Desirable difficulties: retrieval vs rereading

You have just reread the chapter. The sentences flowed. The page looked familiar. The immediate feeling is one of mastery. The next day, faced with an open question or a case with no options, the content disappears. This gap between reading fluency and retrieval performance is the core of desirable difficulties: strategies that make study more costly in the short term and more stable in the long term.

NEXO is a daily medical diagnosis game: one clinical case a day, six progressive clues, and ICD-10 proximity feedback across five levels.

This post replaces the original topic "illusion of competence: thinking you know because you reread". The angle of deceptive fluency already touches on confidence calibration.

It also touches on metacognition and on retention. The focus here is the operational contrast between restudy and retrieval practice, within the framework of desirable difficulties.

The reader of this text is a student or a professional in training in any country. The studies cited come from international designs; the application fits the routine of a local exam, internship or residency, without centering the narrative on a single education system.

Key points

• Fluency in rereading creates an illusion of learning: the text seems familiar, but memory has not yet been forced to produce the answer.
• Retrieval practice beats restudy in medium and long term retention across several undergraduate and residency designs.
• Spacing (distributed practice) and active retrieval belong to the family of desirable difficulties: they cost more effort now and pay off later.
• Health students underuse high-utility strategies and overestimate rereading; subjective preference and objective effectiveness often diverge.
• In NEXO, closing the case with clues and only then reading the explanation is an act of retrieval; rereading the feedback completes the cycle without taking the place of the test.

What desirable difficulties are

Bjork described desirable difficulties as learning conditions that seem to get in the way at the moment (more effort, more apparent errors, less fluency) and, despite that, improve retention and transfer. The opposite is learning that seems easy: quick rereading, abundant highlighting, lectures watched at a comfortable speed, the feeling of "I already know this" when you close the PDF.

Three ideas organize the concept. First, immediate performance is not learning: getting it right right after reading measures familiarity, not future retention. Second, effort is interpreted as a bad sign: many students read the cost of retrieval as proof that they "are not learning", when effort is the mechanism. Third, conditions that vary and space out (mixing topics, spacing reviews, requiring production of the answer) increase useful difficulty.

In anatomy, a study compared repeated reading, partial generation and free recall: the free recall condition, the most demanding, was superior in immediate recall and after one week (Dobson and Linderholm, 2015). The pattern aligns with the idea: the more the session forces production, the more the content stays retrievable.

This does not mean making study chaotic. A desirable difficulty is not confusion without feedback. It is useful cognitive cost with correction afterward. Without feedback, the error consolidates; with feedback after the attempt, the error becomes data for the next session. Cognitive load theory helps to calibrate: desirable difficulty increases germane load (the work that builds schema), not useless extraneous load (bad interface, confusing material, hunger, sleep).

Why rereading deceives

In rereading, the text supplies all the cues. Recognizing the ready-made sentence is easy when it is in front of your eyes. Producing the same association from a vignette with data in a different order is another task. The brain confuses the two and calls both of them "I know this".

Piza and colleagues mapped awareness and use of evidence-based strategies among students and faculty in health professions: there is a gap between what the literature recommends and what routine practice favors (Piza et al., 2019). Rereading and highlighting remain attractive because they produce visible progress (marked pages) and low immediate aversion. The cost appears later, in the exam or on the ward, when no one is holding the book open.

A second mechanism is perceptual fluency. After the second or third pass, processing becomes faster. That speed is interpreted as mastery. In the assessment, without the text, the fluency disappears and only retrieval remains. The error notebook helps when it records retrieval failures, not recognition failures on the page.

Typical signs of fluency without mastery: knowing the chapter with the book open and getting stuck with it closed; recognizing the vignette only when the sentence mirrors the material; avoiding blank questions because you "haven't reviewed enough yet", when the test is the review.

Ziegler and colleagues described, in medical students, the use and perception of high, medium and low utility learning techniques (BMC Med Educ, 2022). The recurring pattern is knowing vaguely that "doing questions is good" and, in the practice of a full week, going back to rereading because it reduces immediate anxiety. The anxiety of the blank page is exactly the price of desirable difficulty.

Retrieval practice in clinical language

Retrieval practice is any act of trying to retrieve information from memory before consulting the source. In Medicine this includes short-answer questions without looking at the summary; explaining the mechanism out loud (teach-back to yourself); drawing the management algorithm from a case and only then checking; closing hypotheses with limited clues; using flashcards with the front showing and the back hidden.

Karpicke and Roediger showed the critical importance of retrieval for learning: testing does not only measure, it consolidates (Science, 2008). Karpicke and Blunt compared retrieval with elaborative concept mapping and found an advantage for retrieval practice (Science, 2011). Maps and diagrams organize; they do not replace the attempt to remember without the map in front of you.

In the clinical clerkship, retrieval appears when the preceptor asks for the next step before opening the full chart, or when you summarize the case to a colleague without reading the progress notes. Case-based learning gains strength when the case requires production, not just reading the already solved vignette. Useful feedback in cases completes the cycle: the attempt comes first, the explanation after.

There are gradients of demand. Recognition (multiple choice) is lighter retrieval. Cued recall (fill in the blank, short answer) weighs more. Free recall (blank sheet: "write everything you remember about adrenal insufficiency") weighs even more. The review in health professions notes that more demanding forms tend to correlate with larger gains, although the ideal design depends on the content and the interval until the exam (2024 review).

What the evidence shows

A systematic review in health professions education gathered studies of distributed practice and retrieval practice. Most experiments showed a positive effect compared with massing, rereading, usual lecture or no intervention; retrieval alone was the most frequent design (PMID 37615780). Desirable difficulties are not confined to the psychology lab; they transfer to health contexts.

In residency, Larsen, Butler and Roediger randomized didactic conference topics to repeated testing with feedback versus repeated study of the same content. After more than six months, the repeated testing group kept performance superior to restudy (Larsen et al., 2009). The long interval matters: the advantage of retrieval usually appears when the final assessment is not immediate.

With electronic flashcards in clinical nephrology, Schmidmaier and colleagues found an advantage of retesting over restudying for one-week retention; at six months the difference disappeared and overall recall was poor (Schmidmaier et al., 2011). Practical reading: retrieval helps, but it is no magic without continuous spacing and without a volume of reviews over the semester. Students' predictions of their own performance also did not match actual performance: another reminder that the post-study impression is a biased judge.

Strategy	Immediate feeling	Typical effect on retention	Sensible use
Rereading	High fluency	Low to moderate if done alone	Only after an attempt at recall, to correct
Highlighting / passive summary	Visible progress	Limited without retrieval	Generate cues to test later
Retrieval (testing)	Effort, errors	Better over medium and long intervals	Core of the session
Spacing	"I forgot a little"	Better than massing	Review at growing intervals
Interleaving	Uncomfortable mixture	Helps discrimination	Mix similar diagnoses

Eva recalls that teaching clinical reasoning must make the process visible, not only the product (Eva, 2005). Retrieval of mechanisms and of next steps is a way to make the process visible to yourself. Epstein, on assessment, reinforces that different formats capture different competencies (Epstein, 2007): studying only in the exam format helps in the short term; mixing retrieval formats helps transfer.

Spacing and massing

Distributed practice is spreading study over time. Massing is concentrating everything on the eve. A health review confirms the benefit of spacing across several designs (PMID 37615780). The student feels massing as productive because familiarity rises fast the night before; the curve drops fast after the exam.

The useful combination is to space the sessions and, within each session, require retrieval. Spacing only rereadings spreads a weak method. Testing only on the same day in a massed block loses the temporal reinforcement. In semester practice, set short blocks with a production goal; leave an interval before revisiting the topic; on revisiting, start with the test, not the PDF; use errors as a queue: what failed comes back before what was answered with ease.

Progress tracking makes sense when it measures correct answers in spaced retrieval, not the number of yellow pages. If your spreadsheet only counts hours, it does not distinguish desirable difficulty from busyness.

How to build a study session

A simple protocol, aligned with the evidence, fits in about one hour. Suggested steps:

1. Activation (3 to 5 min): free recall without material.
2. Test (15 to 25 min): questions, cases or flashcards in recall mode.
3. Feedback (10 min): answer key only after the attempt; classify the error.
4. Surgical rereading (5 to 10 min): only the passages linked to the errors.
5. Closing (2 min): mark the next date to revisit the topic.

This format reverses the classic order "read everything, then if there is time left do questions". The question comes first. Reading becomes correction. For clinical work, replace the question with a case: read only the presentation, write hypotheses and the next step, and only then open the exams and the discussion. Training the differential with cues is retrieval applied to reasoning, not only to facts.

If the session is in a group, the risk is turning retrieval into a collective lecture. Agree on this: each person writes the answer in silence; only then do you discuss. That way the group does not steal individual retrieval. (The post on peer teaching goes deeper into the collaborative design; here the point is to protect the moment of recall.)

Common mistakes when applying retrieval

Turning retrieval into a theater of productivity is easy. Looking at the back too early turns a flashcard into disguised rereading. Staying only with easy multiple choice favors recognition. Training without feedback consolidates false confidence. Testing once on the eve does not replicate Larsen's design with spaced repetitions. Giving up at the discomfort is falling into the misinterpreted-effort hypothesis: fleeing the method that "seems" inefficient (Cogn Psychol, 2019).

Another mistake: using retrieval only for isolated facts and never for decisions. In Medicine, recalling the name of the disease without recalling what to do now leaves a gap at the "know how" rung of Miller's ladder (Miller, 1990). Mix factual items and management items. Alternate "what is the mechanism?" with "what is the next step?" and "what to rule out first?".

Finally, do not confuse retrieval with punishment. If the error rate is so high that you cannot correct it carefully, the study unit is too large. Break up the chapter, add cues, shorten the session. Desirable difficulty respects the zone in which the error still teaches.

On the eve, a short block of retrieval of items already worked on yields more than rereading the textbook for the first time. Without prior foundation, retrieval only reveals the gap: prioritize a minimal core, test, correct. Teachers can open the class with one minute of written questioning before the presentation. If you use NEXO with a question bank, alternate cases with clues and exam items, always writing the why before the answer key.

How NEXO fits in

NEXO offers fictional cases with a real clinical presentation and up to six clues. The move aligned with desirable difficulties is to open the case of the day, answer with the available clues without looking up the text first, record the hypothesis and the why in one sentence, only then read the explanation, and the next day try to retell the case without opening the history.

This does not replace your exam's question bank. It complements it with retrieval under incomplete information, close to clinical reasoning.

Programs such as NEXO Researcher go deeper into method.

NEXO Curator goes deeper into curation.

Limits of use, where applicable, are in Economy.

Active learning describes the pedagogical umbrella; this post details the retrieval mechanism within that umbrella.

Relation to neighboring posts: retention explains forgetting; metacognition teaches monitoring; calibration looks at confidence in the diagnostic act; here the study method makes retention and metacognition possible without repeating those h2s.

An example week (no rigid recipe)

Alternate free recall, management questions, cases with clues and flashcards in recall; retest the errors on the following days. If you have a shift, shorten the blocks, do not abandon the test-then-text order. Rereading stays on the margin, only to correct.

How to talk with your group without sabotaging recall

Agree on one minute of silent writing before any discussion. Then compare hypotheses and only then open the material. This avoids the false consensus of those who only recognize a colleague's speech. Time in testing with feedback comes back as retention; fluency in rereading does not.

Connection with cognitive load and with error

When the material is too new, do a single pass of guided reading, then a short test and feedback. Desirable difficulty starts at the second operation. Cognitive load explains why free recall of an entire textbook fails for the beginner: there is no schema yet.

If the error repeats after feedback, change the retrieval format (explanation out loud, mechanism instead of label). Varying the format is desirable difficulty with a purpose. Erring in the notebook is cheap; deliberate practice reduces reliance on luck on the ward.

Frequently asked questions

Is rereading never useful?

It is useful after the retrieval attempt, to correct and complete. As a sole method, it usually measures fluency.

Do concept maps replace testing?

No. Elaboration helps organize; retrieval consolidates. Use the map as an answer key after trying to redraw it from memory.

How many times do I need to test the same item?

Enough to get it right at increasing intervals. A single session does not replicate studies with spaced reviews.

What if I make too many mistakes and get discouraged?

Errors in practice are the material. Adjust the difficulty, do not abandon retrieval. If test anxiety flares up, seek the help you need and read the text on test anxiety.

Does this apply to practical skills?

Procedures also benefit from deliberate practice and retesting, not just from watching the procedure again.

How does NEXO fit in without becoming autopilot?

One round with a written hypothesis and a classified error is worth more than several rounds without intention.

Final checklist

• Session starts with an attempt without the text
• Feedback comes after, not before
• Rereading only at the points of error
• The next review already has a date
• The metric is recall, not pages
• Cases and facts go into the same method
• The discomfort of effort does not cancel the session

Play the case of the day

Open the case of the day, answer before looking up material, and only then check. That minute of effort is the desirable difficulty in the clinical format.

References

1. Karpicke JD, Roediger HL 3rd. The critical importance of retrieval for learning. Science. 2008. PubMed 18276894.
2. Karpicke JD, Blunt JR. Retrieval practice produces more learning than elaborative studying with concept mapping. Science. 2011. PubMed 21252317.
3. Larsen DP, Butler AC, Roediger HL 3rd. Repeated testing improves long-term retention relative to repeated study. Med Educ. 2009. PubMed 19930508.
4. Schmidmaier R, et al. Using electronic flashcards: retesting versus restudying. Med Educ. 2011. PubMed 21988625.
5. Dobson JL, Linderholm T. Desirable difficulties and anatomy recall. Anat Sci Educ. 2015. PubMed 25227209.
6. Systematic review of distributed and retrieval practice in health professions education. Adv Health Sci Educ. 2024. PubMed 37615780.
7. Piza F, et al. Awareness and usage of evidence-based learning strategies. Med Teach. 2019. PubMed 31407930.
8. Miller GE. Assessment of clinical skills/competence/performance. Acad Med. 1990. PubMed 2400509.
9. Eva KW. What every teacher needs to know about clinical reasoning. Med Educ. 2005. PubMed 15612906.
10. Epstein RM. Assessment in medical education. N Engl J Med. 2007. PubMed 17251535.
11. Misinterpreted-effort hypothesis / effort and learning judgments. Cogn Psychol. 2019. PubMed 31470194.
12. How do medical students learn conceptual knowledge? BMC Med Educ. 2022. PubMed 35387637.