Limits to human movement planning with delayed and unpredictable onset of needed information

In motor tasks with explicit rewards and penalties, humans choose movement strategies that nearly maximize expected gain (Trommershäuser et al. in J Opt Soc Am A 20:1419-1433, 2003). Here, we examine whether performance is still close to optimal when information about payoffs is not available prior to movement onset. Subjects rapidly touched a target region while trying to avoid hitting an overlapping penalty region placed randomly to the left or right of the target. Subjects received rewards and incurred penalties for hitting the corresponding regions. Late responses (>700 ms) were heavily penalized. The penalty region was displayed 0, 200 or 400 ms after the reward region and the subject could not know where it would be until then. Reaction times to begin the movement after stimulus appearance were constant across conditions. Median reaction times were approximately 200 ms, i.e., the time the penalty was first displayed in the 200 ms delay condition. Performance was compared to that of an optimal movement planner that chooses mean end points to maximize expected gain despite movement variability. In the 0 and 200 ms delay conditions, subjects selected strategies that did not differ significantly from optimal, indicating that humans are able to plan their movements well despite delayed and unpredictable onset of information. Performance dropped below optimal in the 400 ms delay condition, with mean movement end points closer to the penalty region than predicted by the optimal strategy (in the high-penalty condition). We conclude that relevant information concerning the reward structure is required between 200 and 400 ms prior to the end of the movement, but can still be integrated into the movement plan after movement initiation.