The purpose of this study was to investigate how elementary school student reader to take information from text and use background knowledge when generating questions by reading information composed of multimodal text in reading to learn. To this end, when reader generated question Type with reading resource reference, the path of the reader’s attention movements in and out of the text was analyzed using an eye tracker. And the generated questions were categorized. this study results are as follows. First, A generated question within the text (Type①) was divided into the case where the question was created from the text and the case where the question was created by integrating the text and the picture. Second, generated questions (Type②) according to the shift of attention from background knowledge to text were discriminated according to the significance of meaning expansion. Third, generated question according to the shift of attention from the text to the phenomenon (Type③) was created while maintaining the possibility of additional text search and a critical perspective. The implications for reading education based on this study results are as follows. First, an individualized instructional design should be made considering that the level of generated questions may vary depending on the level of reading comprehension. Second, even in textbooks, it is necessary to motivate the expansion of meaning in the process of generating questions around the topic and finding answers from the outside. Third, at the level of textbook activity, the teacher and peers’ questions like priming water should be provided as scaffolding.