Nearly 6,000 science questions collected from five different web-based, TV-based and school-based sources were rigorously analyzed in order to identify profiles of K-12 students' interest in science, and how these profiles change with age. The questions were analyzed according to their topic, thinking level, motivation for and level of autonomy in raising the question, the object of interest and its magnitude, and psychological distance of the object in question from the asker. Characteristics of the asker, such as gender, grade level, and country of origin were also considered, alongside characteristics of the data source, such as language, setting (Internet, school, TV), and the potential science-attentiveness of the users. Six meta-clusters of children's and adolescents' interest in science were identified using cluster analysis of their self-generated science questions. A developmental shift in interest from non-classical to classical school science subjects was noted. Other age-related developments, such as an increase in thinking level as reflected by the questions, a decrease in organization level and the psychological distance of the object in question with age were also explored. Advantages and shortcomings of web-based data collection for educational research are discussed, as are the implications of the results obtained using this methodology for formal science education. (Contains 2 tables, 4 figures, and 9 notes.)