Virtual balancing tasks facilitate the study of human motion control: human reaction to the change of artificially introduced parameters can be studied in a computer environment. In this article, the dynamics of human stick balancing are generalized using fractional-order derivatives. Reaction delay sets a strong limitation on the length of the shortest stick that human subjects can balance. Human processing of visual input also exhibits a memory effect, which can be modelled by fractional-order derivatives. Therefore, we hypothesize a delayed fractional-order PD control of the unstable fractional-order process. The resulting equation of motion is investigated in a dimensionless framework, and stabilizability limits are determined as a function of the dynamics’s order. These theoretical limits are then compared with the results of a systematic series of virtual balancing tests performed by 18 subjects. The comparison shows that the theoretical stabilizability limits for controllers with fixed fractional order correspond to the measured data points. The best fit is obtained if the fractional order of the underlying control law is 0.475.