Cells of Escherichia coli are able to swim up gradients of chemical attractants by modulating the direction of rotation of their flagellar motors, which spin alternately clockwise (CW) and counterclockwise (CCW). Rotation in either direction has been thought to be symmetric and exhibit the same torques and speeds. The relationship between torque and speed is one of the most important measurable characteristics of the motor, used to distinguish specific mechanisms of motor rotation. Previous measurements of the torque-speed relationship have been made with cells lacking the response regulator CheY that spin their motors exclusively CCW. In this case, the torque declines slightly up to an intermediate speed called the "knee speed" after which it falls rapidly to zero. This result is consistent with a "power-stroke" mechanism for torque generation. Here, we measure the torque-speed relationship for cells that express large amounts of CheY and only spin their motors CW. We find that the torque decreases linearly with speed, a result remarkably different from that for CCW rotation. We obtain similar results for wild-type cells by reexamining data collected in previous work. We speculate that CCW rotation might be optimized for runs, with higher speeds increasing the ability of cells to sense spatial gradients, whereas CW rotation might be optimized for tumbles, where the object is to change cell trajectories. But why a linear torque-speed relationship might be optimum for the latter purpose we do not know. molecular motor | motility | nanogold | switch M easurements of the torque-speed relationship of the bacterial flagellar motor provide a crucial test of models for motor rotation (1, 2). Previous measurements of this relationship have been made with smooth-swimming [counterclockwise (CCW) rotating] mutants of a variety of species: with the proton-driven motor of Escherichia coli (3, 4), with the sodiumdriven motor of Vibrio alginolyticus (5), and with a sodium-driven chimeric motor in E. coli (6). In all cases, motor torque is approximately constant up to a knee speed, after which it drops rapidly to zero. In E. coli at room temperature, the knee speed is about 170 Hz, and the zero-torque speed is about 300 Hz. It has been assumed that CCW and clockwise (CW) rotation are symmetric and exhibit the same torques and speeds (7).Here, we measured the torque-speed relationship for an E. coli strain locked in CW rotation. This strain is deleted for the genes that encode the response regulator, CheY, and its phosphatase, CheZ, as well as the adaptation enzymes CheR and CheB. Introduction of a plasmid that encodes wild-type CheY that can be induced to high levels with isopropyl β-D-thiogalactoside (IPTG) enables CW rotation. For comparison, we also measured the torque-speed relationship with the same strain lacking the plasmid, which is locked in CCW rotation. The measurements were made by adsorbing 0.356 μm diameter latex spheres to sticky-filament stubs (8) and monitoring rotation rates in motility medium containing diffe...