Selective tenotomy of one or two heads of the quadriceps based on electromyographic criteria improved knee flexion in hemiplegic patients who walked with a unilateral stiff-legged gait. The improvement was greatest in eight patients in whom the rectus femoris was released, either with or without release of the vastus intermedius, with activity in pre-swing stance and initial swing confined to those heads of the quadriceps; in these patients knee flexion improved an average of 20 degrees. Knee flexion improved an average of only 8 degrees in five patients in who activity was present in one head of the quadriceps that was not surgically released. No improvement in knee flexion occurred in seven of eight patients in whom activity was present in two or more of the quadriceps heads that were not surgically released.