Fig. 2.

Predictions from the static neural network model. A: representative schematic of 1 of the models. Two network units, z, are shown connected to muscle outputs, u, via random weightings, w, shown by the thickness of the lines. The output variables, u, drove mono- and biarticular muscles on a 2-joint arm, as shown. The values of the units in z were optimized to solve analogs of the reaching and posture tasks. B: the relationship between each unit's preferred torque direction (PTD) in the posture task and the calculated PTD of its motor field (defined by synaptic connectivity) in 1 network of 1,000 units. The size of the circle for each unit corresponds to its strength of tuning. Note that the muscles, plotted here for reference, must lie on the diagonal as they correspond with their own motor field.