A better understanding of time-varying multimodal human operator control behaviour is required for the development of advanced adaptive control support systems for cases when vehicle dynamics suddenly change, e.g., a stability augmentation system failure. For this research, an approach based on recursive autoregressive exogenous (ARX) models is verified and validated for the identification of time-varying multimodal manual control in compensatory tracking tasks. The multimodal recursive ARX method is an extension of the unimodal recursive ARX method developed in earlier work. The verification of the recursive multimodal ARX method was performed by means of Monte-Carlo simulations. The validation of this method is based on the identification results of a human-in-the-loop experiment. Both for the verification and validation it was investigated which forgetting method of the recursive least squares algorithm resulted in the estimates for the ARX coefficients and human operator model parameters with the best model fit. A forgetting matrix with an infinite memory horizon for the coefficients of the A(q) polynomial and a memory horizon of 5 seconds for the coefficients of the B(q) polynomials was found optimal for both tasks with time-varying as well as linear time-invariant controlled element dynamics. The estimation results for the human operator model parameters gathered during the human-in-the-loop experiment were in line with the results that were found in literature. With this multimodal identification method now verified and validated, the potential for application in the development of online adaptive support systems for the flight deck can be investigated in future work.