Online learning platforms offer students the option of sharing content. They have become common tools in many universities over the last 10 years. But there is little information about how content spreads in the classroom, ie, how information cascades appear and evolve and what factors are relevant for the formation of cascades. This work analyses information cascades in the classroom, bringing new insights on student learning: students do not share much content, they prefer to share the content they find themselves as opposed to professor-given content, they share more data towards the end of the course and they do it in bursts. The paper also reveals different behaviour by high-performing students: their interactions are distributed more evenly over the term, their behaviour is more stable and they tend to share documents faster than low-performing students. Documents with high information tend to be less shared. Documents with fewer well-known entities are also shared fewer times. Paradoxically, high-performing students exchange more documents with high information, compared to mid- and low-performing students.