Many folks (myself included) worry about "pattern matching" in spaced repetition—that they're learning to parrot a response in a narrow context. This experiment tried varying question text to combat that and, surprisingly, got a null result. andymatuschak.org/files/papers/B
Conversation
Replying to
Worth taking with a grain of salt: the learning scenario here is contrived, may not extend to authentic situations.
I should mention, too, the primary result: retrieval practice appears to effectively support transfer of learning, even to distant contexts.
2
13
Replying to
Two thoughts! 1) Maybe we can think of a learned concept’s utility as a function of “connectedness to other things” and “depth”. The fact that the result was null might be because the depth from one method balanced the connectedness from the other method.
1
1
2) Im wondering if asking the “same question” in different ways is really affecting “connectedness” or “depth” more. I guess it depends on how different the prompts are. E.g. “design a shirt” and “design a car” vs. “when did Einstein die” and “where did Einstein die”
1
Show replies
Replying to
Side tracking: it was delightful to see you link to a PDF that contains your annotations, hosted on your own domain, no less! It gives people a raw glimpse into what you took notice while reading an original text, which I think is valuable
3
Replying to
I worry that the narrow pattern matching primarily hurts the "recognition" element of far transfer, i.e. that you match the knowledge to the card rather than to a broader concept of situations in which it'd be helpful. As the authors state, they didn't test that :-(
2
3
Thanks a lot for showing the paper, though! Fascinating read!