Many of the investments decisions facing with uncertainty can be characterized as real options problems. There is evidence of deviation from the predictions derived using such normative models. The proposed research sheds light on the importance of integrating normative models with experimental methods in order to predict and explain such cognitive limitations, in the particular context of R&D alliances. The focus is on appropriate validation of such models on experimental data. We propose a simple design starting from a real options model dealing with alliance timing decisions. We present the decision makers with risky choices formulated as abstract gambling decisions in order to assess their risk propensity and to validate the normative predictions of the model. This paper introduces the basic principles of the use of fuzzy grouping variables in economic analysis. On the survey data gathered to validate the predictive power of the presented model we show that fuzzy sets can be effectively used to partition the experimental data into fuzzy subsets for model verification (e.g. when subgroups cannot be defined in a crisp way). We compare the validation of the model on a full data set with a “refocused” validation on a fuzzy subset of the original sample.