The units are only "comparable across studies" if there weren't any mishaps (eg, clinical or methodological heterogeneity). If there's clinical heterogeneity, then we're probably comparing apples to oranges (ie, either participants, interventions, or outcomes are different among studies). If there's methodological heterogeneity, then that means there's a difference in study design

Paper gives surprisingly good overview of models of learning within the cognitive sciences up to 2008. Attempts to dispel myths and summarize the literature on multimodal learning. Link to paper on Semantic Scholar

A white paper written by Metiri Group commissioned by Cisco in 2008. I came here to fact check some claims on this YT video about a "Feynman Technique 2.0".

The claims were that

1. direct hands-on experience in unimodal learning is (on average) inferior to multi-modal learning that wasn't hand-on. viz., for "basic concepts", a more abstract learning model is better

2. "Once you get into higher-order concepts then hand-on experience is better"

Page 13 was displayed while making these claims.

These claims still need to be verified.