Rectus abdominis muscles of young healthy female volunteers were studied to determine both the maximum strength in isometric flexion and the relation between the surface electromyogram (RMS EMG) and torque. Both the upper and lower portions of the abdominal muscle were studied during graded Increase and decrease of torque. Repeated testing was performed over 6 weeks. The form of the relation between torque and EMG was better described by a quadratic than by a linear regression relationship, but with considerable variability about the best-fit line. The torque-increasing and torque-decreasing parts of each test were different and were analyzed separately. After 6 weeks of repeated testing, maximum voluntary isometric flexion torque Increased 16.8% (P < 0.01). There was a decrease in the ratio of electromyographic activity to torque production, which was statistically significant in the torque-Increasing recordings. These changes in maximum torque and in the EMG-torque relation were attributed to learning through a test-retest effect, rather than to a true change in muscle characteristics. These findings, which show changes in normal subjects undergoing repeated testing, do not support the reliability of isometric strength measurements, or measurements based on RMS EMG recordings, for quantifying abdominal muscle function in patients with back pain, or those undergoing strengthening exercises.