The hand-over-hand stepping mechanism of kinesin at low loads is inadequately understood because the number of molecular steps taken per encounter with the microtubule is difficult to measure: optical traps do not register steps at zero load, while evanescent wave microscopy of single molecules of GFP-kinesin suffers from premature photobleaching. Obtaining low-load data is important because it can efficiently distinguish between alternative proposed mechanisms for molecular walking. We report a novel experiment that records the missing data. We fused kinesin to gelsolin, creating a construct that severs and caps rhodamine-phalloidin actin filaments, setting exactly one kinesin molecule on one end of each fluorescent actin filament. Single kinesin molecules labeled in this way can be tracked easily and definitively using a standard epifluorescence microscope. We use the new system to show that, contrary to a recent report, kinesin run length at low load is independent of ATP concentration in the muM to mM range of ATP concentration. Adding competitor ADP in the presence of saturating ATP decreases both velocity and run length. Based on these data, we propose a simplified model for the mechanism of processive stepping.