A machine learning method was investigated to analyze how syntax affects reading difficulty, first categorising graded reader texts into parts of speech, then using a sliding window method to count different patterns of word types. Next we used clustering to check whether the features could be classified by difficulty level.

Various activities can be conducted in class to maximize fluency building opportunities with graded readers and regular course target language. Speaking and writing activities based on thematically connected graded readers, and reading aloud activities constructed from target language can promote both wide and narrow fluency development.