“Learning plateau” gets used for at least three situations that need completely different responses. Telling them apart is most of the work.

One: the curve was always going to flatten. Skill gains are front-loaded. Heathcote and colleagues reanalysed practice data across a large set of tasks and argued the improvement curve is exponential rather than the power law everyone had assumed. The argument matters to researchers; for you the consequence is the same either way. Each additional hour buys less than the hour before it did. So part of what feels like a plateau is real improvement that has dropped below the level you can notice week to week.

Two: the skill went automatic. This is the one Ericsson’s deliberate practice paper is really about. His argument was never that hours produce expertise. It was close to the opposite: once performance becomes automatic, simply accumulating more experience stops improving it. People who keep improving are the ones who keep deliberately working on what they cannot yet do. Everyone else gets very comfortable at their current level and stays there.

Three: you are practicing the part you are already good at. The most common version, and the most fixable. Nobody designs it on purpose. It happens because the parts you are good at are pleasant to practice and the parts you are bad at are not.

The reason plateaus are disorienting is that your own read on whether practice is working turns out to be unreliable. There is a clean demonstration of this.

Simon and Bjork had people learn movement patterns under two schedules. Blocked, where you drill one thing until you move to the next, and random, where the tasks are shuffled. During practice, blocked was clearly better. Then they asked learners to predict how they would perform the next day. The blocked group predicted they would do well, because their sessions had gone well. They did worse. The random group, whose sessions felt clumsy, retained more.

This is the same effect Shea and Morgan found two decades earlier: shuffling practice makes performance worse while you are practicing and better when you are tested later, including on transfer to new versions of the task.

So the feeling of a good session tracks how fluent the session felt, not how much you learned. A plateau often means the last stretch of practice was smooth. Smooth is the problem.

Raise the difficulty until you are failing again. Guadagnoli and Lee’s challenge point framework makes the mechanism explicit: practice only teaches you something to the extent it generates information, and how much information a task generates depends on its difficulty relative to your current skill. Too easy and there is nothing to learn from. Too hard and there is more than you can use. The optimal point moves up as you improve. A plateau is the signature of practice difficulty that stayed still while your skill rose past it.

Shuffle the order. Blocked practice is the default because it feels productive. Interleaving the things you are drilling costs you visible performance now and buys retention and transfer later. Expect it to feel worse. That is the trade you are making on purpose.

Test yourself instead of reviewing. Karpicke and Roediger ran the cleanest version of this. After material was learned, dropping it from further study barely affected recall a week later. Dropping it from further testing cut recall substantially. Repeated retrieval built retention; repeated review, once you already knew the material, added close to nothing. If your practice is mostly going back over things you already have, that is a plateau by construction.

Change the feedback, carefully. Kluger and DeNisi’s meta-analysis is the one people misquote. Feedback improved performance on average, yes. But more than a third of the feedback interventions they reviewed made performance worse. The pattern in the harmful ones was feedback that pulled attention toward the self rather than the task. “You are not good at this” is a different intervention from “your third step is early.” Only one of them contains information you can act on.

Concretely, if you are stuck right now:

Find the part you have been avoiding. Not the part you are worst at in theory. The one you quietly skip. That is where the information is.

Isolate it and make it harder. Slower, faster, no notes, no reference, worse conditions, more constraints. You want a failure rate you can feel. If you are getting everything right, the session is not teaching you anything.

Shuffle what you drill. Stop doing twenty of one thing. Do them mixed, unpredictably.

Test cold. Start a session by producing the thing from nothing, before any warm-up or review. That is the measurement. What you can do warm is not what you have learned.

Ask for feedback about the task. If you are getting coaching, ask for the specific step. If you are self-teaching, record yourself and watch the failure, not your reaction to the failure.

Give it three weeks. Restructured practice performs worse before it performs better. That dip is the point where most people conclude it did not work and revert.

More hours of the same practice. This is the default response and it is the one the deliberate practice literature specifically argues against.

Motivation. Wanting it more does not add information to a session that has none.

A long break. Rest is worth having, and coming back fresh is real, but a break does not restructure anything. If you return to identical practice you return to the identical plateau, which is why the break so often feels like it worked for about ten days.

Switching to a new skill. The early curve is steep everywhere. Starting over feels like progress because it is the easy part of a new graph, and it costs you the hard-won part of the old one.

I want to be specific about the limits here, because this is a literature people quote much more confidently than they should.

Shea and Morgan and Simon and Bjork are laboratory motor-learning studies with small samples and artificial tasks. Their finding that harder practice retains better is well replicated, but the size of the effect in your particular skill is not something those papers can tell you. Karpicke and Roediger tested verbal learning, not motor skill or performance. Guadagnoli and Lee’s challenge point is a framework for organizing results, not a trial with an outcome.

And Ericsson’s deliberate practice claims have been genuinely contested. Macnamara and colleagues ran a meta-analysis across music, games, sports, education and professions and found deliberate practice accounted for around 12 percent of the variance in performance, far less than the original framing implied, and much less in less structured domains. Practice structure is not the only thing that determines where you end up. It is the part you control.

None of these studies examined your plateau. What they give you is a mechanism and a direction, and that is worth more than another month of repeating what already works.

A learning plateau is an information problem, not an effort problem. Your practice got comfortable, comfortable practice stopped generating anything for you to learn from, and adding hours to it just buys more of the same nothing.

Make it harder. Shuffle it. Test instead of review. Get feedback about the task. Then hold the new structure through the dip, because it will get worse before it gets better, and that dip is the only part of this that most people never get past.

If you want the underlying method, I have written about why testing yourself beats re-reading and about the limits your brain puts on learning. Both go deeper on the machinery this article sits on. Start here if you want the whole system.

Every DOI below was resolved and checked against the published record before it went in.

• →Ericsson KA, Krampe RT, Tesch-Römer C. The role of deliberate practice in the acquisition of expert performance. Psychological Review 1993;100(3):363-406. doi:10.1037/0033-295X.100.3.363

• →Heathcote A, Brown S, Mewhort DJK. The power law repealed: the case for an exponential law of practice. Psychonomic Bulletin & Review 2000;7(2):185-207. doi:10.3758/BF03212979

• →Shea JB, Morgan RL. Contextual interference effects on the acquisition, retention, and transfer of a motor skill. Journal of Experimental Psychology: Human Learning and Memory 1979;5(2):179-187. doi:10.1037/0278-7393.5.2.179

• →Simon DA, Bjork RA. Metacognition in motor learning. Journal of Experimental Psychology: Learning, Memory, and Cognition 2001;27(4):907-912. doi:10.1037/0278-7393.27.4.907

• →Guadagnoli MA, Lee TD. Challenge point: a framework for conceptualizing the effects of various practice conditions in motor learning. Journal of Motor Behavior 2004;36(2):212-224. doi:10.3200/JMBR.36.2.212-224

• →Karpicke JD, Roediger HL. The critical importance of retrieval for learning. Science 2008;319(5865):966-968. doi:10.1126/science.1152408

• →Kluger AN, DeNisi A. The effects of feedback interventions on performance: a historical review, a meta-analysis, and a preliminary feedback intervention theory. Psychological Bulletin 1996;119(2):254-284. doi:10.1037/0033-2909.119.2.254

• →Macnamara BN, Hambrick DZ, Oswald FL. Deliberate practice and performance in music, games, sports, education, and professions: a meta-analysis. Psychological Science 2014;25(8):1608-1618. doi:10.1177/0956797614535810