Researchers conduct gait analyses utilizing both overground (OG) and treadmill (TM) modes of running. The few studies comparing these modes are limited to a 2D analysis of the hip, knee and rearfoot or a 3D analysis of the hip region alone. Therefore, the purpose of this study was to compare 3D kinematics of the hip, knee and rearfoot between overground and treadmill running. In order to make an accurate comparison, we needed to determine the most accurate method of identifying stance during overground and treadmill running. Therefore, there were two aims of this study. In the first aim, we compared five kinematic methods of identifying stance present in the literature and validated them against stance identified with a forceplate. We recruited thirty subjects (15 male, 15 female) and compared methods of determining stance for both overground (20 subjects) and instrumented treadmill running (10 subjects). We found that the most valid and reliable methods for determining footstrike were ones that utilized the change from negative to positive in vertical velocity or minimum vertical position of the distal heel marker. These methods had absolute errors from 22.4 ms to 24 ms Toe-off was best identified using peak knee extension, with absolute errors of 4.9 ms for overground running and 5.2 ms for treadmill running. Our second aim was to compare kinematic curves during stance and joint angles at footstrike and at peak. We utilized trend symmetry (Crenshaw and Richards, 2006) for curve comparisons and intraclass correlation coefficients (ICCs) for discrete variable comparisons of joint angles. We recruited ten male and ten female recreational runners for this aim. The stance phase was determined using the methods identified from aim one. We found that the transverse plane and knee frontal plane curves exhibited the lowest similarity between overground and treadmill running. However, the overall average similarity was high (0.94, where 1.0 is perfect symmetry). While these results suggest that patterns of motion were generally similar, there were some differences in the discrete variables. Our ICC values ranged from 0.42 for the rearfoot frontal plane at footstrike to 0.93 for peak hip internal rotation. Overall, only 8/18 discrete variables exhibited strong correlations (>0.8). However, 17/18 discrete variables exhibited modest correlations (>0.6). These results serve as a reference for comparisons between overground and treadmill running.