A single kinesin molecule can move for hundreds of steps along a microtubule without dissociating. One hypothesis to account for this processive movement is that the binding of kinesin's two heads is coordinated so that at least one head is always bound to the microtubule. To test this hypothesis, the motility of a full-length single-headed kinesin heterodimer was examined in the in vitro microtubule gliding assay. As the surface density of single-headed kinesin was lowered, there was a steep fall both in the rate at which microtubules landed and moved over the surface, and in the distance that microtubules moved, indicating that individual single-headed kinesin motors are not processive and that some four to six single-headed kinesin molecules are necessary and sufficient to move a microtubule continuously. At high ATP concentration, individual single-headed kinesin molecules detached from microtubules very slowly (at a rate less than one per second), 100-fold slower than the detachment during two-headed motility. This slow detachment directly supports a coordinated, hand-over-hand model in which the rapid detachment of one head in the dimer is contingent on the binding of the second head.