This study investigates relationships between self-efficacy, self-regulated learning strategy use and academic performance. Participants were 96 undergraduate students working on projects with three subtasks (idea generation task, methodical task and data collection) in a blended learning environment. Task self-efficacy was measured with self-reports administered during each subtask. Learning strategies were assessed by counting each instance of strategy use as it occurred in peer-to-peer conversations typed into a computer software system. Results showed that for each subtask, learners with higher task self-efficacy had higher task performance. Those who used more learning strategies on each subtask also had higher performance. In turn, high performance was associated with high self-efficacy on subsequent subtasks. Surprisingly, results showed that task self-efficacy and learning strategy use were not significantly related during any subtask. Overall, results imply that task self-efficacy, learning strategy use and past performance are important predictors of task performance.