Optimizing motor-timing decision via adaptive risk-return control

Human’s ability of optimal motor-timing decision remains debated. The optimality seems context-dependent as the sub-optimality was often observed for tasks with different gain/loss configurations: people achieved optimality with symmetric gain configuration but not with asymmetric configuration. In the current study, we designed a temporal decision-making task where participants could adjust the sensitivity parameter (i.e., risk-return trade-off) of the gain function, in order to testify whether people could optimize the responses for asymmetric gain configuration by adjusting the sensitivity parameter. Participants were asked to click a point within a certain spatial region at a specified timing, where the click timing determined the reward whilst the click position determined the sensitivity parameter. We prepared three types of gain functions (symmetric, risk-after and risk-before conditions) and tested whether or not the participants achieved Bayesian optimality irrespective of gain structure. Most participants’ performance reached optimality not only in the symmetric condition but also in the asymmetric condition, albeit some discrepancies from optimality observed in the risk-before condition. This confirmed that people could achieve Bayesian optimality even for asymmetric gain configuration. We argued that the adaptive risk-return is beneficial for the performance optimality.