> The most notable limitation of interleaving, however, is that its benefits are likely limited to difficult discriminations. The learning tasks in the studies cited in this review were obviously chosen to produce discrimination errors (e.g., the artist paintings in Figure 2), and discriminations of this subtlety are rarely encountered in the classroom. If these studies had instead required subjects to distinguish between dissimilar concepts or terms, the size of the interleaving effects would have almost certainly been smaller. This boundary condition sharply constrains the utility of interleaving.

I thought this paragraph was interesting. Most interleaving research (that I’ve read) is within a single subject, and not across multiple subjects. The SRS/Anki proponents frequently recommend having a single “everything” deck, with interleaving being the main justification for that. This paragraph seems to indicate that having such widely different questions would be of limited benefit since the questions aren’t “difficult discriminations”.

I would love to see research that explicitly studies the idea of interleaving across different subjects.