One of the real problems with improving education systems is that there tends not to be much agreement about what education is actually for. I’ve written about this issue before and have made clear my view, education should exist to make children cleverer. Clearly this in part depends on a belief that it is actually possible to make children cleverer , no matter their starting point.

So, what evidence is there that we can become more intelligent? Everyone knows about Carol Dweck’s immensely popular theory of the growth mindset; that we can become cleverer by believing we can become cleverer. This is certainly encouraging, but unfortunately I think it’s a bit more complicated than that. If wishes were fishes we’d all have salmon for supper. In this post I pointed out that Dweck is in danger of making her theory unfalsifiable and therefore unscientific.

It’s probably worth looking at the concept of intelligence in a little more detail. In 1971, the psychologist Raymond B Cattell broke intelligence into two separate components: fluid intelligence (Gf) and crystallised intelligence (Gc). Fluid intelligence is usually defined as the ability to reason and solve problems, whereas crystallised intelligence is the ability to access and utilise information stored in long-term memory.

It turns out that while they’re not the same thing, fluid intelligence correlates surprisingly well with working memory capacity. One of the most important things to understand about working memory is that no mater how clever you are, your capacity to pay attention to different ideas and facts at the same time is strictly limited. But, although everyone’s working memory is fragile, there’s no doubt that some people have greater capacities than others. This confers a real advantage and, if we’re interested in making children cleverer, it seems sensible to investigate how we can improve their working memories. Sadly, despite the claims of various brain-training gurus, it doesn’t actually appear possible to increase working memory capacity: what you get is what you get.

So, what about crystallised intelligence? This is a much more profitable avenue of enquiry. If part of the measure of general intelligence is the ability to access items stored in long-term memory, then the good news is that for all practical purposes there’s no real limit to the amount of stuff you can cram in to your brain. Of course what we know is subject to forgetting, but with practice we can improve our long-term stores of knowledge quite considerably. Other good news is that the ability to remember doesn’t seem to rely on fluid intelligence. No matter how poor you are at reasoning or problem solving, you can still commit facts to long term memory. And, the more you remember, the easier it becomes to remember additional items of knowledge.

The real benefits become clear when we understand that by improving crystallised intelligence we can ‘hack’ our fluid intelligence. That is to say, we can use what we’ve stored in long-term memory to compensate for deficits in working memory. When we store information in long-term memory it gets organised into schemas – interconnected webs – which mean that when we retrieve one item we also bring with it all the information we’ve stored in the same schema. So, for instance, if I know nothing about chess and try to remember the positions in the diagram below, the task becomes a formidable feat, and I’ll most probably give up.


However, if I know the starting set up of a chess board the task becomes much easier. Now I only have to track which pieces have moved or are missing. And if I were a very experienced chess player I might recognise this configuration as an opening called The Dutch Defence. In that case the entire board becomes just a single item in working memory.

There are two ways we can cheat the limitations of working memory:

  1. Extended networks of ideas and information (schemas) take up the same amount of space in working memory as single isolated facts so the more we know about a subject the more space we have to pay attention to novel ideas and interesting combinations of ideas.
  2. Through practice we can automatise various procedural knowledge so that it becomes background knowledge. When a skill or a fact has been automatised, we are no longer consciously aware of it and it takes up very little space in working memory allowing us to concentrate on things we haven’t yet mastered.

Much of teachers’ efforts in the current system is spent trying (and failing) to improve fluid intelligence by drilling students in ‘skills’ like essay writing. Unfortunately, writing essays isn’t really a skill that we ever automatise – in order to write thoughtfully we have to think. Worse, there are still schools and teachers who see themselves as best serving the needs of children and society by concentrating on generic skills like creativity, collaboration and communication. These skills are for the most part dependent on fluid intelligence and if we take this approach we guarantee that those with greater working memory capacities will do well and those who don’t will struggle and often fail.

If instead teachers prioritised encoding and retrieving knowledge then all children would see an increase in crystallised intelligence. Instead of practising skills that don’t really automatise we could spend time automatising the component parts and on factual recall. In this system, the entire bell curve is more likely to move to the right. Admittedly those with higher fluid intelligence would still do better but there wouldn’t the long tail of underachievement we currently have. Everyone would see themselves get cleverer as they passed through the system.

To me, this seems a worthy aim. I understand that some people may not agree with my analysis but I think it’s hard to dispute that making children cleverer is a good thing. If you think I’m wrong about the benefits of increasing crystallised intelligence then I’d be interested in reading what you would propose we do instead.