In ReproLab I try to reproduce or replicate findings from research articles.
This reproduction is for:
DeCaro, M. S., Bego, C. R., Thompson, A. K., & Patrick, R. J. (2026). Considering Cognitive Load When Exploring Before Instruction: The Potential Role of Element Interactivity. Educational Psychology Review, 38(1), 127. https://doi.org/10.1007/s10648-026-10223-7
I tried to reproduce this article. I used R for this but got some support for Claude for formatting the concluding sentences.
The paper’s main results reproduce from the raw data, with the same samples (N = 388 and 795) and cell means. Checking F statistics and Table 2 had a few discrepancies but it doesn’t nominally change the conclusions.
One thing -and kudos again for making the data available anyway- I was interested in is whether findings changed if you would take into account clustering. Note that conditions were assigned to sections and not students. In my view, this means the data are not independent, and you have to take into account clustering with a multilevel model. When I did a quick test on that I would say that headline findings change from statistically significant to inconclusive. In my opinion, the data look like ordinary differences between sections. As there are so few sections, the power actually seems very low. If there really are no differences, then that would actually be more surprising, given the widely communicated (and sometimes exaggerated) conclusions regarding some-approaches-first and task complexity.
Nice study, for the rest, and happy to be corrected.
