A narrow base of gait (BOG) during running, such that the foot crosses contralateral to the body’s line of gravity, has been implicated as a cause of iliotibial band syndrome and patellofemoral pain. BOG has been shown to vary by sex and running speed, but it is unknown if body anthropometric and running kinematic measures predict BOG. Determining predictors of BOG will clarify if BOG is a result of non-modifiable anthropometric factors or biomechanical factors, which could be modified using targeted interventions. PURPOSE: To determine if BOG at midstance during running can be predicted by anthropometric or kinematic measures. METHODS: Whole body kinematics were obtained for 71 Division I cross country runners (30 males) during treadmill running at preferred speed. Athletes were healthy with no history of stress fracture 3 months before testing or any history of lower extremity surgery. Anthropometric measures from each athlete were obtained from whole body dual-energy X-ray absorptiometry scans: greater trochanteric (GTR) and hip joint center width; leg and femur length. Kinematic measures during stance phase included: peak lateral pelvic drop, hip adduction (HADD), knee flexion (KFLEX); vertical excursion of center of mass (vCOM); anterior-posterior distance from heel to COM at initial contact. Correlations between predictors and BOG were calculated, with variables moderately correlated or better (|r| ≥ 0.3) included in subsequent analyses. Data from both limbs were included in a forward, stepwise regression to determine predictors of BOG, controlling for sex and speed. RESULTS: Stride length, vCOM, peak KFLEX and HADD, GTR width, and leg and femur length (|r| = 0.32 - 0.51) were entered into the model. The model with the best overall fit included all predictors except leg length (R2 = 0.383, BIC: -32.1). The strongest predictors were GTR width (β = -0.27), vCOM, (β = -0.18), and peak HADD (β = -0.17). CONCLUSIONS: Biomechanical and anthropometric measures explain less than 40% of the variance in BOG. Given that GTR width is the strongest predictor of BOG and is non-modifiable, BOG appears to be largely influenced by an individual’s anthropometrics. Conversely, vCOM and peak HADD are modifiable through gait retraining strategies such as step rate modification and may be targeted in those where narrow BOG is a concern.