Reconstruction of muscle spindle firing rates by two candidate models.

<p>(A) Two main candidate models for predicting IFRs are shown: the force-related model used a combination of the force developed in the musculotendon and its first time-derivative as input (linearly combined, inputs and model prediction in blue); the length-related model used a linear combination of the muscle length, velocity, and acceleration (inputs and model prediction in red). The instantaneous firing rate of the afferent (black dots) was used to compute the error to optimize model parameters. (B) Three examples of muscle stretch with different stretch speed are shown (10, 20, and 40 mm/s). The stretch was sustained for 1s before being released. From top to bottom, the force-related model output (blue line) is overlaid on the instantaneous firing rate during the stretch; the recorded force (light blue) and the response of the afferent (black dots and raster) are shown directly below the force-related model output; the length-related model output (red line) is overlaid on the instantaneous firing rate and shown with the controlled musculotendon length (light red) below the response of the afferent. (C) The muscle was stretched to the same final length and at the same velocity, but with different initial accelerations (700 and 1400 mm/s², respectively). (D) The muscle was stretched with sawtooth patterns, where stretches were not sustained but immediately released and repeated. In all cases, the force-related model fits were as good or better than the length-related model fits, as indicated by the R² value. The examples shown are for afferents 4 (A-B, D) and 5 (C).