Operating an aircraft requires pilots to continually manage their cognitive and affective states, which depends on pilot self-regulatory skills. To design training systems that support such regulation, we need to understand how affect and cognitive load vary during the learning process. While affect and cognitive load have been studied, their interaction in simulation-based pilot training is poorly understood so we conducted a study ( \(n = 20\) ) to investigate how affect valence impacts the prediction of extraneous and germane load when novices are learning through a simulator. Results indicate that both positive and negative affect predicted germane load, while negative affect was the only predictor of extraneous load. These predictive effects were moderated by individual differences in working-memory capacity. Our findings highlight the importance of integrating affect and cognitive load into the learner models that drive adaptive training systems.