Parallax: Quality Engineering (Step 4b)

Project: Zetetic Astronomy. Earth Not a Globe! (1865) by "Parallax" (S. B. Rowbotham), and Gleason's 1892 "New Standard Map of the World".

What I read: prompt.txt; manifest.json and _textprep.log; all six digests; the deep read; facts.json; claims.md; sites.md and all twelve research notes; physicist.md, mathematician.md and contrarian.md; arbiter.md; primer.md; example.py. I checked the extracted text at PDF pp.16-19, 44, 46, 52-53, 56, 73-74, 78-79, 85-86, 107, 127, 164-165, 171-173 and 178-179. I rendered PDF pp.164-165 as images (/tmp/parallax-qe/). I also checked the image sizes of the two map files, read only.

Tags: [read: p.N] is the book, PDF page N (book page = PDF page − 5). [read: page image p.N] means I read a rendered page. [calc] means I computed it with the project venv. [memory] means recalled, with confidence given.

Verdict: MINOR

The central verdicts hold. The arbiter's numbers reproduce, example.py runs, and the source tags are honest. There are three problems:

1. The eclipse argument has no verdict. It is part of claim 4 in claims.md: lunar eclipses are not caused by the Earth's shadow, with the book's records of "horizontal" eclipses and of the red Moon in totality. The arbiter and the primer give it no verdict.
2. Four checkable arguments were missed by every later file. They are the geodesy arguments (pendulum length and meridian arcs), the "spherical excess" argument, the New Zealand day-length argument and some of the long-distance sightings.
3. Known errors are still in claims.md and facts.json. The arbiter found them, but nobody went back to fix them.

All three can be fixed by updating facts.json, claims.md, arbiter.md (a "QE revision" section) and primer.md. None of them is likely to change a central verdict.

---

1. Coverage

• Digest gaps. These are small and acceptable.
• Digest 01 skips p.44. The text there is the end of the lost-day argument and the Ross quotation [read: p.44], which the digest gives under pp.42-43.
• Digest 03 skips p.127, the last page of the "destruction by fire" passage [read: p.127].
• Digest 03 stops at p.150 and digest 04 starts at p.154. The deep read covers pp.151-153.
• One small error in facts.json. It calls s-l1600.webp a "higher-resolution copy". Both files are 1200 × 1597 pixels (sips), so the webp is a same-size duplicate. Correct the wording. The status does not change.

2. Shortcuts

• Acceptable. These rest on page reads rather than a formal deep read, and the readers say so:
• Bedford Level and horizon: the physicist re-read pp.16-24 and 160-180.
• Sun triangulation: the physicist re-read pp.75-79, and the arbiter read p.79.
• Polaris (p.46), the southern circuit (pp.56-57), air-gun and tubes (pp.73-74) and tides (pp.112-114): read by the arbiter.
• Page images of pp.15, 18, 155 and 164-165: rendered by the reviewers.
• The pendulum section had a full Opus-grade deep read.
• Defect: the lunar eclipse and moonlight section (PDF pp.94-110) rests on the digest alone. It carries part of central claim 4. The physicist says so ("I did not re-read the eclipse, tide and scripture sections"). The contrarian writes "Lancet 1856 thermometer tests … I did not read them in detail". Nobody read the horizontal-eclipse records, the Ward and Walkey totality accounts (19 March 1848) or the De La Hire, Forbes and Lancet moonlight figures from the page. Deep read 1 is requested below.
• Defect: the geodesy passages (PDF pp.49-67) rest on the digest alone. They cover pendulum lengths against latitude, meridian-arc measurements, "spherical excess" and the surveyors' horizontal datum. The book's rebuttal of the main evidence for an oblate Earth is there, and no later file engages it (§3). Deep read 2 is requested below.

3. Missed material

I read every digest and checked the text. The items below are in the book but are missing from the reviews, the arbiter and the primer, or are mentioned without a verdict. Under SOURCE HONESTY they belong in the record whatever their merit.

1. Pendulum length against latitude is put down to temperature (pp.49-54). - The book prints a table of seconds-pendulum lengths: 39.027 in at the equator, 39.139 in at London and 39.197 in at the pole. A seconds pendulum is one whose swing takes one second; the length that does this depends on the local strength of gravity. - The book quotes Phillips: a change of 30 °F alters the pendulum's length by 1/5000. It concludes that the colder north explains the variation "without supposing that they arise from a peculiar spheroidal form of the Earth" [read: pp.52-53]. - It also cites Derham on air density (p.54). - Check from the book's own numbers [calc]:
   • The temperature fall from equator to pole in its table is 84.2 °F. That changes the length by 84.2/30 × 1/5000 = 5.6 × 10⁻⁴ of the whole.
   • The tabulated lengths change by 0.170/39.11 = 4.3 × 10⁻³ of the whole, about 7.7 times more.
   • The quoted standard length is stated "in vacuo … at the temperature of 62°" [read: p.53]. So these are lengths already corrected for temperature and air, which the argument overlooks.
   • None of the three reviews, the arbiter or the primer mentions this argument. It is a clean "Refuted by calculation" case built only from the book's own figures, and it suits the primer.
2. Meridian-arc measurements are said to show an "oblong" Earth (pp.49-52). - The book cites the Ordnance Survey's Arbury Hill to Clifton series, with fathoms per degree rising northward. It also cites the French measurement in Peru (56,653 toises) against the Swedish one in Lapland (57,422 toises), and says the Swedes later found an error of 196 toises. - From this it argues that "no two sets of experiments agree" (digest). - Only the physicist touches it, in passing. It needs one paragraph: the modern geodetic answer, and whether the 18th- and 19th-century spread was larger than the flattening of about 1/300 [memory: high] it was meant to measure.
3. "Spherical excess" put down to theodolite lenses (pp.64-67). - Surveyed triangles sum to more than 180°. A theodolite levelled toward the sea sees the horizon below the cross-hair. - The book says both come from magnification by the instrument's lens, not from curvature (digest). - Spherical excess is a predicted geometric quantity: area / R² radians, by Legendre's theorem [memory: high]. It does not depend on the instrument's optics. No later file mentions the argument.
4. The New Zealand seasons and twilight argument (pp.85-90). - The book says Wellington's "latitude is about the same as in England" [read: p.86]. - It quotes the Cook's Strait Almanack for 1848: longest day 14 h 58 m, shortest 9 h 2 m. For England it gives 16 h 34 m and 7 h 45 m. - It concludes the asymmetry is "altogether incompatible" with a globe. It explains it, and the abrupt southern twilight, by the Sun's light sweeping a larger circle on the plane. - Check [calc]:
   • Wellington is at about 41.3° S and London at 51.5° N [memory: high], so the premise is off by about 10°.
   • Sunrise-to-sunset formula with standard refraction: 15 h 10 m and 9 h 11 m at 41.3°; 16 h 38 m and 7 h 50 m at 51.5°.
   • The book's figures are what a globe predicts for the two real latitudes, to within about 12 minutes.
   • Shorter twilight at lower latitude is also standard globe behaviour [memory: high].
   • No review, the arbiter or the primer mentions this. It needs a verdict: Refuted by calculation (the premise is false; the data fit the globe).
5. Eclipse observations: no verdict in the arbiter or the primer. - Horizontal eclipses. These are eclipsed Moons seen with the Sun still above the horizon: Paris, 19 July 1750; 20 April 1837; 20 September 1717; and Glaisher's list for 1590, 1648, 1666 and 1668 (digest pp.95-96). The physicist and the contrarian answer them with refraction at the horizon. That answer is right, but it rests on [memory: high] and it does not reach the arbiter or the primer. - The red Moon in totality. The Hon. Mrs Ward (Dublin) and Mr Walkey (near Cullompton) both describe the eclipse of 19 March 1848 (digest pp.97-101). Walkey wrote that "the Moon positively gave good light from its disc during the total eclipse". The physicist covers the redness generally. Nobody names these two witnesses after Step 1. - The book's eclipse sub-claim ("eclipses of the Moon are not caused by the Earth's shadow", claims.md claim 4) gets no verdict in the arbiter's verdict table.
6. The Lancet, 14 March 1856: concentrated moonlight "reduced the temperature upon a thermometer more than 8 degrees" [read: p.107]. - The mathematician keeps it as an observation and suggests radiative cooling. - The arbiter and the primer drop it. See Retention, §5.
7. Long sightings in Section 1 and the lighthouse list. - Great Orm's Head seen from Douglas Harbour from 100 ft. The book gives 60 miles; 47 miles beyond a 13-mile horizon; 1,472 ft hidden; summit "872 feet below the horizon" [read: pp.18-19]. Check [calc, coordinates from memory: medium]: Douglas to Great Orme is about 61 miles, and from 100 ft the globe hides about 1,605 ft (k = 0) or 1,347 ft (k = 0.13). k is the refraction coefficient (the curvature of a light ray as a fraction of the Earth's curvature). The book's distance is right. - The Brighton line-of-sight. A line was stretched between two poles 6 yards apart, and a vessel was followed westward for 40 miles (pp.20-21). - Glaisher and Coxwell's balloon reports of the horizon at eye level (pp.21-24). - The full lighthouse list (pp.172-178): Bidston, Calf of Man, Cromer, Whitby, Norwegian, French, Madras, Port Nicholson, Cape Bonavista and Ballycotton. - Only the mathematician mentions Great Orm, and only the contrarian mentions the list. The arbiter's claim 2 covers Eddystone, Southsea-Ryde, Grimsby and the Nab, but not Great Orm, Brighton or the list. They are favourable to the book and should get the same treatment: the refraction each would need, and "Not testable as stated" where conditions are unrecorded.
8. Southern navigation details (pp.26-30, 56-58). - Ross's barometer readings fall southward of the Tropic of Capricorn. The book uses this to say the water stands higher in the south (pp.28-30). The drop itself is a real feature, the low-pressure belt around Antarctica [memory: medium], so the observation is sound and only the inference is in question. - Wilkes turned back at about 70° (digest; check the page). - The book blames southern navigational errors for the loss of the Challenger frigate in 1845 (p.58). - These need a line each.
9. Minor items, for facts.json only. - The Kane "open polar sea" tide argument (pp.114-115). - Spirit-level bubbles that never settle, as evidence of a floating Earth (p.115). - Herschel's 1,900,000-year light time and the 6,000-year chronology (pp.200-202). - The Saros (p.109). - Royal Astronomical Society (1850) and French Academy notes on possible further non-luminous bodies near the Earth (pp.108-109).

10. Investigator leads not carried forward. These are from Schadewald, a critic, via the Investigator's notes:

    • Rowbotham's Burnley no-show at his second lecture, December 1849, after a question about ship hulls.
    • The Teign river walk-out from an agreed repeat of the marker test.
    • The 1849 Trowbridge lectures as "S. Goulden".

    They bear on how Rowbotham handled tests of his claims. The contrarian's Mallove section ("how claims were rejected") should weigh them, stating Schadewald's stake, and the arbiter should list them. They are absent from all of them.

4. Accuracy spot checks

5. Retention

Pertinent material is left out of the arbiter or the primer in four places. In each case the material was either judged negatively at Step 2 or never weighed:

• The Lancet 1856 report of concentrated moonlight lowering a thermometer reading by more than 8 degrees [read: p.107]. The mathematician keeps it, with a possible explanation. The arbiter and the primer drop it while they rule on the moonlight inference. It is an observation that favours the book's "cold Moon", so it must stay in the record next to the verdict.
• The horizontal-eclipse records and the Ward and Walkey totality accounts (§3 item 5). They are reported observations, they were answered by reviewers, and they do not appear in the arbiter or the primer.
• Great Orm, Brighton and the lighthouse list (§3 item 7). These are observations favourable to the book that the arbiter does not carry forward.
• Schadewald's accounts of the Burnley no-show and the Teign walk-out (§3 item 10). They count against the inventor, but they are still pertinent Investigator findings and should be listed with the critic's stake.

The arbiter's own "Retention" note does cover every Investigator lead that the reviewers weighed. The gaps above are book material and Investigator leads that the reviewers did not weigh.

6. Process

• Source tags: present and honest throughout.
• The reviewers mark digest-only reads ("via digest", "Step 1 summary").
• The arbiter marks what it took from reviewers without re-checking.
• Secondary and critic sources (Wikipedia, Schadewald, Proctor, De Morgan, The Field) are named with their stakes.
• Interested parties are flagged: Gleason's patent text against the "As It Is" slogan, Hampden and Carpenter, Blount, and Rowbotham's own religious motive.
• Web checks:
• The physicist reports two checks that failed: the Airy formula and Reich.
• The mathematician lists what it left unchecked.
• The contrarian used 6 of 6 calls.
• The arbiter used 8 of 8 and resolved both of the physicist's failures (arXiv 2502.12230; Reich via Wikipedia).
• Done properly.
• Verdict scale:
• The reviewers stray in places: the contrarian's "Not supported" for Neptune, and the mathematician's "Established as fact" for the Neptune elements.
• The arbiter uses the scale names exactly and splits verdicts properly.
• Gap: the eclipse sub-claim of claim 4 has no verdict (§3 item 5).
• Primer:
• Follows STEP 4 in order, sections 1-9 with the answers at the end.
• Defines every symbol (R, d, h, e, f, k, R_e, P, T, dT/dh, φ, Ω, ω_p, a, b, L, ω₀) and introduces the Coriolis effect from scratch.
• Gives a fully worked example with a table, a check-yourself section and further reading that includes the inventor's own works.
• Adds no verdicts. Fits the READER PROFILE well.
• Small point: §1 tells the Bedford story as Rowbotham's own, in 1838 or 1839. The 1865 text says only "the observer", with no date [read: p.16]. The attribution and dates come from Schadewald and Wikipedia, which the paragraph does cite. Add one clause saying the book itself does not name the observer or give the year.
• example.py: runs cleanly from /tmp with the venv. Parts A-G print every figure the primer quotes; exit code 0.
• Absence-of-evidence analysis: present in the arbiter and summarised in the primer, with competing explanations and deciding records. Good.

7. Remediation

Deep reads (most important first)

Changes to each file

• facts.json
• Southsea to Ryde: "43-mile" becomes 4.5 miles.
• Air-gun: "within a few feet" becomes "within a few inches; twice fell back upon the mouth of the barrel".
• Southsea Common and West Cowes eye height: 2½ ft.
• s-l1600.webp: same-size duplicate (1200 × 1597), not higher resolution.
• Add key_claims for: pendulum length due to temperature; meridian arcs show an "oblong" Earth; spherical excess due to lenses; New Zealand day length and twilight, with the premise that the latitudes are the same; the horizontal eclipses and the red Moon; the Lancet 1856 cooling report; Great Orm from 100 ft; the Brighton line; Ross's southward pressure fall.
• Update how_read after the deep reads.
• claims.md
• Fix "43-mile" to 4.5 miles.
• Under claim 4, name the eclipse sub-claim explicitly.
• Under claim 6, add the geodesy rebuttals (pendulum lengths, arcs, spherical excess) and the New Zealand seasons argument.
• arbiter.md, a "QE revision" section with these changes:
• Verdicts for the eclipse sub-claim. Horizontal eclipses: refraction at the horizon (web-check the size of horizon refraction). The red Moon: sunlight refracted through the Earth's atmosphere. Expected labels: "Tested, not supported" for "eclipses are not caused by the Earth's shadow", with each record kept and named.
• Pendulum length due to temperature: Refuted by calculation, from the book's own figures (5.6 × 10⁻⁴ against 4.3 × 10⁻³).
• Spherical excess due to lenses: Refuted by calculation, or Tested, not supported.
• Meridian arcs: a verdict after deep read 2.
• New Zealand seasons: Refuted by calculation (wrong latitude premise; globe day lengths fit within about 12 min).
• Great Orm, Brighton and the lighthouse list: the refraction each needs, and Not testable as stated where conditions are unrecorded.
• Keep the Lancet 1856 report beside the moonlight verdict.
• Add the book's Tropic of Capricorn framing to the Wilkins item.
• List Schadewald's Burnley and Teign accounts, with the critic's stake.
• primer.md
• Carry the new verdicts into the §6 evidence table.
• Add the pendulum-length check (from the book's own numbers) and the New Zealand day-length check to §7 as instructive mistakes. Either could be a check-yourself question.
• Name the eclipse witnesses and the Lancet report.
• Fix table row 3c to give the book's Capricorn framing.
• In §1, add that the 1865 text does not name the observer or give a year.
• example.py, optional: a Part H reproducing the New Zealand day lengths and the pendulum-temperature ratio.

Recheck

Rechecked against the current claims.md, facts.json, arbiter.md, primer.md, example.py and deep/earth-not-globe-book-c293c9.md. I used grep and one run of example.py (exit 0). I did not re-read every page.

Corrections to this QE. The arbiter found two statements in the first QE that were wrong. They are recorded in arbiter.md under "Where the QE needs correction", and I accept them. - Mrs Ward's account is of the partial eclipse of 7 February 1860, not the totality of 19 March 1848, so §3 item 5 was wrong to pair her with Walkey. - The Douglas to Great Orme distance is 60 to 61 miles (the book gives 60), not a firm 61.

Neither changes a verdict.

Final: PASS