This study explores three assumptions about learning paths, including: (A1) learning progresses in a linear fashion, (A2) all or most learners follow the same learning path, and (A3) teacher-driven sequences, slides, activities, and interactions effectively predict learning paths. We applied previously developed methods to assess students’ learning progressions—explanations of biological phenomena such as allergies—to a large dataset of traces collected from multiple small cohorts of higher secondary school students. Our findings challenge all three assumptions. Learning did not follow a linear path but progressed in a mosaic pattern, varying considerably across cohorts and learners. Further analysis revealed that some areas of the learning goals showed high variability, while others were consistently attained early or late. Added analysis suggested that learning paths were influenced by both internal and external factors, not only by teacher activities, or knowledge structure in textbooks. These results emphasize the need for real-time, fine-grained assessment during learning to inform teaching interventions, to guide students efficiently towards the learning goals, for example helping them identify gaps in explanations and refine the causal explanations required. These findings contribute to discussions on learning heterogeneity, developmental constraints, and help identify “slow spots” that are difficult for learners.