Measuring reading comprehension using eye movements

We investigate eye movement measures and methods for predicting reading comprehension. This builds on previous work on factors affecting reading comprehension, namely perceived familiarity with the text content. We further investigate answer-seeking behavior and present a method for measuring and comparing this behavior. The number of fixations, number of regressions, and total fixation time are an indicator of reading intensity and the intensity of reading is related to comprehension. We show that a feed-forward backpropagation neural network can be used to predict subjective comprehension scores as well as quiz scores. We propose using the degree of answer-seeking behavior to measure how question difficulty and as an implicit measure of how difficult a participant finds a tutorial and quiz. Such information is beneficial to apply in eLearning to create dynamic learning environments that use eye movement to predict implicit question difficulty as well as individual participant difficulty.