SUMMARY

The degree to which muscle activity is controlled by neural selection or determined by biomechanics is unresolved; musculoskeletal redundancy allows for an unknown amount of variation in the muscle activation patterns. Computational musculoskeletal models can help answer questions about redundancy, but contradictory results are found in the literature. Some studies suggest wide ranges of nervous control options, while others suggest biomechanics largely determine muscle coordination. We hypothesized that contradictory results are due to different model complexity, and that models with more realistic complexity allow for more variability in muscle coordination.
We systematically varied the muscle complexity and the number of degrees-of-freedom (DoFs) of a musculoskeletal model and tested the significance of individual muscles in each model by looking at 1) the sensitivity of the feasible force set (FFS) to single muscle loss and 2) the feasible muscle activation ranges (FMARs) at maximum force in the sagittal plane.
We demonstrated that results from many studies with oversimplified models do not generalize to more realistic systems, while more realistic models suggest that very few muscles are constrained by biomechanics. As model complexity increased, the sensitivity of the FFS to single muscle loss decreased. We showed that most muscles have wide FMARs in many maximum force. Only a few muscles (the hip-knee biarticular muscles) were completely constrained in all maximum sagittal plane forces. Further, we showed that the effects of complexity in muscles and DoFs are general for any musculoskeletal system.
An understanding of the degree to which muscle activity is determined by biomechanics or by neural selection has significant implications for rehabilitation. Low levels of biomechanical constraints suggest many different neural strategies or compensations are feasible, indicating rehabilitation efforts should focus on training muscle coordination.