There is considered the adaptive control of a time-varying finite- dimensional stochastic system having an autoregressive moving average representation. The unknown, time-varying parameters of the system are estimated by a stochastic gradient minimum variance algorithm. The aim is to design an adaptive control law to cause the output of the system to track in the mean square sense a given desired output sequence and to ensure that the output and control sequence remain bounded in some sense. Furthermore, the robustness properties of the standard stochastic adaptive control algorithms for time-varying systems having exponentially convergent parameters are analyzed.