Adaptation to novel dynamics has been well explained by a model dissociating two different timescales: a fast process that learns but also forgets quickly and a slow process that learns slowly but retains more of the learning. While this model reproduces the main experimental evidences of motor learning, such as savings, spontaneous recovery and interference; it cannot account for adaptation to multiple tasks. Nevertheless, in the presence of appropriate contextual cues, humans are able to adapt simultaneously to opposing dynamics. Consequently, this model was expanded, suggesting that dual-adaptation occurs through a single fast process and multiple slow processes, but was only tested against experimental data in a single adaptation design. In this study, we challenged the model’s structure using computational modelling and by experimentally testing the ability to predict spontaneous recovery in dual-adaptation. Participants
performed reaching movements to a target and simultaneously adapted to two opposing force fields (adaptation and deadaptation phase), each associated with a contextual cue (workspace visual location). Twelve multi-rate models were then fitted to experimental data and compared with a BIC model comparison. Participants learned both opposing tasks simultaneously, reducing the motor error by learning the appropriate compensation to the perturbation and displayed spontaneous recovery towards the first learnt dynamics. The analysis of model predictions and the BIC model comparison supported the existence of two fast processes, and extended the model timescales to include a third rate: an ultraslow process. Our results on experimental data and model comparison showed that dual-adaptation can be best explained by a two-fast-triple-rate model. This new architecture of model can predict the formation of two independent motor memories at three different timescales for our experimental timeframe, but might be extended to more timescales for longer periods of adaptation.