Four recent, generalized stochastic models of brand choice are empirically compared. The models are the Brand Loyal and Last Purchase Loyal Markov models, the heterogeneous form of the Linear Learning model, and the Probability Diffusion model. The models are first outlined and then compared with respect to their representation of consumers’ brand choice probabilities. Each of the four models is then fitted to each of eight sets of brand choice data. The Learning and Diffusion models were found to be the best, with some advantage;for the Diffusion model in the present empirical case.