Figure 7.

(A) Normalized weights given to memorized location for predicting finger position as a function of the proportion of time through subjects’ movements. We have only shown the weights computed for the last half of the movement because the weights estimated for points in time during the first half of the movement were, for some subjects, extremely noisy. This is reflected in the standard errors of the weight estimates. The standard errors on weight estimates for a few subjects exceeded 10 for times prior to the halfway point of the movements (high standard errors on normalized weights are possible, because normalized weights were constrained to sum to 1, not to lie between 0 and 1; thus, the re-sampled distribution of memory weights used to compute error bars could include negative values and values greater than 1). In part, this was due to the low signal-to-noise ratio for estimating weights early in movements when the finger trajectories to different target positions have not diverged by much. In part, it was due to bad motion recording data for some subjects early in movements. Superimposed on the plots are graphs of the weights computed in the previously published experiment (dashed lines). (B) Average standard errors of the weight estimates in the last half of the movements. Average standard errors earlier in the movement are not shown because they are so high.