Predicting consumer behavior with attitude measures has not always been successful. It is proposed to deal directly with the attitude components. In this article one such approach is discussed. Consumer decisions are studied in simulated choice experiments. It is hypothesized that prediction of choice should be based upon the cognitive elements which are aroused in the choice process, rather than upon single preference or attitude ratings. To illustrate that such predictions can be made, measurements are made, prior to the choices, of the subjective importance of the aroused values, and of the perception of the alternatives to these values. From these scores the attractiveness of alternatives is computed using a multidimensional subjective expected utility-type of model. In the experiments, different choices are studied and predictions based upon different kinds of values are compared. In all the choices highly significant predictions can be made, and the findings suggest that for this a relatively limited number of values are sufficient, so that the procedure is applicable in most consumer choice situations. Major problems with the model relate to the identification of relevant values which are not interdependent, and to the rating techniques applied.