We propose an e-learning scenario where eye tracking is exploited to get valuable data about user behavior. What we look at -- as well as how we do that -- can in fact be used to improve the learning process, revealing information which would otherwise remain hidden. The prototype system we are developing at the University of Pavia takes into account both the user's "emotional status" and the way learning activities are carried out, employing these data to adapt content presentation in real-time. Keywords E-learning, eye tracking, eye communication, emotion recognition, multimodal interaction, perceptive interfaces. ACM Classification Keywords