The answer does depend in part on the nature of the scores, as your query correctly implies.
However, it will also depend on what specific research question one is trying to address, how the students were selected and the data gathered, and the technical quality (e.g., reliability, validity) of the scores being used.
You can use data mining tools such as classification and clustering to group the learners based on thier performance and determine the levels of variations within the clusters by using x-means clustering.