Peer-assisted learning involves having students teach and learn from each other. This can include students from a higher year teaching those in lower ones (near-peer) or students at the same level teaching each other (peer-to-peer). This approach is thought to support learning in two ways. First, students might be more likely to explain ideas in simpler and more direct ways than educators. Educators can often succumb to the curse of knowledge and forget how challenging it was to learn certain information, whereas students are much more familiar with the challenges of learning and so are more appreciative of their peers’ needs. Second, learning from peers is one mechanism thought to drive learning outcomes, but so too is teaching peers. Teaching others requires students to reflect more deeply on what they’ve learned including what they understand and what they don’t. As such, reflecting, preparing, and delivering learning is another mechanism thought to drive learning outcomes as a result of peer-assisted learning.
The most reliable evidence on peer-assisted learning shows that students are just as good as educators when it comes to teaching [1]. The impact of students as teachers is small (➕➕) and not significant, meaning they have the same effect on learning as educations. In another study of lesser quality [2], students-teaching-student had a medium effect on student learning (➕➕➕➕). The effect increased when this approach was used with clinical (➕➕➕➕➕) as opposed to pre-clinical students (➕➕➕), with the latter not being statistically significant. It also increased when used for practical skills (➕➕➕➕➕) compared to theoretical knowledge (➖/➕) and when learning was assessed +4 weeks after peer-assisted learning was used (➕➕➕➕➕) compared to less than 4 weeks (➕). The latter was not statistically significant.
This summary is informed directly by four meta-analyses and supported by an additional two systematic reviews. The first paper [2] that informs this summary is of modest quality (➕➕). It included 27 randomised-controlled trials, but reported a considerably high heterogeneity (93%) and a likely publication bias, and only considered medical students. The second study [3] was of low quality (➕). It included 11 studies that used a pre-post research design. Again, heterogeneity was high (98%) and the primary studies were exclusive to medical students, and no assessment of publication bias was reported. While the effect size was large (ES = 1.26), the confidence interval around it was wide (.58, 1.94) raising questions about the precision of the findings. The third study [1] was of modest quality (➕➕➕). It included 13 randomised-controlled trials. Initial heterogeneity was high, but, when one primary outlier study was removed, the heterogeneity was reduced, but was still high. Removing this study also reduced the effect size and resulted in a non-significant main effect. That is, peer-assisted learning was no better than teacher-led instruction. Analyses revealed limited risk of publication bias for this meta-analysis, but, as before, the sample was restricted to solely medical students. The fourth study [4] was of poor quality (➕). It included a variety of methods (not just randomised-controlled trials) and, again, was exclusive to medical students. No publication bias or analysis of heterogeneity were reported, but the number of studies and width of the confidence intervals were within acceptable limits.
