Abstract
This study investigated learners' attention during interaction with a serious game. We used Keller's ARCS theoretical model and physiological sensors (heart rate, skin conductance, and electroencephalogram) to record learners' reactions throughout the game. This paper focused on assessing learners' attention in situations relevant to learning, namely overcoming errors and obstacles. Statistical analysis has been used for the investigation of relationships between theoretical and empirical variables. Results from non-parametric tests and linear regression supported the hypothesis that physiological patterns and their evolution are suitable tools to directly and reliably assess learners' attention. Intelligent learning systems can greatly benefit from using these results to enhance and adapt their interventions.