Can you train your working memory? What the meta-analyses show

Try this: read the number 7 3 9 1 5 once, look away, and say it backwards. The small mental space in which you just held and rearranged those digits is called working memory. It is involved in following a conversation, doing mental arithmetic and, not least, in solving puzzles where you have to keep several rules in mind at once. No wonder that “train your working memory” has become a popular promise. What does the research say?

A small workspace with a hard limit

In 1974, Alan Baddeley and Graham Hitch proposed a model that still shapes the field: working memory is not a single store but a system, with a component that holds sounds and words, one that holds visual and spatial information, and a “central executive” that directs attention between them.

Its capacity is strikingly small. In an influential review, Nelson Cowan argued that adults can typically keep only about three to five meaningful units – “chunks” – in mind at once, when they cannot rehearse or group them. People who seem to remember much more usually do so by packing information into larger chunks, for example by recognising a familiar pattern.

The hope: train the workspace, improve reasoning

Working memory capacity is correlated with performance on reasoning tests. That raised a tempting idea: if you could enlarge the workspace through practice, reasoning ability might improve as well.

A 2008 study by Susanne Jaeggi and colleagues gave the idea a big push. Participants practised a demanding “n-back” task, in which you constantly compare the current item with one shown several steps earlier. Afterwards, the trained groups improved more on a fluid reasoning test than a control group did. The study received wide attention – and triggered a decade of follow-up research.

What the meta-analyses found

When a question has been studied many times, a meta-analysis combines the results of all available studies. Several such analyses have looked at working memory training:

• Monica Melby-Lervåg and Charles Hulme (2013) found that training produced reliable short-term improvements on tasks similar to those trained (“near transfer”), but no convincing evidence of improvements in other abilities such as nonverbal reasoning (“far transfer”).
• Jacky Au and colleagues (2015) analysed n-back training studies specifically and reported a small positive effect on fluid intelligence measures.
• Melby-Lervåg, Thomas Redick and Hulme (2016) reviewed a much larger set of studies. When training groups were compared with active control groups – people who also practised something, just not working memory – the far-transfer effects largely disappeared.
• Daniel Simons and colleagues (2016) examined the evidence cited by brain-training companies themselves. They found extensive evidence that people get better at the trained tasks, less evidence for improvements on closely related tasks, and little evidence that training improves everyday cognitive performance.

Why do the conclusions differ? Partly because of study design. With a passive control group that does nothing, any training group may improve simply because participants expect to, or because they become more motivated or more practised at taking tests. Studies with active control groups are better at separating these effects, and they tend to find smaller benefits.

What this does not mean

• It does not mean practice is useless. Training reliably improves the skill you practise. If you want to get better at a particular type of puzzle, practising that type of puzzle works.
• It does not mean working memory never matters. It clearly plays a role in reasoning. The open question is whether training it in isolation changes reasoning ability in general.
• It does not settle the question for every group. Most studies involved healthy adults or children. This article is not about rehabilitation after illness or injury; for that, please speak to a qualified professional.

So what helps?

The honest summary is that “far transfer” – getting better at many things by training one narrow thing – is hard to achieve. Broader factors look more promising. For example, a 2018 meta-analysis by Stuart Ritchie and Elliot Tucker-Drob concluded that each additional year of education is associated with gains of roughly one to five IQ points. That is not a quick fix, but it is a reminder that thinking skills grow through rich, varied learning rather than one repeated drill.

How we use this at IQ Arion

In the IQ Arion app you can practise each puzzle area at your own level, and play daily puzzles and duels. Based on the research above, we describe them honestly: they help you become more familiar and more confident with these types of puzzle. They do not promise to raise your IQ.

If you are curious how your working memory copes with keeping several rules in mind, the logic and pattern puzzles in our free short test are a good place to start.