Postoperative knee flexion in patients undergoing Insall-Burstein-II total knee arthroplasty at 2 years was evaluated regarding two basic questions: what groups of patients gain or lose the most flexion and what groups of patients have the best or worst postoperative flexion. Thirteen preoperative variables (maximum flexion, flexion arc, tibiofemoral angle, quadriceps strength, extensor lag, Knee Society score, Knee Society patient assessment, gender, age, height, weight, diagnosis, and surgeon) and four postoperative variable (leg length change, tibiofemoral angle, distance from patella to the joint line, and the tibial prosthesis anteroposterior translation on a lateral radiograph) were used in an attempt to explain postoperative flexion. The analysis was performed on 164 consecutive Insall-Burstein-II total knees in which the data were gathered prospectively on a time oriented medical record database. A regression tree analysis was used to identify several groups of patients, characterized by preoperative factor values, who had markedly above average performance on postoperative flexion. The preoperative factors identified include preoperative flexion, flexion arc, tibiofemoral angle, extensor lag, diagnosis, and age. The only postoperative variable of significance was tibiofemoral angle. Among the potential determinants of postoperative flexion that failed to appear predictive were the Knee Society scores and surgeon. Preoperative flexion is known to be a critical determinant of postoperative flexion in total knee replacement. However, in the current study, preoperative flexion accounted for only half of the difference between the best (122 degrees) and the worst (88 degrees) group, as determined with regression tree analysis.