Impulse buying is usually described as a sudden, compelling, hedonically complex purchasing behavior without deliberate consideration of all information and choice alternatives. Since 1950’s, evidence has indicated impulse buying to be a pervasive and distinctive aspect of consumption phenomena, contributing a large portion of sales. It is not surprising that this topic has attracted considerable attention from multiple disciplines, including economics, psychology, and marketing. With the proliferation of electronic commerce, consumers can buy around the clock. They are exposed to a variety of multimedia stimuli that may trigger impulse behavior. Indeed, several reports based on large-scale surveys have reported evidence of online impulse buying. Stimulating impulse has become a focal phenomenon in online consumer behavior. Surprisingly, very few studies investigated online impulse buying in general and the role of system design effects in particular. Most prior research on online shopping focused on intentional behavior, using attitude-intention models (e.g., theory of planned behavior and technology acceptance model). Impulse behavior, however, is characterized by the lack of planning and the dominance of emotions, precluding the formation of intention. New theories are therefore necessary to explain online impulse buying. The main objective of this thesis is to address this theoretical void. Drawing upon the “stimulus-organism-response” framework, I develop and empirically validate a model explaining the relationship between system design characteristics and online impulse buying. Impulse formation and enactment are modeled as a response to four organismic factors, i.e., telepresence, social presence, pleasure and arousal. These organismic variables are determined by system stimuli that are characterized as interactivity and vividness features. The research model is validated by an experiment with a 2×2 full factorial design involving 151 undergraduate students. The empirical results provide very strong support for the model. The theoretical and practical implications of these results are discussed.