Abstract
Abstract
Multiple external representations (MERs) are central to learning abstract scientific concepts, yet most evidence on their effective use comes from teacher-guided settings. As online and blended learning environments expand, how students navigate and make sense of MERs without direct instructor support remains underexplored. This study aimed to examine how undergraduate students engage with, interpret, and coordinate static images, videos, and interactive simulations during self-regulated, inquiry-based online science learning, and to identify the conditions under which these representations support or constrain conceptual understanding. Grounded in the Predict–Observe–Explain–Evaluate (POEE) scaffolding framework, the study adopted a qualitative data-dominant design involving 30 first-year science students at a large Australian university, who completed two online learning modules (
Heat
and
Phase Change
). Data were drawn from real-time observations, screen recordings, and stimulated recall interviews. Findings indicate that MERs supported students in correcting misconceptions (e.g., molecular continuity across phase changes), strengthening spatial reasoning, and bridging everyday perceptions with scientific explanations. Under guided scaffolding, students displayed representational fluency in extracting meaning across visualisations. However, overconfidence with familiar representations often produced surface-level learning, while open-ended exploration without sufficient guidance led to cognitive overload and fragmented understanding. Simulations encouraged molecular-level reasoning but required scaffolding to manage complexity, whereas videos conveyed spatial depth and contextualised molecular behaviour more effectively. Static images functioned as accessible visual anchors that clarified abstract processes. The study demonstrates that the effectiveness of MERs in online science learning depends on instructional scaffolding that supports self-regulation, mitigates overconfidence, and balances interactivity with clarity. Implications are drawn for designing online environments that support deeper engagement with MERs in independent learning, and for understanding how prediction and metacognitive monitoring operate in representation-rich learning.