Abstract The selection of a subset of attributes is a very important step in sensory profiling as it consists in reducing the list of variables. This reduced list can be used in subsequent studies thus saving fatigue and time to the members of the panel. In this paper, we discuss statistical procedures that may be used to select a subset of attributes. These procedures are based on principal component analysis, the minimization of a procrustes criterion, or a clustering of variables approach. The various procedures are compared on the basis of a real data set and a simulation study.