The aim of this paper is to give an overview of statistical, spatial and cognitive image characteristics which could influence the performance of gamut mapping algorithms. In addition to reviewing these characteristics, a general method for evaluating their importance in the above context is given whereby this method is based on the creation of sets of images which either share or differ in terms of given characteristics in known ways. The psychophysical evaluation of the reproductions of such image sets can then yield information about their relative impact on the performance of the gamut mapping algorithms used. Having such information would be of great benefit when trying to improve the performance of colour reproduction systems in general and gamut mapping in particular.