<p>What makes an answer a good explanation cannot be only that it is correct in the light of accepted school knowledge; it must be framed by the knowledge structure of the discipline (its paradigm), which calls for discipline-specific assessment criteria. In the currently taught biology paradigm, good answers should provide causal explanations of underlying mechanisms. However, influences from other paradigms (including popularized medicine), school tradition, and psychological construals often lead to simple descriptions or teleological, anthropic answers. Assessing answers that contain no errors yet fall short of disciplinary explanatory expectations remains a challenge in both education and research. We extend previous work by elaborating a new, paradigm-based, domain-independent method to assess student text answers for causal mechanistic explanations, expressed in terms (M/CM scores), referred to as paradigmicity. The method can generate fine-grained, just-in-time evidence—informing teacher intervention, improving design, and supporting comparisons across learning designs. To explore its applicability to generate usable evidence for learning and design, we applied the method to a multi-cohort dataset of student responses. Mapping students’ paradigmicity progress onto reference learning goals yielded fine-grained insights into how understanding developed during learning, and revealed areas where explanatory reasoning progressed unevenly. The analysis also opened new research avenues regarding the causes of persistent difficulties and equity challenges in helping all students master the explanatory power of scientific reasoning for understanding the living world. We discuss potential applications for iterative design improvement, for grounding formative assessment, and for adapting the approach across biology domains and to other scientific disciplines.</p>