Adaptive task allocation (ATA) provides a new solution for human-machine interaction in a highly automated system. Previous research demonstrates that psychophysiological signals yield sensitive information about human functional states, which can be used to build a closed loop for human-machine interaction to reallocate the tasks upon the status of human operator. The present study investigates the feasibility of adaptive task allocation in a field driving context. Driver’s mental workload was evaluated by electroencephalogram (EEG) in real-time and this result was used to dynamically adapt a secondary task allocated to driver. The results showed that ATA has a potential benefit to maintain a driver’s workload in a moderate level. However, generally, no significant increases in task performance were found between ATA and without ATA conditions.