revision item 11)"; item 11 is the claim 5 web check, so the number matches.

Final: PASS