SilverTooth: Science Contrarian review (Step 2)

Reviewer role: Science Contrarian (judges by experimental evidence, applies the same rigour to the inventor and to the critics). Date: 2026-10-04. Sources: facts.json, claims.md, deep/Experimental-Detection-of-the-Either-dfc881.md, the Investigator's sites.md and research/ notes, and the saved 1976 Applied Optics paper (research/files). Tags: [read] = I read it myself in the folder or saved files; [web] = URL given; [memory] = recalled, with confidence; [calc] = my own arithmetic (Python, .venv).

Source gap: the project rests on one 5-page paper, "Experimental detection of the ether" (E. W. Silvertooth, received May 1986), a modern retyping with typos. It has no raw data, no error bars, no run count, no dates, no place. The RADC and DARPA reports named as sponsors are not in the folder and the Investigator found no trace of them. Every statement below about what happened in the experiment rests on this paper, on Silvertooth's own later statements, and on other people's reports of it. Where I rely on [memory] alone, I say so.

0. The short version

The paper contains one real piece of logic (a round-trip null cannot fix a one-way quantity), one unsupported step (carrying the Sagnac shift of a rotating loop over to a segment in straight-line motion), and one number (378 km/s) whose measurement is only described, never documented. The claim is important because it is concrete: a bench-top instrument said to read the solar system's absolute velocity from a half-cycle phase flip over 0.25 mm of slide travel. That is a testable claim, and it has been tested only partly. The one careful independent replication I know of (Marett 2012, [web], read here only through the Investigator's note) reproduced a sidereally aligned 12 h signal of about the right size, then found it was equally present in local time, had a large offset and the wrong sign, and tracked room temperature. Both halves belong in the record. Mainstream critics (Sherwin, Phipps) had named specific instrumental artefacts (AC coupling; a lambda/4 twist) years earlier. The original data were never published, so nobody can re-analyse them.

1. Corrections to Step 1

1. Symbol v-bar in eq. (10). The deep read left it undefined and noted that, if it is a velocity, Lv-bar/c^2 has units of time and cannot be added to a dimensionless term. The Investigator's lead is right, and I checked it in the saved 1976 paper: the extra Doppler factor is written with "the radial acceleration of the source", and the quasar example is "r = domega^2" for a quasar a distance d from the centre of a galaxy rotating at angular velocity omega [read: research/files/Silvertooth-1976-AppliedOptics-15-1100-anomalous-redshifts.pdf, p.1102]. So v-bar in 1986 is very probably an acceleration (the retyping lost the dot). The units problem then goes away. The deep read's concern should be restated as: eq. (10) is dimensionally fine if v-bar is an acceleration; the 1986 paper still never defines it.
2. Reference 8 (Gunn and Tinsley). The deep read's table says "page possibly 454" against the printed 434. Checked: "An accelerating Universe", Nature 257, 454-457 (1975) [web: https://www.nature.com/articles/257454a0]. The printed 434 is a typo (original or retyping). Their paper argued for a positive cosmological constant from the Hubble diagram and is described by the search summary as marginally significant and exposed to evolution corrections. It is not the 1998 supernova result and is not about the Doppler mechanism of eq. (10).
3. Reference 7 (Muller). Confirmed to exist: Scientific American 238 (May 1978) 64-74, U-2 radiometer, hotter toward Leo [web: https://www.scientificamerican.com/article/the-cosmic-background-radiation-and/ and https://ui.adsabs.harvard.edu/abs/1978SciAm.238e..64M/abstract]. The number about 390 km/s appeared only in a secondary web summary of the search; I did not read the article. The Forbes note gives the modern CMB dipole as 371 km/s [web: Investigator note on Koberlein 2018]. Retag: the Muller number stays [memory/secondary web, medium].
4. Timing of the paper. The 1986 paper says "with Leo on the horizon" but not whether rising or setting. This matters (see claim 4): a fixed east-west axis sees the sign of the motion component flip between rising and setting, but a half-cycle (180 degree) reversal cannot show the sign. Add this to the list of things the paper does not state.
5. Wesley 378 +/- 8 km/s (Investigator lead). The central value is identical to Silvertooth's single 378. Silvertooth's own 1989 article says 378 km/s was the value "on the day when this particular test was performed" and that all Delta values "ranged within +/-5%" [read: research/2026-10-04-spirit-science.md, which is the Investigator's transcription of the PDF]. +/-5% is about +/-19 km/s, so +/-8 km/s is a tighter figure than the inventor himself claims. Where Wesley's error bar comes from is unknown. I treat it as a re-analysis of Silvertooth's number by a sympathetic author until the original is read.
6. Interested parties. The record of "support" is mostly from claimants of absolute-motion detection (Marinov, Wesley, Whitney, Aspden) who have their own programmes. The record of "critique" includes Phipps (also a dissident-physics author, [memory, medium]) and Marett, who says he hoped the effect was real (Investigator note). Neither bloc is neutral. I report each as that party's characterisation.
7. Funding line. Step 1 quotes "sponsored in part by" RADC and DARPA [read: p.4]. No report trail was found. This is an unexplained absence, not a refutation (competing explanations in sec. 7).
8. Minor: Forbes says Academy award "1942"; atogt (Investigator) confirms the 1941 award. Not load-bearing.

Everything else in the deep read's algebra checks I re-derived and agree with (eq. 5 from 3 and 4; eq. 6-7 exact; v/c = lambda/(2*Delta) gives 377.7 km/s [calc]).

2. What the instrument claims to do (so the verdicts are about something specific)

Light of fixed frequency nu travels in the lab at c1 = c/(1+v/c) one way and c2 = c/(1-v/c) the other, so the wavelengths differ (lambda/(1+v/c), lambda/(1-v/c)) [read: eq. 5, p.2]. A micrometer slide moves the source assembly toward one mirror and away from the other, so one path gets longer by the same amount the other gets shorter. A thin photocathode (D1) senses the standing wave of the two oppositely directed beams. A second interferometer (D2) counts rest wavelengths. After a slide travel Delta, D1 and D2 are 180 degrees out of phase; then v/c = lambda/(2*Delta) [read: eq. 8-9, p.4]. Worked check [calc]: lambda 0.63 um, Delta 0.25 mm, so about 397 wavelengths of travel, v/c = 1.26e-3, v = 377.7 km/s.

Sizes that matter for judging artefacts [calc]: - A half-cycle error in 397 wavelengths of travel is a fractional wavelength mismatch of about 1.3e-3. A helium-neon laser's frequency range is of order 1e-6 of the optical frequency (gain bandwidth about 1.5 GHz against 4.7e14 Hz, [memory, high]), so simple laser frequency drift is about a thousand times too small to fake it. This is evidence in the inventor's favour on the "simple wavelength drift" objection, and should be said. - A cosine (tilt) error on the slide is second order: to give lambda/4 over 0.25 mm it would need a tilt of about 0.035 rad (2 degrees) [calc]. Not plausible for a good slide. - But first-order geometric errors are not excluded: a lateral beam walk, a twist or an Abbe offset of the moving assembly of a quarter wavelength over 0.25 mm. This is Sherwin's objection ("of the order of lambda/4" over about 500 lambda, [web: Investigator note, naturalphilosophy wiki]). Marett's version (beam exit angle from the laser follows temperature) is a first-order pointing error of this type. The paper gives no arm lengths, mirror mounts or thermal control, so none of this can be ruled in or out from the paper alone. - The phase comparison was made by eye on a dual-trace oscilloscope [read: p.4].

3. Claim-by-claim

Claim 1. Michelson-Morley-type experiments cannot detect an ether because a direction-dependent light speed appears as a direction-dependent wavelength, and the round-trip count is unchanged

a. Mathematics. Internally consistent but circular. Eq. (6)-(7) are eq. (3) rewritten after substituting eq. (5), so "n and t independent of v" is the assumption A2 (exact round-trip null) returning as the result [read: deep read, Reader's notes]. Also the speeds c/(1 +/- v/c) are not the Galilean c -/+ v; they are exactly the form used in the standard conventionality-of-simultaneity literature (Reichenbach; Winnie 1970), where any one-way speeds that average correctly over a round trip are allowed [memory, high; Winnie cited in Forbes note]. So the mathematics reproduces a known result in different words. b. Experiments. The round-trip null results are not in dispute: modern rotating-cavity tests give no anisotropy of round-trip c at about 1e-17 [web: https://arxiv.org/abs/1002.1284]. The paper's weak statement (a round-trip experiment cannot measure the one-way speed) agrees with all of them. The strong statement (that a real, physical wavelength anisotropy exists in the lab and the null result "requires neither isotropy in c nor the Lorentz contraction") is not tested by any of them; it is the same claim that claim 4 tries to test. c. Tested? The weak form is a theorem-level point (Winnie, Tyapkin 1973, both cited by the paper's critics and by the paper's own reference 5). Silvertooth's own 1972 JOSA one-way test is recorded by the SR FAQ (Roberts and Schleif) as a null consistent with SR [web, Investigator note: math.ucr.edu; interested party: FAQ authors are mainstream SR teachers]. I have not read the 1972 paper. d. If true (strong form): there is a preferred rest frame (here the CMB frame) in which light travels at c, and measuring devices that see only round trips would be blind to it. That is the Lorentz-ether position, with the difference that it asserts a first-order, lab-scale one-way effect that SR denies.

Verdict, split: - Weak form (round-trip null does not fix one-way speed or wavelength): Established (as a statement of conventionality), with the note that the paper arrives at it by a circular route. - Strong form (a real anisotropy of wavelength in the lab, with no length contraction needed): Not testable as stated at the level of this argument; it becomes testable only through the instrument of claim 3 and the result of claim 4, which are judged separately.

Claim 2. The Sagnac shift of a straight segment "in pure translation" gives one-way wavelengths lambda/(1 +/- v/c), first order in v/c

a. Mathematics. The algebra of fig. 1 is right (4Aomeganu/c^2 = 2Lv/(clambda) with A = Lr/2, omega = v/r) [read/calc]. The independence from r is trivial: r cancels because v = omegar was held fixed; it just says the phase is the line integral of v along the segment. The decisive step A3 is that this local line integral is a physical phase shift of that segment alone, preserved in straight-line translation. For a closed loop in uniform translation the integral of v around the loop is zero, which the paper itself uses to cancel the shift on the other two sides [read: p.2]. So the claim is that an open segment carries a phase that is physically meaningful but cancels on every closed path. The 1976 paper says the same thing ("it is simply cancelled in the closed path") [read: 1976 p.1101-1102]. That is a statement that the shift cannot be seen by any closed-path (single-source interferometer) measurement, which is exactly what makes it equal in effect to a synchronisation convention. A hidden assumption: that "absolute time" synchronisation along the segment is the physically correct one (the offset is vL/c^2 between synchronisation conventions; this is [memory, high], standard SR kinematics, not in the paper). b. Experiments. Sagnac effect for rotation is well established (ring laser gyros, fibre gyros, the 1976 Applied Optics page before Silvertooth's own letter describes a fibre gyroscope [read]). That a translation produces the same segment shift is not established by any experiment I can cite. Silvertooth's own 1976 test of the mechanism (two free-running lasers, two photocathodes 50 cm apart on an air-bearing granite block, 79 cycles expected for delta v/c = 1e-4) reports "No phase shift was observed" [read: 1976 p.1102]. He reads this as showing phase invariance and a wavelength change that is a result of frequency and velocity changes. Whatever reading one prefers, that test as printed does not show the wavelength anisotropy. c. Tested? Only by Silvertooth himself (1976) with a null as printed, and indirectly by claim 4 and the replications. d. If true: the Sagnac phase would be a local, absolute-velocity-dependent property of a path segment in inertial motion; the "relativity principle" at first order in v/c would fail.

Verdict: Tested, not supported as a physical lab effect (by the inventor's own 1976 null as printed; no positive test elsewhere), and the key extrapolation A3 is an unsupported step, not a derived one. I flag that I read the 1976 text layer only (an OCR-quality text layer, not page images), so the exact wording of the null should be re-read by the Arbiter.

Claim 3. The one-way wavelength can be measured with a slide-mounted two-beam standing-wave interferometer: v/c = lambda/(2*Delta)

a. Mathematics. The working formula follows from eq. (8)-(9): ndelta = lambda/2 with n = Delta/lambda and delta = lambdav/c gives Delta(v/c) = lambda/2 [read/calc]. Reconstruction of the shift (move of lambda changes the two paths by 1 +/- v/c wavelengths; the standing wave moves by half the phase difference, so delta = lambdav/c) is the deep reader's, not the paper's. Two letters (n, delta) do double duty in the paper, and the labels 1 and 2 are swapped between eq. (5) and fig. 2 (and again in the 1989 article) [read]. These do not change magnitudes. b. Consistency with experiment. Two points against: (i) any physical standing-wave position is a lab observable, so the claim is not a convention but a prediction; SR predicts zero shift. (ii) One-way experiments with no clock synchronisation problem, such as the guided-wave one-way tests (Gagnon, Torr, Kolen, Chang, Phys. Rev. A 38, 1767 (1988) [web: search result via https://journals.aps.org/pra/issues/38/4]) and a long list of Mossbauer-rotor and fast-beam Doppler tests [memory, medium], found nothing at first order in v/c; I have not read their results here. These bound the same first-order one-way effect and are the strongest mainstream evidence against the principle. c. Tested? By Silvertooth (his own, undocumented in this paper), by Marinov in January 1987 in a "substantially different arrangement" reporting 386 +/- 38 km/s (declination -22 +/- 6 degrees, RA 12.5 +/- 0.5 h) [web: naturalphilosophy wiki abstract, via Investigator], and by Marett 2011-12 with a thin-film standing-wave sensor from the Julich Research Centre [web: http://www.conspiracyoflight.com/Silvertooth/Silvertooth.html, via Investigator note; not read by me]. Marinov is himself a claimant of absolute-motion detection (interested party). Marett's result, in his own words as quoted: "it is in fact only a very elaborate thermometer!" and he also cites Tyapkin (1973) and Lorentz's theorem of corresponding states for a one-way wavelength test being impossible in principle [web, Investigator]. Marett's own earlier single-laser standing-wave test (2011) gave a null (title only, not read). d. If true: the Lorentz transformations would not hold at first order for one-way light propagation; standard clock-synchronisation convention (Einstein) would not be a convention.

Verdict: Tested, not supported (one careful independent replication that did not support the interpretation, plus the mainstream one-way tests; Marinov's partial repetition is the exception and is not neutral). Not "dismissed without testing": there was a replication attempt. But there has never been a neutral, blinded, instrument-controlled test with the original apparatus, and the original apparatus data do not exist in the record.

Claim 4. East-west travel with Leo on the horizon gives Delta = 0.25 mm, so v = 378 km/s; no effect north-south or with Leo 6 or 18 hours away; agrees with Muller

a. Arithmetic is right: 377.7 km/s [calc]. In 1989 Silvertooth gave the same 378 from a typical Delta = 0.025 cm with different bookkeeping ((lambda1 - lambda2)/lambda = lambda/Delta, 2v/c = 0.00252): 0.000063 cm / 0.025 cm = 0.00252 is consistent with the 1986 value [read: Investigator transcription]. But 0.025 cm is 0.25 mm, so these are one value reported twice (1986, 1989), not two measurements. Units and notation: no unit errors found in the calculation. Geometry check [calc, with the CMB dipole direction RA about 11.2 h, declination about -7 degrees [memory, medium], and latitude 48.6 degrees N for Olga, which the paper does not say is the site]: such a source rises at azimuth about 100 degrees (about 10 degrees south of east), so a fixed east-west axis has projection cos(10 degrees) = 0.98 and a north-south axis about 0.18; it rises about LST 5.7 h and sets about 16.6 h; at transit it is due south at altitude 34.5 degrees, so the east-west projection is zero and the north-south horizontal projection is 0.82. This matches Marett's observed peaks "at about 5 h and 17 h sidereal" and the east-west pattern (zero at transit), so the pattern Silvertooth reports is internally consistent with a dipole-like source. It also shows that the north-south null "with the apparatus rotated 90 degrees" is expected at rising/setting and not at transit; the paper does not say when the north-south run was made, and Marett argues the north-south null contradicts Silvertooth's own theory because Leo has a north-south component at 33 degrees elevation [web: Marett, via Investigator]. Without the time stamp neither side can be checked. b. Experiments. The dipole speed itself (about 370 km/s, from CMB observations) is established and Muller's U-2 result is real [web]. What is not established is that a laboratory standing-wave instrument responds to it. SR predicts a null, and the precision one-way and round-trip tests (see claim 3) say any such response is far below 1e-3 for c unless it is of a kind they cannot see (the very conventionality loophole). An ability of the instrument to see it would contradict the claim 1 argument itself, which says the one-way quantity is not measurable by round-trip means but is measurable by this single-path means; so it is a claim about this device only. c. Tested? Silvertooth reports the result in one paper (no data), then "repeated in a variety of configurations over the past several years. Values of Delta measured have all ranged within +/-5% of the cited value" (1989) [read: Investigator transcription of the PDF]. Marinov 1987 and Wesley 1987 (see corrections). Marett 2012 found: for the inventor: a 12 h diurnal change aligned with the sidereal Leo directions and a magnitude on average near the CMB motion after normalising offsets; against: a large non-zero offset when the beam was perpendicular to the motion, large day-to-day scatter, wrong sign (the sensor frequency always lower), the same 12 h pattern in local solar time, the same pattern from a single Michelson detector with no standing-wave sensor, and correlation with room temperature and laser lock voltage [web: Investigator note]. A reanalysis by Sherwin (1989, Physics Essays 2(2) 125) says AC coupling to the oscilloscope makes nodes look like "inverse phase" and that "a very small misalignment (of the order of lambda/4) in a total traverse of approximately 500" wavelengths could produce the observed effect [web: naturalphilosophy wiki abstract]. Note this critic's claim is a possible mechanism, not a measurement of the original apparatus. I did not find a published test that showed Sherwin's mechanism to be the cause. How it was rejected: no evidence of suppression of the experiment in the sources I have. The sidebar in the 1989 Wireless World article speaks of suppression of dissent (Essen), but it is an unsigned sidebar whose author is unknown. The Nature piece is described by Koberlein as a one-column letter that promised details to follow [web: Forbes note]. The detail never reached a refereed journal with data (Marett: none of Silvertooth's papers has data graphs). So the original result was neither tested under controlled conditions nor properly reanalysable. Funding: RADC and DARPA sponsorship has no report trail (sec. 7). d. If true: the ether (or a preferred frame at the CMB velocity) is detectable by a table-top optical instrument at first order, SR's relativity principle fails for one-way light propagation, and GPS-grade clock transfer would show the effect unless engineered around it.

Verdict, split: - The existence of a sidereally modulated 12 h signal of roughly the stated size in this kind of apparatus: Plausible, reproduced once with caveats (Marett), but with the replicator's evidence that it is thermal and not kinematic. I give this part "Tested, partly reproduced as a signal, not supported as a velocity". - The interpretation as the solar system's absolute velocity (378 km/s): Tested, not supported. - The original run itself (0.25 mm, one value, no errors): Never properly tested at the level of the original data, which are not available.

Claim 5. The same velocity-dependent wavelength count explains the accelerating universe, anomalous quasar redshifts and GPS errors; extra term L*v-bar/c^2

a. Mathematics. Eq. (10) is given with no derivation, and v_obs, v_source and v-bar are not defined in the 1986 paper [read]. The 1976 paper supplies the derivation sketch and the meaning of v-bar as radial acceleration [read: 1976 p.1102]. First term: the paper's printed form has v^2/c^2 inside both roots; the standard longitudinal Doppler factor is sqrt((1-v/c)/(1+v/c)) [memory, high]; the 1976 text layer shows (1/gamma)(1 +/- v/c), which equals that for the sign choice. So the 1986 first term is probably a typing error; the Mathematician should check. My own order of magnitude [calc]: with L = 1e25 m (the paper's 1e27 cm), a galaxy radius d = 10 kpc (3e20 m) and omega = 1e-15 rad/s (period about 200 million years, [memory, medium]): a = domega^2 = 3e-10 m/s^2 and La/c^2 = 0.033. The ordinary Doppler shift from the rotation speed omegad = 300 km/s is 1e-3. So the extra term is about 30 times the ordinary rotation shift and of the size (Delta z of a few hundredths) that the discordant-redshift literature discusses [memory, low-medium; not verified]. So the claim is not an absurd magnitude; it is a tunable one ("depending on the numbers"). Conceptual problem: in ordinary kinematics the observed Doppler shift depends on the source velocity at emission. A term proportional to La/c^2 is the velocity change a source with acceleration a would make over the light-travel time L/c; it would appear if the observation compared the source's velocity at emission with one at reception. This is my reading of where such a term could come from and I have not seen the author say it. Hedge accordingly. b. Experiments. Accelerating expansion: Gunn and Tinsley 1975 [web: https://www.nature.com/articles/257454a0] inferred a cosmological constant from the Hubble diagram; the 1998 supernova discovery is a separate, later result [memory, high]. No mechanism connects either to eq. (10). If eq. (10) held, high-redshift rotation curves would show differential line-of-sight redshift between the near and far sides of rotating galaxies amplified by about L*omega/c (about 30 in the numbers above); I am not aware that this is observed [memory, low; unverified]. GPS: the first-order Sagnac correction (up to about 133 ns, or about 40 m, from Earth's rotation) is a standard, documented part of GPS processing and is applied, and the system works at metre level with it [memory, high; Ashby, Living Reviews in Relativity 2003 not re-fetched]. Silvertooth's own 1990 paper treats "the position error when the Sagnac contribution is ignored" [web: naturalphilosophy wiki abstract, via Investigator], which is therefore an error in a system that does not ignore it. I have not seen the paper. c. Tested? No. Silvertooth cites no data for quasars or GPS in 1986, and the Investigator found no test. d. If true: redshifts would partly measure acceleration times distance, not velocity, which would change distance-ladder and dark-energy inference. It would also require re-deriving GPS corrections.

Verdict, split: - Quasar redshift anomaly mechanism (eq. 10, v-bar as acceleration): Not testable as stated in the 1986 text (no definition, no derivation, no numbers); plausible-sounding in magnitude per my arithmetic, but never tested. - Accelerating universe: Not testable as stated (one sentence, no calculation). - GPS errors: Refuted by existing practice if the claim is that GPS is wrong because it ignores the Sagnac term (it does not ignore it); otherwise Not testable as stated.

4. Evidence the mainstream reviewers tend to leave out (kept in the record whatever the verdict)

• The null of Silvertooth's own 1976 two-laser test is a point against his mechanism, and a point for his honesty: he printed it. [read]
• Marett's replication found the predicted sidereal alignment, and its size on average near the CMB value, from an independent experimenter who was hoping for the result. That is real evidence that something at 12 h sidereal periodicity is in the instrument; it is also evidence for a thermal mechanism. [web, via Investigator]
• Laser frequency drift cannot by itself produce the signal size (about 1e-3 against about 1e-6) [calc]. The competing artefacts are geometric (twist, walk, pointing) or electronic (AC coupling).
• The positions of Leo's rise and set at Olga fit the reported East-West maxima and zero at transit [calc].

5. Evidence against, stated plainly

• No data, error bars, run count or date in the paper that carries the headline. One value reported twice (1986, 1989). [read]
• Phase reversal judged by eye on an oscilloscope, with Delta adjusted until the reversal appears ("an arbitrary choice of the experimenter") [read: p.4]. This is a procedure where the experimenter's expectation can steer the outcome.
• The Michelson-Morley argument is circular (claim 1).
• Reference errors in the retyping (volume/page swaps, 1913 vs 1915) show the text is not a faithful copy of the original; the deep read already flagged it.
• Eq. (10) undefined symbols and a likely typed Doppler factor.
• Precision first-order one-way tests (Gagnon 1988 and others) and the general success of GPS time transfer do not show the effect.

6. Does the claim need the mainstream to change?

Yes for claims 2-4: a preferred frame detectable at first order by one-way optical means. That would be a larger change than Lorentz-ether theory (which is built to be undetectable at first order). For claim 1, no. For claim 5, a different treatment of redshift, with no calculation provided.

7. Absence of evidence: sponsorship and data

The 1986 paper says the work was sponsored in part by RADC and DARPA [read: p.4]. The Investigator found no report in DTIC (via Google; direct DTIC search not done), no university report beyond a pointer to the Arizona OSC FY75 report, and no obituary. Competing explanations: 1. Informal or in-kind support with no report written (the Wikipedia-type biography mentions classified Air Force think-tank work, but that is an unsourced wiki statement). 2. Reports exist but are not indexed online or have restricted distribution. 3. The sponsorship was a minor or contested contribution described generously. 4. A reviewer-side explanation: a reviewing RADC physicist (Newburgh, 1977, AO comment on the 1976 paper, [web: Investigator]) was critical, so institutional support may have been limited. What would decide: a direct DTIC search, AFRL Rome history office, DARPA archives, a FOIA request, and the Arizona OSC FY75 report. Not-yet-read papers: the 1986 Nature letter, Silvertooth 1972 JOSA, 1992 Whitney and Silvertooth Physics Essays, 1990 Galilean Electrodynamics, Sherwin 1989, and Wireless World Nov 1986 and Nov 1989.

8. Verdict table

9. What would settle it

A blinded, automated run of the original design with the oscilloscope eyeballing replaced by lock-in readout, a thermally controlled and pointing-monitored laser, and a rotating table to separate sidereal from local-time and thermal effects, run at a known latitude with the north-south and east-west axes at declared times. If the 12 h sidereal signal survives with a constant sign and no thermal correlation, claim 4 would become a real anomaly; if it tracks room temperature (as Marett found) and shows the same pattern in solar time, it is closed. Independent of that, the Arbiter should ask for the 1972 JOSA and 1992 Physics Essays papers.