Model state update.

A. The state update includes error-independent decay (top) and error-dependent learning (bottom). The four model states (top 1st plot) correspond to the two rates (superscripts, 1 and 2) and two movement directions (subscripts, o = Out and r = Return). The retention factor (top 2nd plot), which can be different for the two rates, determines the error-independent decay leading to decayed states (top 3rd plot). Error-dependent learning (bottom) is determined by the error experienced (eo(n) for an Out trial, bottom 1st plot), the learning rate, which can be different for the two adaptation rates (bottom 2nd plot), and a coupling factor (bottom 3rd plot). The coupling factor is always 1 for the states (both rates) corresponding to the current movement, but can have non-zero and different values for the states corresponding to the reciprocal movement. The error update (bottom 4th plot) is the product of the error, the learning rate and the coupling factor. The updated state vector (top far right) is the sum of the decayed states and error update (yellow bar shows the previous state vector). Finally, the motor output (bottom right plot) is the sum of the states corresponding to the the given movement as determined by the binary selection vector.