All articles
5 min readlearning / training

Novices need the worked solution, experts need the problem

Most cohorts I train ask for the same thing on the first morning: send us the finished prompts, we will adapt them. My reflex was to refuse and hand out exercises, on the theory that people who work something out themselves remember it better. That reflex costs beginners time and accuracy. It starts paying a few hours later, with the same people, on the same material.

What the research shows.

John Sweller and Graham Cooper ran five experiments with algebra students in 1985. One group practiced by solving problems. The other studied fully worked solutions to those same problems. On a later test with matched problems, the worked-example group solved them in about half the time and made around a fifth of the errors, and the problem-solving group had needed several times longer to get through the training sequence 1. The finding held up under scrutiny: a 2023 meta-analysis covering 55 studies and 181 effect sizes put the worked-example effect on mathematics performance at g = 0.48 2.

Sequence matters inside a session. Tamara van Gog, Liesbeth Kester and Fred Paas gave secondary students four training tasks on troubleshooting electrical circuits, arranged four ways: examples only, example then problem, problem then example, problems only. Examples only and example-problem pairs came out ahead on both learning and efficiency. Putting the problem before its example scored no better than practice alone 3.

The picture then flips. Slava Kalyuga, Paul Chandler and John Sweller trained apprentices on industrial control tasks. Early in training, studying worked examples beat solving problems. After the same trainees had drilled the domain, solving problems beat studying worked examples on the harder material 4. Same content, same people, opposite verdict a few hours apart. The pattern earned a name, the expertise reversal effect 5, and a 2025 meta-analysis sized it across 176 effect sizes, 60 experiments and 5,924 participants. Learners with low prior knowledge did better with high-assistance instruction, d = 0.505. Learners with high prior knowledge did better with low assistance, d = -0.428. The two halves are uneven: supporting a beginner buys more than withdrawing support from someone experienced 6.

The case for struggling first is narrower than it sounds.

Manu Kapur's productive failure runs the other way. Students attempt a problem they have never been taught, generate several wrong or partial solutions, and only then get the canonical method. Tanmay Sinha and Kapur pooled 166 comparisons from 53 studies and found g = 0.36 in favor of problem solving before instruction, rising to d = 0.58 in the studies that followed the design closely. The trend reversed for second to fifth graders and for domain-general skills 7.

Greg Ashman, Kalyuga and Sweller tested the order directly with Year 5 students learning about light energy efficiency. Explicit instruction followed by problem solving beat the reverse order on problems resembling those taught (N = 64). When they raised the number of interacting elements in the material (N = 71), explicit instruction first won on transfer problems as well 8.

Both results can stand. Kapur's design gives learners a set of contrasting cases and a teacher who consolidates the attempts afterwards, and his effects concentrate in conceptual understanding. Ashman's material asks a child to juggle several interacting variables with nothing to hold them together. The deciding variables are how much the learner already knows and how many parts of the problem have to be held in mind at once.

The protocol.

  1. Measure prior knowledge in 2 minutes. Put one representative task on screen, give 20 seconds, and ask for the first step only. Not the answer, the first move. If more than 80% of the room gets the first step right, cut the guidance. Below that, run worked examples.
  2. Pair each example with a matched problem, example first. One worked solution, then one problem of the same type solved alone. Repeat. The reverse order performed like unguided practice 3.
  3. Attach a self-explanation prompt to every example. After each step, ask learners to name the rule it applies. A meta-analysis of 69 effect sizes puts self-explanation prompts at g = 0.55 9.
  4. Fade the solution backwards. Remove the last step and ask for it, then the last two, then the last three, until only the problem statement is left. Robert Atkinson, Alexander Renkl and Mary Margaret Merrill combined backward fading with self-explanation prompts and got medium to large effects on near and far transfer, without adding time to the session 10.
  5. Re-test the first step after each block. Prior knowledge moves in hours, so the mode has to move with it. Once the group clears the first-step check, hand them problems and stop narrating solutions.
  6. Reserve struggle-first for one concept with few moving parts. Give it 15 to 20 minutes, ask for several solution attempts rather than one, and consolidate in the same session. Skip it for procedures with many interacting steps taught to a room of beginners.

The answer to the cohorts who ask for the finished prompts on the first morning is yes, with the reasoning written next to each line, followed by a task of the same shape. Run the 20-second first-step test at the start of your next session and pick the mode from what comes back.

Sources.

  1. Sweller, J., & Cooper, G. A. (1985). The use of worked examples as a substitute for problem solving in learning algebra. Cognition and Instruction, 2(1), 59-89. doi.org/10.1207/s1532690xci0201_3
  2. Barbieri, C. A., Miller-Cotto, D., Clerjuste, S. N., & Chawla, K. (2023). A meta-analysis of the worked examples effect on mathematics performance. Educational Psychology Review, 35, 11. doi.org/10.1007/s10648-023-09745-1
  3. Van Gog, T., Kester, L., & Paas, F. (2011). Effects of worked examples, example-problem, and problem-example pairs on novices' learning. Contemporary Educational Psychology, 36(3), 212-218. doi.org/10.1016/j.cedpsych.2010.10.004
  4. Kalyuga, S., Chandler, P., & Sweller, J. (2001). Learner experience and efficiency of instructional guidance. Educational Psychology, 21(1), 5-23. doi.org/10.1080/01443410124681
  5. Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23-31. doi.org/10.1207/S15326985EP3801_4
  6. Tetzlaff, L., Simonsmeier, B. A., Peters, T., & Brod, G. (2025). A cornerstone of adaptivity: a meta-analysis of the expertise reversal effect. Learning and Instruction, 98, 102142. doi.org/10.1016/j.learninstruc.2025.102142
  7. Sinha, T., & Kapur, M. (2021). When problem solving followed by instruction works: evidence for productive failure. Review of Educational Research, 91(5), 761-798. doi.org/10.3102/00346543211019105
  8. Ashman, G., Kalyuga, S., & Sweller, J. (2020). Problem-solving or explicit instruction: which should go first when element interactivity is high? Educational Psychology Review, 32, 229-247. doi.org/10.1007/s10648-019-09500-5
  9. Bisra, K., Liu, Q., Nesbit, J. C., Salimi, F., & Winne, P. H. (2018). Inducing self-explanation: a meta-analysis. Educational Psychology Review, 30, 703-725. doi.org/10.1007/s10648-018-9434-x
  10. Atkinson, R. K., Renkl, A., & Merrill, M. M. (2003). Transitioning from studying examples to solving problems: effects of self-explanation prompts and fading worked-out steps. Journal of Educational Psychology, 95(4), 774-783. doi.org/10.1037/0022-0663.95.4.774

Keep reading