The primary error metric we used was the absolute angle by which a velocity command would have missed the target, taking into account the cursor and target radii. Because task success requires hitting the target (i.e., cursor-target overlap), we define all commands that would result in cursor-target overlap as having zero angular error. Mathematically, this corresponds to any velocity command that points within degrees from the target center, where is the distance between target center and the position from which the velocity command originates, and and are the cursor and target radii, respectively. A velocity command that would not hit the target is given an error, , equal to the angle by which the cursor would have missed the target. Equivalently, we can consider the cursor-target overlap zone defined by a target-concentric circle with radius , and define angular error, , to be the smallest angle between the velocity command and the perimeter of the cursor-target overlap zone. (A) Consider an example in which we assess the error of velocity commands (blue and green arrows) originating from a position mm from the target center (the distance between workspace center and target center in a typical experiment). Here, the cursor radius, , and the target radius, , are both 7 mm (typical values from experiments). Any velocity command that points within of the target center would result in cursor-target overlap and thus would be evaluated as having zero angular error. The green arrow points in the direction farthest from the target center such that movement of the cursor (dashed blue circle) in this direction would result in cursor-target overlap. A velocity command (blue arrow) pointing from the target center would miss the cursor-target overlap zone by . (B) Consider a similar example, but with the velocity command originating from a position mm from the target center. Because the cursor-target distance has decreased, the zero error window increases to . As a result, a velocity command that points from the target center (blue arrow; same as in panel A), is now evaluated as having a smaller error, . The difference between the error angles, , in panel A and panel B, reflects the task goals, because a wider range of velocity commands would result in task success in panel B compared to panel A, and thus the same velocity command is more task-appropriate in panel B than in panel A. The metric was used extensively throughout this work (Figure 2A,C, Figure 3B,C, Figure 4C, Figure 6A, Figure 3—figure supplement 3, Figure 3—figure supplement 4, Figure 3—figure supplement 7, Figure 3—figure supplement 8, and Figure 4—figure supplement 2). We repeated those analyses using as the error metric (i.e., ignoring the distance to the target, cursor radius, and target radius) and found qualitatively similar results. In Figure 2C and Figure 3—figure supplement 8, velocity commands were evaluated as originating from a range of lagged cursor positions. Since cursor positions later in a trial tend to be closer to the target than earlier positions, velocity commands will tend to have smaller when originating from these later cursor positions. We controlled for this distance-to-target effect to ensure that it did not influence our results (see Materials and methods).