By around month three or four of learning something new, most learners hit a period where continued practice stops producing visible progress. The scores don't move. Vocabulary you drilled last week is gone. The effort continues and the return on it drops to nearly zero.
William Bryan and Noble Harter documented this in 1897 while tracking telegraph operators learning Morse code. They called the flat stretches “plateau periods” — extended phases during which performance showed no measurable improvement despite continued practice, followed by sudden gains once the plateau passed. They were the first researchers to establish that the plateau is a structural feature of skill acquisition, not a failure of effort.
Most learners never learn that. So when the plateau arrives, they assume they've found their limit.
What's Happening Inside the Plateau
The cognitive explanation for plateau periods came with Fitts and Posner's 1967 model of skill acquisition.1 They identified three distinct stages all learners pass through. The first is explicit: deliberate, effortful, rule-based. You are consciously thinking about every step. The third is automatic: the skill runs without conscious attention. The middle stage — which they called the associative phase — is where the plateau lives.
In the associative phase, the learner is transitioning from explicit processing to automatic processing. The brain is reorganizing what it knows, building connections between pieces it has acquired separately. This reorganization takes time, doesn't produce immediate performance gains, and feels like stagnation from the inside. It is not stagnation. The work is just happening below the surface.
Cognitive psychologist John Sweller's research on working memory adds another layer.2 At the intermediate stage, learners hold more information in mind than beginners do — they know enough to perceive complexity they couldn't see before. But that additional complexity competes for working memory capacity, leaving less cognitive resource available for performance. The intermediate learner often performs worse on familiar tasks than they did as a beginner, precisely because they now understand the task well enough to see everything that can go wrong.
This is Dunning-Kruger in reverse. The beginner doesn't know what they don't know, so they feel competent. The intermediate learner knows exactly what they don't know, so they feel incompetent. Both feelings are inaccurate. The intermediate learner is doing more cognitive work. It just doesn't show up yet.
What a Plateau Feels Like From the Inside
There is a specific kind of discouragement that hits around months two or three of learning a new language. The early gains stop. You've passed through the phase where everything was new, where five new words a day felt like real progress. Now you have a foundation and the gaps in it are all you can see.
I went through this with Japanese — more than once, at different stages. Around the midpoint of building my kanji vocabulary, my recall scores on new characters were measurably worse than they'd been three months earlier. The cognitive load had increased: I now had enough kanji to perceive interference between similar characters, and resolving that interference used resources that had previously gone toward simple recall.
The plateau broke eventually, as they tend to — the underlying work had been accumulating the whole time without showing up in the scores.
What the Research Suggests Doing Differently
Two findings from the learning science literature have direct bearing on plateau periods.
The first is retrieval practice — forcing yourself to recall information from memory before reviewing it.3 During a plateau, the temptation is to return to easier, more comfortable study habits. This tends to make the plateau last longer, not shorter. The discomfort of struggling to retrieve something is the mechanism by which deep encoding forms.
The second finding concerns variation. Research by Kornell and Bjork in 2008 found that interleaving different types of practice — mixing problem types, switching between topics, varying the review format — produces better long-term retention than blocked practice of a single type.4 The mixed session feels harder and produces worse immediate performance. That's the signal it's working.
If you've hit a wall, reducing difficulty is unlikely to help. The plateau is a transition state. Making the transition harder — more retrieval, more variation, less passive review — tends to shorten it.
The People Who Stop Here
Most people who stop learning something stop during a plateau. Not because the goal was wrong. Because the plateau feels like failure, and there's nothing in most people's education that tells them plateau periods are a structural feature of skill acquisition, not a sign of inadequacy.
Bryan and Harter noted that operators who took their plateau periods as evidence they'd reached their limit mostly didn't progress beyond them. Operators who kept practicing — even with nothing to show for it — eventually broke through. The ceiling, as it turned out, was a landing on a staircase.