Documents in some collections have been categorized automatically into "levels" corresponding to document difficulty, where 1 is the easiest level. The classification was done using a supervised classifier trained on manually "leveled" documents. Difficulty levels should be considered a noisy, experimental feature.