The article presents an attempt of applying stochastic Hidden Markov Models (HMM) to the analysis of human visual activity. The input data for simulation experiments were taken from the empirical investigation concerned with pairwise comparisons of diverse product virtual packages. Simulations involving multiple estimations of HMM parameters were conducted for all distinctive types of comparisons and assuming different quantities of hidden states. The analysis of simulations’ results allowed to determine the optimal, in terms of AIC measure, numbers of hidden states. The best models were next analyzed and discussed in the perspective of visual strategies used during pairwise comparisons.