We tested the hypothesis that kinesin moves parallel to the microtubule's protofilament axis. We polymerized microtubules with protofilaments that ran either parallel to the microtubule's long axis or that ran along shallow helical paths around the cylindrical surface of the microtubule. When gliding across a kinesin-coated surface, the former microtubules did not rotate. The latter microtubules, those with supertwisted protofilaments, did rotate; the pitch and handedness of the rotation accorded with the supertwist measured by electron cryo-microscopy. The results show that kinesin follows a path parallel to the protofilaments with high fidelity. This implies that the distance between consecutive kinesin-binding sites along the microtubule must be an integral multiple of 4.1 nm, the tubulin monomer spacing along the protofilament, or a multiple of 8.2 nm, the dimer spacing.