In this paper we make use of a particular technique of data analysis to empirically study the effect of joint attributes presentation in a multi-step process of choice, in which consumers first use a simple method for shortlisting, then proceed with a closer inspection of a restrict number of alternatives. Shortlisting is based on an incomplete description of the attributes of an alternative. We focus, in particular, on the presentation couples of attributes. The mathematical framework we used is the generalized spectral analysis. We tested this method on data collected through an ad hoc survey. Thanks to this powerful machinery we were able to identify the attraction single attributes have, from the effect of their combination. The use of generalized spectral analysis to decompose data on preferences is totally new. The decomposition allows us to underline two effects: the first and second order effect. The first order effect measures the average attraction that a single feature has when it is coupled with a second one. The second order effect detects the positive (or negative) power of combination of two coupled attributes. We present here a particular case, the choice of a car, among the ones we studied, to show how the method can be used, and its power. A particular emphasis will be given to gender differences in the evaluation of car attributes in the choice process.