Retrieval practice: immediate fluency and later retention
Compare study conditions across delays and distinguish immediate performance from durable retrieval.
Question and prediction
Can the condition that looks best immediately perform worse after a delay?
Predict whether easy rereading and successful delayed recall must produce the same ranking.
Mechanism and evidence
Retrieval practice requires bringing information to mind rather than only seeing it again. Retrieval can strengthen later access, but its effects depend on success, feedback, spacing, material and the final test. Repeated study can increase immediate familiarity and confidence without producing the same delayed performance. These are empirical comparisons, not a rule that reading or explanation has no value.
Roediger and Karpicke’s 2006 experiments compared study and retrieval conditions using prose material. Repeated study favored an immediate test, while prior testing produced an advantage on delayed retention measures. The key inference concerns an interaction between learning condition and delay. The finding does not mean every test format is equally useful or that performance transfers automatically to any new task.
In the fictional table, restudy scores 18/20 immediately and 10/20 after a week; retrieval practice scores 15/20 immediately and 14/20 later. The ranking reverses. Immediate scores and final scores come from separately assigned groups in this case, so the immediate test does not itself become extra practice for the delayed group. That design detail is important.
An unfamiliar application should require explaining or using the material, not merely repeating the original wording. A retention benefit on practiced facts is not yet evidence of broad understanding. Specify equal total time, comparable initial learning, feedback exposure and the exact outcome. The task is to audit an evidence-based learning claim, not prescribe one technique for every learner and subject.
Inspect the relationships
| Practice condition | Immediate | One week |
|---|
| Restudy | 18/20 | 10/20 |
| Retrieval | 15/20 | 14/20 |
| Retrieval minus restudy | −3 | +4 |
Worked example
Immediate retrieval-minus-restudy contrast=15−18=−3 items; delayed contrast=14−10=+4. The contrast changes by 7 items across delay. These fictional data illustrate a ranking reversal, not the original study’s scores or proof of broad transfer.
Your investigation
Plot the two condition means at each delay and write the contrast beside each point. Use the authentic article card to explain why delay belongs in the claim. Design an unfamiliar application question and a scoring rule separating factual retention from explanatory transfer.
Unfamiliar transfer
A practice test includes answer feedback and a comparison lacks it. What must the interpretation distinguish? Explain your reasoning before checking feedback.
Reading
Roediger and Karpicke (2006): Test-enhanced learning. The records below are original fictional teaching cases, not research results from this source.
Fictional condition-by-delay table Restudy: immediate 18/20, one-week 10/20. Retrieval practice: immediate 15/20, one-week 14/20. Delay groups are separate random groups with equal initial study and total practice time.
Fictional transfer limitation Final tests ask for the practiced facts. No unfamiliar explanation or application task is included.
Authentic evidence and its limits
Published Experiment 1 retention values — Henry L. Roediger III and Jeffrey D. Karpicke (2006). Experiment 1 reported recall of 81% after restudy versus 75% after testing at five minutes; at one week, recall was 42% after restudy versus 56% after testing. Practice recall had no answer feedback.
These are attributed published results, unlike the separately labeled 20-item teaching table. The sample comprised undergraduates, learning condition was within participants and final-test delay was between participants. Percentages describe recalled idea units, not percentage of students passing.
Read the original source. Original LearnAI paraphrase and attributed factual results; no reproduced figure or substantial verbatim passage. Original publication retains its rights.
Original-data task: Calculate testing minus restudy at five minutes and at one week using the published percentages. Explain why these are percentage-point differences in idea-unit recall, not percentages of students who passed. The contrasts are −6 and +14 percentage points; the change in contrast is 20 points.
