Science Contrarian review: Verlinde-Erik

Reviewer: Science Contrarian (Step 2). Project document: E. Verlinde, "On the Origin of Gravity and the Laws of Newton", arXiv:1001.0785 (6 Jan 2010; published JHEP 04 (2011) 029). Source text was readable (29 pages, no source gap). Tags: [read: p.N] from the paper text layer (I re-read pp. 8-10, 22-23, 28 myself); [deep] = Step 1b deep read of pp. 12-22 (equations checked there against page images); [web] with URL; [memory] with confidence.

Standard I apply: judge by experimental evidence, separate "tested and failed" from "never properly tested" from "dismissed without testing", and be as hard on the author as on his critics.

Headline

This is an unusual case for a contrarian: the 2010 paper is not an anomalous-experiment claim and was never dismissed without testing. It is a reinterpretation that reproduces Newton's law and (in sketch) Einstein's equations, so it cannot be distinguished from them by any experiment on the paper's own terms. The testable content lives in the successor programmes (the 2016 "dark gravity" model; the 2019+ fluctuation model), and those have been tested, with mixed and partly negative results. The most honest case for the paper is as a heuristic of real conceptual interest in the lineage of Jacobson 1995; its strongest weaknesses are circularity (the author admits the reversal of logic), the failure of the screen assumptions beyond spherical symmetry, and the neutron-interferometry objection.

Corrections to Step 1

1. claims.md lists Sections 1-3 and 6 as digest-only. I re-read pp. 8-10 and 22-23 in full text and found nothing that changes the claims, but two things deserve more weight than claims.md gives them: - p.8: "For the following it is best to forget about the Unruh law (3.8), since we don't need it. It only served as a further motivation for (3.6)." Yet (4.17) and (5.28) reuse the Unruh form for T [deep]. The paper is internally loose here; claims.md does not flag it. - p.9: "These equations do not just come out by accident. It had to work, partly for dimensional reasons, and also because the laws of Newton have been ingredients in the steps that lead to black hole thermodynamics and the holographic principle. In a sense we have reversed these arguments." This is the author's own admission that the derivation is a consistency reversal, and claim 3 in claims.md ("Newton's law follows from holography") is stronger than the author's own wording. The paper says "practically from first principles" (p.9) and that the 2 pi is "picked precisely" to give F = ma (p.8).
2. claims.md says claim 3 is "[read: p.8-16]" and the author caveats list "equipartition ... least obvious". Confirmed (p.10: "Perhaps the least obvious assumption is equipartition, which in general holds only for free systems").
3. The deep read states sign looseness in (4.25) and (5.27)/(5.30) with magnitudes right [deep, calc]. I accept this; the inventor's 2026 note says the sign works because S is an adiabatic invariant and that the naive F = -T dS/dx reading is "not correct" [web] https://www.worldscientific.com/doi/10.1142/S0217751X26470020 (per Investigator notes). So the 2010 presentation of "force from entropy increase" is superseded by its own author.
4. sites.md says the Investigator found that "the record does not go silent". I agree; no absence-of-evidence anomaly exists here.
5. Claim 6 in claims.md is correctly reported as non-predictive in 2010 (p.28 per claims.md); I could not find in the text I read (the p.28 page is references) where this appears, so I rely on claims.md [read: via claims.md]. Not corrected, flagged for the QE to confirm the page number: the page 28 in the text layer is Acknowledgements/References, so the force-fluctuation remark is probably on p.25-27.
6. Investigator wording check: "Brouwer 2017 ... EG prediction is in good agreement" is the authors' own phrase [web] https://arxiv.org/abs/1612.03034; the Investigator did not list Lelli, McGaugh & Schombert 2017 (see Leads below), which is the main contrary test of the 2016 model.

Claim-by-claim

Claim 1. Gravity is an entropic force, F = T dS/dx

a. Mathematics: Internally consistent as algebra. The polymer analogy (2.4) and the microcanonical form (5.31) follow correctly [deep, calc]. Sign conventions are loose (the deep read found the (4.25) integral equals the work done against gravity; (5.27) has the hovering observer's acceleration with the free-fall sign) but magnitudes match. The author later acknowledges the naive entropic-increase picture is not right (2026 note, above). b. Experiments: The decisive tension is with quantum gravity experiments. [memory: high] Colella-Overhauser-Werner (1975) neutron interferometry showed a coherent gravitational phase shift on neutrons, and ultra-cold neutron bound states in Earth's field have been seen (qBounce/GRANIT, 2002 onwards). Kobakhidze argues an entropic (thermal, decohering) force would spoil these: "experiments with ultra-cold neutrons in the gravitational field of Earth disprove recent speculations on the entropic origin of gravitation" [web] https://arxiv.org/abs/1009.5414 (abstract only; I could not retrieve the detailed argument). Chaichian, Oksanen and Tureanu found it "does not necessarily contradict" bound neutron states, but spontaneous decays of excited states "cannot be explained" without extra ingredients [web, via Investigator] https://arxiv.org/abs/1104.4650. Verlinde's 2026 answer is that in the Born-Oppenheimer regime the slow system keeps its coherence [web, via Investigator]. So the neutron objection is contested between two published camps, not settled. c. Tested? Not by a neutral controlled test of the specific claim. The neutron data exist, but the dispute is about what the entropic picture predicts, not about the data. Replication: the neutron states are replicated; whether they disprove the claim is the argument. Wang and Braunstein 2018 show the first law of thermodynamics on holographic screens fails away from horizons except under spherical symmetry, "undermining the key thermodynamic assumption" [web, via Investigator] https://www.nature.com/articles/s41467-018-05433-9. This is a calculation, not an experiment; a reader comment (A. Peach) says it targets area-law screens and not the 2016 volume-law model. Visser 2011 finds "real and significant problems" for general-potential entropic force constructions, but not for Jacobson/Padmanabhan versions [web, via Investigator] https://arxiv.org/abs/1108.5240. d. If true: gravity is not a fundamental interaction; no graviton as an elementary quantum; Newton's G becomes a count of microscopic degrees of freedom. Mainstream quantum gravity programmes that quantise the metric would have to change.

Verdict: Never properly tested (as stated); the neutron-interferometry objection is a contested argument by two published groups and not a controlled test of a quantitative entropic prediction. Not "dismissed without testing": it was debated in the literature within months. Part: the strong form (a thermal decohering force) is Tested, not supported only if Kobakhidze's reading is correct, which Chaichian et al. dispute.

Claim 2. Inertia is entropic: dS = 2 pi k_B (mc/hbar) dx gives F = ma

a. The 2 pi is chosen to produce F = ma (p.8). That makes the "derivation" an identification: given Unruh T and the postulated dS, F = ma is a one-line consequence, but the postulate was built to give it. Consistent, circular as a derivation. Hbar cancels, which the author uses as a feature (p.10: hbar "can be chosen at will"). b. Consistency with experiment: F = ma is of course in agreement with all experiment; but so is any theory with it as input. The Unruh effect itself has no direct laboratory confirmation as of my knowledge [memory: medium]; the claim leans on it as motivation only. c. Tested? Nothing distinguishes it from F = ma. Not testable as a difference. d. If true: inertia would be the response to information change, with no new prediction I can find in the paper. Inertia from Unruh-type heat baths was proposed earlier by others [memory: medium] (the author cites Jacobson, and in 2026 names inertia as his addition).

Verdict: Not testable as stated (interpretation with no measurable difference from F = ma).

Claim 3. Newton's law follows from holography, N = Ac^3/(G hbar), equipartition

a. The algebra is correct and I checked the structure of pp. 8-9: eliminating E and T gives F = GMm/R^2 [read: p.9]; the Gauss-law step (4.20) reproduces by the deep read's symbolic check; Earth numbers give g = 9.82 m/s^2, T = 3.98e-20 K, N = 1.95e84 bits and (1/2) N k_B T = M c^2 by construction [deep, calc]. Circularity: G is defined by N (p.8: "one can simply regard this equation as the definition of G"), M is input, the 2 pi is tuned, Phi is "a device to describe the local acceleration" (4.17) [deep]. The Poisson step (4.21) rests on a one-sentence requirement for arbitrary screens, while screens are defined as equipotentials [deep]. Dimensional analysis forces the form, which the author concedes (p.9). b. Experiments: it reproduces Newtonian gravity in the Newtonian regime, nothing more. It cannot be contradicted by any Newtonian data, nor does it predict a new number. It also implies Bekenstein-Hawking entropy on the screens, which the author notes appears to violate the Bekenstein bound [per claims.md, p.22-23]. c. Tested? The identity with Newton is a consistency check, not a test. Critiques are theoretical (Visser; Wang-Braunstein). d. If true: holography and equipartition would be foundational, not derived from gravity; the status of the Bekenstein-Hawking entropy would have to be re-argued.

Verdict: Plausible, untested as a derivation (it is consistent with Newton by construction; no discriminating test). Split: the equipartition-on-screens ingredient fails the first-law check away from spherical symmetry (Wang and Braunstein, a calculation); I class that part as Internally inconsistent beyond spherical symmetry only if their calculation holds, which I did not verify.

Claim 4. Relativistic version: Komar mass and the Einstein equations (sketch)

a. (5.30), (5.31), (5.34) follow by algebra; I confirm the Schwarzschild check in the deep read gives M at every r > 2GM [deep, calc]. The matter side of (5.37) is posited ("presumably fixed by comparing properties on both sides", p.20 per deep read) and in Wald's account is obtained using the Einstein equation, so it is input [deep]. Approximate Killing vectors in small regions do not satisfy (5.36) exactly [deep]. The cosmological-constant term is not discussed [deep, memory: medium]. The author calls it "a sketch". b. Experiments: if it reproduces GR it agrees with GR's tests (Mercury, light bending, gravitational waves) by construction; no independent check. c. Tested? Not independently derived or tested. The 2026 note relates the Jacobson argument to Wald entropy as "the reverse argument" [web, via Investigator]. Jacobson 1995 (PRL 75, 1260) is the prior and more careful version of the null-screen case [read: p.28, ref 7]. d. If true: Einstein's equation would be an equation of state, as Jacobson already argued; geometry would not be fundamental.

Verdict: Plausible, untested (a sketch resting on Jacobson's earlier argument; not independently verified or falsified). The extension to time-like screens at constant redshift is the part with no neutral check, and Wang-Braunstein indicates the thermodynamic assumption fails off spherical symmetry.

Claim 5. Space, geometry and gravity are emergent; redshift is a consequence of entropy gradients

a. Not mathematics but a programme; supported by analogy to AdS/CFT and open/closed string duality (pp.22-23). The "other way" admission (p.19, deep): "we should have started from the microscopics". The argument that string theory's non-Wilsonian UV/IR connection supports emergence is an analogy. b. Experiments: none directly. The only quantitative cosmological extension is the 2016 model (area law plus a volume term in de Sitter space, extra "dark gravity" tied to a_0 = c H_0) [web, via Investigator] https://arxiv.org/abs/1611.02269. Tests of the 2016 model, in the record: - For: Brouwer et al. 2017, 33,613 isolated galaxies (KiDS x GAMA), prediction "in good agreement" with lensing, "only a first step" [web, via Investigator] https://arxiv.org/abs/1612.03034. Brouwer et al. 2021 (KiDS-1000; Verlinde a co-author, an interested party) found the RAR "agrees well" with MOND and EG predictions [web, via Investigator] https://arxiv.org/abs/2106.11677. - Against: the same 2021 paper found a difference of at least 6 sigma between early- and late-type galaxies that modified-gravity theories "would be unable to explain" unless extra gas is invoked [web, via Investigator]. Lelli, McGaugh and Schombert 2017 found EG consistent with the observed RAR only if stellar mass-to-light ratios are substantially decreased (values in tension with other estimates), and that EG predicts radius-correlated residuals that are not seen [web] https://arxiv.org/abs/1702.04355 (search-result summary of the abstract; I did not read the full paper). Pardo 2020: for 452 isolated dwarfs EG "severely underpredicts" velocities above ~100 km/s and overpredicts below [web, via Investigator] https://arxiv.org/abs/1706.00785. Nieuwenhuizen 2017 reports Abell 1689 lensing rules out EG without eV neutrinos [web, via Wikipedia per Investigator; primary not opened]. This record is mixed: the 2016 model has been tested and is contested; these are tests of the successor, not of the 2010 paper. c. Tested? The 2010 emergence claim: not testable as stated. The 2016 successor: tested by independent groups, with results both ways; no clean replication or refutation. The fluctuation model: see Claim 6. d. If true: dark matter phenomena might be an elastic response (not particles), and the direct-detection searches would stay null; a massive change in cosmology and in quantum gravity.

Verdict: Not testable as stated for the 2010 paper's emergence claim. The 2016 extension: Tested, not clearly supported (mixed evidence, listed above, with each side kept).

Claim 6. Gravitational force fluctuations, larger for weak fields between small bodies

a. No mathematics in the 2010 paper; the author says better theory is needed to turn it into a prediction. b. Experiment: Verlinde and Zurek made the idea quantitative as "geontropic" spacetime fluctuations in interferometers [web, via Investigator] https://arxiv.org/abs/1902.08207. Carney, Karydas and Sivaramakrishnan: standard graviton effective field theory predicts only Delta L ~ 1e-35 m, so a detected large signal would mean a breakdown of effective field theory [web, via Investigator] https://arxiv.org/abs/2409.03894. The Holometer set an experimental constraint (alpha) on geontropic fluctuations, and GQuEST projects probing alpha < 0.6 at 3 sigma in 60 hours [web] https://arxiv.org/abs/2404.07524 (search summary). I found no published GQuEST result. c. Tested? Not yet properly; an experiment is being built (an experimental test is designed by interested parties; Zurek is Verlinde's co-author). The Holometer limit is a partial constraint on one parameter of the model, per the GQuEST abstract; I did not read the Holometer paper. d. If true: effective field theory of gravity would break down at low energy; a clean signature.

Verdict: Never properly tested (testable, with a dedicated experiment designed and no result found; not dismissed without testing).

Honest case for the claims

• Jacobson 1995 and Padmanabhan 2010 show gravitational field equations can be read thermodynamically; Verlinde adds inertia and a "mechanism". Strominger, as quoted by Wikipedia (per Investigator), said reactions ran from "can't be right" to "right and we already knew it" [web, via Investigator], which fairly sums the situation.
• Prize and publication record: Spinoza Prize 2011, JHEP publication, the programme continues in mainstream journals up to 2026. It was neither suppressed nor ignored: there is no absence-of-evidence anomaly.
• The successor programme made a falsifiable low-acceleration prediction, and it was exposed to data (Brouwer; Lelli; Pardo), which is what a science-by-evidence standard asks for.

Against the claims (same rigour on the inventor)

• The derivation is a reversal of known results, admitted on p.9. "Cause, a mechanism" (p.22) is a claim of explanation that the equations do not deliver; there is no microscopic model in the 2010 paper.
• Several free choices (2 pi, G as bit density, T from the Unruh form after saying it was not needed, screens as equipotentials) each tuned to recover the known result.
• Typo: (4.17) as printed has an extra k [deep]. Sign slips in (4.25), (5.27)/(5.30) [deep].
• The Bekenstein-bound tension and the non-equilibrium of equipotential screens are admitted but unresolved [per claims.md].
• Bad-data check: I found no experimental data in the paper to audit; the numbers in the deep read (Earth example) are the reviewer's own and consistent.

How it was received (rejection analysis)

Not a case of blocked funding or publication: posted to arXiv, appeared in JHEP, won the Spinoza Prize [web, via Investigator]. The criticism that stuck is technical (Visser; Wang-Braunstein; neutron coherence), and the author answered some of it in later work. Open: the Volkskrant 2009 announcement, Woit's contemporary critique and the NYT 2010 article were not read by the Investigator (blocked or not reached), so the early press reception is unchecked.

Leads weighed (kept in the record)

Kept in full: Verlinde 2026 IJMPA note; 2016 EG paper; Verlinde-Zurek; Brouwer 2017 and 2021; Chaichian et al.; GQuEST; Visser; Wang-Braunstein; Kobakhidze; Pardo; Nieuwenhuizen; Carney et al.; Lelli et al. (added by me). Not read by me: Strings 2011 slides, Tower Wang 2012, Mäkelä 2010.

Verdict summary

What would settle it

A GQuEST (or equivalent) null or positive result at sensitivity beyond the Holometer limit; a neutron-state decay measurement of the kind Chaichian et al. say would discriminate; and a derivation of the paper's identifications (T, N, 2 pi, equipartition) from an explicit microscopic model.