The Speed of Gravity: Revisiting Tom Van Flandern’s Challenge to Einstein
Sunlight reaches Earth more than eight minutes after it leaves the Sun. Why doesn’t gravity’s travel time produce a comparable lag in the motion of the planets? In 1998, astronomer Tom Van Flandern turned that apparent contradiction into a provocative claim: gravitational influences must travel at least 20 billion times faster than light. His argument opened a dispute about what stable orbits reveal, how fields transmit information, and what an experiment would actually have to measure to distinguish almost instantaneous attraction from light-speed causation.
An Astronomer’s Eight-Minute Problem
At Earth’s average distance from the Sun, light crosses approximately 149.6 million kilometers in 499 seconds. During that interval, our planet travels almost 15,000 kilometers along its orbit at roughly 29.8 kilometers per second. These numbers help explain the appeal of Van Flandern’s question. If an interaction takes time to travel, the bodies involved move before it arrives. It seems reasonable to expect a gravitational pull to become slightly misaligned, changing an orbit a little on every revolution. Over thousands of years, even a small systematic error could become enormous.
There’s a subtlety in that picture. Earth’s motion while sunlight is in transit isn’t the same as the Sun moving 15,000 kilometers away from an old position. Apparent directions depend on the observer’s motion and the chosen reference frame. Astronomers already account for such effects when converting observations into positions and velocities. Van Flandern’s concern was whether the gravitational interaction should display an analogous aberration: a deflection in its effective direction caused by finite propagation speed. The absence of a large cumulative torque was the observational fact he wanted explained.
Van Flandern knew celestial mechanics professionally. He earned his doctorate in astronomy at Yale in 1969 and worked at the U.S. Naval Observatory from 1963 to 1983, contributing to lunar observations, orbit calculations, and practical astronomical methods. His paper, “The Speed of Gravity—What the Experiments Say,” appeared in Physics Letters A on December 21, 1998. Its eleven journal pages developed a challenge to the usual interpretation of gravity, arguing that the apparent immediacy of orbital interactions exposed a physical problem beneath the successful mathematics. The argument belonged to a broader mechanical model he’d already described in a 1996 paper.
When I interviewed Van Flandern for American Antigravity in March 2005, he described six lines of evidence that he believed ruled out light-speed gravitational force. He also emphasized a distinction that’s often lost when his argument is retold: gravitational waves, he said, could still travel at light speed. Nearly two decades later, Rupert Sheldrake and Mark Vernon returned to the argument in a January 2024 dialogue, showing how readily the orbital puzzle survives outside specialist discussions. Understanding it requires following both the original inference and the physical explanation offered by Van Flandern’s critics.
What Twenty Billion Times Faster Actually Meant
Van Flandern assumed that finite-speed gravity would acquire an aberration analogous to the effect of sunlight on moving matter. Under his prescription, a light-speed interaction would include a tangential attraction about one ten-thousandth as strong as the radial attraction, steadily increasing Earth’s orbital angular momentum. He calculated that Earth’s orbital radius would double in about 1,200 years. That’s spectacularly incompatible with observed planetary motion. The result depends on his assumed response to the target’s motion; merely retarding the position of a stationary Sun wouldn’t produce that angle. General relativity doesn’t predict the expansion, and the adequacy of his interaction law became the central issue in the dispute.
Binary pulsars supplied his strongest numerical argument. PSR B1913+16 and PSR B1534+12 complete their orbits in approximately 7.75 and 10.10 hours, respectively. Their radio pulses allow astronomers to reconstruct orbital motion and measure exceptionally small changes in orbital periods. Van Flandern used those observations to constrain the unwanted orbital evolution produced by his assumed propagation behavior. The headline lower limit of 20 billion times light speed came from B1534+12. Nobody had timed a gravitational pulse crossing the binary. The number expressed how fast the interaction would need to be, under his assumptions, to avoid contradicting the timing data.
That qualification matters because his numerical conclusions changed sharply with his assumptions about extrapolation and field regeneration. The headline limit depended on treating momentum transfer as necessarily producing the assumed aberration. In a response dated August 2000, he enumerated six supposed tests, four of them astronomical: Earth’s orbital angular momentum, binary pulsars, eclipse geometry, and radar-ranging interpretations of the direction of gravitational acceleration. Several therefore shared the same inferential foundation. They weren’t independent stopwatches all recording the same extraordinary transit speed.
The remaining categories involved neutron interferometry and work on oscillating near fields. Neutron interferometers detect the phase difference acquired by matter waves taking different paths through a gravitational field. A mass-dependent phase doesn’t, by itself, show that different materials fall with different accelerations or reveal how fast gravity propagates. Walker–Dual work examined near-field phase behavior and proposed extraordinarily sensitive measurements. Here another distinction becomes essential: the motion of a phase pattern needn’t equal the speed at which new information arrives. Establishing superluminal communication would require a detectable new influence to reach an observer before light could cover the same separation.
The Physical Model Behind the Claim
Van Flandern wanted a mechanism that made gravitational attraction intelligible as a physical interaction. In his April 30, 1999 presentation to the First International Conference on Future Energy in Bethesda, Maryland, he described a version of Le Sage’s pushing-gravity model. Space was filled with an approximately isotropic flux of extremely small, fast-moving particles. Ordinary matter absorbed a tiny fraction of that flux or scattered it with a loss of energy. Two bodies partially shadowed one another, leaving less incident momentum from the direction between them than from the surrounding directions. The imbalance pushed them together and appeared macroscopically as attraction.
His classical carriers, which he called C-gravitons, weren’t the spin-two quanta invoked in quantum-gravity discussions. Their proposed speed and interactions belonged to his own model. To reproduce familiar gravity, the flux had to penetrate matter almost completely while transferring enough momentum to generate the observed attraction. It also had to make the interaction scale appropriately with mass and distance. These demands pull against one another. Stronger absorption can make shadowing easier to produce, but it also raises questions about heating, shielding, and composition dependence. Extremely weak interactions avoid some of those effects while requiring a correspondingly substantial incident flux.
He therefore distinguished the fast carriers from a separate light-carrying medium. In his account, gravitational force shaped this slower medium, rather as gravity establishes a density gradient in an atmosphere. Light and gravitational waves were associated with disturbances traveling through it, while the much faster carrier flux supplied the underlying attraction. He sought to account for light bending, gravitational redshift, radar delay, and perihelion advance through this architecture. The construction explains why he didn’t regard a light-speed gravitational wave as a contradiction: he had assigned force transmission and wave propagation to different physical processes.
Van Flandern coupled this picture to Lorentzian relativity, with a physically preferred frame and a preferred ordering of events. Such additional structure can change the causal implications of superluminal propagation; the familiar special-relativistic construction of signals returning to their own past assumes different symmetry conditions. But adding a preferred frame creates further obligations. Its effects must agree with precision observations or remain undetectable for a specified physical reason. A viable alternative must explain how clocks, matter, and both proposed media interact, while recovering the many measurements that ordinary relativity already describes. The mechanical picture makes those questions concrete without answering all of them.
How a Causal Field Can Look Almost Instantaneous
The first difficulty with the delayed-force argument is that a moving source’s gravitational field contains more information than a snapshot of its former location. A relativistic interaction includes motion-dependent contributions as well as the part that resembles Newtonian attraction. Introducing a time delay into the Newtonian term while leaving the rest unchanged creates a hybrid law whose behavior needn’t resemble the complete theory. The resulting orbital instability is real for that hybrid. Applying its enormous speed requirement to general relativity requires demonstrating that the terms omitted from the calculation can’t remove the instability.
Electromagnetism offers an instructive comparison. The electric field of a uniformly moving charge lies along the line joining the observation point to the charge’s present position, even though changes in the field propagate at light speed. The field’s retarded structure contains information about both the charge’s earlier position and its velocity. Those contributions combine to give a direction that looks immediate. Michael Ibison, Harold Puthoff, and Scott Little developed this comparison in their 1999 preprint, “The Speed of Gravity Revisited,” before examining a gravitational counterpart. The example shows why force direction alone is an unreliable clock: a causal field can contain enough information about established motion to produce an apparently instantaneous alignment.
Steven Carlip’s 2000 paper, “Aberration and the Speed of Gravity,” addressed an accelerating source using an exact solution in which anisotropic radiation accelerates a mass. He agreed that the naive delayed-Newtonian prescription would demand an enormous propagation speed, then showed how motion-dependent interactions in general relativity nearly cancel the offending aberration. In the weak-field coordinates used for the comparison, the response can be expressed using the source’s retarded position, velocity, and acceleration. Every input comes from its past. If its motion changes unexpectedly, the distant field doesn’t immediately acquire that information. The apparent extrapolation therefore doesn’t amount to precognition, and the cancellation follows from the field dynamics rather than an independent instruction to make gravity instantaneous.
This cancellation doesn’t erase relativistic corrections or make every force exactly radial. Conservative effects, including orbital precession, remain. For slowly moving binaries, the leading radiation-reaction acceleration appears at 2.5 post-Newtonian order: it is suppressed relative to the Newtonian acceleration by five powers of the ratio of orbital speed to light speed. Gravitational radiation can carry away energy and angular momentum, producing a gradual orbital decay. General relativity therefore accommodates both the absence of catastrophic planetary expansion and the measurable shrinking of compact binaries. Those observations test different terms within a consistent dynamical description.
A Published Dispute About Causality
The argument generated a published exchange. Gerald Marsh and Charles Nissim-Sabat criticized Van Flandern’s interpretation in Physics Letters A in 1999, and he replied in the same volume. Carlip’s paper followed in 2000, while the Ibison–Puthoff–Little response circulated as a preprint. The range of participants complicates a simple division between establishment physicists and alternative researchers. Puthoff, familiar to many readers through unconventional propulsion and vacuum-physics research, coauthored an analysis concluding that the observations didn’t require superluminal gravity. The disagreement concerned the interaction law and its interpretation, not merely whether an unusual idea deserved consideration.
Van Flandern and Jean-Pierre Vigier developed the challenge in their 2002 Foundations of Physics paper, “Experimental Repeal of the Speed Limit for Gravitational, Electrodynamic, and Quantum Field Interactions.” Its abstract argued that a field interpretation allowing faster-than-light propagation could preserve causality where a geometric interpretation allegedly failed. Van Flandern’s earlier, publicly available response had also disputed the distinction critics drew between force direction and propagation speed. He maintained that they confused gravitational force with gravitational waves. The disagreement concerned physical causation: he wanted a mechanical account of momentum transfer, while his critics argued that the complete causal field already supplied it.
In his response to Carlip, Van Flandern raised an objection involving tidal friction. If relativistic terms canceled the apparent lag responsible for an orbital torque, he argued, why wouldn’t they also cancel real tidal effects? The answer depends on identifying where energy and angular momentum go. A deformable body can dissipate energy internally and transfer angular momentum between rotation and orbital motion. Gravitational waves provide another route for energy and angular momentum to leave an orbit. These processes differ from the spurious torque introduced by a delayed central-force prescription. Conservation of the complete system doesn’t demand that every individual orbit or rotating body keep its own angular momentum unchanged.
His black-hole objection was similarly concerned with physical causation. If nothing can escape an event horizon, how could two black holes continually update the gravitational fields that govern their mutual motion? In general relativity, the exterior field is already part of the physical system and evolves locally according to the field equations. It doesn’t have to be continually replenished by messages emerging from matter inside the horizon. This also exposes the limits of the familiar rubber-sheet illustration. Curved spacetime isn’t a passive surface requiring an external downward pull; its geometry has dynamical equations, constraints, and propagating disturbances.
LAGEOS and the Engineering of Tiny Forces
Van Flandern’s model also suggested an observable departure from ordinary gravity: sufficiently dense intervening matter might partially shield an attraction. He saw a possible example in LAGEOS, a satellite designed to be unusually insensitive to environmental disturbances. LAGEOS-1 is a passive sphere about 60 centimeters across and 407 kilograms in mass, with 426 corner-cube reflectors embedded in its surface. Of those reflectors, 422 are fused silica and four are germanium. Orbiting roughly 5,900 kilometers above Earth every 225 minutes, it returns laser pulses to ground stations so that researchers can reconstruct its trajectory with exceptional precision.
Precision tracking revealed average along-track deceleration on the scale of a few trillionths of a meter per second squared, together with eclipse-season variations. Van Flandern proposed that Earth might partly screen the Sun’s gravity, illustrating the effect with a shielding coefficient of 2 × 10⁻¹⁸ square centimeters per gram. He acknowledged that the data allowed shielding without requiring it. But lunar laser ranging had already placed a 1990 upper bound of 1 × 10⁻²¹ square centimeters per gram on a universal shielding coefficient, 2,000 times below his choice. That comparison assumes the same attenuation by intervening mass, including matter inside each body, with unchanged inertial response and no compensating effects. His illustration therefore needed to satisfy existing Earth–Moon constraints as well as explain satellite residuals.
Thermal explanations were already in the literature. A satellite absorbs energy from sunlight and Earth’s infrared radiation, then emits heat unevenly and with a delay. The emitted photons carry momentum, generating a small recoil. Spin orientation, surface reflectivity, and charged or neutral particles further complicate its motion. A 1991 analysis by R. Scharroo and colleagues modeled solar thermal effects and asymmetric reflection in addition to other drag contributions, reducing the standard deviation of residual accelerations by 61%. Their result depended on physical and fitted assumptions, including satellite spin behavior. It nevertheless showed that ordinary mechanisms could account for substantial portions of the supposedly anomalous signal.
Later work made the engineering more explicit. In 2006, J. I. Andrés and colleagues modeled each LAGEOS satellite with 2,133 thermal elements, including the reflector assemblies, shells, and core. Inputs included eclipse-modulated sunlight, Earth’s reflected sunlight, and terrestrial infrared radiation. Simulated component temperatures showed typical variations of roughly 30–100 kelvin, while instantaneous thermal accelerations reached the order of 70 picometers per second squared. These were simulation results; applying them to refined orbit calculations was left for follow-up work. The changing forces aren’t equivalent to the much smaller long-term average drag, and their magnitude shows why an orbital residual requires detailed modeling before it can identify gravitational shielding.
Heat and Drag in a Mechanical Universe
A flux capable of pushing planets must do more than produce the correct attraction. Its particles exchange momentum with matter, and that exchange can deposit energy or resist motion through the background. An isotropic flux may exert no net push on an isolated stationary object while still delivering energy from every direction. Producing a small imbalance between two bodies can therefore involve a much larger underlying energy traffic. The relevant questions concern the incident flux, particle energies, interaction cross sections, scattering directions, and the fate of absorbed energy. Those quantities must agree with both thermal measurements and orbital dynamics.
Van Flandern addressed these difficulties in his “Meta Cycle” discussion, considering separate absorption and scattering behavior. He proposed that matter absorbed carrier energy and returned energy through radiation, with photons eventually transferring energy back to the carrier medium during their passage through space. He used a historical Earth heat-flow estimate equivalent to about 32 terawatts and an upper bound on Earth-orbit drag of approximately 7.5 × 10⁻¹⁴ meters per second squared. These numbers constrained his proposed mechanism; they didn’t detect it. A successful energy budget would also have to establish the proposed return process and its consistency with observed radiation.
The model’s finite-range claim added another layer. Van Flandern suggested a characteristic range of about two kiloparsecs, or roughly 6,500 light-years. Collisions among carriers would progressively refill the shadow between widely separated bodies, weakening their direct interaction. His proposed explanation of galactic rotation then relied on attraction among neighboring stars and an assumed spatial distribution, producing an effective large-scale behavior different from the ordinary inverse-square calculation. Thus he wasn’t simply proposing stronger pairwise gravity beyond a certain distance. The model needed to reproduce collective galactic dynamics and light deflection using its specified interactions.
These mechanisms motivated the technological possibilities discussed at COFE: gravitational sails, usable energy from the background, and communication through a superluminal interaction. But a carrier’s proposed speed doesn’t determine whether engineers can manipulate it. A sail would need a material with a measured response and a reproducible momentum transfer. An energy device would need a demonstrated power balance, including heat rejected and energy consumed. A communication system would need controllable modulation and a detector that recovered new information. The conference presentation offered implications of a model; translating them into hardware required physical properties that its astronomical arguments hadn’t established.
Jupiter and the Meaning of a Measurement
On September 8, 2002, Jupiter passed approximately 3.7 arcminutes from the line of sight to the quasar J0842+1835. Ed Fomalont and Sergei Kopeikin used the Very Long Baseline Array and the Effelsberg telescope at 8.4 gigahertz to measure the deflection of the quasar’s radio signal. Their predicted radial deflection was about 1,190 microarcseconds. A much smaller component associated with Jupiter’s motion was approximately 51 microarcseconds. By combining widely separated radio telescopes, the experiment could investigate an angular effect far below the resolving power of ordinary optical observation.
Using two calibration quasars, the investigators reported a root-mean-square positional error of no more than 10 microarcseconds. They fitted the smaller component at 0.98 times its predicted amplitude, with a one-standard-deviation uncertainty of 0.19, and interpreted the result as a test related to the propagation speed of gravity. Their analysis reported a subtle effect consistent with their calculation. The ensuing dispute concerned both the identification of that component and what the agreement established. A radio-wave deflection caused by a moving gravitating body includes light propagation, gravitational geometry, and velocity-dependent corrections. Extracting a separate gravitational speed requires showing exactly where it enters the observable.
Clifford Will examined that dependence in a 2003 Astrophysical Journal paper. Allowing gravitational propagation speed to differ from light’s, he found that the measured correction wasn’t directly sensitive to that speed at first order in Jupiter’s velocity relative to light speed. Direct dependence entered at higher orders, while other possible changes involved theoretical parameters associated with preferred-frame effects. Hideki Asada derived the same leading correction while attributing it to light propagation. Stuart Samuel went further, arguing that the apparent enhancement arose from the angular variable used in the analysis and that the relevant physical velocity corrections were too small for the experiment to detect. Their arguments differed, but both challenged the claim that this observation isolated gravity’s propagation speed.
The Jupiter episode illustrates a difficulty that also affects Van Flandern’s argument. An instrument measures a position, phase, arrival time, or acceleration; a physical model connects that quantity to an inferred parameter. Agreement with a prediction doesn’t automatically establish every interpretation attached to it. Equally, criticism of the interpretation doesn’t erase the underlying observation. Van Flandern’s skepticism toward the Jupiter announcement overlapped with objections from mainstream relativists, but their agreement on that limited point didn’t establish his faster-than-light carriers. They reached it through different accounts of the same measurement.
What Gravitational Waves and Pulsars Add
The observational situation changed dramatically on August 17, 2017. The neutron-star merger GW170817 produced gravitational waves and the gamma-ray burst GRB 170817A at a distance of approximately 40 megaparsecs, or 130 million light-years. Gamma rays arrived 1.74 seconds after the merger time inferred from the gravitational waves, with a timing uncertainty of about 0.05 second. In the collaboration’s analysis, the relative speed difference was constrained to between three parts in a quadrillion slower than light and seven parts in ten quadrillion faster. A journey lasting millions of years had produced a relative arrival difference measured in seconds.
Those limits included assumptions about emission timing. The analysis used a conservative lower distance of 26 megaparsecs and assumed that the gamma rays were emitted between zero and ten seconds after the gravitational-wave peak. That interval was an astrophysical assumption, not a measured emission lag. The result tightly constrained the observed radiative propagation. Van Flandern, however, had already assigned gravitational waves a light-speed role in his separate medium. The observation alone therefore doesn’t dispose of every hypothetical additional force channel. Such a channel would still need a defined coupling to matter and predictions that could be compared with other gravitational observations.
Binary timing tests the dynamical account from another direction. In 2021, Michael Kramer and colleagues reported 16.2 years of observations of the double pulsar PSR J0737−3039A/B, whose two neutron stars orbit every 2.45 hours. Their study measured six post-Keplerian timing parameters, with an earlier eclipse-based measurement of spin precession bringing the system’s total to seven relativistic observables of this kind. The predicted gravitational-radiation damping agreed with observations to 0.013%, at 95% confidence. The radiation-associated orbital shortening is approximately 39 microseconds per year. Extracting that result required accounting for other contributions, including external accelerations and a pulsar’s mass loss. Its precision concerns the predicted damping, rather than a direct gravitational transit time.
An alternative interaction must consequently explain more than the absence of an obvious lag. It must reproduce orbital precession, the timing of radio signals through a companion’s gravitational field, clock effects, and the measured loss of orbital energy. It must do so without introducing excluded drag, additional radiation, or composition-dependent accelerations. Van Flandern’s separation of force and waves makes the 2017 comparison less decisive against his particular architecture than a simple headline might suggest. The combined dynamical evidence creates a broader challenge: demonstrate that the proposed mechanisms reproduce the same connected set of quantitative observations.
The Experiment That Would Move the Debate Forward
Laboratory gravity measurements show how far sensitivity has advanced. In a 2021 Nature paper, Tobias Westphal and colleagues measured gravitational coupling between gold spheres of approximately 90 milligrams and one-millimeter radius. They periodically moved the source at 12.7 millihertz and accumulated 350 hours of data with a miniature torsion balance. The sphere surfaces approached to about 400 micrometers. Across the combined measurements, the statistical acceleration uncertainty was approximately three trillionths of a meter per second squared, while the systematic uncertainty was about ten times larger. Even tiny laboratory masses can produce a detectable gravitational response when disturbances and detector behavior are carefully controlled.
Measuring propagation presents a different problem. Light takes approximately 3.3 nanoseconds to cross one meter, while a one-hertz disturbance traveling at light speed has a wavelength of about 300,000 kilometers. A tabletop apparatus operating at that frequency sits deep in its near field. Its oscillating response can’t be interpreted simply as a delayed copy of the source’s position. The source motion, the gravitational interaction, mechanical coupling, and the detector’s own response all affect phase. Greater force sensitivity helps, but it doesn’t automatically turn a slow torsion balance into a nanosecond propagation clock.
The useful next step is a prediction that distinguishes the models before data are collected. For a transient experiment, each theory must specify the observable waveform, amplitude, and distance dependence produced by a controlled source change. A previously unpredictable modulation could test information transfer, provided the complete apparatus and ordinary signal paths were modeled. A shielding proposal must first satisfy the lunar-ranging constraints applicable to its interaction law, then predict an additional detectable effect for specified materials and geometry. Electrostatic forces, vibration, thermal effects, and the ordinary gravity of the shield itself would all require control.
Van Flandern’s question remains useful because it exposes how misleading a simple picture of delayed attraction can be. Stable orbits rule out that picture at light speed, while relativistic field dynamics explain how orbital forces can look nearly immediate without transmitting new information instantaneously. His broader particle model supplies additional claims, some with numerical parameters, that must be assessed on their own predictions. The observation that would change this debate is therefore specific: a reproducible gravitational response that the complete relativistic calculation fails to explain and that an independently specified alternative predicts. That’s the point at which an appealing mechanism earns empirical support beyond what the existing theory can provide.
References
Tom Van Flandern’s Papers and Gravity Model
- The Speed of Gravity—What the Experiments Say — Tom Van Flandern. Physics Letters A, 250, 1–11 (1998).
- Complete Gravity Model and Free Energy — Tom Van Flandern. First International Conference on Future Energy, April 1999. Conference abstract.
- Possible New Properties of Gravity — Tom Van Flandern. Astrophysics and Space Science, 244, 249–261 (1996). Author-hosted reprint.
- The Meta Cycle — Tom Van Flandern. Meta Research discussion of gravitational carriers, heating, drag, and energy transfer.
- Experimental Repeal of the Speed Limit for Gravitational, Electrodynamic, and Quantum Field Interactions — Tom Van Flandern and Jean-Pierre Vigier. Foundations of Physics, 32, 1031–1068 (2002).
Critiques and Responses
- Comment on “The Speed of Gravity” — Gerald E. Marsh and Charles Nissim-Sabat. Physics Letters A, 262, 257–260 (1999).
- Reply to: “Comment on The Speed of Gravity” — Tom Van Flandern. Physics Letters A, 262, 261–263 (1999).
- Aberration and the Speed of Gravity — Steven Carlip. Physics Letters A, 267, 81–87 (2000).
- The Speed of Gravity Revisited — Michael Ibison, Harold E. Puthoff, and Scott R. Little. Preprint (1999).
LAGEOS and Thermal Recoil
- LAGEOS — NASA Goddard Space Flight Center. Mission overview and satellite specifications.
- On the Along-Track Acceleration of the LAGEOS Satellite — R. Scharroo, K. F. Wakker, B. A. C. Ambrosius, and R. Noomen. Journal of Geophysical Research, 96, 729–740 (1991).
- Numerical Simulation of the LAGEOS Thermal Behavior and Thermal Accelerations — J. I. Andrés, R. Noomen, and S. Vecellio None. Journal of Geophysical Research, 111 (2006).
The Jupiter Experiment and Its Interpretation
- The Measurement of the Light Deflection from Jupiter: Experimental Results — E. B. Fomalont and S. M. Kopeikin. The Astrophysical Journal, 598, 704–711 (2003).
- Propagation Speed of Gravity and the Relativistic Time Delay — Clifford M. Will. The Astrophysical Journal, 590, 683–690 (2003).
- The Light-cone Effect on the Shapiro Time Delay — Hideki Asada. The Astrophysical Journal, 574, L69 (2002).
- On the Speed of Gravity and the v/c Corrections to the Shapiro Time Delay — Stuart Samuel. Physical Review Letters, 90, 231101 (2003).
Gravitational Waves and Binary Pulsars
- Gravitational Waves and Gamma-rays from a Binary Neutron Star Merger: GW170817 and GRB 170817A — LIGO Scientific Collaboration, Virgo Collaboration, Fermi Gamma-Ray Burst Monitor, and INTEGRAL. The Astrophysical Journal Letters, 848, L13 (2017).
- Strong-Field Gravity Tests with the Double Pulsar — Michael Kramer and colleagues. Physical Review X, 11, 041050 (2021).
Laboratory Gravity and Near-Field Studies
- Propagation Speed of Longitudinally Oscillating Gravitational and Electrical Fields — William D. Walker and Jürg Dual. Preprint (1997).
- Classical Underpinnings of Gravitationally Induced Quantum Interference — Philip D. Mannheim. Preprint (1996).
- Measurement of Gravitational Coupling between Millimeter-Sized Masses — Tobias Westphal, Hans Hepach, Jeremias Pfaff, and Markus Aspelmeyer. Nature, 591, 225–228 (2021).
Interviews and Discussion
- The Speed of Gravity: A Conversation with Rupert Sheldrake — Rupert Sheldrake and Mark Vernon. Sheldrake–Vernon Dialogue 84, January 26, 2024.

