Abstract. The Ph.D. thesis is devoted to the study of two economic phenomena, namely reference dependence and social interactions, applied to financial markets. In Chapter 1 we study S-shaped utility maximization for the standard portfolio selection problem. We derive a mean-variance criterium of choice, which preserves reference dependence and the reflection effect. Subsequently we study diversification possibilities and obtain the demand for the risky asset. The model of a financial market presented in Chapter 2 assumes heterogeneity of agents with respect to both their reference points and their beliefs. We analyze the impact of the former layer of heterogeneity on the asset return and wealth dynamics. The results reveal that different reference points of the agents can lead to market instabilities and persistent trading volume even if their beliefs are kept homogeneous. In Chapter 3 we propose a model of a financial market populated with heterogeneous belief agents. Every period each agent chooses a predictor of the future price on the basis of the past performance of own and alternative strategies of the directly connected to her agents. Using the rewiring procedure we construct four types of commonly considered networks and investigate the effects of their topology on the asset price dynamics. We show that network structure does influence price dynamics due to the different speed of information exchange.