One of the more surprising results from research on learning (and this appears to be true for almost any type of learning, from physical skills, to mathematical skills, to memorizing facts) is that practicing in a way that makes you better today, doesn’t necessarily make you better tomorrow. In fact, you’ll learn more and retain what you’ve learnt longer if you practice in a way that causes more mistakes.
And it’s not just a few studies, this is a robust, highly replicated result of decades of research [Footnote 1]. Surprisingly, it’s not a well known “pop science” fact, which is surprising given the popularity of the neuro- and psychological science of learning — everyone and their dog has heard about the dubious "10,000 hour rule” [Footnote 2]. Then why hasn't this important result reached critical mass in the media? I have some ideas about why, but first let’s talk about the effect itself. There’s a lot to learn about learning here.
First of all, it’s important to realize that your performance isn’t a good indicator of learning. In the context of the research we’re discussing here, performance means “how good you are at something while you practice it”, and learning means “how good you are at something when you try it again some time later” — be it a day, a week, or longer. Remember these definitions as you keep reading. Let’s look at some examples of this from a 2015 review article by Soderstrom and Bjork.
For instance, you can train a mouse on a maze, and deny them any sort of reward for reaching the end. As you keep training them, it looks like they aren’t learning anything… they just keep exploring the maze somewhat randomly. However, when you start giving them a reward at a particular location, they suddenly return to that location about as often as mice who have gotten the reward right from the first time they explored the maze — meaning the mice have been learning the maze all the while. This effect is referred to as “latent learning”, and it means learning is occurring even though you don’t see a performance improvement.
Then there’s another effect called “over learning”. If you keep practicing a skill that you’ve already reached peak performance on, of course you’re not going to see performance improvement — you can’t get any better! However, the more you “over learn”, the longer it takes before the memory fades. This is another good reason to keep practicing the fundamentals of any skill, even if you think you’re beyond it.
Now for what I think is the most fascinating research: practice scheduling. That is, how should you structure practices, how often, how many repetitions, in what order should you train each skill, etc. First, when people use "spaced practice", spreading their training over days and weeks instead of “cramming”, they retain more information later (Footnote 3). Cramming leads to greater performance during practice — but you end up forgetting most of it in the long run.
Second, “random practice” works much better than “blocked practice”. Let’s say you are working on three different baseball pitches — skill ‘A’ (a fastball), skill ‘B’ (a curve ball), and skill ‘C’ (a changeup). Blocked practice means going through practice like this: A-A-A-B-B-B-C-C-C. Random practice means randomizing these, so it looks something like this: C-A-B-A-B-A-C-C-B. Performance during practice typically improves the most with blocked practice… but your improvement the next day (or later) is far greater with random practice.
Third, practice variability. Let’s say you’re a basketball player working on your three-point shot. Should you stand at the three point line every time you practice, or should you try different distances from the net? Turns out that practicing at different distances leads to better learning. What’s more, studies have shown that even if you never practice at the three point line and keep changing positions in front or behind, you actually perform better on that line the next day compared to someone who just only practiced on that line.
Of course there are many more examples – and how effective each strategy is depends on what you’re learning, how complex it is, your baseline level of skill, and countless other variables.
What’s going on here?
Why these effects work the way they do is a hot area of research. There seems to be some solid science explaining why spaced practice is best [Footnote 3]. Less well understood is why random scheduling and variable practice seems to be best — it might be that it forces you to “reload” the skill more often, compare the essential features between them, and understand the most influential factors. Whatever the reasons are, they're probably not as simple as “make more mistakes and you’ll learn faster”.
The important point here is this: just because you’re performing well at practice, it doesn’t mean you’re learning more. Understanding this is powerful. Why? Because numerous studies have also shown that learners judge the quality of their practice by their performance. This lack of awareness or “metacognition” about ones own learning can lead to hours and even years of sub-optimal practice.
And that’s probably one of the reasons why barely anyone talks about this research — to most people who practice any sort of skill, it just doesn’t seem “right”. Another reason is that it’s kind of frustrating — you have to practice in some challenging ways that might make you feel like you’re not progressing as quickly. However, I suggest we start framing these results in a more encouraging, positive way.
Just because you’re having a hard time at practice, it doesn’t mean you’re not learning. If you’re making mistakes, you’re learning. It’s just another example of the payoff of hard work — a fact most of us know and appreciate. And that’s not so surprising.
Footnote 1: to avoid bloating this blog post with numerous references I’ll point the interested reader to a couple good reviews: Soderstrom and Bjork 2015 PMID: 25910388, and for a review focused more on motor learning check out Kantak and Winstein 2011 PMID: 22142953.
Footnote 2: Yes, dubious. Not that it isn’t important to practice a lot, and in a deliberate manner, but if you read about this “rule” for 10,000 hours the only thing you’ll become an expert in is oversimplification. Until we have a full article on the topic, check out these references: Macnamara et al. 2016 PMID: 27217246; Tucker and Collins 2012 PMID: 22535537.
Footnote 3: Smolen et al. 2016 PMID: 26806627.