Against

The evidence and the practice that cut against the learning principles, one case per section.

Learning styles

  • Pashler 2008 (R): no adequate evidence that matching instruction to a learner's style improves learning.
  • The meshing hypothesis needs a crossover interaction; the reviewed studies did not show it.
  • Pictures help every learner (Noetel 2022), so a visual design needs no style argument.

Controversies and failed replications

  • Laptop deficit: Mueller 2014 found worse conceptual recall with typing; Urry 2021 (preregistered replication) found negligible effects.
  • Testing effect and complexity: van Gog 2015 argue the effect shrinks with element interactivity; Karpicke and Aue 2015 reply that a small effect remains.
  • Worked examples vs testing: for a new procedure, studying examples wins on an immediate test (Yeo 2019). A week later testing equals it with varied problems, and wins only with problems identical to those practised.
  • Expanding gaps: Latimier 2021 find no advantage over uniform gaps.
  • Cheat sheets: Dickson 2005 find no exam gain from a prepared sheet.
  • Concreteness fading: Fyfe 2014 review is positive; problem-first wins in school maths (Sinha 2021).
  • Own highlighting: comprehension gain not significant; instructor highlighting works (Ponce 2022).
  • Erroneous examples: small (g = 0.14).
  • Self-explanation prompts added to worked examples lowered the gain in one pool (Barbieri 2023).

Weak or absent evidence

  • Atomic notes and linked-note systems: one small trial, bundled with spaced repetition (Malashenko et al. 2023).
  • Cornell notes: evidence is inconclusive.
  • Animation over static pictures: d = 0.37 and 0.23 with high heterogeneity; benefits depend on what is shown.
  • Hypertext restrictions: the best data are from small samples.

Limits of the evidence base

  • Short lab materials dominate; classroom courses are fewer.
  • Several effect sizes were read in summaries (marked secondary).
  • Surface gains exceed deep gains, so recall tools do not replace problem practice.

Practice that contradicts the evidence

  • Belief in learning styles is 89.1% in a review of 37 samples (Newton 2020) and persists after the evidence is shown (Newton 2017: 32% of 114 UK academics).
  • Edtech efficacy pages report association or engagement; Oreopoulos 2026 finds 0.06 to 0.08 SD per year, small.