When mixtures of dissimilar tasting compounds are perceived, specific taste sensations are integrated into a percept. Four models are evaluated that were designed to predict the total taste intensity of a mixture percept from the total taste intensities of the unmixed components or from the specific taste intensities of the components within the mixture percept. The predictions of the two dominant component models are inferior to those made by the two other models. Predictions made by the vector summation model are accurate, but the model contains theoretical weaknesses that make it an invalid psychological integration model. The sum of sensations model predicts total intensity with an accuracy similar to the vector summation model, but it is more parsimonious. Furthermore, it can account for all of the empirical phenomena observed.