Brain-training exercises compared: what each one actually measures

Why comparison is hard

People talk about brain training as if it were one activity, in the way that running is one activity. It is not. The exercises grouped under that label load very different things, and a gain on one rarely predicts a gain on another. Comparing them fairly means asking three separate questions of each: what does it demand while you are doing it, what does practice reliably improve, and what does practice not improve. Almost all the confusion in this area comes from answering the first question and then quietly assuming the answer to the second and third.

Dual n-back: continuous updating

Dual n-back asks you to hold two rolling windows and rewrite both every couple of seconds while deciding whether the current item matches the one N steps back. Nothing is stored for you and nothing stays still. It is demanding in a specific way: not the size of what you remember, but the rate at which you must replace it. Practice reliably improves your n-back scores and improves performance on closely related updating tasks. Whether it improves anything further is contested, and covered in detail in the companion guide on the evidence.

Digit span: simple storage

Digit span asks you to repeat back a growing list of numbers, forwards or backwards. It is the oldest and simplest working-memory measure, and it is still used clinically. What it demands is storage and, in the backwards version, a little manipulation. What it does not demand is continuous updating, because the list stops before you answer. People improve at digit span quickly, often by chunking, and that improvement is largely specific to the strategy rather than to memory itself.

Complex span: storage while doing something else

Complex span tasks alternate remembering with a processing step: read a sentence, judge whether it makes sense, remember a word, repeat. These are used heavily in research because they correlate more strongly with reasoning measures than simple span does. What they demand is protecting stored items from interference while your attention is elsewhere. That is closer to what working memory does in daily life than either digit span or n-back, which is why they are common as outcome measures rather than as training tasks.

Stroop: resisting the obvious answer

In a Stroop task you name the ink colour of a word that spells a different colour. Reading is automatic, so the correct answer has to be produced against interference. What it demands is inhibition rather than memory. Practice makes you faster at Stroop, and the effect is well documented, but there is little reason to expect it to change your memory, and the two are frequently confused because both feel mentally effortful.

Reaction time: speed, not thinking

A simple reaction-time task asks you to respond as fast as possible to a signal. A choice reaction-time task asks you to pick between responses. These measure processing speed and, in the choice version, decision time. They are useful as a baseline and they are extremely sensitive to sleep, caffeine and alertness, which makes them a good check on your state and a poor measure of anything stable over weeks.

Task switching

Switching tasks asks you to alternate between two rules, for example sorting by colour then by shape. The cost of the switch is measurable and consistent. What it demands is flexible control of what you are attending to. Like Stroop, practice reduces the cost for the trained pairing more than it changes anything general.

Commercial brain-training suites

Most commercial products bundle several of the above with a scoring layer and a progress narrative. The individual exercises are usually reasonable. The problem has historically been the claims made around them. In 2016, a large review led by Daniel Simons examined the evidence behind these products and concluded that it supports improvement on the trained tasks, offers limited support for closely related tasks, and does not support broad claims about improved everyday cognition.

What all of them share

Every exercise in this list improves with practice on itself. That is not a trivial finding, but it is the least interesting one, because it is also true of learning a card game. The question that separates a useful exercise from an expensive one is whether the improvement travels, and for this whole family the honest answer is that it travels a little to very similar tasks and not reliably beyond.

Where the evidence is strongest

The strongest and least disputed claims are the narrow ones. Practise n-back and your n-back improves. Practise Stroop and your Stroop improves. Meta-analytic work by Soveri and colleagues in 2017 found exactly this pattern for n-back training specifically: clear improvement on the trained task, some on close relatives, little beyond. If you want an exercise whose benefit you can actually verify, choose one that reports an honest number and judge it on that number.

Limits of broader claims

Controlled reviews do not establish a general promise of higher intelligence, better grades or improved everyday functioning from playing a particular training game. Findings depend on the task, comparison group and outcome. Read the evidence for the specific claim rather than assuming every exercise has the same benefits.

Choose by the activity you want to practise

Choose n-back to practise remembering and updating a sequence, or a reaction task to practise its particular response rule. For a study goal, practise retrieving and using the material you need to learn. A free download does not mean every feature is free: check the app’s purchase information.

Combine tasks for a clear reason

You can alternate tasks for variety, but a mixture does not automatically produce broader cognitive benefits. Keep separate records for each task and identify the reason for including it. Do not treat a short reaction game as a validated check of your fitness for another activity.

Assess the evidence behind a result

Before comparing a leaderboard score or a before-and-after account, ask about the task, settings, measurement and comparison group. An anecdote can suggest a question but cannot isolate the effect of training. A percentage improvement on a game should never be relabelled as the same percentage increase in intelligence.

A reasonable position

The reasonable position is neither that brain training is a scam nor that it rebuilds your mind. These are legitimate cognitive tasks with a real and measurable learning curve, wrapped for two decades in marketing their own evidence does not support. Practise one because you like the difficulty and want an honest measurement of it, and treat every broader promise as unproven until someone shows you a controlled trial.

What Dual N-Back does

Dual N-Back implements the exercise itself and reports your accuracy per session, with additional streams available when you want a harder version. It does not claim to raise your intelligence, and the guides on this site set out what the research does and does not support, because a training tool that overstates its evidence is not worth using even when the exercise underneath it is sound.

Further reading

• Simons et al. (2016), Do “Brain-Training” Programs Work?, Psychological Science in the Public Interest
• Soveri et al. (2017), Working memory training revisited: a multi-level meta-analysis of n-back training studies, Psychonomic Bulletin & Review
• Melby-Lervåg, Redick & Hulme (2016), Working memory training does not improve performance on measures of intelligence or other far transfer, Perspectives on Psychological Science
• Kane et al. (2007), Working memory, attention control, and the n-back task, Journal of Experimental Psychology: Learning, Memory, and Cognition