Treadmills are often used in research projects to simulate overground locomotion, assuming that locomotion is similar on a treadmill and overground. The purpose of this investigation was to determine whether a treadmill could be used to simulate overground locomotion. Twenty-two subjects ran on four different surfaces: overground and three treadmills that differed in size and power. The kinematics of the right leg and foot were studied using two high-speed Locam cameras (lateral and posterior view). The subjects ran in two different shoes at four different speeds (3.0-6.0 m.s-1). The differences in the kinematics between treadmill and overground running could be divided into systematic and subject dependent components. Subjects systematically planted their feet in a flatter position on the treadmill than overground. Most of the lower extremity kinematic variables, however, showed inconsistent trends for individual subjects, depending on the individual subject's running style, running speed, and shoe/treadmill situation. The differences were substantial. It is not yet understood how the human locomotor system adapts to a particular treadmill running situation. However, it is concluded that individual assessment of running kinematics on a treadmill for shoe or shoe orthotic assessment may possibly lead to inadequate conclusions about overground running.