Blondlot "N rays": arbiter's report

Inputs: facts.json, claims.md, physicist.md, mathematician.md, contrarian.md. Tags: [read] from the book (via the reviewers' page-image reading), [web] URL, [memory] with confidence, [calc] my own arithmetic (run in the project venv).

Source gap (repeated from Step 1). The folder holds one document: a Google Books scan of the 1905 English translation of "N" Rays (161 PDF pages) with no usable text layer. Step 1 and the reviewers read it from rendered page images and scratch OCR. Not in the folder: the French Comptes Rendus originals (1903-1904), Blondlot's later papers, Wood's 1904 Nature report, Rubens' and Lummer's reports, and any replication or critique papers. All statements about Wood, Kelvin, Crookes, Lummer, Rubens and the "physiological" explanation rest on [web] https://en.wikipedia.org/wiki/N_ray (re-fetched by me), not on the folder; I could not read any primary account. I did not re-open the book's page images myself. Conclusions that rest on the book rest on the reviewers' readings, which agree with each other on every page cited.

How the verdicts were set (consistent rule)

• "Tested, not supported" is used only where a test on record bears on that specific claim. Here the only such record is Wood's visit (prism removed, file swapped for wood) plus failures to replicate by Kelvin, Crookes, Lummer and Rubens [web]. Wood's test is one visit, reported by Wood, and bears on the prism-spectrum (claim 3) and emitter (claim 6) claims directly; for the eye-brightness claim (5) it bears indirectly. I note that limit per claim.
• "Refuted by calculation" requires accepted laws applied to Blondlot's own stated numbers. It is used for claim 4 and for the penetration and dispersion parts of claims 2 and 3. It does not need Wood.
• "Never properly tested" is used where no neutral, blinded, instrumented test exists and none of the sources shows the claim was tested.
• Nothing in the record shows suppression: the Academy gave Blondlot the Leconte prize in 1904, citing his whole body of work rather than N rays [web] https://en.wikipedia.org/wiki/Ren%C3%A9_Blondlot; about 120 scientists published about 300 articles [web] N_ray page. The community was over-credulous first, then corrected itself. So "dismissed without testing" applies to none of these claims.

Corrections the arbiter adopts to Step 1

1. Claim 1 was reassigned by Blondlot himself from X rays to N rays (Preliminary Notice, Complementary Note 1: "the experiments are correct", only the identification changes) [read, per all three reviewers]. claims.md should not present it as a standing proof of X-ray polarization.
2. "X rays essentially unpolarized" is wrong. Barkla showed X-ray polarization by scattering in 1904-05 [web] https://en.wikipedia.org/wiki/Charles_Glover_Barkla (cited by all three reviewers; I did not re-fetch). Tube radiation (bremsstrahlung, "braking radiation") is also partially polarized [memory, high].
3. The wavelength unit is micrometres, and this is derivable, not guessed: 200 lines/mm gives spacing 5 micrometres, so a wavelength of 8.15 micrometres could not diffract at all; the Newton's-ring check counts about 0.589/0.0085 = 69 rings per yellow ring [read p.59; calc]. So 8.15 to 17.6 nm, photon energy 1240 eV nm / lambda = 152 down to 70 eV [calc].
4. The quartz indices 2.93, 2.62, 2.436, 2.29 belong to Blondlot's first (long-wavelength, Rubens residual-ray) interpretation, not to X rays. In January 1904 he reversed to wavelengths about 6000 times shorter [read note 7]. The same radiation family is thus claimed at both ends.
5. Copper "65 cm thick" is real (note 5, p.78) [read, contrarian]. Wood's file swap bears on claim 6, not just claim 3.
6. A seventh central claim is missing from claims.md: the photographic "objective" registration (pp.62-67; about 40 trials, one failure). Added below.

Disagreements reconciled by recomputation

Aluminium attenuation: the web-check result

From the CXRO table [web] https://henke.lbl.gov/optical_constants/sf/al.nff (f2 column, solid density 2.70 g/cm^3) [calc]:

So aluminium is a usable EUV filter only as a film of order 100 nm. The most generous window gives a transmission of 10^-8900 through the plate Blondlot says N rays pass; the physicist's example of exp(-150) was a vast understatement. The conclusion does not depend on any assumption except that the rays are photons in this energy range.

Claim-by-claim arbitration

Reviewers' verdicts are summarised as P (physicist), M (mathematician), C (contrarian).

Claim 1: spark gap near an X-ray tube changes with orientation; read as polarization

• Agreement: All three: the claim was reassigned by Blondlot to N rays; Barkla's X-ray polarization (scattering, ionization detectors) is real but unrelated to Blondlot's detector; no independent replication of the spark-orientation effect is on record in what anyone could read.
• Disagreement: P: "Never properly tested" in isolation. M: weak form "Plausible, untested", strong form (rotation by quartz and sugar) "Internally inconsistent" and "Not testable as stated" (the 360-degree rescue). C: weak form "Plausible, untested", strong form "Tested, not supported" and "Historically superseded".
• Stronger argument: M's, on the rotation sub-claims: quartz rotates only along its optic axis (across the axis it is a birefringent plate), a lump of sugar is polycrystalline, and "40 + 360k" degrees can absorb any measured angle. The missing control is a gap rotated with the tube off, or rotated with leads fixed (leads act as antennas). C is right that nobody tested the strong form as such, so "Tested, not supported" does not fit. P and C are right that the weak form was never tested neutrally.
• Tested / never tested / dismissed: NEVER PROPERLY TESTED (spark-by-eye as a polarimeter). Not dismissed without testing: Blondlot retracted the X-ray attribution himself. The reassigned version inherits claim 2's record.
• Final verdict (split):
• X rays can be plane-polarized: Established (Barkla), but not by this experiment.
• A tube-correlated, orientation-dependent change in a hand-tuned spark: Plausible, untested (confounders: coil leads, stray fields, humidity; no blind control).
• Spark-gap by eye measures polarization, and rotation by quartz or sugar: Internally inconsistent (rotation across the quartz axis, sugar lump) and Not testable as stated (modulo-360 rescue).
• Would settle it: rotate the gap with the tube off and with leads fixed, and with a coded and blinded tube-on/off sequence; record spark counts with a photodiode rather than the eye.

Claim 2: ordinary sources emit N rays that pass metals, wood, black paper but not water

• Agreement: All three: no equations to check; classification changes after the fact (rock salt opaque then transparent, lead opaque then "oxide", salt water transparent but water opaque, cold platinum opaque but hot transparent); Wood, Kelvin, Crookes, Lummer, Rubens failed [web]; the literal numbers are impossible.
• Disagreement: C cautions that Wood's test is one visit, with the account coming from Wood; P and M treat it as decisive. All three give the same verdict, so this is a matter of weight.
• Stronger argument: The calculation. It needs neither Wood nor trust. Aluminium transmission is 10^-8900 or less (table above); a 2000 K lamp emits these photons at 10^-176 to 10^-383 of the Boltzmann factor. I weigh Wood as real but modest evidence: the removal of the essential element is a genuine blind manipulation, but it is a single reported visit.
• Tested / never tested / dismissed: TESTED (several labs, a covert manipulation), not supported. Not dismissed untested. Not never tested.
• Final verdict: Tested, not supported; with the wavelengths of claim 4 the penetration part is also Refuted by calculation. If one ignores claim 4's wavelengths and says only "some unknown penetrating radiation", Wood's and the other labs' null results are the sole evidence against it.
• Would settle it: a coded blind trial, sources and screens hidden from the observer and randomised, scored by an independent photodetector; any real physical radiation passing 1.5 cm Al and 65 cm Cu would show in a thermopile or film. Blondlot's own thermopile and resistance tests found nothing (pp.29-32) [read, M].

Claim 3: refraction, reflection, dispersion (aluminium indices 1.04-1.85), diffraction

• Agreement: Snell's-law arithmetic and prism deviations are correct (I rechecked p.56). Wood's removal of the aluminium prism with unchanged reports is the direct test. Wavelength rising with index is opposite to ordinary dispersion, and Blondlot admits it (p.61).
• Disagreement: P: Tested, not supported, plus refuted and "internally inconsistent" (quartz 2.9 vs aluminium 1.04-1.85). M: Internally inconsistent (dispersion sign) plus Tested, not supported. C: Tested, not supported; not independently tested by instrument.
• Stronger argument: C and P on the label; M on the physics, but corrected by my CXRO calculation: n for Al can reach about 1.04 at the L edge, not 1.85, and no 8-18 nm absorption-free window exists. The quartz-2.9 vs aluminium-1.04-1.85 conflict is a change of interpretation (long to short wavelength), not an equation contradicting itself, so "internally inconsistent" is not warranted. The mathematician's remark that blanks in the lens table follow image-existence thresholds exactly is interesting: it is what an observer who knows the expected index would produce, consistent with expectation bias [read, calc by M].
• Tested / never tested / dismissed: TESTED (Wood removed the prism). Instrument measurement by anyone else: NEVER DONE.
• Final verdict: Tested, not supported (Wood prism test) and Refuted by calculation for the claimed magnitudes and dispersion at 70-150 eV (Al n at most about 1.04, absorption length 35 nm to 0.7 micrometre).
• Would settle it: repeat the prism spectrum with the prism randomly in or out, observer blind, spark position scored by a photodiode array.

Claim 4: wavelengths 8-18 nm by gratings and Newton's rings, error under 4 per cent

• Agreement: Fringe-angle arithmetic is correct and the method is self-consistent on paper; the unit is micrometres; the claimed 4 per cent error is violated by the n=1.19 row; the radiation, if photons, is 70-152 eV EUV; no independent measurement exists.
• Disagreement: Only in numbers (resolved above) and in which sub-label to emphasise.
• Stronger argument: The calculation (absorption and Boltzmann tail), now backed by the CXRO data. Newton's rings at 8 nm would also need surfaces smooth to a few nm and EUV normal-incidence reflectance of order 1e-3 or lower [memory, medium, not verified] so the ring count of 69 per yellow ring is not credible.
• Tested / never tested: The grating measurement was NEVER independently repeated. The consequences were tested via claims 2, 3 and 6.
• Final verdict: Refuted by calculation (for any electromagnetic reading of the table) with a minor Internally inconsistent (error bound vs own table). The suggestion that the unit could be read otherwise is closed.
• Would settle it: a grating or interferometer with a photodetector, no eye; any real 8-18 nm radiation would need vacuum and would not cross air.

Claim 5: detection by the unaided eye as increased glow (normal view), decreased (tangential view)

• Agreement: All three: the detector is the eye at threshold; the protocol (do not look at it, silence, no smoke, "some people never succeed", cannot be shown to an audience) makes failure the observer's fault by construction; the physiological explanation is supported [web, one-line from Wikipedia: "work in the period clearly showed the change in brightness to be a physiological phenomenon"]. I did not find the underlying papers.
• Disagreement: P gives a split verdict including "Plausible, untested" for the psychological effect; C and M give "Tested, not supported" with "Not testable as stated" for the protocol. M adds a point not in the others: the thermal tests (pp.29-32) show no temperature rise, so an incandescent-wire glow increase has no thermal mechanism and the radiative power needed (about 1.3 mW per M) would exceed what a filtered lamp provides.
• Stronger argument: C's: Wood's visit tested the prism and the emitter, not the brightness effect itself, so "Tested, not supported" rests mainly on failed replications and the contemporary physiological finding [web, indirect]. It still holds, but at lower weight than claims 2, 3, 6. P is fair that a real psychological effect of expectation on dim-light perception is plausible and was never isolated here.
• Tested / never tested: TESTED by others (Kelvin, Crookes, Rubens, Lummer failed [web]); the viewing-position and observer-selection rules are AS PROTOCOLLED close to unfalsifiable.
• Final verdict (split):
• A new radiation raises the glow seen by the eye: Tested, not supported.
• As Blondlot specifies it (only the sensitive observer, not demonstrable to an audience): Not testable as stated.
• The perceived brightness changes as a human psychological effect: Plausible, untested here.
• Would settle it: a fully blind, randomised sham/real trial with several untrained observers and a photometer reading the same phosphor.

Claim 6: bricks, stones, sea water, ox eyes store N rays; strained bodies emit indefinitely

• Agreement: All three: energy conservation does not rule it out (about 1.3 nW for a blade over 12 centuries); the problem is the quantum energy, 70-150 eV per photon against about 2e-4 eV of strain energy per atom; Wood's file-for-wood swap bears on emission; storage by bricks, sea water and ox eyes has no separate test; the 12-century Merovingian knife is a single unblinded anecdote (p.51).
• Disagreement: P and M add "Refuted by calculation" to the emission part only via the quantum energy; C declines that label without a stated N-ray quantum, noting the hardened/annealed comparison is cleanly testable.
• Stronger argument: M and P hold if the rays are 8-18 nm photons (Blondlot's own claim); C's caution applies only if one drops that claim, in which case the tool-steel claim reverts to a blind-testable comparison that nobody ran.
• Tested / never tested / dismissed: EMISSION: tested once (Wood, file vs wood) [web], not supported. STORAGE (brick, stone, sea water, ox eye, hyposulphite) and the 12-century duration: NEVER PROPERLY TESTED; not dismissed, but swept away when the programme collapsed.
• Final verdict (split):
• Emission by hardened or strained steel: Tested, not supported (Wood), plus Refuted by calculation if the photons are 70-150 eV.
• Storage in brick, stone, sea water, ox eye: Never properly tested.
• Indefinite (centuries) emission: Never properly tested; no decay law is given (storage times quoted range from 20 minutes to over 1200 years).
• Would settle it: a coded set of hardened and annealed chisels, observer blind, with a photodetector in vacuum or a thermopile.

Claim 7 (added by contrarian): photographic registration of N-ray action on a spark

• Agreement: Only C addressed it fully; M touched it under claim 5 (no densitometry, photogravures that do not render the originals, 1 failure in about 40 with no success criterion; the spark is tuned until it responds). P noted the photographs show the spark, not the rays.
• Stronger argument: C's: the alternation method with a metronome, several physicists present, and an admitted failure are real strengths; the weaknesses are an operator who knows which exposure is which, no densitometry, a screen carried by hand in the dark, and success defined as a responsive spark. Spark behaviour depends on humidity, draughts and sound (Blondlot cites Macé de Lépinay) [read, C].
• Tested / never tested / dismissed: NEVER PROPERLY TESTED (no blinded, independent photographic replication found; whether anyone attempted one is unknown). Not shown to have been dismissed untested; it fell with the rest.
• Final verdict: Never properly tested. It is the one claim where a mundane confounded effect (spark sensitivity to environment) may have been real; a positive finding would not establish N rays.
• Would settle it: an assistant who does not know the screen position randomises the exposures, plates are developed and scored by densitometer by a third party; add sham screens that look identical.

Contrarian's tested / never tested / dismissed distinction (kept visible)

Where the contrarian and critics each get the benefit of the doubt

• For Blondlot: the early optics arithmetic is correct; he had a record of real work (first measurement of the speed of radio waves, 1891, within 1 per cent, via Lecher lines [web] https://en.wikipedia.org/wiki/Ren%C3%A9_Blondlot); he reported null thermal tests and one photographic failure openly; nothing shows fraud; about 120 scientists published about 300 articles, mostly French [web].
• Against Blondlot: every result depends on an observer who knows the expected answer and is told to look away; failures are explained by bad observers; he never obtained an instrument signal, and his own table contradicts his error claim.
• Against the critics: Wood's account is the sole source of the covert-removal test in what I could read; the Berlin failures (Rubens, Lummer) are known to me only by Wikipedia. The conclusion does not hinge on them because the calculation stands on its own.
• The episode is the reverse of the suppression story: the Academy honoured Blondlot (Leconte prize 1904, for his total work [web]). The claim failed through lack of independent replication, not through blocked publication.

Corrected claim list

1. The effect of a hand-tuned spark near an X-ray tube changes with gap orientation. Blondlot first read this as X-ray polarization (Feb 1903), then (preliminary notice) reassigned it to a different radiation, N rays; the quartz and sugar "rotations" are internally inconsistent.
2. Many ordinary sources (Auer burner, Nernst lamp, flames, sun, strained steel, even sound and a clenched fist) emit "N rays" that cross aluminium, wood, black paper and up to 65 cm of copper but not water, wet paper or (at first) lead and platinum. The classification was adjusted after the fact (salt water, polished rock salt, lead oxide).
3. N rays are refracted by quartz (first indices 2.93-2.29, read as long wavelengths) and by an aluminium prism (eight discrete indices 1.04-1.85, read as short wavelengths), with wavelength rising with index. The printed quartz-lens example gives n of about 1.72, not the quoted mean 2.93 (QE revision).
4. N-ray wavelengths are 8.15 to 17.6 nm (0.00815-0.0176 micrometre, a unit derived from his grating geometry), measured with 50-200 lines/mm gratings and Newton's rings, with errors he believes under 4 per cent (his own table: up to 6.5 per cent half-range in one row).
5. N rays are seen as an increase of glow of a faint spark or phosphor viewed normally (decrease tangentially) by the eye, under rules (look away, silence, no smoke, "some never succeed") that make it impossible to show to an audience.
6. Bricks, stones, sea water and ox eyes store N rays; compressed or hardened bodies emit them continuously (a Merovingian knife for 12 centuries).
7. (Added) N rays can be registered "objectively" by the alternating-exposure photographs of a spark (about 40 trials, 1 failure).
8. (Added by QE revision) The velocity of propagation of the radiation, first measured in 1897-98 as "X" rays and reassigned to N rays in 1904, equals that of Hertzian waves and light. Verdict: Never properly tested (see QE revision).

Source gap statement in verdicts

Verdicts on claims 4 and 3's magnitudes rest on the book's own numbers plus accepted physics (web-checked CXRO data) and need no outside source. Verdicts that rely on "Tested, not supported" (claims 2, 3, 5, 6 emission) rest on [web] https://en.wikipedia.org/wiki/N_ray alone, a secondary source, because Wood's Nature note and the Berlin reports are not in the folder and I could not read them. Verdicts "Never properly tested" rest on an absence of evidence in what could be read and searched; web search was exhausted, so absence of a replication is "none found", not "none exists". Remaining [memory] items that do not carry a verdict: quartz rotation about 21.7 deg/mm (medium), mica birefringence about 0.04 (medium), steel stress about 1 GPa (medium), EUV normal-incidence reflectance of glass (medium), Reststrahlen identification of the first N-ray interpretation (the book's own p.13, [read]).

What would settle it

• One experiment (all claims): a randomised, double-blind trial of the central visual phenomenon: a source or sham, an aluminium screen or an identical-looking sham, behind an opaque panel switched by a third party; several observers, including Blondlot-style "sensitive" ones chosen beforehand, record their reports; a calibrated photometer reads the same phosphor continuously. If the observers' reports track the sham as well as the real source and the photometer shows nothing, the claim is a perception effect. This is the test Wood performed informally in 1904 and no neutral study has repeated formally [web, indirect].
• One calculation (claims 2-4): the absorption of aluminium at 70-150 eV, done here: transmission through the 1.5 cm plate is below 10^-8900. Any radiation of 8-18 nm is excluded; only a different wavelength, or no radiation, is left.
• Claim 1: tube off and leads fixed controls for the spark-orientation effect.
• Claim 6: coded hardened and annealed chisels, measured blind with an instrument.
• Claim 7: densitometer scoring of blind-randomised exposures.

Open questions

1. Was there a real physical effect behind the spark photographs (humidity, draughts, acoustic or thermal coupling) that was misattributed? Nobody tested it separately; the weak form of claim 1 and claim 7 are where it would lie.
2. How large is the expectation effect on dim-light perception in the exact protocol (averted gaze, silence, "passive" viewing)? The physiological explanation is asserted in secondary sources [web]; I could not find the primary experiments.
3. (CLOSED by QE revision) The printed lens pair does not reproduce 2.93; it gives about 1.72. See QE revision.
4. Did anyone besides Blondlot (Charpentier, Sagnac, Becquerel, others [web, C]) record N-ray data with something other than the eye? Unknown.
5. What exactly did Wood's test control? Details (number of observers, whether Blondlot knew) come only from Wood's own account, which I could not read.
6. Why do the prism-index blanks in the lens table follow the image-existence thresholds exactly (mathematician)? It is a hint that the observer knew where the image "should" be, but this is an inference, not demonstrated.

Sources

• [read] The book, via the reviewers' page-image reading (pages cited above are book pages; PDF page = book page + 25).
• [web] https://en.wikipedia.org/wiki/N_ray (fetched by me: Wood's prism removal and file-for-wood swap, failures of Kelvin, Crookes, Lummer, Rubens, "physiological phenomenon", about 120 scientists and 300 articles, Blondlot convinced in 1926).
• [web] https://en.wikipedia.org/wiki/Ren%C3%A9_Blondlot (fetched by me: 1891 speed-of-radio-waves measurement, Leconte prize 1904 for his whole work, retired 1910, "no one outside Nancy believed" by 1905).
• [web] https://henke.lbl.gov/optical_constants/sf/al.nff (fetched by me: aluminium f1 and f2 for 60-158 eV; attenuation lengths and refractive index computed from it [calc]).
• [web] https://en.wikipedia.org/wiki/Charles_Glover_Barkla (as cited by reviewers; not re-fetched).
• [web] https://en.wikipedia.org/wiki/Optical_rotation (as cited by the mathematician; not re-fetched).
• [calc] Boltzmann factors, grating spreads, lens-formula index, prism deviation, strain energy budget, Al absorption and index: Python in .venv.

QE revision (after qe.md, verdict MINOR)

Tags: [read] = extracted OCR text (_text/N-rays-00bf93.txt) or deep/N-rays-00bf93.md, [web], [calc]. No central verdict changes; one claim is added and two statements are corrected.

New evidence

1. Quartz-lens index (closes open question 3). [read: PDF pp.37-38, book pp.14-15, OCR text] The lens has f = 12 cm for yellow light, slit-to-lens distance l = 26.5 cm, focus l' about 13.9 cm. The text then says that "by varying the value of l, other values of l' are obtained", and that substituting these values in the lens formula gives 2.93 as "the mean value". Only this one pair is printed (a second lens, f = 33 cm, gave 2.942, no pairs printed). [calc] 1/f' = 1/26.5 + 1/13.9 gives f' = 9.12 cm, and n' = 1 + (1.544 - 1) x 12/9.12 = 1.72. So the one printed pair does not reproduce 2.93, and the other pairs are not printed. The OCR agrees with the mathematician's figures, so this is not a misreading. Status: a genuine gap in the book as printed (the mean cannot be checked). It affects the interpretation of claim 3's first (long-wavelength) phase only; no verdict depends on it.
2. Velocity of propagation (new claim 8). [read: PDF p.17, book p.viii] The Preliminary Notice says Blondlot's earlier measurements of the velocity of "X" rays (footnote: Comptes Rendus t. cxxv, pp. 666, 721, 763) "apply in reality not to X rays, but to N rays", that the velocity "was the same as that of Hertzian waves, and consequently of light", and that the determination "is now nothing more than a verification of an assured fact". Correction to the QE text: the footnote reads t. cxxv (125), not t. 135. Volume 125 is the 1897 volume [memory, medium], so this work predates the 1903 N-ray papers by about six years. [web] check failed: https://en.wikipedia.org/wiki/Ren%C3%A9_Blondlot has only the 1891 radio-wave (Lecher line) speed, 297,600 km/s, and nothing on X-ray velocity. I did not find the C.R. originals. - Reviewer questions: (a) maths: no numbers in the folder to check. (b) Established results: the speed of X-rays equals c, which is established by other work; agreement with c therefore does not support Blondlot's detector. (c) Tested: no independent replication of his method is on record that I found. (d) If true: nothing new. - Verdict: Never properly tested (a measured speed equal to c carries no weight when the detector is the spark-by-eye of claims 1 and 5, since nothing in the folder shows that the detector responded to X-rays or N rays at all; Blondlot himself calls it a "verification", i.e. a result expected beforehand). Not dismissed without testing; it was simply absorbed into the reassignment. Settles it: the original C.R. notes, plus a timing measurement with an instrumented detector.
3. Rubens letter, correct reading (corrects both the mathematician and the QE). [read: deep/N-rays-00bf93.md section 1; PDF p.55, book p.30] The sentence "M. H. Rubens made the same observation, as he was kind enough to write me, his apparatus being much more sensitive even than mine" follows both Blondlot's null thermopile result and his CaS brightening observation. The antecedent is ambiguous: Rubens could have confirmed the null deflection, the brightening, or both. It is a private, undated, unquoted second-hand report. Rubens' own later failure to find N rays is [web] via https://en.wikipedia.org/wiki/N_ray (already used above); the deep read's Rubens-Taylor detail is [memory, medium], unverified. For claim 2's replication column: do not cite the letter as independent confirmation of anything.
4. Named Nancy corroborators. [read: deep/N-rays-00bf93.md section 2] Bichat (ethyl ether under tension emits N1 rays, PDF p.104; lead opacity is due to oxide and carbonate, C.R. t.138 p.548, 29 Feb 1904, note 5); Charpentier (physiological work, C.R. t.137 p.1049, 24 Dec 1903; wet cardboard transmits rays from a knife blade, notes 11-12); Gutton (magnetic field and EMF act on a faint luminous surface, p.106); Macé de Lépinay (sound brightens a phosphor, p.106; the printed date "January 11, 1902" is probably a misprint for 1904, since t.138 is the 1904 volume); Sagnac (a competing long-wavelength reading, p.85). Becquerel appears only as the discoverer of radioactivity, not as an observer. None of these gives numbers that independently replicate a Blondlot measurement, and Gutton and Macé de Lépinay report effects on luminous surfaces (not N rays) that Blondlot says are analogous. So "no independent replication" stands, but the record shows colleagues at Nancy reporting visually based confirmations (Bichat, Charpentier), consistent with the over-credulous first phase described above.
5. Voice and fist. [read: PDF pp.113-114 per QE; p.106 per deep read] Blondlot says the spots become distinct when one speaks aloud, whistles, or brings the clenched fist near the screen (plus the p.106 statement that the audience would see contradictory effects). This is the textual basis for the corrected claim 2 list ("sound and a clenched fist"); the book does not say "the human body" as such. Not a verdict change.
6. Aluminium index at the L edge. Largest Al refractive index in 64-153 eV is 1.037 at 72.6 eV [calc from the CXRO table already cited, as in the primer and example.py], not "about 1.02" (1.023 at 72.0 eV). Edited in place above. The conclusion is unchanged: nowhere near 1.85, and the band is strongly absorbing.
7. Smaller items. The 3 July 1904 sun thermopile test is inside a note dated 1903, so the translation contains later additions [read: p.55]. The thermal limit is 1/60 K (1/4 division at 15 divisions per degree), not the digest's "0.025 C". N1 rays (indices 1.004-1.041, wavelengths about 0.003-0.0081 micrometre, PDF pp.97-104) continue the same rising n-versus-wavelength trend, which strengthens the dispersion anomaly. The page offset between PDF and book page is +23 for book pp.3-16, +25 for pp.17-62 and +27 to +33 after p.63; the "+25" in the Sources list below holds only for the middle range.

Source gap, updated

The folder text is not "without a text layer": Step 0 produced a 200,936-character OCR text which reads well (QE spot checks). What is missing remains outside the folder: the French C.R. originals, Wood's Nature note, the Berlin reports, and the 1897-98 velocity notes. "Tested, not supported" verdicts still rest on https://en.wikipedia.org/wiki/N_ray alone.

Changed verdicts