The number was wrong. Not dramatically wrong — not the kind of wrong that crashes a probe into a planet or sends a mission spiraling into the void. Just wrong by 3.92 millimeters per second. A whisper of velocity. The kind of discrepancy you'd expect to chalk up to measurement noise and move on. But the engineers at the Jet Propulsion Laboratory who were poring over Doppler telemetry from the Galileo spacecraft in late 1990 couldn't move on. The number kept coming back. Clean. Consistent. Impossible to explain.
Galileo had swung around Earth on December 8, 1990, using the planet's gravity as a slingshot toward Jupiter. Gravitational assists are textbook physics — elegant, well-understood, used dozens of times across the history of spaceflight. You know exactly how much speed you're supposed to gain. JPL knew. And Galileo arrived moving faster than it should have. By exactly 3.92 mm/s. The Deep Space Network's Doppler data showed a 66 mHz frequency shift at perigee that matched no known force, no known error, nothing in the model.
Nobody published anything immediately. They checked the data. Then they checked it again.
---
To understand why this anomaly burrowed so deep into the minds of physicists and aerospace engineers, you have to understand what gravitational assists represent in the culture of deep space exploration. These maneuvers are not approximations. They are among the most precisely calculated events in human engineering — the product of decades of refinement in orbital mechanics, general relativity corrections, and telemetry analysis. The Deep Space Network, the global array of massive radio antennas that tracks every NASA deep-space mission, measures spacecraft velocity to sub-millimeter-per-second precision using Doppler shifts in S-band and X-band signals. When the DSN says a spacecraft is moving at a certain speed, it is not guessing.
By the late 1990s, the internet had given physicists and space enthusiasts a new kind of collective nervous system. Preprint servers, mailing lists, and early forums meant that anomalous data didn't stay locked in institutional corridors. When something strange turned up in spacecraft telemetry, word spread. And among the people paying closest attention — orbital mechanics researchers, amateur astronomers with serious technical backgrounds, the emerging community of people who tracked space missions the way others tracked sports — the flyby anomaly became a slow-burning obsession.
---
The second Galileo flyby, in December 1992, showed no clear anomaly, though drag estimates from atmospheric interactions carried error bars wide enough that a small anomalous acceleration couldn't be definitively ruled out. Researchers filed that uncertainty away. Then, on January 23, 1998, the NEAR spacecraft — the Near Earth Asteroid Rendezvous probe — performed its own Earth flyby on the way to asteroid 433 Eros. The Doppler data came back, and this time the discrepancy wasn't a whisper. NEAR had gained 13.46 mm/s beyond what the models predicted. The largest anomalous velocity increase ever recorded.
That number was impossible to dismiss. Cassini-Huygens flew past Earth in August 1999 and gained 0.11 mm/s. Small, but there. Rosetta, the European Space Agency's comet-chasing probe, swung past Earth in March 2005 and gained 1.82 mm/s. Each new flyby added another data point to a pattern that had no business existing.
Then the pattern started breaking. MESSENGER, NASA's Mercury orbiter, performed an Earth flyby and showed no significant anomalous velocity increase — but MESSENGER's trajectory was nearly symmetric about the equator, approaching and departing at similar angles relative to Earth's equatorial plane. Rosetta's second flyby in November 2009, tracked with extraordinary care specifically to measure the anomaly, showed nothing. Juno's 2013 flyby of Earth on its way to Jupiter showed nothing. The anomaly was appearing and disappearing, and nobody could say why.
---
J. D. Anderson at JPL had been studying the data long enough to notice something. In 2008, he and his colleagues published an empirical equation — not a physical theory, but a mathematical description — correlating the anomalous velocity change with Earth's rotation and the angles at which spacecraft crossed the equatorial plane on their way in and out. Probes that arrived from high latitudes and departed toward opposite latitudes showed anomalies. Probes that crossed the equator symmetrically showed nothing. The pattern was suggestive of something tied to Earth's rotation, to the geometry of approach and departure. But Anderson's equation described the phenomenon without explaining it. It was a map without a territory.
The MESSENGER result fit the equation. The null results from Rosetta's second flyby and Juno fit it too — or at least didn't contradict it. But fitting an empirical formula to a handful of data points is not physics. It's pattern recognition, and pattern recognition without mechanism is where science starts to feel uncomfortable.
Analysis of the Juno flyby added another layer of unease: researchers found that using a low-precision gravity field model — specifically a 10×10 coefficient model rather than a higher-fidelity version — could produce a velocity error of 4.5 mm/s on its own. The implication was uncomfortable. Were some of the earlier anomalies artifacts of imprecise modeling? Or had the modeling improved enough by Juno's time to mask a real effect?
---
Investigations at JPL, Goddard Space Flight Center, and the University of Texas produced no satisfying answer. The anomaly appeared in both Doppler and ranging data, which ruled out certain classes of instrumentation error — you'd need correlated failures across independent measurement systems to fake it. Atmospheric drag, solar radiation pressure, thermal recoil forces from the spacecraft itself — all were examined and found insufficient, either in magnitude or in the wrong direction.
The proposed STE-QUEST satellite mission, which could have been designed specifically to probe the anomaly and separate real physics from measurement artifacts, was not selected for launch in 2014. The community noted the loss. Without a dedicated experiment, researchers were left mining data from missions designed for other purposes, working with error bars that sometimes swallowed the effect they were trying to measure.
---
What investigators confirmed is that the anomaly is real in at least several documented cases, that it appears in independent data streams, and that no conventional physical explanation has survived scrutiny. General relativistic effects, including frame-dragging and gravitomagnetic phenomena, have been investigated and found unable to account for the magnitude or the pattern. What remained contested is whether the null results from later flybys reflect improved modeling, different trajectory geometry, or evidence that the anomaly is intermittent in ways nobody understands.
The speculative territory is wide. A dark-matter halo around Earth has been proposed. The Céspedes-Curé hypothesis invokes variations in the speed of light through regions of variable gravitational energy density. A topological torsion current has been suggested, predicting anomalies only for retrograde spacecraft trajectories. None of these have moved from speculation to testable prediction to confirmation. The community came to believe, with varying degrees of conviction, that the equatorial geometry pattern in Anderson's 2008 equation pointed toward something real about Earth's rotation — but what that something might be remains unanswered.
When ʻOumuamua, the first confirmed interstellar object to pass through the solar system, was found in 2018 to be decelerating more slowly than solar gravity predicted, some researchers drew comparisons to the flyby anomaly. The parallel was noted carefully, not claimed. ʻOumuamua's excess velocity was attributed to outgassing, though no outgassing had been detected at the time.
---
The flyby anomaly remains open. No mission has been dedicated to resolving it. The data from the 1990s and early 2000s sits in archives, analyzed and re-analyzed, yielding the same stubborn residuals. The probes that showed the anomaly are long past Earth now — Galileo deliberately crashed into Jupiter in 2003, Rosetta descended to comet 67P in 2016. Their telemetry records are all that remain.
The question Anderson's equation leaves hanging is the one that refuses to close: if the anomaly correlates with how a spacecraft crosses Earth's equatorial plane, then something about the geometry of Earth's rotation is doing something to the fabric of space, or to our measurement of it, or to physics itself in ways we haven't accounted for. Three millimeters per second. Thirteen millimeters per second. Numbers so small they barely register against the scale of interplanetary space — and yet, after three decades, still without an explanation.
SOURCES