A discussion of methods for identification of dynamic systems is presented. Problems and methods for determining model structures and estimating unknown parameters are considered. The maximum likelihood (ML) formulation for parameter estimation is discussed in detail due to its generality and its success in numerous applications. An outline is given of the steps and the computational considerations involved in a system identification problem. The benefits of identifying the process and observation noise sources and then applying the ML approach as opposed to the classical least-squares technique are discussed. Present and potential applications in the nuclear industry are reviewed.