In this paper, an analysis of the tracking and noise performance of several adaptive algorithms is carried out for the case of model structures with fixed pole positions. Such structures have recently been proposed as an efficient generalisation of the common FIR model structure. The focus of this work is to analyse the associated tradeoff between noise sensitivity and tracking ability in the frequency domain by illustrating how it is influenced by such things as input and noise spectral densities, step size and, what is the main focus of this paper, the choice of the fixed pole locations. The latter influence is not described by pre-existing analysis, but is shown here to be amenable to attack by a particular class of orthonormal bases.