Prep
What desirable difficulties are
Desirable difficulties is a term from the psychologist Robert Bjork (1994), later developed with Elizabeth Bjork (2011). It describes a counter-intuitive pattern: some ways of studying make learning feel harder and look worse in the moment, yet produce better long-term memory and better transfer to new problems.
A 2023 review by de Bruin et al. sums up the problem: these strategies feel effortful, and because effort feels like poor learning, learners tend to resist them. The authors propose a framework for helping learners start using desirable difficulties and stick with them.
The key idea is the gap between performance during study (how well you do right now) and learning (what you can still do weeks later). Easy methods inflate the first. Desirable difficulties build the second.
| Difficulty | What you do | Evidence | When it backfires |
|---|---|---|---|
| Spacing | Return to a topic after days, not minutes | Classroom meta-analysis d = 0.54 (Mawson et al., 2025) | Gap so long you must relearn from scratch |
| Retrieval practice | Recall or answer questions instead of rereading | Classroom meta-analysis d = 0.56 (Schwieren et al., 2017) | Material never learned and no feedback |
| Interleaving | Mix similar problem types | g = 0.42 overall, wide variation (Brunmair & Richter, 2019) | Dissimilar categories, word lists; some maths classrooms show no effect |
| Generation | Produce an answer before seeing it | Generation effect .40 across 86 studies (Bertsch et al., 2007) | Wrong answers left uncorrected |
Why easy studying fools you
Two studies show the trap clearly.
In Kornell & Bjork (2008), people learned to recognise painters' styles from examples shown either grouped by artist or mixed together. Mixing led to better classification of new paintings. Yet participants rated grouping as more effective, even after their own test results had shown the opposite.
In Kornell (2009), spacing flashcard practice beat massing it for 90% of participants, but after the first session 72% believed massing had been more effective.
Rereading shows the same illusion from the other side. Seban et al. (2024) found that students who reread a text felt more confident after each reading and ended up overconfident, especially when tested a day later. Fluency feels like learning. Often it is just familiarity.
This has a direct consequence for exam prep: a score taken during or straight after a study session is a measure of performance, not learning. If you take a practice test right after reviewing its topics, a high score mostly tells you the material is fresh. The score that predicts exam day is the one you get days later, cold. That is also why desirable difficulties can look like a bad deal in the short term: they lower the first number to raise the second.
The four best-studied desirable difficulties
1. Spacing
Spreading practice over time instead of massing it. When you return after a gap, recall is harder, and that effort pays off. The meta-analysis by Cepeda et al. (2006) (317 experiments) found the optimal gap grows with the time until the test. In real classrooms, Mawson et al. (2025) found a moderate effect of d = 0.54 for distributed over massed practice. An effect size d is the difference between groups in standard deviations: 0.54 means the typical spaced learner scored about half a standard deviation higher.
2. Retrieval practice (testing)
Recalling information instead of rereading it. A review by Roediger & Butler (2011) concluded retrieval practice often produces large gains in long-term retention compared with repeated studying. In real psychology courses, Schwieren et al. (2017) found d = 0.56 in favour of testing.
3. Interleaving
Mixing different types of problems instead of practising one type at a time. A meta-analysis by Brunmair & Richter (2019) of 59 studies found a moderate overall effect, g = 0.42, but it varied a lot: larger for paintings (g = 0.67), small for maths tasks (g = 0.34), unclear for expository texts, and reversed for word lists, where blocking was better (g = -0.39). Interleaving helps most when categories are similar and easy to confuse, as Carvalho & Goldstone (2014) also found.
4. Generation
Producing an answer yourself before seeing it. The meta-analysis by Bertsch et al. (2007) of 86 studies found an average generation effect of .40, almost half a standard deviation, over simply reading.
Example: the same topic studied the easy way and the desirable way
Say you are learning the main AWS storage services for the Cloud Practitioner exam.
The easy way: read the S3 section, then the EBS section, then the EFS section, highlighting as you go. Reread the highlights on Sunday. Every session feels smooth, and by Sunday the pages look very familiar.
The desirable-difficulty way:
- Before reading, write down what you think each service is for (generation). Some guesses will be wrong; that is fine.
- Read, then close the book and write a one-line use case for each service from memory (retrieval).
- Practise with mixed scenario questions where you must choose between the three, rather than ten S3 questions in a row (interleaving similar, easily confused categories).
- Come back to the same questions three days later and again the next week (spacing).
The second plan produces more visible mistakes in the first week. It also produces the skill the exam actually tests: picking the right service from a short description, after weeks of other topics have piled on top. Judge the two by a test a week later, not by how the sessions felt.
When difficulty is not desirable
"Harder is better" is the wrong lesson. Difficulty helps when it makes you do the mental work that builds memory and you can eventually succeed. It does not help when you lack the background to make any progress.
- Retrieval with no feedback on material you never learned mostly produces guesses. Roediger & Butler note that feedback enhances the benefits of retrieval practice.
- Interleaving is not universal. Two large classroom experiments by Rowlandson et al. (2025) repeatedly failed to find an interleaving advantage for defined mathematical categories. Brunmair & Richter found blocking better for word lists.
- Gaps can be too long. Cepeda et al. (2009) found the benefit of a gap rises sharply and then gradually declines; if you have forgotten everything, the review becomes relearning from scratch.
- Difficulty that is just friction (a confusing book, a noisy room, poor sleep) adds effort without adding learning.
A useful test: does this difficulty force me to recall, distinguish or produce something the exam will ask for? If yes, it is probably desirable. If it just makes studying unpleasant, it is not.
How to stick with them
Knowing about desirable difficulties is not enough; the feeling that they are not working persists. Zepeda, Martin & Butler (2020) suggest five motivational levers: find the value in the strategy, reduce its cost, reframe how you appraise the struggle, keep the challenge at an appropriate level, and give yourself some choice.
In practice:
- Judge by delayed results, not by how a session felt. Retest a week later and compare.
- Lower the cost. Prepare question lists in advance so retrieval sessions start instantly.
- Rename the struggle. A miss in practice is information, and the effort is the point.
- Calibrate the challenge. If you get almost nothing right, go back and learn the basics first; if you get everything right, space further or mix more.
Learning about these strategies can change habits. In Brown-Kramer (2021), students who completed an assignment on evidence-based study strategies used fewer low-utility and more high-utility strategies, and improved by about a third of a letter grade compared with a control semester.
How to use this for a certification exam
- Generate first. Before reading a chapter, skim its headings and try to answer each as a question. Then read.
- Retrieve, don't reread. End each session by writing what you remember from memory, then checking.
- Space your reviews. Revisit each exam domain about weekly if your exam is one to three months away; see our study schedule template.
- Interleave where confusion lives. Mix question types that are easy to confuse: similar AWS storage services, similar attack types on Security+, the PMP process groups. Free practice tests mix domains by design; try the AWS practice test or the Security+ practice test.
- Measure with a delayed retest. Retake the practice test after a week or two, not the same day. That score reflects learning; same-day scores mostly reflect memory of the answers.
Go deeper into each method: spaced repetition, interleaving, the testing effect, or start with the study science overview.