The article presents a description and results of the experimental study doing in order to determine variability of expository text comprehension by learners. The materials of the investigation are expository texts from geography textbooks for middle school (7th grade). The texts differ in their emotiogenic degree in terms of interest evoking. The experimental method of the investigation is the keywords method which allows to determine a relevance degree of text idea units for a recipient. The subjects of the experimental study are 40 schoolchildren (7th grade) and 26 university students of geography department. The article contains a comparison and an analysis of data collected in the experimental study (individual keywords sets). The comparison evidences that expository text comprehension by learners is variable to a high degree. The analysis of the experimental data allows to conclude that variability of expository text comprehension is due to two main reasons. Firstly, learners do not structure the content of expository texts during reading. Secondly, they consider a text idea they must to remember or know to be of higher value. Further, these comprehension features are due to undeveloped structure strategy of expository text comprehension in learners. In addition, the article also contains a description of some expository text characteristics which may affect variability of expository text comprehension by learners.