Current educational theory emphasizes active, student-centered learning, particularly with personally relevant content. This is supported by some experimental evidence, primarily from university students in controlled settings. However, these studies typically lack realistic classroom contexts and ignore potential serial interaction effects between lessons. This study utilizes the recent introduction of evolution into the UK primary school curriculum to address these limitations. By employing a 2x2 randomized controlled trial design, the study investigates the optimal teaching approach for evolution, considering both teaching style (teacher-centered vs. student-centered) and subject matter focus (human-centered vs. non-human-centered). The researchers hypothesized that student-centered, human-centered learning would be most effective but acknowledged the need to test this hypothesis within the complexities of real-world classroom settings.

The introduction extensively reviews the literature on active vs. passive learning, highlighting the lack of large-scale, in-situ studies, especially those considering sequential lesson interactions. It points to the challenges of conducting such research, including recruitment, teacher compliance, and the need for large sample sizes to detect interaction effects. Existing research mainly focuses on university students, leaving a gap in understanding effective teaching methods for school-age children. The review also addresses the challenges of assessing children’s understanding of evolution, citing the dearth of appropriate quantitative assessment tools for primary-aged children.

The study employed a 2x2 randomized controlled trial (RCT) design with two tranches (test and replicate). Tranche 1 included 1152 students from 17 schools, and Tranche 2 included 1505 students from 28 schools. The four teaching schemes involved four lessons taught in a fixed order. Lessons 1 (variation) and 3 (deep time) were invariant across all schemes. Lesson 2 (selection) was either student-centered (hands-on moth hunting activity) or teacher-centered (PowerPoint presentation). Lesson 4 (homology) was either human-centered (pentadactyl limb) or non-human-centered (trilobites). A quantitative assessment tool, previously developed and validated by the researchers, was used for pre-test, post-test, and retention testing. Teacher training was provided to ensure consistent delivery of the schemes. Qualitative data were collected through teacher and student feedback to gain insights into the effectiveness and engagement aspects of each scheme. Statistical analysis included Wilcoxon rank-sum tests, Kruskal-Wallis tests, LOESS regression, and analysis of interaction effects. Covariates such as gender, age, and teacher-assessed student ability were also considered in multivariate models.

All four schemes significantly improved student understanding of evolution in both tranches, with effect sizes above the implementation threshold (0.4). However, the teacher-centric/non-human-centered scheme (SoW 3) unexpectedly outperformed the others, often being the most or second most effective. This contradicts the initial hypothesis and conventional wisdom favoring student-centered and human-centered approaches. Analysis revealed significant heterogeneity between the schemes, indicating interaction effects between lessons. The interaction effects were highly replicable. Individual comparisons of student-centered vs. teacher-centered methods (Lesson 2) and human-centered vs. non-human-centered methods (Lesson 4) showed weak and inconsistent results. Multivariate analysis, controlling for gender, age, and ability, confirmed SoW 3’s superior performance. Teacher-level analysis revealed that improved teacher confidence was a significant predictor of student performance but wasn't strongly replicated between tranches. School-level analyses did not identify robust predictors.

The findings challenge the prevailing belief that active, human-centered learning is universally optimal. The unexpected success of the teacher-centric, non-human-centered scheme highlights the importance of sequential lesson planning and the potential for interaction effects. The study emphasizes the limitations of parallel testing designs in capturing these interactions. The results suggest that engagement, while important, doesn't necessarily require active, human-centered methods. The use of trilobites, for instance, proved engaging due to their novelty and intrigue. The study notes limitations in generalizability due to sample demographics and school selection bias. Despite these limitations, the strong and replicable findings suggest that SoW 2 and 3 could be valuable resources. The study encourages a reevaluation of learning policies based on dichotomous views of active vs. passive learning and human-centered vs. non-human-centered approaches.

This large-scale, replicated RCT demonstrates the importance of sequential lesson design and reveals the unexpected effectiveness of a teacher-centric, non-human-centered approach to teaching evolution. The results challenge current pedagogical assumptions and call for a more nuanced understanding of effective learning strategies. Future research should explore the underlying mechanisms of the observed interaction effects and investigate the generalizability of these findings across different populations and subject matters. The freely available resources developed for this study provide a valuable tool for teachers.

The study's generalizability may be limited by the predominantly white, European-descent sample from the Southwest of England, and the non-random selection of participating schools. The lack of full randomization in allocating schemes of work and the potential for non-compliance by some teachers, although low, are other limitations. The study also lacks temporal effects due to the simultaneous delivery of the SoWs. Finally, the reliance on teacher assessment of student ability introduces potential bias.