Television(TV)audience research can provide various accurate and reliable approaches to investigate TV ratings and viewer behavior.However,extant research ignores the issue of how the episodic nature of TV watching influences viewer behavior and program selection.The purpose of this paper is to elucidate the impact of an audiences dynamic decisions under two interactive behaviors: learning behavior and forgetting behavior.Learning behavior refers to updating drama quality assessment using information,while forgetting behavior disturbs the use of information.Based on a deep exploration of viewing behavior in a longitudinal context,we construct an innovative methodology employing Bayesian rules to model the dynamic process,and test its performance by comparing with three benchmark models.Through empirical work of model calibration based on the people meter data from the Hong Kong TV industry,we show that the learning and forgetting behaviors are two important factors determining the fit and predictive ability of our model.Some managerial implications are then discussed in order to provide useful insights to guide advertiser and TV broadcaster decision-making.