Modeling and Measuring Tree-Reading Skills in Undergraduate and Graduate Students

Evolutionary trees are central to learning about evolutionary processes, yet students at all educational levels struggle to read and interpret them. The synthetic tree-reading model (STREAM), based on published and not yet empirically tested models, was tested to determine whether the assumed hierarchy of the model could be substantiated and how far students’ skills could be distinguished empirically. We developed a tree-reading test instrument based on STREAM and assessed it with 592 undergraduate and graduate biology students. Following item response theory, we conducted a dimensional analysis and evaluated item difficulty. Investigating item difficulty and the resulting Wright map showed that skill levels displayed a broad scatter of overlapping item difficulty. Furthermore, the skill level assumed easiest was actually the third most difficult. No conclusive evidence of the hierarchical nature of the model was obtained. Dimensional analysis showed that a five-dimensional model outperformed all other reasonable models, corroborating that the skills could be arranged in empirically differentiable groups. Consequently, we revised the STREAM by discarding the hierarchical organization, using a five-dimensional organization instead. Comparison of the revised STREAM with another recently published approach showed that, although these two instruments have a different focus, they are supplemental approaches that show comparable results.