This paper proposes an Entity-based Narrative Graph (ENG) to model the internal-states of characters in a story. We explicitly model entities, their interactions and the context in which they appear, and learn rich represen- tations for them. We experiment with different task-adaptive-pretraining objectives, in-domain training, and symbolic inference to capture dependencies between different decisions in the output space. We evaluate our model on two narrative understanding tasks, predicting character mental states, and desire fulfillment, and conduct a qualitative analysis.