Failing to decide when acting under time constraints can be detrimental, such as a driver failing to decide where to steer a car and causing a crash. Decision-making research has explored how humans select motor plans to maximize reward given sensorimotor delays and uncertainties. Yet studies imposing time constraints have not examined indecisive behaviour, where humans fail to decide before a deadline. Here we test the idea that optimal motor planning that accounts for sensorimotor delays and uncertainties will result in indecisions. Participants were shown two targets and were required to reach the same target as a computer agent. They received one point for selecting the same target as the agent, zero points for selecting the opposite target, or zero points if they were indecisive and failed to reach a target by 1500 ms. Agent movement onset time was drawn from a normal distribution. In a repeated measures design, we manipulated the mean (1000, 1100, or 1200 ms) and standard deviation (50 or 150 ms) of the normal distribution that determined the agent movement onset time. We developed a model that finds the decision time that maximizes expected reward. The model predicts less indecisions in the 1200 ms conditions than the 1000 ms conditions. However, participants made more indecisions in the 1200 ms condition than the 1000 ms condition. Further, participants made more indecisions in the 1200 ms condition compared to the model. Our results suggest that humans suboptimally account for sensorimotor uncertainties, leading to an excessive number of indecisions.