Muscles: Non-linear Transformers of Motor Neuron Activity

Predicting movement from neural activity requires quantitative understanding of muscle response to motor neuron input. Muscles are sufficiently complicated that fulfilling this goal requires computer simulation. We therefore first explain in considerable detail one approach to modeling muscle. We then provide multiple examples of how muscle intrinsic properties and muscle diversity make straightforward predictions of how muscles transform neural input into movement impossible, including the dependence of muscle velocity on sarcomere number, the inadequacy of mean data in muscle modeling, the effects of muscle low-pass filtering, spike-number vs. spike frequency coding for contraction amplitude, how the role of passive muscle force in movement generation varies as a function of limb size, how muscles produce forces greater than their ‘maximum force’, energy conserving mechanisms, muscles that brake rather than produce movement, and how muscles can generate restoring responses (preflexes) to perturbing input in the absence of sensory feedback.