Abstract
In carrying out a movement, there is uncertainty due to motor variability. In past work (Trommershauser et al., 2003, JOSA, 20, 1419), we asked actors to rapidly touch a reward object while avoiding a nearby penalty object. As we varied reward, penalty, and object locations, actors modified their movement plans to maximize expected gain (EG). They effectively considered each possible outcome (penalty only, reward only, both, neither) paired with the probability that the planned movement would lead to that outcome. Such a list of paired outcomes and probabilities constitutes a ‘lottery’, We found that actors chose the lottery with maximum EG. In contrast, humans typically fail to maximize EG when choosing lotteries presented in pencil-and-paper form (Kahneman & Tversky, 2000, Cambridge Univ. Press). In our task, probabilities are due to the actor's intrinsic motor variability, while in pencil-and-paper tasks they are chosen by the experimenter and conveyed symbolically. We modified our task so that, in some conditions, the reward and/or penalty regions were stochastic: when the actor hit a stochastic region, s/he would receive the reward or penalty with probability 0.5. We compared performance for all four combinations of certain or stochastic rewards and penalties. Each condition was run in a separate session and the first and last sessions were always fully certain. The actor was explicitly told the probabilities before each session. Five naive actors participated. We found that all actors maximized EG in the fully certain conditions but that three of the five were markedly sub-optimal in one or more of the stochastic conditions. While movement planning takes into account implicit probabilities due to motor variability, the introduction of explicit probabilities led to sub-optimal performance.
NIH EY08266; HFSP RG0109/1999-B; DFG (Emmy-Noether-Programm)