Kilner-Walter: Arbiter (Step 3)

Project: Walter J. Kilner, The Human Atmosphere, or the Aura Made Visible by the Aid of Chemical Screens (Rebman, 1911), the dicyanin ("spectauranine") screen and the complementary-colour (C.C.) band. Inputs weighed: facts.json, claims.md, physicist.md, mathematician.md, contrarian.md, deep/, sites.md, research/ notes. I re-read the table pages of the book as images (PDF pp.80-81) and recomputed the numbers on which the reviewers differ. Printed page = PDF page minus 19, as in the reviews. Tags: [read] = the book or saved files; [web] = URL given; [memory] = recalled, confidence given.

Revised after QE (2026-10-04). Sections 1-10 are the first arbiter pass. Section 11 ("QE revision") adds the new deep reads and the QE findings and governs wherever it differs from sections 1-10; the places it changes are marked "[Revised: 11.x]" below, and the Corrected claim list (section 7) has been rewritten to match.

Source gap note. The book's OCR text is readable throughout, so there is no source gap for the 1911 text. Three things bear on verdicts but were never read by any role: the BMJ review of 6 Jan 1912 (PMC2344221; blocked for the Investigator, the Contrarian and for me, which returned an HTML challenge page instead of the PDF), the 1911-12 Archives of the Roentgen Ray papers (read only by the Investigator in a browser; no raw table saved), and the NY Herald, NYT and Annales reports of the Chicago (O'Donnell) trial. Where a conclusion rests on these it rests on the Investigator's notes or on Wikipedia alone, and I say so.

---

1. Corrections to the Step 1 and Step 2 record

1. "Case 88 and Case 21 appear twice" (claims.md). The Physicist found that "Case 21 (continued from page 189)" is a continuation, not a reuse. Drop the item unless someone finds a real duplicate. [read, via Physicist] [Revised: 11.1. Half right: "Case 21 (continued from page 189)" is a real continuation, but "Case 88" is an OCR misreading of the digit 3, so it is two different cases (33 and 38).]
2. Claim 4 overstates Case 40 (Mathematician; I accept this). In Case 40 the text says "The C. C. band showed nothing unusual"; the dead-foetus finding rested on the texture of the Inner Aura plus history and later palpation, and the dead infant was delivered two months later [read: PDF p.321-323, via Mathematician]. Case 40 is evidence offered for claim 3, not for the C.C. band in claim 4. "Dead foetus" should be removed from claim 4's wording or marked "Kilner's text attributes it to texture".
3. Claim 2's "outside the ordinary visible spectrum" is hedged in 1911 ("perhaps, but this is only supposititious", p.107-108) and firm only in the 1920 preface ("an ultra-violet phenomenon"). Both forms are judged below. (Physicist; agreed.)
4. The observer's own visual system. facts.json says it "is not on his list". More exactly, it is not among the four listed causes, but Kilner wrote that the C.C. colours are "entirely subjective" (p.121/PDF p.140), that eye movement spoils the afterimage, that the eye fatigues in minutes, and that the aura's blue look is "most likely" due to the screen and its long action on the retina (p.138). So it is "noticed but not controlled". (Physicist, Contrarian; agreed.)
5. "No blinding" is too strong for the Appendix. In Experiments 5 and 6 the location was hidden from the observer. What was not blinded: colour, the subject's cue channel (she announced when the spot was ready and could see the observer), and there were no sham trials. (Contrarian; agreed.)
6. Experiment count. Case 42 has six experiments, not five; Experiments 5 and 6 test location, not colour; "G.2 = B.G." is an inference from Table II. (All three agree.)
7. Claim 6 / Table I. See section 3 (heredity): the Mathematician's count of 18 parent-child pairs treats F.X. as the parent of G.X. and E.X.; the table labels them "Granddaughter(s)". My recount differs.

---

2. Recomputed numbers where the reviewers differ

2a. The chromatic-focus paper (Kilner 1911-12, +1 to +6.5 divisions; 1 division = 1/1000 in = 25.4 micrometres, µm)

[Revised: 11.3. The instrument assumed below (a 1/4-inch objective, from an OCR "14-inch") is most likely wrong: the 1920 account of the same experiment says 1.5 inch. Depth of focus is also the wrong yardstick for a setting task. The "3 to 17 depths of focus" conclusion is withdrawn.] The reviewers use the same formula (a stage shift Δz moves the intermediate image by M²Δz, which the eyepiece of focal length f converts to a change in vergence, in dioptres D, of M²Δz/f²) but different assumed instruments, and the Contrarian's depth-of-focus remark is off by an order of magnitude. - Physicist: M = 10, f = 25 mm gives 4.1 D per division. Recomputed: correct, but a 1/4-inch objective has M of about 25 (tube length 160 mm / focal length 6.35 mm), not 10. The M = 10 figure is the low bound. - Mathematician: 1 D corresponds to about 0.4 to 4 µm of object focus for M = 25 to 40 and f = 25 to 50 mm. Recomputed: M = 25, f = 50 mm gives 6.4 D per division (4 µm per D); M = 25, f = 25 mm gives 25 D per division (1 µm per D); M = 40, f = 25 mm gives 65 D per division. So 2.5 divisions (64 µm) is 16 to 160 D, in line with the Mathematician. [calculation, this review; instrument values are assumptions, since the paper's objective and eyepiece were not seen] - Depth of focus (diffraction, about wavelength/NA², wavelength 0.55 µm): NA 0.25 gives about 9 µm; NA 0.17 gives about 19 µm. These are 0.35 to 0.75 of one division. The Contrarian's "within the plausible depth of focus ... tenths of a millimetre" [memory, medium] is not right for a 1/4-inch objective; tenths of a millimetre belongs to a low-power objective. Retract that sentence. - Net result, on which all three reviewers' substance agrees: a reported shift of 1 to 6.5 divisions (25 to 165 µm) is about 3 to 17 depths of focus. It is far too big to be a change of the eye's refraction (the eye's chromatic spread is about 2 D over the visible range and its whole accommodation 10 to 15 D [memory, high]; tens of dioptres are needed). It is also many times the instrument's own focus tolerance. So it is neither "just noise" at the level of one depth of focus nor a credible optical change in the eye. What it could be (setting criterion drift after staring at a bright light through a dark screen, accommodation shifts, selection of 12 of 30 observers) cannot be decided without the paper's table. Open, needs the primary paper (pp.344-352).

2b. Visibility by scattering (Mathematician)

The Mathematician required optical depth τ of at least 0.02 × 0.03 = 6×10⁻⁴ for a 2% contrast against a black ground of reflectance 0.03. Single scattering with an isotropic phase function (1/4π per steradian) against a Lambertian background (radiance ρE/π) gives τ ≈ 4 × contrast × ρ = 2.4×10⁻³, four times larger. For a 0.25 m layer, the scattering coefficient is then about 1×10⁻² per metre, about 740 times clean-air Rayleigh scattering (taken as 1.3×10⁻⁵ per metre [memory, high]), equivalent to a visibility of about 0.4 km by Koschmieder's rule. The Mathematician's figures (170 times; 1.6 km) are low by that factor of 4, but the conclusion is unchanged: a visible reflecting haze would have to be mist-like in its particle load, and Kilner himself says it does not behave like vapour (p.91-92). The calculation tests nothing about Kilner's claim; it is a plausibility bound under stated assumptions.

2c. Chance of the two blind location trials (Experiments 5 and 6)

• Physicist: about 3% per trial (abdomen 30 × 25 cm, a hit within 2.5 cm). Mathematician: about 1% abdomen, 0.5% thorax (hit within 0.5 in). Contrarian: 1.2% (8 × 8 in, 0.5 in radius). Recomputed: 2.7% for a 2.5 cm radius over 750 cm²; 0.7% for a 1.27 cm (0.5 in) radius over 750 cm²; 1.2% for 0.5 in over 413 cm². All three are right for their stated assumptions; they differ only in tolerance and area.
• None of these is the probability of the result, because (i) Kilner searched a field of dark patches rather than guessing at random (he says small dark patches are common, p.147, and subject D. already had a sacral patch), so the right per-trial chance is about N times the single figure for N candidate patches (N = 5 gives about 3.5% to 14%); (ii) the subject saw the aura, was in view, announced when to look, and so offered a cue channel; (iii) the number of discarded or unreported trials is not given ("No more experiments of this nature were possible", p.338); (iv) the near miss (0.5 in off, a diffuse patch about 2 in across) was scored as a hit. Two trials cannot support a statistical claim either way. The reviewers' numbers are reconciled: about 1% to 3% per trial for one random guess; unknown, and probably several times larger, for the procedure actually used.

2d. The heredity tables (the one place with counts)

[Revised: 11.4 and 11.5. It is not the only place with counts, and the sibling count was incomplete: Table I holds sibling sets too.] I read Tables I and II as page images (PDF pp.80-81). Classes: W = wide, A = average, N = narrow. - Table II (brothers and sisters), 9 pairs: B.T./N.U. (W,W) same; N.D./E.D. (W,A) differ; L.N./Q.N. (A,W) differ; G.B./I.B. (A,W) differ; N.G./B.G. (W,A) differ; N.N./I.N. (A,A) same; Q.I./O.I. (A,A) same; O.N./N.N. (W,W) same; H.K./E.E. (W,W) same. So 5 of 9 agree. This matches the Mathematician. - Table I (parents and children): my count is 15 parent-child pairs, all in the same class (12 of 12 families). The Mathematician has 18 pairs with 16 agreeing, because he counted G.X. (A) and E.X. (W) as children of F.X. (N). The table lists F.X. (12½), G.X. (9½) and E.X. (7½) as "Granddaughter, dull", "Granddaughter", and "Hysterical-Case 14", so they look like three grand-daughters of the X family rather than a parent and two children. If so, they belong with the siblings: three girls, three different classes (0 of 3 agree). The table layout does not fully decide this; I report both readings. - Chance agreement from the pooled class frequencies (48 entries: A 24, W 15, N 9; probability two random people agree 0.38; independence assumed, which does not hold since parents recur): 15 of 15 agreeing has probability about 6×10⁻⁷; 12 of 12 families about 1×10⁻⁵; 5 of 9 siblings about 0.23 (not unusual). A one-sided Fisher exact test of parent-child (15 of 15) against siblings (5 of 9) gives p = 0.012; with the three X girls added to the siblings (5 of 12) p = 0.0009. [calculation, this review] - A detail neither reviewer reported: footnote 1 to Table I says the mother N.N. "would have been considered wide except that it was narrowed by the thighs and legs", and her daughter's aura "is on the border of being wide". So one of the 15 agreements is a post-hoc classification; without it the pair would read W against A-or-W. The result is robust to dropping that pair (14 of 14). - Reading. Parent-child agreement is far above chance; sibling agreement is not, although siblings share as much ancestry. Candidate explanations that do not need a hidden variable: an unblinded observer who knew the relationships and had no stated scale for infants (3 of the 15 children are 4 months old) or for children generally ("decided on by experience only", p.78-79); post-hoc footnotes. Against that reading: Kilner states the tables hold "all the examples ... investigated" (p.79), which argues against cherry-picking, and he did publish the discordant sibling pairs. A heredity explanation would predict sibling agreement comparable to parent-child agreement, so the asymmetry favours a method effect, but 9 sibling pairs is too few to say so firmly. The datum is kept in the record, as the Mathematician urged.

2e. Other numbers checked

• Etheric Double angular size (Physicist): 1/8 to 3/16 in at 2 m is 4 to 5 arcminutes, close to the eye's resolution limit; the 2 m distance is the Physicist's assumption, not in the text. Accepted as stated.
• Air-boundary-layer refractive index change Δn of about 1×10⁻⁵ (Physicist): n−1 = 2.9×10⁻⁴ at room temperature times ΔT/T = 10/300 gives 1×10⁻⁵. Correct.

---

3. Web checks done by the arbiter (budget 8; 5 used)

---

4. Claim-by-claim

Claim 1: A faint structured haze (Etheric Double, Inner Aura, Outer Aura) surrounds every body and becomes visible after viewing through a dicyanin screen

Agreement. All three reviewers split the claim into a weak form (a screen-primed observer in dim light reports a faint haze or edge band round a body) and a strong form (an objective, body-specific, structured emanation seen through the screen). All three give the same two verdicts: weak form plausible and untested in a controlled way; strong form never properly tested. All three keep the same evidence: H. C. S. and Goodspeed (a haze "more pronounced with an inanimate hand-towel than with the human being"), Kilner's reply (never seen a haze round an inanimate object except a magnet or radio-active substance, p.240), the Paris failures, the Graz positives, the Chicago second-hand report, and the publisher's journal.

Disagreement and what is stronger. - The Physicist leans to a perceptual cause (afterimage, chromatic adaptation, rod vision, Mach bands, veiling glare, Troxler fading) and supports it with Kilner's own words (p.138: the blue look is "most likely" due to the screen and its action on the retina; p.112: the explanation is "completely hypothetical and without proof"). The Contrarian builds the best case for the strong form: independent groups reported seeing something, Kilner reported failures, and no neutral test was run. The Mathematician adds that a filter that only removes wavelengths cannot add a band, so the screen can act only on the eye's state, and that a reflecting haze would need mist-like scattering (section 2b). - Which argument is stronger. The Contrarian's point that "something is seen by some people" has informal support in more than one place is right, and the Physicist concedes it. But that support is exactly what the perceptual reading predicts, and the one independent datum with a control object (the towel) points against body-specificity. The perceptual reading therefore explains more of the record at no cost; the strong form has no independent support beyond Kilner's own reports. This is a judgement of fit, not a proof: nobody has tested a person against a towel or an empty background, blind, with a sham screen. - Interested parties stay marked: Rebman (publisher, and its Archives of the Roentgen Ray, which sold the screens), Altmann (publisher of Surya and Feerhow), Boddington (sold Aurospecs). The Chicago report (O'Donnell, about 20 physicians, NY Herald 15 Jun 1911) is documented in two secondary mentions and neither primary source has been read; it is evidence, not dismissed, and not verified. The Paris trials are an informal failure, called by their authors "a reservation, not a denial", not a controlled test. The BMJ replication is unverified (section 3).

Final verdict (split). - Weak form: Plausible, untested (consistent with known vision effects; no sham-screen comparison exists). - Strong form (objective, body-specific emanation): Never properly tested. It was tried informally in 1911-12 in Paris, Philadelphia, Chicago and Graz, so it was not dismissed without any trial, but none of the trials was blinded or counted.

What would settle it. A blinded detection experiment: several observers prepared with either a dicyanin screen or a sham filter of similar colour; behind a curtain, a person, a warm towel or mannequin at body temperature, or nothing, in randomised order; each observer says "present or absent" and "which"; at least 50 trials each; pass criteria fixed in advance. Human-specific haze requires person-versus-towel discrimination above chance and sham-screen failure. In addition, Kilner's chromatic-focus table (Archives v.16 pp.344-352) should be recomputed with the objective and eyepiece used.

Claim 2: The aura is a force from the body acting on the ether (1 aF and 2 aF), lies outside the visible spectrum, and has counterparts round magnets, uranium and galvanic circuits

Agreement. All three split it. Ether-force and "two forces": the Physicist says Historically superseded but instructive; the Contrarian says Not testable as stated (and superseded); the Mathematician says Not testable as stated (and calls the two-force statement a bookkeeping device by Kilner's own words, p.101). Outside-visible-spectrum: all three find an internal tension: the aura is said to be seen by reflected light through red, yellow, green and blue screens (p.109) and also to lie outside the visible spectrum (p.110), and a screen that only subtracts visible light cannot reveal a non-visible band. Magnet, uranium, galvanic hazes: all three say Never properly tested (single observer, no control objects).

Disagreement. The Physicist adds "Tested, not supported" for the ultraviolet form, based on Kilner's own failed photography (1911: "negative, owing to the inordinate exposures", p.110 PDF; 1920 preface: "to a slight extent"). The Mathematician leads with "Internally inconsistent". Which is right? The two are not rivals. The 1911 form (not luminous, seen by reflection, yet outside the visible spectrum) is internally inconsistent as stated. The 1920 ultraviolet form is a different, testable statement. The photographic result is a failure in 1911 and a weak, unspecified partial success in 1920, with no exposure data in the sources I have; the NYT headline of 5 Feb 1911 ("The human aura has at last been photographed") conflicts with the book, but its text was not read. A single investigator's failed photography with unknown exposures is a weak test, so "Tested, not supported" is too strong; I downgrade it to "inconsistent, and not supported by the only photographic attempts on record, which were Kilner's own". My web check adds that the lens absorbs below about 390 nm, so a real ultraviolet source would be invisible to the eye and would need to be detected by plate or quartz optics; Kilner's partial attempts used ultraviolet-sensitive plates, with no success reported in 1911. - Uranium: the Physicist's alternative (uranyl fluorescence, a yellow haze, stronger through a yellow screen and removed by blue, p.97) is now web-confirmed (section 3). It is a plausible ordinary explanation that no one tested with a control crystal. The Contrarian's alternative (uranium radiation too weak to see) is consistent with it. - N-rays: Kilner's N-ray period explained erratic needle deflections as rays falling on the needle only at times (p.101-102 / PDF p.120-121) [read]. This is a rescue argument that cannot lose, and N-rays failed R. W. Wood's 1904 test [web: https://en.wikipedia.org/wiki/N_ray]. Nickell's link of Kilner to N-rays is a sceptic's characterisation and is kept as such. - Wimshurst charging (1920 edition, 1915): aura vanishes on charging, returns 3 in wider; eyeballed, single observer, unread beyond the Investigator's note; a charged person sheds ions and moves air. Kept as an unreplicated report.

Final verdict (split). - Ether-force / "aF" forces: Historically superseded but instructive (and the "two forces" are a bookkeeping device by Kilner's own words). - 1911 statement "outside the visible spectrum yet seen by reflection through a visible-light screen": Internally inconsistent. The 1920 ultraviolet form: Never properly tested (the only attempts, Kilner's own photography, were negative or "slight" with no data). - Hazes round magnets, uranium and galvanic circuits: Never properly tested. Ordinary explanations (veiling glare, fluorescence of uranyl salts) are not excluded.

What would settle it. (i) Photograph and spectrograph the haze round a person and round a uranyl-nitrate crystal and a control yellow salt, with UV-sensitive plate or camera and quartz optics, using the prepared-eye routine for a human observer simultaneously; (ii) place a magnet or dummy wires behind a screen and ask blind observers to say which side has the object. Failure of (i) with an unaided instrument and success of the prepared human observer would locate the effect in the eye.

Claim 3: Aura size, shape, texture and colour reflect sex, age, heredity, temperament and disease

Agreement. All three: Never properly tested. All three note that classes were fixed "by experience only" (p.78-79), that the observer knew the case, and that no blind classification or second observer exists. The Archives of the Roentgen Ray review itself declined to assess the disease claims.

Disagreement. The Mathematician and Physicist differ on emphasis: the Mathematician says the disease and temperament parts are Not testable as stated until classes are defined in advance; the Physicist says the claim is testable and untested, and notes that physique alone (wider female trunk, a hemiplegic's posture) could explain many descriptions without new physics. The Contrarian adds that the absence of any clinical follow-up in 1912-20, despite published case lists and charts, is an anomaly (section 6). The stronger argument: the width classes are testable in principle (width in inches, blind second observer), so "Never properly tested" is right for width, sex and age; for "temperament", "intelligence" (the claim that the most intelligent have the finest auras, p.72-78) and "disease type" the categories were defined with the answer in view, so those parts are Not testable as stated.

Heredity (kept in the record; see 2d). Parent-child agreement 15 of 15 (or 16 of 18 on the Mathematician's reading); sibling agreement 5 of 9. The pattern is unlike what pure inheritance would give and consistent with an unblinded-classification effect, but the sample is small.

Final verdict (split). - Width and form against sex, age and parentage: Never properly tested (the heredity table is a real, small dataset worth a blind repeat). - Temperament, intelligence, disease-specific types: Not testable as stated until classes are fixed beforehand; individual predicted outcomes (for example the epileptic whose fit followed "a few days later") are anecdotes without denominators.

What would settle it. Blind width measurement by two independent observers on the same persons on different days (including children of known and unknown parentage), compared with a family-relationship list the observers do not see; the same on groups with and without a clinical diagnosis. Inter-observer agreement beyond chance is the first test, before any claim about heredity or disease.

Claim 4: The C.C. band turns lighter or darker over local disease, tenderness, impending menstruation, early pregnancy or a dead foetus, so it can diagnose

Agreement. All three: Never properly tested as a diagnostic method. All three note the kernel that is real: a negative afterimage is a real, well-established perceptual effect whose apparent brightness depends on the surface it falls on, so differences in skin colour, sweat, inflammation, shading and lighting will make patches lighter or darker. The Physicist and Mathematician add that Kilner's list of four causes (skin, thickness, texture, colour of aura) omits the observer's retina, expectation and body curvature; the Mathematician adds that in the decision logic every outcome is assigned a cause (same change in all bands = granularity; opposite changes = colour; no change = granularity "at times has no effect"; light spot = gap; dark spot = small gap or dense surround), so no result can count against it; and that the sign flipped with fatigue in Experiment 1.

Disagreement. The Mathematician offers "Internally inconsistent" for the combination of rules and "Never properly tested" for the diagnostic claim; the others use only "Never properly tested". I adopt the split below. The Mathematician's observation that his sign rules (a red ray darkens both blue and yellow bands, a yellow ray darkens blue) do not follow from opponent-colour theory [memory, high] is kept; whether the rules are wrong cannot be said, only that they are stated and not derived.

Outcomes on record. [Revised: 11.6. This list mis-scored two items and mis-numbered one; the corrected scoring is in 11.6.] The few reported predictions: menstruation in 4 days (Case 18); 6 to 7 days predicted for a 38-year-old, which occurred 7 days later (one case, a hit); a young woman expected in 3 to 4 days, whose period came a week late (a miss); Case 21, pregnancy ruled out, menstruation 7 days later. With a 28-day cycle, "about a week" is right by luck a good fraction of the time, and there are no denominators. Kilner reports his misses, which counts in his favour.

Final verdict (split). - Diagnosis of disease, tenderness, menstruation, pregnancy by the C.C. band: Never properly tested. - Kernel (afterimage patches track skin and lighting differences): Plausible, untested as a crude sign; the afterimage itself is Established. - The decision logic as a whole (every outcome has an assigned cause): not falsifiable as written; the combination of rules is Internally inconsistent only in the narrow sense described above. - "Dead foetus" belongs to claim 3 (texture), not to the C.C. band (Case 40).

What would settle it. A blind trial: an observer who does not know the diagnosis examines patients and marks patches; a second observer or a clinician records pregnancy, menstrual date and tender areas independently; sensitivity and specificity computed against a naive baseline (for example "no change" for every woman). Include a skin-colour-only control (a patient with the same rash in a non-diagnostic area).

Claim 5: C.C. band changes, where texture does not explain them, come from the aura's colour, which the naked eye cannot see

Agreement. All three: Never properly tested. All three observe that the colour hypothesis is introduced as "the only [agent] that will explain all the remaining cases", that its "presumptive evidence" is the clairvoyants' assertions (circular), that the list of causes is not shown to be exhaustive, and that Kilner's own footnote (p.156) points to the Appendix as his evidence, whereas the Appendix's only blind trials (Experiments 5 and 6) used the blue band alone, which on his own rules darkens for red and yellow alike (p.158-159), so they cannot discriminate colour. The only colour-specific pattern (Experiments 1 and 4) was gathered with the observer knowing both colour and site.

Disagreement. The Contrarian lists the real controls in the Appendix (left-right comparison in Experiment 1; turn-round check) and Kilner's candid failures (the willed colour never matched the cue book; no true yellow in Experiment 3). I keep these in the record. They exclude an illumination asymmetry, not observer expectation. The Contrarian and Mathematician add that the colour hypothesis is awkward: a colour that affects the eye's response through an afterimage should also be visible to the same eye; Kilner resolves this with the extra hypothesis of a ray pointing at the observer.

Final verdict. Never properly tested, with this added: the experiments Kilner offers as his test do not test it (they test location, not colour), and as an elimination argument it cannot be wrong. The interpretation "aura colour" is Not testable as stated in the Appendix design because no condition distinguishes it from other causes of a dark patch.

What would settle it. Repeat Experiments 4 to 6 with a blind observer, several C.C. bands (blue, yellow, red, green) in random order per trial, the subject choosing the colour and site by lot and unseen by the observer, and sham trials with no willing. A colour-specific pattern (dark on blue and yellow, light on red for a red ray) predicted before the trial, and reproduced across observers, would be the first real evidence.

Claim 6: The will can project rays from the Inner Aura and change the aura's colour; in two blind trials the observer located the willed spot (exact centre; 0.5 in off)

Agreement. All three: Never properly tested. All three note one observer, one paid subject (D., an artist's model, hysterical temperament, "able to see the Aura easily ... sometimes even before the observer could"), a procedure the observer had demonstrated to her (Case 41), and no sham trials. The Physicist and Contrarian offer a second verdict for the location trials: they test location, not colour.

Disagreement and chance. The reviewers' chance figures are reconciled in 2c: about 1% to 3% per trial for a random single guess, but the observer scanned a patterned field and the subject could cue (she announced when to look and saw the observer). The Physicist stresses that the near-miss was scored as a hit and that the sign of the effect changed with fatigue in Experiment 1. The Contrarian's point that Kilner partly sees the confound himself (p.329: concentrating on a distinct part is easier) is kept. The Contrarian notes too that Kilner's sentence that the Appendix "proves the truth of the theory previously advanced" states the conclusion before the experiments; an observation about his reasoning, kept.

Final verdict (split). - Willed rays and willed colour change as objective phenomena: Never properly tested (observer-known sites, a subject who sees the aura, no sham). - The two hidden-site trials: Never properly tested; as evidence for colour specificity, Not testable as stated in that form, because the blue band does not discriminate red from yellow. They are described, not dismissed: one exact hit and one near-miss, no statistical conclusion possible.

What would settle it. The single design that settles claims 5 and 6 together: a subject behind a screen that blocks the observer's view, the site and colour chosen by lot by a third person, many trials with sham (no willing) trials interleaved, the observer scoring by finger placement before the answer is revealed, and hit and miss counts reported. About 20 trials give a usable test: at a 3% single-guess chance, five or more hits in 20 would be very unlikely by chance if all candidate patches counted, whereas the same score with a cue-blocked subject would be a real anomaly.

---

5. Where the arbiter's weights differ from a reviewer's

• The Contrarian's strongest honest case (Kilner published his data, named the dye, reported failures, and several groups saw something) is kept in full; I do not weight it as evidence of the aura itself, but it does show that the screen routine reliably produces a visual impression in some people, and that the Paris failure is itself not a controlled test.
• The Physicist's "Tested, not supported" for the ultraviolet form is downgraded (section 4, claim 2).
• The Mathematician's heredity count is corrected (section 2d); his chromatic-focus arithmetic stands; his scattering numbers are corrected by a factor of 4 (section 2b).
• The Contrarian's depth-of-focus remark is corrected (section 2a).
• Nothing was dropped from the record because a reviewer judged it negatively: the H. C. S. towel report, Kilner's reply, O'Donnell, Surya, the Paris trials, Werner and Caan's photo-activity of tissue (a different claim: plate fogging by contact, reported in Annales 1912, not read), the NYT headline, and the BMJ review are all kept as leads with their stakes.

---

6. Absence-of-evidence analysis: why the record goes dark after 1912-1920

What is silent. After the 1911 first edition and the 1912 reviews, there are no independent controlled trials, no clinical follow-up of the C.C. diagnostic method, and no further work by Kilner except the 1915 Wimshurst experiments, published in the 1920 second edition (Kegan Paul). Kilner died in June 1920. Later material is by sellers or devotees: Boddington (1931, Aurospecs goggles), Bagnall (1937, pinacyanol screens). This is an anomaly to analyse: a hospital electrician-physician at a leading London hospital published case lists and sold screens; yet there are no hospital trials, no Lancet or BMJ follow-ups found, and no student of his.

Competing explanations, with evidence for and against.

These explanations are not mutually exclusive. The record does not distinguish them. The simplest reading supported by the existing evidence is a combination: a method that depended on one observer's training and a particular dye, poor repeatability in the hands of others, and an interruption by the war and by Kilner's age and death. That is a reading, not a finding.

---

7. Corrected claim list (revised after QE; changes from the first pass are marked [changed] or [new])

1. A faint, structured haze (Etheric Double, Inner Aura, Outer Aura) is seen round the human body in dim diffuse light after the observer has looked through a dicyanin screen. Weak form (observers report it): Plausible, untested. Strong form (an objective, body-specific emanation): Never properly tested. Evidence added by QE: one informal, unwitting, uncounted sham-screen observation (methylene blue substituted three times, 1920), the Paris negative trials now documented, and an unperformed blind design by de Fontenay (11.7).
2. (a) The aura is a force acting on the ether (1 aF, 2 aF; in 1920 I.A.F. and O.A.F.): Historically superseded but instructive. (b) The 1911 argument that it lies outside the visible spectrum yet is seen through a visible-light screen: Internally inconsistent; the 1920 ultraviolet form (Kilner had moved from an infra-red origin in 1908 to "an ultra-violet phenomenon"): Never properly tested (Kilner's own colour-through-screens test exists but is confounded; no UV-opaque filter or quartz test was ever done). (c) Hazes round magnets, radio-active uranium and galvanic circuits: Never properly tested; ordinary explanations (veiling glare; uranyl fluorescence, web-confirmed) not excluded. [new] (d) 1920 effects on the aura of static charge (Wimshurst), magnets near the body, and fumigation with iodine, bromine, chlorine and other vapours: Never properly tested (single observer, unblinded, no counts; coloured halogen vapours are an ordinary candidate for the colour changes). [new] (e) Dicyanin shifts the eye's focus by about 1 dioptre (range 0.4 to 4.5; 1920 microscope series): Plausible, untested by neutral observers; Kilner's inference from it to ultraviolet sensitivity: Never properly tested (a focus shift in visible light does not by itself bear on ultraviolet).
3. Aura width, form, texture and colour reflect sex, age, heredity, temperament and disease. [changed] Width against sex, age and parentage: Never properly tested; the heredity tables are kept and recounted (parent-child 15 of 15; siblings 5 of 9 in Table II, 9 of 13 pairs when Table I's sibling sets are added; unblinded). The blue / blue-grey / grey colour series (40, 36, 17 of the first 100, 7 unrecorded) against mental fitness and diagnosis: was Not testable as stated, now Never properly tested (the scheme is definite and the counts are published; the classifier knew the diagnoses). Left-side narrowing in epilepsy ("all ten" left-sided; three individually documented): was Not testable as stated, now Never properly tested (a definite, blind-testable prediction; Kilner retracted one 1911 "healthy symmetric" case that turned out to be an epileptic). Temperament and intelligence ratings, and disease types without a stated rule: Not testable as stated. The dead-foetus case (Case 40) is evidence for texture, not for the C.C. band; its history (three weeks without foetal movement) already pointed to foetal death.
4. The C.C. band can be used for diagnosis (disease, tenderness, menstruation, pregnancy): Never properly tested; kernel (afterimage varies with skin and lighting) Plausible, untested; the afterimage itself Established. [changed evidence, same verdict] The corrected record has about four readings that ran against the patient's own expectation or answer (11.6) and several candid misses, and no denominators for either.
5. C.C. band changes not explained by texture come from the aura's colour: Never properly tested; Not testable as stated in the Appendix design (blue band cannot discriminate red from yellow).
6. The will projects rays and changes colour; the observer located two willed spots blind: Never properly tested; the two trials (one exact, one 0.5 in off, one observer, one subject who sees auras, no sham) allow no statistical conclusion. [new] The 1920 claim that "some women, no man or boy" can change aura colour at will rests on one model, two hypnotised women (Dr Douglas) and unnamed patients who were not tested: Never properly tested.

8. What would settle it (all claims in one programme)

1. Blinded detection test (claim 1): person, warm towel or mannequin, empty ground; sham versus dicyanin screen; several observers; counts fixed beforehand.
2. Instrument test (claim 2): plate, camera and spectrograph (including quartz and UV) on a person, a uranyl crystal and a control salt; a magnet or dummy wires behind a screen.
3. Inter-observer width test (claim 3): two observers, blind, repeated days; relationship list hidden.
4. Blind diagnostic test (claim 4): clinical record against C.C. patch maps, against a naive baseline.
5. Cue-blocked location and colour trial (claims 5 and 6): site and colour by lot, observer unseen, 20 or more trials, sham trials.
6. Calculation and measurement (claim 2e): [Revised: 11.3] recompute the chromatic-focus shift with the objective and eyepiece actually used (1920 text: 1.5-inch objective; eyepiece not given), from the Archives v.16 pp.344-352 table or a scan of the 1920 Table 3; and, cheaply, measure the eye's refraction before and after dicyanin viewing with an autorefractor (a refraction-measuring instrument), against a sham dye.

9. Open questions

• What does the BMJ review of 6 Jan 1912 say, and did any BMJ writer run a trial (Wikipedia says "attempted to replicate ... negative"; nobody has read the review)?
• What did O'Donnell actually do in Chicago in 1911 (Herald 15 Jun 1911; Chicago paper 21 Jun 1911; Annales 1 Sep 1911 pp.264-268)? What counted as "success"?
• What is in the NYT article of 5 Feb 1911, "The human aura has at last been photographed"? It conflicts with the book's report of photographic failure.
• Does the chromatic-focus table (12 of 30 observers; clairvoyant largest, over-60s smallest) show a shift that is larger than the repeat scatter of the setting, and why were 12 of 30 chosen?
• Why do parent-child pairs agree in aura class (15 of 15) while sibling pairs do not (5 of 9)? Is the F.X. group parent-child or siblings? Could a blind classifier reproduce it?
• Does a person at body temperature produce more or less haze than a warm towel? (H. C. S. found the towel stronger; Kilner denies it. Nobody has measured it.)
• What did Werner and Caan's "photo-activity" of human tissue (Annales 1912 p.189) show? A different effect, but it is the only instrumental report from the period.
• Where did Kilner die (Isle of Wight or Bury St Edmunds), and what did Rebman sell after 1912?
• Why did the 1920 edition report "slight" photographic success with no exposure data?

10. Sources not read by any role (carried as leads)

BMJ 6 Jan 1912 review (PMC2344221); Archives of the Roentgen Ray v.16 pp.156, 199-200, 240, 344-352 (read by the Investigator in a browser; no raw copy saved); Annales des Sciences Psychiques 1911-12 (saved, but read only for pp.74-84); NY Herald 15 Jun 1911; NYT 5 Feb 1911; Rebman 1912 booklet (hvd.hnxbgt); Feerhow 1913 chs. II-III and appendix; the 1920 second edition (OCR saved in research/files/, not digested in full); Boddington 1931.

---

11. QE revision (2026-10-04)

Trigger. The QE (qe.md, verdict MINOR) found that claims 3 and 4 rested on digests alone, that several accuracy errors had crept in (OCR case numbers, an incomplete sibling count, two mis-scored menstrual cases), and that the Investigator's saved files (1920 edition, Boddington 1931, Annales 1912) had never been read. Five deep reads were run. I read all five in full and checked them against my own earlier reading of the Table I and II page images (PDF pp.80-81) and the QE's recount.

Inputs. (a) Chapters V-VII, every diagnosis or prediction (deep/humanatmosphere00kilngoog-d2c3dd.md, last section, PDF pp.166-325). (b) Chapters I-II and the start of III: protocol, other observers, Etheric Double experiments, heredity text (same file, "addition", PDF pp.20-108). (c) The 1920 second edition (deep/kilner-1920-human-atmosphere-2nd-ed_djvu.md). (d) de Fontenay, Annales des Sciences Psychiques March 1912, and the Werner/Caan note (deep/annales-des-sciences-psychiques-v22-1912.md). (e) Boddington c.1931 (deep/boddington-1931-aura-kilner-screens_djvu.md). Page tags: PDF page for the 1911 book (printed = PDF minus 19); printed page for the 1920 book, as the 1920 deep reader gave them.

Limits that remain. The BMJ review of 6 Jan 1912 is still unread. No scan of the 1920 book exists locally, so its Tables 3, 4 and 6 (microscope readings, colours through screens, hypnosis) are too garbled to use; I use only numbers the prose states. The Archives of the Roentgen Ray paper (v.16 pp.344-352) is known only through the Investigator's note; its table is not saved. Web checks: none new (3 of the 8 remain unused). Every new verdict below rests on [read] or on high-confidence [memory]; the one medium-confidence suggestion (aphakic UV vision, 11.11) feeds a proposed test only, not a verdict.

11.1 Corrections to the record

1. Case numbers: OCR reads the digit 3 as 8. [read: page images p.169, 176, 182, 208, 246, 303, per the deep reader] "Case 18" (I. N., dressmaker) is Case 13; "Case 28" (E. E., cystic breast tumour) is Case 23; "Case 80/82" are Cases 30/32; "Cases 81-89" are 31-39. So my correction 1 in section 1 was half right. "Case 21 (continued from page 189)" is a real continuation (the pregnancy question is on PDF p.317). But "Case 88" in claims.md was two different cases, 33 (stomach ulcer) and 38 (asthma). The book runs Cases 1-42, as its contents page says. Case numbers are not stable between editions either: the 1911 Appendix "Case 42" (the model) is Case 25 in 1920, and 1920 "Case 42" is an epileptic woman.
2. Case 30 is not pregnant. She is a stout 58-year-old woman whose abdomen made her look six months pregnant; Kilner wrote "we are unable to explain this remarkable Aura" [read: PDF p.235-238]. Her figure is Fig. 24. Digest 03 and the first deep read were wrong on both points.
3. Digest 03's page labels drift from PDF to printed numbering. I use the deep readers' PDF pages. No content was skipped.
4. The first pass said Tables I-II are "the one place in the book with counts". Wrong. The first-hundred colour series and the epilepsy counts (11.5) are counts too.
5. Page convention. The Mathematician's "p.91-92" (cloud-banner and vapour argument) is PDF p.110-111. The primer should follow.

11.2 Verdict table: what changed

No verdict moves to a stronger or weaker label than the evidence already supported: two parts of claim 3 move from "Not testable as stated" to "Never properly tested", because the new reads show the categories are definite enough to test. The QE's expectation that no central verdict would change holds.

11.3 The chromatic-focus numbers, recomputed (replaces 2a)

Terms. A dioptre (D) is a unit of lens power, one over the focal length in metres; an eye focused for distance that gains 1 D of power is best focused at 1 metre. One mi is one division of Kilner's fine-focus wheel, 1/1000 inch = 25.4 micrometres (µm) of stage travel.

What the 1920 text says about the instrument [read: 1920 p.61-63]: a microscope whose fine adjustment moves 1/100 inch per turn with the wheel in ten divisions; "the lowest eyepiece, inch-and-a-half objective"; object a blow-fly bristle; screens were glass cells of weak dye solutions (carmine, yellow, methylene blue, gentian violet); the observer refocuses with the fine screw alone after each screen, then repeats after looking at light through the dark dicyanin screen for about 30 seconds. Observer "A": before dicyanin red -1, blue +1, violet +3 mi; after dicyanin white light needed +2.5 mi, then red +1, yellow +1, blue +2, violet +3.5 mi; tested on three occasions with "no variations" [read: 1920 p.62-63]. This is the same experiment as the 1911-12 Archives paper: the Investigator's note of that paper (research/2026-10-04-hathitrust.md, A1, read in a browser; I did not see the paper) gives the same worked example for observer A (white +2.5 mi after dicyanin; before it red -1, blue +1, violet +3) and a table of 12 observers "taken consecutively from thirty cases", the clairvoyant largest at +6.5. The note gives the objective as "1/4-in [OCR '14-inch']"; the 1920 text says "inch-and-a-half", and "14-inch" is most likely an OCR misreading of "1½-inch". I take 1.5 inch as the better reading. The 1920 account differs in counts (50 accepted series, 11 observers listed plus a mean).

Recalculation. A stage shift Δz moves the intermediate image by about M²Δz, where M is the objective's magnification; an eyepiece of focal length f_e turns that into a change of vergence of M²Δz/f_e². A 1.5-inch objective has a focal length of about 38 mm. At a 160 mm tube length M is about 4.2; at the British 250 mm tube length it is about 6.7. With f_e between 25 and 50 mm this gives about 0.2 to 1.8 D per mi (M = 4.2, f_e = 50 mm: 0.18 D; M = 4.2, f_e = 25 mm: 0.72 D; M = 6.7, f_e = 25 mm: 1.8 D). [calculation, this review; M and f_e are assumptions]

A check from Kilner's own data. Observer A's red-to-violet spread of 4 mi (100 µm) should equal the eye's chromatic focus difference, which is about 2 D across the visible spectrum [memory, high]. That fixes the scale at about 0.5 D per mi, in the middle of the range above. At that scale the dicyanin shift of +2.5 mi is about 1 D (range 0.4 to 4.5 D). Kilner's own opera-glass and reading-glasses anecdotes, which describe a temporary myopic shift that helps mild presbyopes, are the same size (1920 p.60-61; the Archives editorial repeats "a sort of artificial or temporary myopia", research/2026-10-04-hathitrust.md).

What this retracts. The first pass said the shift was "3 to 17 depths of focus", "far too big to be a change of the eye's refraction", needing "tens of dioptres". Both halves are withdrawn. (i) It assumed a 1/4-inch objective; the 1920 account says 1.5 inch. (ii) A microscope's depth of focus (diffraction term about 0.55 µm over NA squared; at NA 0.06 to 0.1 that is 55 to 150 µm, with a larger geometrical term for a low-power objective [memory, medium]) is the wrong yardstick for a task where the observer sets best focus to a fraction of it. The right yardstick is the observer's repeat scatter. Kilner reports no scatter figure, but he does describe one real control: during each series he secretly displaced the focus two or three times and rejected the series unless the observer returned to the original reading, "not more than one series out of every three or four accepted", leaving 50 accepted series [read: 1920 p.63]. That control selects steady observers. It also means that roughly 100 to 150 rejected series are unreported (50 accepted at one in three or four implies about 150 to 200 run), so this is a filtered sample. The controls were comparative: the other dye screens went through the same procedure, and Kilner reports that in no case where the white-light shift after dicyanin exceeded the shift from the blue screen alone was there "the slightest difficulty in seeing the aura" (p.63-64), a correlation reported without numbers.

What it now means. A shift of about 1 D is the size of ordinary changes in an eye's focus (accommodation, the eye's tendency to drift to an intermediate focus after staring at a bright or dim field, pupil size, a change in the observer's focus criterion after a bright exposure) [memory, medium]. It is therefore plausible and cheap to test, and it neither requires nor shows a change in the lens's chemistry or any ultraviolet sensitivity. Kilner's own text draws the line: "no test has been devised, the question must be treated entirely theoretically" [read: 1920 p.68]. His step from a visible-light focus shift to "an ultra-violet phenomenon" (preface; p.67-68) is not supported by this datum. Boddington's heading "Dicyanine cures short sight" runs the other way: a temporary myopic shift would not help a short-sighted eye, and Kilner reported no effect in myopia or emmetropia [read: Boddington p.18, p.21; Kilner 1920 p.61].

11.4 The heredity tables, recounted (replaces 2d's last two bullets)

The QE is right that Table I contains sibling sets. From the page image of PDF p.80 and the deep read: parent-child pairs 15 (all concordant, unchanged). Siblings inside Table I: X.X, B.X and C (all narrow; 3 persons, so 3 pairs but only 2 independent comparisons), and K.M and M.M (both average; 1 pair). The X family's grand-daughters F.X (narrow), G.X (average) and E.X (wide) are labelled "Granddaughter" and appear to be three sisters or cousins, not children of F.X, so they are siblings or near-siblings with 0 of 3 pairs concordant. Table II (page image of PDF p.81) has 9 pairs, 5 concordant.

[calculation, this review; the QE's 9 of 13 and 9 of 16 reproduce] The class frequencies (p = 0.383) come from pooling Tables I and II (A 24, W 15, N 9 of 48 entries). Independence does not hold: the same person sits in both tables for two cases, parents recur, and the narrow class is concentrated in one diseased family. Caution about the Fisher test: pairs share members, so the p-values are indicative.

Reading. Parent-child agreement is far above chance. Sibling agreement is not below chance: depending on how the X grand-daughters and trios are counted, it ranges from not unusual (5 of 9, 9 of 16) to mildly above chance (9 of 13, 8 of 12). So the first pass's statement that "siblings do not agree beyond chance" and that "the asymmetry favours a method effect" is withdrawn. The asymmetry is weaker than stated and its direction is not clear. Two ordinary explanations stay open and neither is excluded: a family resemblance of body build and health (heredity or shared environment; Table I's narrow group is mostly the diseased X family and the neurotic, epileptic and Bright's-disease patients), or an unblinded classifier who knew the relationships and the diagnoses. Further facts from the deep read: the standard for each class was fixed "by experience only, and not by any scientific plan" (women 8 to 10 inches, men 3.5 to 4.5, children 2.5 to 3, adolescents "each upon its own merits"), with no statement that the classifier was blind to relationships [read: PDF p.78-79]. The case labels in Table I's condition column do not match the text's case list (a 30-year-old labelled Case 2 whose text-case is an infant; unresolved in the book). Footnote 1 (the mother N.N.) is a post-hoc classification; without it the result is 14 of 14. Table I's class mix (A 16, W 5, N 9) and Table II's (W 10, A 8, N 0) are quite different, which is itself a reason to treat the tables as different samples. The datum stays in the record; the blind repeat in claim 3 settles it.

11.5 New counts for claim 3

Colour series of the first 100 cases [read: PDF p.247-248]. Blue 40 ("no individual below ... the average in mental power"); blue with more or less grey 36 (1 hemiplegia, 2 epileptics, 1 meningitis); grey 17 (2 eccentric, 6 epileptics, 1 general paralysis, 3 mentally dull); no colour recorded 7. Kilner: the table "shows conclusively that the owners of the blue Auras are the most mentally fit", and he adds that the proportion of healthy people in this first hundred was larger than later. The 1920 edition keeps the table "unaltered" and prints the three series as 93 (the seven unrecorded are not in the OCR) [read: 1920 p.245].

Calculation. All 8 epileptics in the series fall outside the blue series. If epileptics were placed at random among the 93 coloured cases, the chance of none in the 40 blue ones is C(53,8)/C(93,8) = 0.009; counting the seven unrecorded as non-blue, C(60,8)/C(100,8) = 0.014. [calculation, this review] What the number cannot say: the epileptics came as epileptics and the classifier knew it; "grey" is the observer's judgement; Kilner admits the preparation of the eyes "renders it impossible to appreciate any great variation in range of colour", with colours "for the most part limited to blue and grey"; "mental power" has no criterion. It is a real count with a definite scheme, which is why these parts move to "Never properly tested".

Epilepsy [read: PDF p.178-179, 194, 302]. "In all the cases (ten in number) we have seen the diminution has been on the left side" (four males, six females, ages 12 to 45). Only three are described individually: Case 15 (X.X., 23), Case 16 (B.X., 18), Case 17 (X.T., 13), each with head Outer and Inner Aura narrower on the left. The other seven are "so similar in their essential points that nothing is to be gained by describing them". Against the finding, in Kilner's own words: a mother reported more spasm on the right side; the one fit he saw showed no side difference; in none "has the C.C. band showed any change that can be typical of the complaint" (p.194); non-epileptic patients with brain disturbance showed the same narrowing (p.178, 195). In 1920: "the auras of every one enjoying good health are invariably symmetrical"; the exception, and an explicit retraction: two 1911 "symmetrical healthy" cases included one who "had previously had an epileptic fit, which was not at the time admitted"; a later healthy-looking woman (Case 51) was found to have had epilepsy 27 years earlier, and Kilner says he would not have suspected it from the history alone [read: 1920 p.162 footnote, p.189-190]. Two reads of this, both kept: (i) a uniform, one-sided, easily stated finding with a published retraction of his own counter-instance, which a blind second observer could test in an afternoon; (ii) an observer who looked for a one-sided narrowing in people he already knew or suspected to be epileptic (Case 18 himself says he "considered him to be epileptic, even before we could perceive the Aura"). A patient's concealed history is a documented source of error for the diagnostic claim in both directions.

11.6 Claim 4: diagnoses and predictions, rescored (replaces the "Outcomes on record" list)

The deep read lists 16 cases where Kilner states a reading and later facts can be compared [read: PDF p.169-323]. I score them by what was known beforehand and by what the text actually says was announced. Terms: the C.C. band is the afterimage band described in claim 4; "against expectation" means the reading ran against what the patient said or believed.

Misses, nulls and limits in Kilner's own words (deep read section 6): the C.C. band showed no change typical of epilepsy; "a fair proportion" of painful spots are invisible; the band "sometimes gives rise to no change whatever, when a predication of a marked alteration would seemingly be a certainty"; the greater right-groin tenderness in Case 34 had no band change; "the C.C. band does not show any diagnostic mark of hysteria"; no distinctive case of incipient consumption ("disappointed"); granular kidney: no change in two or three cases, including a man inspected shortly before a fatal hemiplegia; and "the limitations of the Aura as a diagnostic agent are great" [read: PDF p.194, 266, 282, 284-286, 295-298, 309]. Kilner's usual method was to take the history first ("if the history of the patient be known ... greater and earlier attention can be given", p.251). The 1920 edition adds a second refusal to examine first: Dr Merrick brought a woman Kilner "had intentionally not investigated in the ordinary manner" [read: 1920 p.258-259].

Weighing. The QE is right that the first pass left out the readings that ran against what the patient said or believed, and that these are the items that most weaken the objection that the patient cued the observer. I now count four such items (the 38-year-old, Case 21, the young woman if her reading is read as a call, and Case 34's cervical patch). Three limits remain, and they keep the verdict at "Never properly tested": each rests on a single observer who was usually the treating physician; the confirmation is usually the patient's own later recollection, as recorded by Kilner, with no contemporaneous notes reproduced; and there are no denominators, so the hits cannot be compared with how often a reading at that resolution would be right by chance or how many readings failed without being reported. The one reading with an independent event as outcome, Case 40, also matched the patient's fear and a suggestive history.

11.7 Additions to claims 1 and 2

Claim 1, evidence added. - An unwitting sham screen (1920). A lady handed a methylene-blue screen "inadvertently" in place of the dark dicyanin one "was quite unable to see the aura"; she was not told, "because this was regarded as a good test"; with the right screen she then saw it. It happened "twice again" unwitting, and "several times" by intent, "always found inert" [read: 1920 p.58]. No numbers beyond three. This is the nearest thing on record to a blind sham-screen trial, so my statement that "no sham-screen comparison exists" (section 4, claim 1) is corrected to: no controlled one, but one informal, unwitting and uncounted. It supports "the dicyanin screen matters, and expectation alone is not enough", and does not test whether the haze is body-specific. Methylene blue and dicyanin also differ in which colours they pass, so the difference is expected on any theory. - How many people can see it? Three statements, none with numbers behind them. Kilner 1911: with a dark-screen preparation "all will be able to see it - some immediately and the rest within a minute" [read: PDF p.129-130]. Kilner 1920 footnote: "ninety-five per cent" of people with normal sight can see it, and "one gentleman states that only one person out of four hundred ... was unable" [read: 1920 p.2]. Surya (Feerhow 1913, from the Investigator): about half of first-time viewers see nothing. Against all of them: de Rochas and de Fontenay with four screens, de Rochas's untrained subjects and de Fontenay's friends "saw nothing relating to the descriptions of the English savant" [read: Annales p.86; French quoted in the deep read]. Boddington (seller) halves the 95%: "because I eliminate normal clairvoyance, optical illusion and suggestion" and quotes an editor that "large numbers of people will obligingly see anything they are asked to see" [read: Boddington p.20]. These are not reconciled and no one measured. Kilner also lost the effect after a holiday: "the cumulative action is not quite permanent" [read: PDF p.130-131]. - Other observers inside the book. Case 1: the infant's mother and nurse "were able to see the cloud around him, if they looked through the screen" (a conditional sentence; no description of what they saw) [read: PDF p.41]. A ray between two hands that changed from "bright yellow" to "liquid ruby" (who saw it is not stated) [PDF p.106]. In the Etheric Double work nobody besides Kilner is named, and Experiments 1-7 are written as recipes with no data, trials or second observer [PDF p.87-95]. The observation protocol has no light level in any unit and no observer-to-subject distance; the only distances are: subject one foot in front of the background, background about eight feet from the window [PDF p.27-29]. That is itself a reproducibility problem, and de Fontenay says so ("le point délicat du système"). - de Fontenay's Paris trials (Annales, March 1912). He and de Rochas saw nothing with the screens; no numbers, dates or light levels are reported; he is an interested party who wants the aura to be real, and he says one failure with a book is "a reservation and not at all a denial" [read: Annales p.80-81, 86]. He asked for two blind tests that Kilner could have run and nothing in this article says either was run: (1) hide strong magnets among demagnetised ones painted alike and ask a seer which show an aura; (2) 25 pairs of wires on a black board, a cable to a distant room, a battery switched to a group every 5 minutes for two hours, the seer calling the lit group. At 1 in 25 per call the chance rate is 4%, and 24 calls gives a decisive test (he does not compute it) [read: Annales p.88-90]. This is a better-designed test than anything on record, and it is the sharpest answer to "what would settle claim 1 and the magnet haze in claim 2c". - O'Donnell, Chicago (Mercy Hospital, June-July 1911). Documented in the Annales through clippings: the New York Herald of 15 June 1911 says Kilner's experiments were repeated before "about twenty physicians" and "succeeded"; four young women were the subjects, with a screen of "various chemical agents between two glass plates" and the doctors "saw a luminous trail round the bodies"; then a dying man was watched for half an hour and "the radiation disappeared" at the moment of death. O'Donnell has a stake (his own announcement), the headline "photographed the last sigh" does not match the text (he only saw it), the Annales commentator says O'Donnell had not completed the information since July, and no photograph is shown [read: Annales p.80, 90-91]. Kilner's 1920 statement that after death the aura "disappears entirely" and that on "several occasions" he saw no trace hours after death fits O'Donnell's report in direction and differs in timing (O'Donnell: at the moment; Kilner: hours later; no number of corpses) [read: 1920 p.137-140]. Both are kept. - Seller-selected support (Boddington c.1931). Dr Drysdale Anderson's letter of 19 September 1929 (quoted by the seller) says he and his wife "can make out the aura (through the glasses), but it appeared to be only some three inches thick", thinner than Kilner's description; the Daily Chronicle article of 14 March 1930 that Boddington says "confirmed Dr. Kilner's researches" is not quoted and not read [read: Boddington p.6, advertisement]. A. E. Mundy, "consulting chemist", who says he "compounded the original Kilner Screens twenty years ago", wrote 6 June 1929 that the aura "is easily seen by those who persevere" and that "what the nature of the aura is I am not prepared to say" [read: Boddington Foreword, advertisement]. The testimonials (X-ray-like transparent hands, sparks, "snow" over objects) were printed by the seller, who sells at one guinea and refills at 5 shillings, objects to selling "to people who do not accept the conclusions of psychic science", and prints no failure. One testimonial volunteers doubt ("may have been due to imagination"). Boddington himself writes "after image effects are purely subjective" and warns that "suggestion and optical illusions have a definite place" [read: Boddington p.14-15, 40]. His claimed Scatcherd and Baraduc colour photographs are a single paragraph without date, plate or reference [read: Boddington p.11-12]; I keep them as a lead with that gap. All of this is evidence of continuing belief and of one chemist's willingness to put his name to the method; none of it is a trial.

Claim 2, evidence added. - Kilner's own test of the ultraviolet idea (1920 p.75). He states it: if the blue is ordinary reflected 4200 Å light it "should follow the ordinary laws when examined through the coloured screens; if the latter some departure ... might not unreasonably be expected." He reports the departures: the natural blue stays blue through a yellow screen (instead of turning green), and yellow appears through dark blue or violet screens "an impossibility under ordinary circumstances" [read: 1920 p.77-78]. Two confounds, both in his text: the screens pass "a large amount of white light" (p.62), and observers disagree on colour names (p.76); in a dark-adapted, screen-primed eye the apparent colour of a faint grey haze is not given by the ordinary laws of reflected light [memory, medium]. He never tried a UV-opaque filter or a quartz element, and the 1920 book has no spectrograph, lamp or wavelength measurement [read: 1920 p.58-79, searched]. Boddington (seller) proposes quartz lenses (£5 5s) and photographic use "may conceivably" give colour or spirit photographs; no result [read: Boddington p.20]. So the ultraviolet form stays "Never properly tested". Photography: 1911 "negative, owing to the inordinate exposures"; de Fontenay (1912) says Kilner tried panchromatic, ultraviolet and infra-red plates "and obtained nothing"; 1920 preface "to a slight extent but not by any means satisfactorily", no data [read]. The QE's note that the NYT headline of 5 Feb 1911 ("has at last been photographed") is unresolved still stands. - Origin shift. The 1920 text says the early needle experiments located his two forces "in the infra-red portion of the spectrum" (p.3), while the 1920 conclusion is ultraviolet. In 1911 the aura was "most probably" outside the visible spectrum with UV hedged. The eye's lens absorbs below about 390 nm (web-confirmed, section 3), so a normal eye cannot see UV directly; Kilner's route is a changed eye (11.3). - Electrification (1920 p.91-102, 198-200). About ten described cases (some 1915, M.W. in 1918, Case 48 in 1917); negative charge on an insulated stool through a Wimshurst machine makes the Inner Aura vanish in under a minute (confirmed by "a void space" seen through a dark carmine screen), the Outer Aura mass and contract, and after charging the aura returns "enlarged ... often an additional fifty per cent", lasting 24 to 48 hours in M.W. Kilner is the only observer, measures by eye in inches, knows the expectation, and gives no counts of failures beyond "the large Wimshurst only occasionally failed". He corrects himself: "it was at first taken for granted that the aura soon resumed its natural state, but this was found to be an error". The June 1915 repeat on the artist's model is one subject, two measurements by Kilner. Ordinary effects of charging a person (ions, convection of the air layer, the observer's own expectation) are not excluded. Case 48 (a tremor patient who felt better after charging) is a single uncontrolled case. - Magnet near the body (1920 p.88-91). A ray joins body and magnet within seconds; no polarity. The thunderstorm episode is one of the few where the subject's ignorance was a control (eyes shut, magnet covered, "never once making a mistake" at pain); three or four days later no sensation; autumn 1916 a second patient felt "absolutely nil". No count of trials. - Fumigation (1920 p.109-118, from July 1912). Iodine on the left side makes the aura "red-brown and opaque", with a sharp boundary at the body's median line; bromine green, then blue-green, then blue by dose (a belief he corrected: variation came from dose, not from the person); chlorine "yellow green"; ammonia, carbon dioxide and hydrochloric acid fumes green. Controls: the model dressed in her shift, then disrobed, gave an identical result; a spinal pass of iodine gave results "corresponding to the extent of fumigation". Against it: Kilner is the only observer and chose the side fumigated; iodine, bromine and chlorine gas are themselves coloured (iodine vapour violet-brown, bromine red-brown, chlorine yellow-green) [memory, high], and a visible vapour in front of a black background would change a faint haze's colour without any property of the body. For: the sharp median boundary and the dressed control argue for a body-linked effect; ammonia, CO2 and hydrochloric acid are colourless, so a vapour's own colour does not cover those. The question is left open by this reviewer. - Dicyanin supply. In 1920 Kilner himself writes: "From the commencement of the recent calamitous war, it has been impossible to obtain any dicyanin ... formerly made in Germany", and that further work "must be made by other workers, as the author feels on account of age and growing infirmities unable to undertake them" [read: 1920 p.82-83, preface]. This is a primary statement, not the seller's. See 11.9. - Werner and Caan (Annales 1912, p.183 printed). A two-sentence relay from L'Année Électrique: Werner found normal human tissue has a "weak and uncertain" photo-activity, confirmed by Caan of Heidelberg, strongest from the brain; ashes of most organs make air conduct electricity. The reviewer adds a parenthesis that on the living body "special photographic screens" would be needed "for Kilner". No dates, quantities, exposures or sample size. Different effect from the aura. Tissue potassium-40 is an ordinary source of ionisation in ash [memory, high]. Kept as a lead.

11.8 Additions to claims 5 and 6 (1920)

• Colour willing carried over unchanged. The model's Experiments 1-6 are reprinted as Case 25 with the same two blind-location trials (one at the centre, one "about half an inch" off), and "Mrs. A." now named in Exp. 6; whether she is the "K" of Case 11 or the Case 25 widow is unresolved [read: 1920 p.121-126]. The Outer Aura, which the 1911 text called "entirely independent of the will power", becomes "far less under its owner's control" in 1920, with an exception for a hypnotised lady, unflagged as a change [p.135-136].
• "Some women, no man or boy." The preface statement rests on: the one model; two women under hypnosis (below); and "several women" patients whom Kilner says could have given "invaluable assistance" but "none of them were artist's models" and were unavailable [read: 1920 preface p.v, p.78]. No count of men or boys tested is given. A statement with no denominator.
• Hypnosis with Dr A. Douglas, two women, 17 experiments [read: 1920 p.126-133]. Miss X: awake she could not change colour ("a dismal failure"); in hypnosis "after some time" a grey-ochre instead of red, blue "tolerably well", a yellow chest patch invisible to the eye but shown by the blue C.C. band. Mrs Y: some change awake, "not very definite"; purple instead of red in hypnosis; a red chest spot detectable with the bands. In Exp. 8 Dr Douglas induced a hypnotic state "without the writer's knowledge" and Kilner saw the change and the deeper state "instantly" [p.128-130]. That is a single trial in which the hypnotist, not Kilner, knew the state. It is a better design than the location trials, but one trial. Dr Douglas observed his own subjects. Verdict: Never properly tested.
• Willed rays by other women. A girl just under twenty, and "the same experiment repeated by different women always with the like result" [read: 1920 p.118-121]. No counts.
• Chance and cue (2c, unchanged). The figures stay as reconciled in section 2c. The 1920 text adds that the model "signalled readiness" (p.125) and that Kilner "was not told" the position. That confirms the cue channel.

11.9 Absence-of-evidence update (section 6)

New evidence for each explanation, with stakes:

The combined reading in section 6 stands: war, the dye, one trained observer's method, poor repeatability, age and death, with imitation screens as a possible reputational factor. It remains a reading, not a finding.

11.10 Retention check (items the QE found missing, now in the record)

Pertinent material that the first pass and the primer did not carry, with its stake: the counter-expectation predictions and the pre-history readings (11.6); the colour-series and epilepsy counts (11.5); Case 1's mother and nurse and the "all can see it" passage (11.7); Dr Drysdale Anderson (seller-reported) and Mundy; the Scatcherd and Baraduc photographs (Boddington's paragraph, undated and unreferenced); the Archives review's own caution about "distinguishing between sight and suggestion" (research/2026-10-04-hathitrust.md); the opera-glass and 24-hour spectacles anecdotes (they are the origin of the focus work: 11.3); Boddington's heading "Dicyanine cures short sight" against his own accurate sentence (presbyopia only); the 1920 "women change colour at will" statement; the June 1915 repeat on the model; the corrections Kilner published (the retracted symmetric-healthy case; the bromine and enlargement-duration admissions; the shift from visual purple to a focus change; the softened statement on the Outer Aura and will; the closing sentence that "some of the hypotheses ... will have to be rejected", which names none). None was dropped because a reviewer judged it negatively; they were not read.

11.11 What would settle it: additions

1. Claim 1 and 2c (magnet haze): de Fontenay's two blind tests (magnets hidden among demagnetised ones; the 25-pair wire board with a distant switching operator), run with a sham-dye control and counts fixed in advance. For the sham, repeat the 1920 methylene-blue substitution deliberately, with numbers.
2. Claim 2b (ultraviolet): a UV-opaque but visible-transparent filter and a quartz element between subject and observer, with a spectrograph or UV-sensitive camera on the haze; an observer without a natural lens (aphakic eyes transmit more near-UV) [memory, medium] would sharpen the test of the "second spectrum" colours.
3. Claim 2d (fumigation): repeat iodine, bromine, chlorine and the colourless gases blind (an assistant picks the gas and side; the observer names the colour); measure the absorption of the vapour against a black background with a spectrometer.
4. Claim 2e (focus): an autorefractor before and after the 30-second dicyanin routine, with methylene blue as the sham; then the 1920 microscope series with a pupil-fixed (artificial pupil) eye.
5. Claim 3 (epilepsy, colour): blind classification of left-right narrowing by two observers on epileptic patients, other neurological patients and healthy volunteers; colour (blue, blue-grey, grey) assigned before the diagnosis is revealed.
6. Claim 4: the design in section 4 with a naive baseline; the cases with a withheld history (Case 17, Merrick) are the model.

11.12 Internal inconsistencies in the book (kept for the reader)

All from the deep read of Chapters V-VII [read: PDF p.246, 244/257, 259/264/266/308, 313, 80]. (i) Page 246 describes "Case 30" as having an enlarged granular Inner Aura; the description fits Case 40, and the page image prints "30". (ii) Page 313 says "Case 31 is an analogous instance where the stomach troubles were due to pregnancy", while Case 31 was diagnosed as nervous vomiting and recovered with treatment. (iii) The same stiff-neck patient's spot is "over the first lumbar spine" on p.244 and "over the second and third lumbar vertebrae" on p.257. (iv) Case 31's spinal strip is "lighter" (p.259, 308), "dark" (p.266, in the print, checked on the page image) and "upper one light, lower dark" (Fig. 27 caption). (v) The "ten epileptics" figure against 8 epileptics in the first-hundred table; reconcilable (the ten is probably a later total) but unexplained. (vi) The Table I condition column's case labels do not match the case list. These are Kilner's inconsistencies, not OCR errors, except (i) where a printer's error cannot be ruled out.

11.13 Consequences for the other files

• facts.json and claims.md: add the OCR case-number table, the colour-series and epilepsy counts, the counter-expectation scoring, and the three new document entries (1920 edition, Boddington as secondary and interested, Annales); replace "Case 88 / Case 21 appear twice" with 11.1; mark "dead foetus" in claim 4 as texture (Case 40); add the 1920 colour-at-will statement to claim 6.
• primer.md: drop "the one place in the book with counts"; replace the heredity sentences with 11.4; rescore the menstrual rows (11.6) and fix their citations (PDF p.253-254, 317-318); replace the worked chromatic-focus example's assumed 1/4-inch objective with the 1920 1.5-inch figure and the 0.5 D per mi calibration (11.3), and change "far larger than the eye's refraction" accordingly; add the epilepsy and colour rows to the evidence table; correct the case numbers; change "p.91-92" to PDF p.110-111; add Anderson, Mundy and "sight and suggestion" to further reading with the seller stake.
• example.py: add the sibling recount (9 of 13: 0.024 and 0.035; 9 of 16: 0.11 and 0.004; 8 of 12: 0.044 and 0.028; 8 of 14: 0.12 and 0.006), the epileptics-outside-blue calculation (0.009, 0.014), and the 0.18 to 1.8 D per mi range with the 0.5 D per mi calibration.