The exoplanet business is booming, with astronomers rolling out newfound planets outside the solar system seemingly every week. There are now more than 400 candidates, or proposed planets, in the extrasolar planet catalogue, and NASA's Kepler spacecraft, which launched in March, began piling on Monday when the mission's first results were announced. At a meeting of the American Astronomical Society, mission scientists unveiled five hot, massive planets discovered by Kepler, a space telescope that should in the coming years be able to turn up potentially habitable, Earth-like worlds.

But very few exoplanets have actually been directly seen—a planet's presence and properties are usually inferred by its subtle effects on the host star. So it is perhaps not surprising that some discoveries have not entirely proved out, with some exoplanet candidates turning out to be larger objects such as stars or brown dwarfs (celestial bodies larger than planets but too small to fuse hydrogen as stars do) and some seeming to disappear entirely on closer examination.

As the exoplanet population continues to swell, so does the list of disputed or rejected contenders. Alongside the 400-plus viable exoplanet candidates now listed in the online Extrasolar Planets Encyclopaedia sit more than 60 classified as "unconfirmed, controversial or retracted". A number of recent additions to that list of the disputed or retracted underline the uncertainties in exoplanet discoveries as well as the importance of painstaking follow-up research to determine a candidate planet's true properties.

In one of the most recent cases, a planetary prospect straddling the planet–brown dwarf border was found to be solidly in the latter's more massive camp. (Any object of 13 or more Jupiter masses is considered to be a brown dwarf and not a planet, according to the International Astronomical Union.) HD 136118 b, discovered in 2002 by Debra Fischer, then at the University of California, Berkeley, and her colleagues, was initially identified as having a minimum mass of 12 Jupiters. The exoplanet naming convention appends lowercase letters to the host star's name; HD 136118 b orbits a star called HD 136118.

But a paper in the January 1 issue of Astrophysical Journal finds that the true mass is much higher than that: some 42 times that of Jupiter. When HD 136118 b was first discovered, Fischer's team took care to note that it was likely a brown dwarf companion to the star HD 136118, and Fischer, now at Yale University, says she is not surprised by the new result. (And indeed, although brown dwarfs bear little similarity to the cool, habitable Earth-like worlds astronomers soon hope to find, they are interesting in their own right.)

A more dramatic re-identification came in 2007, when the candidate planet HD 33636 b, initially tagged with a minimum mass of 7.7 Jupiter masses, was found to be much larger. A team led by Jacob Bean, then a graduate student at the University of Texas at Austin (U.T.), found that HD 33636 b was not only too large to be a planet but too large to be a brown dwarf, as well. (Bean also co-authored the more recent paper on HD 136118 b.) In fact, it appeared to be a star, some 142 times the mass of Jupiter, in a binary pairing with a larger stellar sibling.

Both of these cases bring to light a soft spot in exoplanet research: The most prolific tactic for finding exoplanets, the radial velocity approach, can only reveal a lower bound to their masses. Radial velocity searches look for Doppler shifts in a star's light induced by the gravitational tug of an orbiting companion. But those shifts only reveal the star's motion along our line of sight, and without knowing what angle the orbit of the planet takes with respect to Earth's line of sight, its true mass cannot be pinned down. The cases above illustrate what happens when the planet turns out to be highly inclined, with orbital planes nearly perpendicular to our line of sight.

"The issue boils down to simple geometry," Fischer says. "We only measure the velocity along the line of sight." Knowing that velocity, two unknowns remain—the planet's mass and its inclination—and neither can be solved for conclusively. "We've never been able to measure inclination with radial velocities…and we never will," Fischer says.

To find the masses of HD 136118 b and HD 33636 b, research teams meshed ground-based radial velocity observations with astrometry, or the precise measurement of the position of stars in the sky, taken from the Hubble Space Telescope. Whereas radial velocity data show if a star is moving nearer and farther, astrometry reveals if a star is moving laterally. "The astrometric measurements detect motion in the plane of the sky, and therefore provide the full 3-D orbit," Fischer says.

Fritz Benedict, a U.T. astronomer who worked with Bean and others on the papers presenting large masses for both HD 136118 b and HD 33636 b, says that astrometry is not used as a cross-check more often for several reasons: One is that few instruments boast the required precision; another is that faraway exoplanets induce planar motions too small to be detected. "As systems get farther from us, the apparent size of the perturbation shrinks," Benedict says. "Bottom line—we run out of nearby stars on which to use this technique."

Contradictory data from astrometry and radial velocity observations were the source of a planetary impasse last month—one group found a planet to be present, whereas another team did not. In May Steven Pravdo and Stuart Shaklan of the NASA Jet Propulsion Laboratory in Pasadena, Calif., announced that they had detected a candidate exoplanet, known as VB 10 b, by ground-based astrometry. (Ground-based observations, which are subject to atmospheric effects, are generally thought to be less trustworthy than space-based observations such as those from Hubble.) But in December Bean, now at the University of Göttingen in Germany, and his colleagues presented radial velocity data, also taken from the ground, that did not reveal the orbiting companion. "With the precision we were getting, it actually would have been very easy to detect this planet," Bean says. "So it was very easy to rule it out."

The new data led Bean and his colleagues to conclude that Pravdo and Shaklan's claim for VB 10 b was "spurious." But Shaklan disagrees, noting that Bean's "work does not entirely rule out the existence of the planet." Without more data, he says, neither of the conflicting results can be confirmed. And Bean notes that a planet with somewhat different parameters than those derived by Pravdo and Shaklan could still exist, even in light of the new research.

All three revisions within the ranks of proposed planets involved cross-checking a possible detection with a different observation method, something the Kepler team used close to 100 nights of ground-based telescope time to do before announcing the spacecraft's first results. But follow-up studies consume time and resources, and in some cases more powerful instruments than are currently available would be necessary to solidly pin down the details of a planetary detection. The European Space Agency's Gaia spacecraft, scheduled for launch in 2012, could fill in some of these details with precision astrometry measurements.

"The application of complementary techniques is really important," Bean says, noting that his willingness to examine new discoveries from different angles is what leads him into the uneasy position of revising or even refuting other scientists' work. "I guess that has sort of evolved into my specialty—using different methods combined," Bean says. "So that's probably how I get into this mess."