A radial velocity measurement gives you a planet mass only if you already know the orbit's tilt. Astrometry measures that tilt directly but pays for it in calibration noise. When the two methods disagree, the gap usually traces back to procedural choices rather than a single faulty dataset. Catalog masses inherit those choices, so reading them carefully means knowing which ruler produced them.

Two Rulers, One Planet

Radial velocity watches a star's spectrum slide back and forth as an unseen companion tugs it along the line of sight. The size of that Doppler shift sets a lower bound on the companion's mass. Astrometry instead tracks the star's position on the sky, looking for a tiny wobble traced out over months or years. Both methods infer a planet nobody has photographed.

The two measurements constrain different quantities. Radial velocity is sensitive to mass times the sine of orbital inclination, which is why the result is a minimum. Astrometry is sensitive to mass directly, because the wobble's amplitude scales with the companion's pull regardless of viewing angle. In principle the pair should agree. In practice they often do not.

Disagreements of a factor of two are common; larger gaps appear for faint or crowded stars. Each method also has a different failure mode. Radial velocity gets confused by anything that shifts spectral lines, including the star's own surface. Astrometry gets confused by anything that shifts a centroid, including a nearby companion star.

The precision requirements are brutal. A Jupiter-mass planet at one astronomical unit from a solar-mass star induces a stellar reflex motion of roughly a few milliarcseconds. For a star at ten parsecs, that is a wobble of a few microarcseconds on the sky. The European Space Agency's Gaia mission was designed to reach microarcsecond astrometry for bright stars, which is why it was expected to open a new channel for planet detection.

To make the scale concrete, consider the nearest stars. Proxima Centauri, at 1.3 parsecs, is the closest. A Jupiter-mass planet at one astronomical unit would induce a wobble of about 3 microarcseconds. For a star at 10 parsecs, that drops to roughly 0.3 microarcseconds. Gaia's expected end-of-mission accuracy for the brightest stars is in the tens of microarcseconds, so even a Jupiter at one astronomical unit around a nearby star is a challenging signal. For a planet at Jupiter's actual distance of 5 astronomical units, the wobble is five times larger, but the orbital period stretches to over a decade, demanding a long baseline.

The Astrometry Mass Controversy

The Hipparcos catalog, collected by a European Space Agency satellite in the 1990s, produced astrometric wobble candidates around bright stars. Some implied companions in the brown dwarf or giant planet range. When radial velocity surveys followed up, the companions often turned out lighter, or the signal vanished entirely.

The HD 114762 system became a useful case study. Its companion was originally reported with a mass near the brown dwarf boundary, then revised downward as more data accumulated. The revisions were not a single correction but a slow accretion of better calibration, longer baselines, and improved models of the star itself.

Statisticians spent years arguing about whether Hipparcos errors were underestimated. The catalog's quoted uncertainties came from a pipeline that had to solve for the whole sky at once, and correlated errors between nearby stars could inflate a wobble that was not there. That debate shaped how later astrometric missions reported their own error bars.

The dispute left a methodological mark. Astrometric planet claims now routinely come with independent checks: radial velocity follow-up, transit searches, or detailed modeling of the star's own motion. The field learned to treat a single-method astrometric detection as provisional until another technique weighs in.

Radial Velocity's Hidden Assumptions

The biggest assumption is that the star is quiet. A rotating star with dark spots carries those spots across the visible disk, shifting line profiles in ways that mimic a small orbiting companion. Magnetic cycles produce long-term trends that look like a distant giant planet on a decades-long orbit.

Without a transit, inclination stays unknown, so the reported figure is a minimum mass. A companion whose true mass is ten Jupiter masses could be reported as two, if the orbit happens to be tilted steeply away from the line of sight. The word minimum does a lot of work in those tables.

Activity indicators help. Observers track emission in calcium lines, or the shape of the cross-correlation function, to flag stellar contamination. The indicators are imperfect. A related piece on this site about how spectrograph throughput sets survey costs makes the point that signal-to-noise budgets constrain how aggressively you can filter.

There is also a selection effect worth naming. Radial velocity surveys find the easiest signals first: massive planets on short-period orbits around bright, quiet stars. The minimum-mass problem is mild for those systems, because a hot Jupiter on a nearly edge-on orbit has a sine of inclination close to one. The trouble grows for long-period planets, where the inclination is unconstrained and the sine can be small. That is exactly the population where astrometry should be most valuable, and where its calibration problems bite hardest.

Stellar activity adds another layer. A star's rotation period, spot coverage, and magnetic cycle can all imprint periodic signals that masquerade as planets. Distinguishing a genuine planet from an activity-induced signal often requires long-term monitoring and multiple activity indicators. Even then, some cases remain ambiguous, and the reported mass carries an asterisk that the table may not show.

Gaia's Astrometric Promise

Gaia was built to measure positions for a billion stars at microarcsecond precision for the brightest ones. That precision should, in theory, reveal the wobble of a star orbited by a giant planet. The mission has released successive catalogs, each with revised astrometry and revised uncertainties.

Early releases produced few unambiguous planet detections from astrometry alone. The reasons are procedural. Bright stars saturate the detectors, crowding in dense fields biases the centroid, and the scanning law means a given star is observed at particular angles rather than uniformly. Each of those effects must be modeled before a wobble can be trusted.

Gaia's own error model has been revised repeatedly, and the revisions sometimes moved candidate masses by large factors. The mission remains the best astrometric dataset available, but extracting a planet mass from it requires assumptions about calibration that are not shared by everyone in the field.

Independent analyses have produced candidate giant planets around nearby stars using Gaia astrometry, with masses in the range of a few to several Jupiter masses. Those claims have been contested, and the debate has centered on whether the astrometric residuals are truly planetary or artifacts of the star's own motion or of the catalog's correlated errors. The pattern is familiar: a promising signal, a follow-up campaign, and a slow convergence toward a number that both camps can live with.

The scale of the challenge is worth restating. For a Sun-like star at 10 parsecs, a Jupiter-mass planet at 1 astronomical unit produces an astrometric wobble of about 0.3 microarcseconds. Gaia's per-observation precision for bright stars is in the tens of microarcseconds, and the end-of-mission combined precision improves with the number of observations, reaching perhaps a few microarcseconds. Even then, the signal is at the edge of detectability for all but the most favorable systems. That is why astrometric planet detection from Gaia remains a careful, assumption-laden exercise rather than a routine catalog product.

Cross-Checking the Catalog

Transits break the inclination degeneracy. When a planet crosses its star's disk, the orbit is edge-on by definition, so the radial velocity minimum mass becomes the true mass. Systems with both a transit and a radial velocity curve are the anchor points that calibrate everything else.

55 Cancri e is a useful example. Its mass has been refined over the years as transit data, radial velocity data, and improved stellar parameters were combined. The refinements were modest but consequential, because the planet sits near the boundary between rocky and volatile-rich compositions.

Community infrastructure matters here. The NASA Exoplanet Archive aggregates masses, uncertainties, and the method that produced each value, and it flags which entries are minimum masses. A related piece on this site about how replication studies stall where grant cycles end describes a similar dynamic: the follow-up work that resolves a discrepancy rarely fits a funding category.

The archive also records the reference for each mass, which lets a careful reader trace a number back to its source. That provenance is not just administrative. It tells you whether the mass came from a single radial velocity curve, a transit plus radial velocity combination, or an astrometric orbit. Each method carries its own systematic uncertainties, and the archive's flags are the first place to look for them.

What a Careful Reader Should Do

Check whether the quoted mass is a minimum or a true mass. A table entry that says m sin i is telling you the orbit's tilt was never measured, and the true value could be substantially larger.

Look for an independent transit confirmation before treating a mass as settled. A single-method detection carries the assumptions of that method; a transit plus radial velocity pair carries far fewer.

Note the instrument and its era. A mass from a 1990s astrometric catalog and a mass from a modern high-resolution spectrograph are not the same kind of number, even when the digits look similar.

Prefer catalogs that publish uncertainty flags and method provenance, and follow updates as Gaia data mature. The archive entries will change. A mass you read today may be revised within a few years, and the revision will usually be a correction rather than a discovery.