Current models of localized muscular fatigue are capable of predicting performance in isometric tasks with reasonable accuracy. However, they do not account for the effect of continuously-varying task intensities on muscular recovery from a fatigued state. In this work, we propose and evaluate three continuous functions for modelling recovery to replace a dichotomous step-function in the three-compartment controller (3CC-r) model of muscle fatigue (Looft et al., 2018) and validate their predictions with previously collected data in the literature for intermittent and sustained isometric tasks of the ankle joint performed at different intensities. When compared to experimental data the accuracy of one of the three proposed models of recovery is found to be nearly the same as that yielded by the original step-function, but this seemingly-identical accuracy may be a limitation of the dataset used. A superelliptical curve relating recovery factor to task intensity is proposed to be the closest replacement for the step function as it depicts both the elevated value of recovery factor for near-rest activities as well as a nearly-constant value for low-to-high-intensity tasks.