ANCHORAGE—In the menagerie of known extrasolar planets, there are hot Jupiters, super-Earths, exo-Neptunes. The terminology astronomers apply to their distant finds rests heavily on the few analogue planets in our own solar system. And they have good reason to believe that planets elsewhere follow the general trends that define our solar system—massive planets, for instance, tend to have large diameters and low densities, just as Jupiter does. But as astronomers have begun to locate worlds without a local analogue, they are often left to guess what those planets might look like.

In particular, researchers have very little idea what kinds of worlds fill the gap between Earth-size and Neptune-size bodies. Every planet in the solar system smaller than Earth—Mercury, Venus, Mars—is rocky, with a solid surface. Every larger planet, from Neptune on up, is a gassy world without a well-defined surface. But quite a bit of variety could occur in between: Neptune is four times the diameter of Earth, and 17 times as massive as our planet. If the development of life proceeds more smoothly on solid worlds such as our own, the question of where the size cutoff for terrestrial planets falls becomes a huge one.

Perhaps optimistically, astronomers have taken to calling any exoplanet a few times the diameter of our home planet a "super-Earth." But Geoff Marcy of the University of California, Berkeley, a leader in the exoplanet field since it began in earnest in the 1990s, thinks that term may not be entirely appropriate. At a semiannual meeting of the American Astronomical Society last week, he outlined some preliminary research indicating that planets can get pretty close to Earth size before they become fully rocky.

Marcy and his colleagues have been observing three stars that NASA's Kepler spacecraft has indicated may host exoplanets intermediate in diameter between Earth and Neptune. Kepler monitors more than 150,000 stars and records their brightness. Periodic dips in a star's apparent brightness can reveal the presence of an orbiting planet, which passes in front of its star every so often and blocks a small fraction of its light. Just how small that fraction is depends on the planet's diameter, so Kepler's observations can pinpoint the size of its exoplanets discoveries.

But astronomers cannot guess at a planet's composition unless they also know its mass, and hence its density. Kepler usually provides no information about how massive a given exoplanet might be. (In some cases, however, when multiple planets orbit a single star, their mutual gravitational interactions can be used to estimate each object's mass.) So Marcy and his colleagues use a time-tested approach, called the radial-velocity method, to estimate masses for the smallish Kepler objects. With the 10-meter Keck I telescope in Hawaii, they trace the stars' velocity over time as measured by spectral changes in the starlight due to the Doppler effect. A star moving away from Earth will look slightly redder; a star moving toward Earth somewhat bluer. The best spectrometers in the world are capable of detecting velocity changes so subtle—roughly the walking speed of a human—that they can reveal the presence of an orbiting planet tugging ever so slightly on its star.

A very small planet produces a very small Doppler shift in its host star's light, but as Marcy later explained to me, Kepler gives astronomers a leg up in identifying it. Roughly speaking, the star's velocity follows a sinusoidal pattern, oscillating in a periodic manner as the planet proceeds through its orbit, first tugging the star slightly toward Earth, then pushing it away. Because Kepler's data reveal the planet's orbital period as well as the phase of its orbit—where the planet will be found along its orbital path at any given time—the researchers know both the timing and the period of the velocity oscillations. All that’s missing is the amplitude—how much the star's velocity fluctuates over the course of the planet's orbit.

The Keck data have started to fill in that missing piece. By measuring the three stars' velocities at various times and fitting those data to the assumed oscillation pattern, the astronomers can begin to pin down the masses of the putative planets. (None of these planetary "candidates" has been officially confirmed by the Kepler team.) Again, the data are unpublished and preliminary, but they are promising. A planetary candidate 3.39 times the diameter of Earth looks to boast a density of 1.8 grams per cubic centimeter—close to that of Neptune. A somewhat smaller world, 2.57 times Earth's diameter, has a correspondingly greater density of 2.7 grams per cubic centimeter. "This is a density intermediate between that of the Earth and Neptune," Marcy said. The implication is that a world of that density might be somewhat rocky, with perhaps a gas-rich outer layer.

An even smaller body, just 1.7 times the diameter of Earth, turns out to be much denser at 7.6 grams per cubic centimeter. That is a bit higher, even, than Earth's density, implying a predominantly rocky makeup. Marcy cautioned that more measurements from Keck are needed to nail down the planet's mass. But the trend among the three Kepler planet candidates, and among the few known small planets in general, is intriguing. "There's an early suggestion that somewhere between one and two Earth radii is where the transition occurs, and one begins to see purely rocky planets," Marcy said.

That is not to say that there is a hard and fast cutoff. Even among the smallest exoplanets, considerable diversity could exist. "We don't even know for the one-Earth-radius planets that Kepler is finding, if they're all rocky," NASA's Natalie Batalha, a Kepler co-investigator, said in a talk at the conference. "Can you have a giant cometlike object at one Earth radius? We don't know."