I've noticed that people often manage to become moderately good at something, but seem to reach some sort of peak at that level and then rapidly start getting worse again.

I've seen this happen with videogames (most notably Starcraft 2), pool games, badminton, and even drawing/painting. People can get from complete beginner to a mediocre level in all these things, and then at some point they will start getting worse at them again. Not all the way back to a beginner level, but far enough to notice a serious difference in results.

If there are no physical changes that are causing this, is there any explanation for this problem with learning?

  • 1
    $\begingroup$ Yes, too many possible explanations. What else can you think of that might be relevant? Are your expectations changing? How about your level of interest in the given task? $\endgroup$ Jul 22, 2014 at 11:51
  • $\begingroup$ Since they're all just hobbies, my interest in them remained about the same overall until this decline started eating away at my motivation to keep doing them. My expectations stay the same as well, with some I actively play to get better, others are casual and have no real goal. Is it possible to rely too much on a perceived level of competence? $\endgroup$
    – Valyrion
    Jul 22, 2014 at 12:03
  • $\begingroup$ It's possible to rely too much on anything. Performance would probably decrease with motivation, but it sounds like the decline started first. Anyway, don't just follow my lead for ideas of what else might be going on – I can only guess, and you're probably in a better position to do that. Any other associated changes that might be relevant? I suppose I could be taking this in the wrong direction anyway...Questions tend to do better here when they're more general than personal. $\endgroup$ Jul 22, 2014 at 12:10
  • 2
    $\begingroup$ Self-help questions are discouraged on the site. This question was only marginally in that category, but I have edited out the personal aspects to make it more general and to discourage voting to close. Feel free to make any changes to my edits as needed. $\endgroup$ Jul 22, 2014 at 19:46
  • 1
    $\begingroup$ Search for >S Shaped learning curve. It will give you a detailed Answer. $\endgroup$ Jul 23, 2014 at 18:06

1 Answer 1


S-shaped learning curves

As per A Umar Mukthar's comment, the phenomenon is known as an S-shaped learning curve. They have been a known phenomenon in psychology for many decades, and were originally attributed to trial-and-error learning sets (Harlow, 1949). Harlow defined a learning set in the following manner:

The monkeys learn how to learn individual problems with a minimum of errors. It is this learning how to learn a kind of problem that we designate by the term learning set.

Meta-cognition and intra-individual performance variability

Harlow's explanation has fundamentally not been changed too much, except perhaps that in modern cognitive research, we now call these meta-cognitive strategies rather than learning sets.

When first encountering a novel problem, participants will adopt one strategy and work to improve their performance on that, but eventually they will discover other strategies. When participants shift strategy thus to an unpracticed strategy, their performance declines, then improves again.

Because all strategies for the same problem tend to share at least some overlap, as participants explore an increasing proportion of the space of possible meta-cognitive strategies, shifting to a novel strategy requires decreasing amounts of new practice. The magnitude of the performance decline associated with strategy shifts is therefore gradually eliminated with prolonged practice (Murre, 2014).


  • Harlow, H. F. (1949). The formation of learning sets. Psychological review, 56(1), 51.
  • Murre, J. M. (2014). S-shaped learning curves. Psychonomic bulletin & review, 21(2), 344-356.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.