Muscle coordination studies repeatedly show low-dimensionality of muscle activations for a wide variety of motor tasks. The basis vectors of this low-dimensional subspace, termed muscle synergies, are hypothesized to reflect neurally-established functional muscle groupings that simplify body control. However, the muscle synergy hypothesis has been notoriously difficult to prove or falsify. We use cadaveric experiments and computational models to perform a crucial thought experiment and develop an alternative explanation of how muscle synergies could be observed without the nervous system having controlled muscles in groups. We first show that the biomechanics of the limb constrains musculotendon length changes to a low-dimensional subspace across all possible movement directions. We then show that a modest assumption—that each muscle is independently instructed to resist length change—leads to the result that electromyographic (EMG) synergies will arise without the need to conclude that they are a product of neural coupling among muscles. Finally, we show that there are dimensionality-reducing constraints in the isometric production of force in a variety of directions, but that these constraints are more easily controlled for, suggesting new experimental directions. These counter-examples to current thinking clearly show how experimenters could adequately control for the constraints described here when designing experiments to test for muscle synergies—but, to the best of our knowledge, this has not yet been done. How the brain and spinal cord control the body is a fundamental question of critical scientific and clinical importance. The preferred experimental approach to answer this question has been to infer the neural control strategy by analyzing recordings of muscle activity and limb mechanics collected while animals and people use their limbs. This has led to a popular, but not yet proven, hypothesis that the brain and spinal cord simplify the control of the numerous muscles by grouping them into few functional units called neural synergies. Our detailed experiments and simulations challenge the utility of this approach and the validity of its interpretation. We point out that mechanical constraints can also explain those experimental recordings. In particular, the anatomy of the limb combined with the type of tasks studied and analysis used, suffice to give the appearance of neural synergies. To be clear, we do not disprove the neural synergy hypothesis. Rather, in the tradition of scientific debate, by showing an alternative explanation to the available data we challenge the community and ourselves to design novel experiments and analyses to conclusively test that hypothesis by ruling out the confounds we point out.

Numerous observations of structured motor variability indicate that the sensorimotor system preferentially controls task-relevant parameters while allowing task-irrelevant ones to fluctuate. Optimality models show that controlling a redundant musculo-skeletal system in this manner meets task demands while minimizing control effort. Although this line of inquiry has been very productive, the data are mostly behavioral with no direct physiological evidence on the level of muscle or neural activity. Furthermore, biomechanical coupling, signal-dependent noise, and alternative causes of trial-to-trial variability confound behavioral studies. Here we address those confounds and present evidence that the nervous system preferentially controls task-relevant parameters on the muscle level. We asked subjects to produce vertical fingertip force vectors of prescribed constant or time-varying magnitudes while maintaining a constant finger posture. We recorded intramuscular electromyograms (EMGs) simultaneously from all seven index finger muscles during this task. The experiment design and selective fine-wire muscle recordings allowed us to account for a median of 91% of the variance of fingertip forces given the EMG signals. By analyzing muscle coordination in the seven-dimensional EMG signal space, we find that variance-per-dimension is consistently smaller in the task-relevant subspace than in the task-irrelevant subspace. This first direct physiological evidence on the muscle level for preferential control of task-relevant parameters strongly suggest the use of a neural control strategy compatible with the principle of minimal intervention. Additionally, variance is nonnegligible in all seven dimensions, which is at odds with the view that muscle activation patterns are composed from a small number of synergies.

Coherence analysis has the ability to identify the presence of common descending drive shared by motor unit pools and reveals its spectral properties. However, the link between spectral properties of shared neural drive and functional interactions among muscles remains unclear. We assessed shared neural drive between muscles of the thumb and index finger while participants executed two mechanically distinct precision pinch tasks, each requiring distinct functional coordination among muscles. We found that shared neural drive was systematically reduced or enhanced at specific frequencies of interest (~10 and ~40 Hz). While amplitude correlations between surface EMG signals also exhibited changes across tasks, only their coherence has strong physiological underpinnings indicative of neural binding. Our results support the use of intermuscular coherence as a tool to detect when coactivated muscles are members of a functional group or synergy of neural origin. Furthermore, our results demonstrate the advantages of considering neural binding at 10, ~20, and >30 Hz, as indicators of task-dependent neural coordination strategies. NEW & NOTEWORTHY It is often unclear whether correlated activity among muscles reflects their neural binding or simply reflects the constraints defining the task. Using the fact that high-frequency coherence between EMG signals (>6 Hz) is thought to reflect shared neural drive, we demonstrate that coherence analysis can reveal the neural origin of distinct muscle coordination patterns required by different tasks.

Computational models of the neuromuscular system hold the potential to allow us to reach a deeper understanding of neuromuscular function and clinical rehabilitation by complementing experimentation. By serving as a means to distill and explore specific hypotheses, computational models emerge from prior experimental data and motivate future experimental work. Here we review computational tools used to understand neuromuscular function including musculoskeletal modeling, machine learning, control theory, and statistical model analysis. We conclude that these tools, when used in combination, have the potential to further our understanding of neuromuscular function by serving as a rigorous means to test scientific hypotheses in ways that complement and leverage experimental data.