Blondlot: Science Contrarian review

Reviewer role: Science Contrarian (judge by experimental evidence, not consensus; separate "tested and failed" from "never properly tested" from "dismissed without testing"; apply the same rigour to the researcher and to his critics).

Source gap (repeated from Step 1). The folder holds one document: the 1905 English translation of Blondlot's "N" Rays (Google Books scan, 161 pp., no text layer). I re-read key passages myself from page images (title/Preliminary Notice pp. vii-viii, spark apparatus p.3, wavelength table p.58, Newton's rings pp.59-61, photographic method pp.62-67, storage/emission pp.45-52, viewing instructions p.82-83, notes pp.77-78) [read: page images]. Everything about Wood, Rubens, Lummer, Kelvin, Crookes and later history is NOT in the folder and rests on [web] Wikipedia (https://en.wikipedia.org/wiki/N_ray) plus [memory]; I could not read Wood's Nature note or the Berlin reports themselves.

Corrections to Step 1

1. Claim 1 as worded is not what Blondlot ended up claiming. The book's Preliminary Notice (p. vii) says the first paper's title ("Polarization of X rays") is a misattribution: the radiation studied was N rays, not X rays; Complementary Note 1 says "the experiments are correct, and the rectification only applies to the nature of the rays". Claim 1 should read: the spark effect near a tube is caused by a new radiation (N rays), not X rays, and that radiation is plane-polarized. Step 1's facts.json does partially say this but claims.md hides it. [read: p.vii, note 1]
2. Unit of the wavelength table (claim 4): Step 1's "presumably micrometres" is right, and can be shown, not just inferred. With a 200 line/mm grating the spacing is d = 5 micrometres. A wavelength of 0.00815 mm (8.15 micrometres) would exceed d, so no diffraction order exists (sin theta = lambda/d > 1). In micrometres, sin theta for the 10th-order fringes Blondlot says he measured (p.58) is 10 x 0.00815/5 = 0.016 (about 0.9 degrees), measurable. The Newton's-rings check (p.59) counts how many times the N wavelength fits in yellow light (0.589 micrometres): 0.589/0.0085 = about 69 rings, consistent with micrometres. So 8.15 to 17.6 nm. Photon energy E = 1240 eV nm / lambda gives about 152 eV down to 70 eV. [read: pp.58-59; calculation]
3. Error-bar wording (claim 4). Step 1 says "claimed relative error under 4 per cent". The text (p.60) is weaker: "I cannot with certainty indicate the degree of approximation... I believe, nevertheless, that the relative errors do not exceed 4 per cent." Also the three gratings disagree among themselves by up to about 6 per cent (index 1.04: 0.00795 to 0.00839; index 1.19: 0.0093 vs 0.0106 is about 14 per cent). [read: p.58 table]
4. Missing central claim. claims.md lists no claim for the photographic "objective" registration (book pp.62-67, about 40 trials, one failure), which is Blondlot's strongest evidence. Added below as claim 7. Step 1 mentions it in facts.json but states it as "the spark photographed, not the rays".
5. Unsupported-looking figure confirmed. The "copper 65 cm thick" in facts.json is real: note 5, p.78 says copper transmits N rays "even when used in thicknesses of 65 cms" [read: page image p.78]. Step 1 gave no page; this is the single most damaging number in the book for claim 2.
6. Wood details. Step 1 says Wood "swapped the supposed source for an inert piece of wood". Wikipedia says: removed an essential prism, and swapped a large file (a claimed emitter, claim 6) for a piece of wood, with N rays still reported [web] https://en.wikipedia.org/wiki/N_ray. The file swap bears directly on claim 6, which Step 1 did not note.
7. No other factual errors found in facts.json; quartz indices 2.93/2.62/2.436/2.29 not re-checked [unverified here].

Claim 7 added: A spark's glow, photographed on a plate exposed alternately with and without N rays (a wet screen interposed), is darker/lighter by an amount that shows the rays objectively, about 40 trials, one failure [read: pp.62-67].

Shared facts used below

• Detector: a tiny spark tuned by hand until it is "sensitive" (p.65), or a dim phosphor, judged by eye; rules: do not look at the source, silence, no smoke, look like an impressionist painter, "some people never succeed" (pp.82-83); audience demonstration "illusory" (p.73) [read].
• Mainstream outcome: Lord Kelvin, Crookes, Lummer and Rubens failed to reproduce; Wood's visit, Nature, 29 Sept 1904; about 120 scientists / 300 papers claimed N-ray detection; by 1905 no one outside Nancy believed; Blondlot reportedly still convinced in 1926 [web] https://en.wikipedia.org/wiki/N_ray.
• X-ray polarization is real but was shown differently: Barkla, scattering from gases, Nature 1904, Phil. Trans. 1905 [web] https://en.wikipedia.org/wiki/Charles_Glover_Barkla.
• Energy scale: lambda 8-18 nm is extreme ultraviolet / soft X-ray. Thin aluminium foils (order 100 nm) are used as EUV filters; such photons are absorbed within tens of nanometres to micrometres of any solid [memory, medium; the web check (henke.lbl.gov) returned nothing on aluminium, so unverified].

Claim 1: spark brightness depends on orientation near a tube; the radiation is polarized

• a. Mathematics. Almost none. The inference "brightness depends on gap orientation, therefore polarized" is a logical step, not a calculation. Blondlot's own rotation-by-quartz readings are later admitted (note 2) to be possibly remainders modulo 360 degrees, which is a rescue assumption [read: note 2]. Not inconsistent, but unconstrained.
• b. Experimental consistency. A tube does influence a spark gap: X rays ionise air and photo-eject electrons; ultraviolet lowers the sparking threshold (Hertz effect) [memory, high]. So an orientation-dependent change near a coil-driven tube is not absurd, since coil leads, tube geometry and stray fields are orientation-dependent. What is inconsistent is calling it X-ray polarization measured this way: tube X rays are only partially polarized, Barkla's result came from scattering [web]. Blondlot retracted the X-ray attribution himself.
• c. Tested? Not by a neutral, blinded protocol that I can find. Blondlot's April 1903 photographs of polarization (Figs. 8-9) were explicitly NOT done with the alternation method (p.67) [read]. Failures to reproduce by others concern N rays generally [web]; I have no source on anyone replicating this specific polarization experiment.
• d. If true. A new polarized radiation emitted by X-ray tubes distinct from X rays; would add a spectral band, not break Maxwell.
• Verdict (split). Weak form (a tube-correlated change in spark behaviour): Plausible, untested in the form described, with confounders unaddressed (the coil, the induction-spark and the observer share one apparatus; no blind control). Strong form (N rays, plane-polarized, a new species of light): Tested, not supported (see claim 2) and Historically superseded but instructive.

Claim 2: ordinary sources emit "N rays" which pass metals/wood/black paper but are stopped by water and wet paper

• a. Mathematics. Internally uneven but no equation fails. Self-contradiction in the physical story: wet paper (a thin film of water) stops the rays, yet a 4 cm vase of salt water passes them "without sensible weakening", and the eye (98.6 per cent water) is explained by salt (p.45) [read]. That is an ad hoc patch, not an inconsistency of equations.
• b. Experimental consistency. Conflicts with the measured world if the stated wavelengths are real. Using the table (claim 4) the photons are 70-150 eV. Real aluminium, copper, wood and black paper do not transmit EUV in centimetre thicknesses (they absorb within roughly nanometres to micrometres) [memory, medium]. Rough bound: if the attenuation length were even as large as 1 micrometre, transmission through 1.5 cm of Al (book p.50) would be exp(-15,000), about 10^-6500; through 65 cm of copper, vastly less. Blondlot also notes air does not absorb his rays while Schumann's short-wave rays are absorbed by air (p.61) and postulates absorption bands in between [read]; air is in fact opaque to EUV [memory, high]. No photodetector, thermocouple, or film (other than via a spark) registered N rays independently.
• c. Tested? Yes, by others: Kelvin, Crookes, Lummer and Rubens failed to see them; Wood's 1904 visit removed the prism and still got reports [web]. Replication in Nancy and by about 120 authors [web] is the positive side; Wikipedia attributes the change in brightness to a physiological phenomenon [web]. Caveat from my reading of the critics: Wood's test, as reported, is a single visit with an uncontrolled, unblinded-by-protocol design and its record rests on Wood's own account; Rubens' and Lummer's papers I could not read. I would not treat Wood's anecdote as a full controlled study, but its design (a covert removal of the essential element) is a genuine blind manipulation, and it is the right kind of test.
• d. If true. Matter transparent to a 10-nm band, contradicting the absorption spectra of Al, Cu and wood; X-ray and EUV absorption tables would be wrong at the heart of their data. Extremely costly, so extraordinary evidence is needed.
• Verdict. Tested, not supported (not dismissed without testing, since several labs tried and a covert test was performed). Combined with the stated wavelengths and transparency claims, the strong form is also Refuted by calculation (attenuation figures above, tagged [memory, medium]).

Claim 3: N rays are refracted, reflected, diffused, dispersed into a spectrum (aluminium indices 1.04-1.85), diffracted

• a. Mathematics. The grating arithmetic is consistent: wavelengths from three gratings agree within a few per cent for the first row, within up to about 14 per cent in the second (p.58); ring method 0.0085 vs 0.0081, 0.017 vs 0.0176 (p.59). Prism angle 27 deg 15 min is not independently checked here. A real difficulty: wavelength increases with index (p.61, acknowledged as "contrary to luminous radiations"), i.e. an anomalous-dispersion pattern across the whole spectrum with no stated physical mechanism, and indices above 1 for aluminium at 70-150 eV, where real refractive indices of metals are slightly below 1 [memory, medium]. Not formally self-contradictory.
• b. Experimental consistency. Refraction in quartz at index about 2.9 (first-series claim) is far from quartz's visible index of about 1.54 [memory, high]; it does not match any quartz measurement I can name.
• c. Tested? Not independently in the sense of a measured angle that someone else's instrument recorded. The only detector was the spark/phosphor eye judgement. Wood's prism removal is a direct test of the refraction claim: the "dispersed" lines were still reported without the prism [web].
• d. If true. A new optics regime in a band where mainstream expects no refraction beyond n slightly under 1.
• Verdict. Tested, not supported (Wood's prism removal). The prismatic spectrum measurements are Not properly tested by independent instrument.

Claim 4: wavelengths 0.008-0.018 (micrometres)

• a. Mathematics. Grating geometry works only for micrometres (see Correction 2). Errors: three-grating spread, up to about 14 per cent in one row (not the "under 4 per cent" Blondlot believes). The 10th-fringe locations were read by eye from the same spark.
• b. Experimental consistency. At 8-18 nm the energies (70-150 eV) are in the EUV band. Transmission through 1.5 cm Al, 3 cm oak and 65 cm copper is contradicted by every X-ray/EUV absorption measurement [memory, medium]. Also: a 200 line/mm grating at 8 nm and fringes tens of orders wide require a coherent narrow beam, which a Nernst lamp does not emit at 150 eV (a lamp at about 2000 K emits negligible photons at 150 eV: Planck factor exp(-E/kT) with kT about 0.17 eV is exp(-880) [calculation; temperature [memory, medium]).
• c. Tested? Not independently. No one else's grating measurement is reported.
• d. If true. The Nernst lamp, flames and the sun would have to emit strongly in a band the thermal spectrum can barely populate.
• Verdict. Refuted by calculation (energy budget: Planck exponent about -880 for a lamp at about 2000 K; absorption lengths above) for any real electromagnetic reading of the table; as an observation the numbers are Tested, not supported with the underlying detector.

Claim 5: N rays detected as a small increase in glow of a spark or phosphor seen normally (decrease tangentially), by the unaided eye

• a. Mathematics. Angular dependence (increase normal, decrease tangential, nothing in between, p.72-73) is stated without a functional form; the "inverse" N1 rays are defined to mirror it. A symmetry-convenient rescue, not a derivation.
• b. Experimental consistency. Eye thresholds near the dark-adapted limit are notoriously variable; apparent brightening of faint patches by fixation, attention or small eye movements is well documented in vision research [memory, medium]. Wikipedia reports the change was physiological [web].
• c. Tested? By Wood, and reportedly by Kelvin, Crookes, Rubens, Lummer [web]. Blondlot's own rules make the claim hard to falsify: viewing position matters, "only some succeed", an audience cannot see it, tobacco smoke masks it (pp.73, 82-83) [read]. These are exactly the properties of an observer-dependent effect; a negative outcome can always be explained by a bad observer.
• d. If true. A new physiological response to a new radiation; no mainstream physics changes.
• Verdict (split). Eye-detection form: Tested, not supported (Wood; Berlin) and, as protocolled, close to Not testable as stated because of the escape clauses. Photographic form: see claim 7.

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

• a. Mathematics. No energy balance is given. Blondlot argues the energy is "most likely borrowed from the potential energy of the strained state" and "doubtless very slight" (p.52) [read]. A Gallo-Roman blade emitting "for more than twelve centuries" (p.51) is the extreme case. No inconsistency in equations because there are none; but "indefinite" emission with "no abatement" without a source of energy is an unquantified claim, and the contrast between hardened steel (active) and annealed (inactive) is a clean, in principle checkable hypothesis.
• b. Experimental consistency. No known radiation from hardened steel, Rupert's drops or compressed wood. Piezo/triboluminescence and fractoluminescence are real but short-lived and brighter [memory, medium]. Nothing records 12-century passive emission.
• c. Tested? Wood swapped a claimed emitting file for wood and observers still reported N rays [web], which tests this claim directly. Not tested otherwise as far as I know; no instrument-based test exists.
• d. If true. A new non-radioactive, non-chemical perpetual emission; strain-energy storage would need re-treatment.
• Verdict. Tested, not supported (Wood file/wood swap), with the specific "indefinite for 12 centuries" and "ox eye / sea water accumulator" parts Never properly tested; the sources are Blondlot's anecdotes and nobody else's measurement.

Claim 7: photographic registration of N-ray action on a spark (about 40 trials, one failure)

• a. Mathematics. None required; it is a comparison of two half-plate exposures.
• b. Experimental consistency. Spark discharge is exquisitely sensitive to humidity, air currents, temperature, electrode surface, vibration and acoustic effects (Blondlot himself cites Macé de Lépinay on sound, p.73) [read, memory medium for general claim]. The wet-paper-wrapped lead screen is carried by the same frame the experimenter moves in the dark (p.63-64): the "with rays" and "without" halves also differ by the screen's position, wetness and the motion. A sham screen exists only in variation (zinc screen transparent to N rays, p.66), not a full blinded sham. The one failure is explained post hoc as an insensitive spark (p.67); with success defined by tuning the spark until it responds (p.65), that makes the protocol self-selecting.
• c. Tested? Alternation of exposures every 5 seconds (p.64) and "several eminent physicists witnessed" (p.66) are real strengths. But the experimenter knew which source was on; no independent photographic replication in the sources I have; whether anyone else tried the photographic version I could not verify (unknown). Plates are shown only as photogravures: "the striae and most of the spots do not exist in the original" (caption p.66), so readers cannot check the density difference.
• d. If true. Same as claim 2.
• Verdict. Never properly tested (no neutral, blinded, independent photographic replication found). Not "dismissed without testing" generally; the visual claims were tested, this one apparently set aside after Wood.

Contrarian assessment

Strongest honest case for Blondlot. He was a serious experimentalist (earlier measured the speed of electric waves, [memory, medium]); the alternating photographic protocol with metronome timing is more careful than most 1903 work; about 40 trials with admitted failure; independent labs (Nancy colleagues, Charpentier, Becquerel, Sagnac and others) reported effects [web]; and he warned against pitfalls such as polished vs ground surfaces. A real, mundane effect (humidity/thermal/acoustic influence on a delicate spark, or the eye's attention-dependent response) may well have been present, and was misattributed to a new radiation. Nothing in the book proves fraud.

Where the claims fail even on the contrarian's own standard (evidence, not consensus). 1. The effect never left the observer's perception: no thermocouple, film (other than via a spark), counter or spectrometer registered N rays. 2. The experiment's response to protocol: failures are explained by a poor observer, an insensitive spark, smoke, wrong viewing angle (pp.67, 73, 82-83). 3. Wood's covert tests (prism removal, file/wood swap) are conditions where a real physical effect would have vanished; it did not [web]. 4. The numbers are physically impossible if taken literally (Planck exponent about -880, transmission through 65 cm copper). 5. Blondlot's own admission that the effect cannot be shown to an audience.

Was it dismissed without testing? No. It was tested (Kelvin, Crookes, Lummer, Rubens, Wood) and failed [web]. What was not done, as far as I can find, was a formal blinded trial that Blondlot agreed to in advance and a published instrumental photographic replication; that gap is a fair complaint, but it does not rescue the claims. The rejection was not blocked by suppression: the Academy awarded him a prize [memory, medium; unverified here], and about 300 papers were published [web]. The social story is opposite to cold fusion's: here the community was over-credulous first, then corrected itself.

Lesson for the reader. A faint, subjective signal near the threshold of perception, a protocol with escape clauses, and an expected outcome known to the experimenter is the classic signature of pathological science; the way to settle it is a blind, quantitative test (see claim 7).

Summary table

Source key: [read] book pages as cited; [web] https://en.wikipedia.org/wiki/N_ray and https://en.wikipedia.org/wiki/Charles_Glover_Barkla; [memory] confidence marked inline; the web check on EUV absorption of aluminium failed (page had no aluminium data; search budget exhausted).