Position

What is measured, what is not, and where practice runs ahead of the evidence.

Known (measured)

  • Retrieval and spacing are the best-supported study techniques, and gains grow after a delay.
  • Worked examples help novices; guidance should fade as expertise grows.
  • Constructing a map beats studying one.
  • Words belong beside the picture or code they explain.
  • Visual cues (bold, headings, arrows) help, most for low prior knowledge.
  • Case comparison transfers better than studying cases one by one, if the reader is prompted.

Known with caveat

  • Skeletal notes: positive, from older randomised experiments.
  • Pre-training on key terms: large effect, mostly from one lab.
  • Hypertext links: load is reasoned from theory and a few small studies.
  • Mastery gating: older meta-analysis, costs time.

Not known

  • No study tests a numeric section-length threshold such as 80 statements.
  • No study tests the depth of layering across several linked notes.
  • Definition-first versus example-first has no deciding study.
  • Atomic notes, maps of content and progressive summarisation have no controlled study.
  • Spaced-repetition schedulers are benchmarked on recall prediction, not on exam scores.
  • Most effects come from short materials and near transfer.

What this repo adds

  • It restates each field result as a general claim about readers, with its grade.
  • It keeps measured claims apart from conventions, so a rule can be traced to its level of evidence.
  • It does not claim that no one has tested a rule, only that the review below found no test.

Practice against the evidence

  • Practice runs ahead: schedulers publish a public benchmark, where neural models score lower loss than FSRS-6, but it predicts recall, not learning.
  • The project's own 20-30% fewer reviews figure (below the quality bar: an example, not evidence) is a simulation, so it grades nothing.
  • Company efficacy pages are owners' claims; the only academic RCT found gave small gains.
  • Practice ignores the evidence: most educators surveyed believe in learning styles, and some keep using them when told.
  • The standard relies on no style, so nothing changes; readers may still expect style-based design.