The objective of this study was to explore relationships among constants used in musculo-skeletal models predicting torque generated about the knee by the quadriceps muscles. A model was developed and matched to data collected from individuals with spinal cord injury performing quadriceps contractions evoked using neuromuscular electrical stimulation. After fitting tendon slack lengths to the quadriceps muscles, the model was able to accurately match experimentally measured knee extension torques using previously reported values for the moment arm-knee angle and force-velocity relationships. Fitting new constants to these relationships did not improve the match between measured and modelled knee extension torques. There was significant interaction between variables used within the model. Using a narrower active force-length relationship for the muscles required the model to have smaller moment arms about the knee to accurately match measured torque across the full range of motion. Reduced moment arms, however, lowered the model's linear velocity of muscle shortening for each angular velocity of the knee, requiring different constants within the force-velocity relationship to predict the appropriate amount of torque decline. The present study demonstrates that, when a model does not fit the observed data, it is difficult to determine exactly which components are responsible because of the interdependent nature of parameters.