Animals must generate forces against their environment to overcome external forces and locomote in a controlled manner. In legged locomotion, limbs generate forces against the ground to support bodyweight and to accelerate bodymass. Ground reaction forces (GRF) are a measure of these forces. Researchers often separate GRF into vector components (vertical, fore-aft, and lateral) for easier analysis. To investigate the relative importance of each component of the GRF during human running, I devised a series of experiments to test their unique roles. The metabolic cost of running was previously believed to be determined by the much larger vertical GRF while the metabolic cost of generating horizontal forces was considered negligible. I show that horizontal propulsive forces constitute -30% of the total cost of running and are more expensive per unit force than vertical GRF. Vertical and horizontal GRF are often suggested to support bodyweight against gravitational force and accelerate bodymass against inertial force, respectively. I demonstrate that in order to maintain orientation of the resultant GRF gravitational forces are the major determinant of both vertical and horizontal GRF magnitudes. Passive forces also contribute to the GRF. Passive collision of the foot with the ground creates impact forces distinct from the active forces generated by the leg. Despite the dominant role of the vertical GRF to support bodyweight, I show that vertical impact forces in human running are influenced independently by an applied horizontal force. The third (lateral) component of the GRF is negligible during straight path. In nature, however, animals rarely run on straight paths. Thus, substantial lateral GRF are often required for maneuvering. A physiological upper limit to leg force has been suggested to explain the apparent trade-off between lateral maneuverability and forward speed. I demonstrate that this theoretical leg force limit is never reached during curve running. The inside leg seems to limit maximum speed as it generates smaller resultant forces with more curved paths. Better understanding of general biomechanical principles of locomotion will require considering the determinants and actions of the resultant GRF rather than investigating the independent roles of its vector components.