Self-regulated learning (SRL) is essential for academic success yet few studies have explored how individual-level variables (e.g., prior academic achievement, self-reported SRL skills) relate to both adaptive (positive) and maladaptive (negative) SRL behaviors across different types of tasks. This study investigated the extent to which self-reported SRL skills and prior academic achievement predict both positive and negative SRL behaviors captured through think-aloud protocols, as well as task performance, across three cognitively distinct academic tasks (reading, oral analysis, and written analysis) within a repeated-measures design. Results showed that higher self-reported positive SRL and prior academic achievement predicted greater use of positive SRL strategies and better performance, whereas negative SRL behaviors appeared more sensitive to task demands than to individual traits. These findings highlight the value of distinguishing between positive and negative SRL and of integrating self-report and process data to better understand the dynamics of SRL and inform targeted educational interventions. Educational relevance statement: This paper is educationally relevant because it shows that learning strategies should adapt to different tasks and stages providing evidence that self-regulated learning is context-dependent and dynamic, varying across tasks and individuals. Its findings inform the design of more responsive pedagogical interventions and valid assessment tools that capture students’ regulation processes in real learning contexts, emphasizing the role of prior achievement and strategy use in adaptive regulation.