Originally, scientists predicted small asteroids to be hard and rocky, as any loose surface material (called regolith) generated by impacts was expected to escape their weak gravity. Aggregate small bodies were not thought to exist, because the slightest sustained relative motion would cause them to separate. But observations and computer modeling are proving otherwise. Most asteroids larger than a kilometer are now believed to be composites of smaller pieces. Those imaged at high-resolution show evidence for copious regolith despite the weak gravity. Most of them have one or more extraordinarily large craters, some of which are wider than the mean radius of the whole body. Such colossal impacts would not just gouge out a crater – they would break any monolithic body into pieces. In short, asteroids larger than a kilometer across may look like nuggets of hard rock but are more likely to be aggregate assemblages – or even piles of loose rubble so pervasively fragmented that no solid bedrock is left. The rubble hypothesis, proposed decades ago by scientists, lacked evidence, until the planetologist Shoemaker realized that the huge craters on the asteroid Mathilde and its very low density could only make sense together: a porous body such as a rubble pile can withstand a battering much better than an integral object. It will absorb and dissipate a large fraction of the energy of an impact; the far side might hardly feel a thing. At first, the rubble hypothesis may appear conceptually troublesome. The material strength of an asteroid is nearly zero, and the gravity is so low one is tempted to neglect that too. The truth is neither strength nor gravity can be ignored. Paltry though it may be, gravity binds a rubble pile together. And anybody who builds sandcastles knows that even loose debris can cohere. Oft-ignored details of motion begin to matter: sliding friction, chemical bonding, damping of kinetic energy, etc. We are just beginning to fathom the subtle interplay of these minuscule forces. The size of an asteroid should determine which force dominates. One indication is the observed pattern of asteroidal rotation rates. Some collisions cause an asteroid to spin faster; others slow it down. If asteroids are monolithic rocks undergoing random collisions, a graph of their rotation rates should show a bell-shaped distribution with a statistical "tail" of very fast rotators. If nearly all asteroids are rubble piles, however, this tail would be missing, because any rubble pile spinning faster than once every two or three hours would fly apart. Recently, several astronomers discovered that all but five observed asteroids obey a strict rotation limit. The exceptions are all smaller than about 150 meters in diameter, with an abrupt cutoff for asteroids larger than 200 meters. The evident conclusion – that asteroids larger than 200 meters across are rubble piles – agrees with recent computer modeling of collisions. A collision can blast a large asteroid to bits, but those bits will usually be moving slower than their mutual escape velocity (the lowest velocity that a body must have in order to escape the orbit of a planet). Over several hours, gravity will reassemble all but the fastest pieces into a rubble pile.