Novelty is a source of creativity. Acceptance of novelty, however, depends on receiver’s emotions. As MAYA(most advanced, yet acceptable) principle of Raymond Loewy suggested, extremely advanced(novel) design may not be accepted. Our motivation of this study is to mathematically model emotions for novelty and explain how novelty affects emotions. By doing so, we aim to provide a fundamental knowledge to guide innovative design in a systematic manner. The most dimensional models of emotion incorporate arousal (or intensity) and valence (positivity and negativity) dimensions. Novelty provides new information to organism. A large amount of information gain by experiencing a novel event elicits intense emotion such as surprise. We hypothesize that the information gain corresponds to arousal of emotions. Based on Berlyne’s arousal potential theory, we hypothesize that an organism positively accept a novel event providing appropriate amount of information gain which can be cope with. Based on above mentioned hypothesis, we mathematically formalize emotional dimensions with respect to novel event (stimuli). We applied Bayesian theorem to renew subjective probability distributions from prior to posterior by experiencing a novel event. We used Kullback-Leibler divergence between posterior and prior, i.e. information gain, as an index of arousal level. By conducting computer simulation using Gaussian distributions for subjective distributions, we found that prior variance (representing uncertainty) and difference between prior peak and posterior peak (prediction error) interactively affect information gain. While low expectation discrepancy, the higher uncertainty, the more information gain. However, as expectation discrepancy increased, the difference of information gain for different uncertainties switch-over. The lower uncertainty, the more information gain. We conducted an experiment with participants using short films where different percussion instruments are tapped. We synthesized different tapped sound to manipulate expectation discrepancy. To manipulate prior uncertainty, we prepared both familiar and unfamiliar percussion instruments. We obtained subjective responses regarding degree of surprise and ERP(event-related potential) P300 representing higher cognitive response to unexpected stimuli. The experimental result showed the interaction effect on both subjective and biological responses so as to support the computer simulation using proposed formulation. This result suggests that information gain can be used as a mathematical index of arousal.