To answer the question why such large differences in stress fracture morbidity rates (2% to 64%) exist in different countries, we prospectively evaluated 312 recruits for possible risk factors for stress fractures. Prior to training, each recruit underwent an evaluation including the following: orthopaedic examination, foot and tibial radiographs, measurements of tibial bone width, bone mineral content, bone density, aerobic physical fitness and leg power, assessments of somatotype and smoking habits, and evaluation of sociological and psychological factors. Using a multivariate analysis, two risk factors were identified: recruits with stress fractures had significantly narrower tibiae (P less than 0.001), and a higher degree of external rotation of the hip (P = 0.016). These two variables were independent and cumulative. Stress fracture morbidity was 17%, 29%, and 45% when neither, one, or both risk factors were present, respectively (P less than 0.001). Identification of these risk factors might explain the susceptibility of some people to stress fractures.