Based on the idea of emotional embeddedness, this paper studies the possible influence of emotions on decision in the ultimatum game. Given the two possible emotion-cueing environments, we propose two emotion indexes, called the nay-based emotion and the reference-based emotion, as measures of the possible emotion state of the subjects in the game. We characterize subjects' behavior as a stochastic choice model with the assumption that their decisions (the offer rate made by the proposed, in our case) can be affected by the triggered emotions. This assumption is then examined using an ordered logit model under different settings of subjects' characteristics. Our estimation based on the Monte Carlo simulation then shows the effects of both kinds of emotions, and that effect can differ by gender and by the employed exchange medium (money or chocolate, in our case).