Self-assessment and task selection are important self-regulated learning skills for secondary school students. More specifically, selecting new tasks based on self-assessments is very important for them, because teachers are not always present or able to select tasks for them individually. However, little is known about the processes underlying these self-regulated learning skills, and thus no guidelines exist for teaching self-assessment and the selection of subsequent learning-tasks. We propose a model for self-regulated learning-task selection (SRLTS) which represents a possible pathway for the task-selection process, and which students could use as a norm when making task selections. The model could help students to decide what possible new tasks might be suitable for their current skill level, based on self-assessments. The aim of this study is to evaluate to what extent secondary school students select learning tasks according to this model, and whether they use self-assessments to this end. Secondary school students (N = 15) selected learning tasks in the domain of genetics from a structured task database. The tasks varied in difficulty and amount of support provided (i.e., completion problems vs. traditional problems). We used eye tracking, performance estimates, estimates of mental effort, judgments of learning, and open questions to gain more insight in what students focus on and think about when selecting a task. Results suggest that students roughly follow the SRLTS model, but they base their decisions on inaccurate self-assessments. This implies that students might benefit from self-assessment and task-selection advice, which could provide feedback on self-assessments and stimulate students to use self-assessment information as input for task selection in the way the model prescribes to optimize their learning.