Bullet point plan for The Rokers

• Introduction

  • Introduce testing effect

  • What about when test performance is poor?

  • Description of Potts and Shanks results

  • Potts and Shanks may be due to familiarity, not associative learning

  • Cued recall shows associative learning

  • Follow up Potts and Shanks with cued recall

• Experiment

  • Method as per protcol document

  • Results as per presentation

• Discussion

  • Summary of findings

    • No effect in either condition

    • Bayesian evidence for null in hard text, but floor effect

    • Performance better in easy test, but inconclusive Bayes.

  • Relation to literature / Implications

    • Different to Potts and Shanks results

    • This likely due to multi-choice vs. cued recall testing

    • Perhaps errorful generation boosts familiarity, not associative learning.

    • In applied contexts, associative learning is the critical aspect.

  • Limitations / extensions

    • Weakness of evidence: Floor effect / Bayesian inconclusive

    • At d = 0.34 effect size (estimated from easy test), sample size should be 55+ for adequate power.

    • Run a large-scale replication of easy test

    • Presentation time of word-definition pairs not matched between conditions

      • Also true from Potts and Shanks, but less of a problem for them than for us, given their results.

      • Run a version that equates pair presentation time between conditions

  • Brief conclusion