How to Improve Working Memory: Interventions That Work

Why Working Memory Matters

Working memory — the cognitive system that holds information in mind while you use it — is one of the most practically important cognitive abilities we measure. It predicts reading comprehension, mathematical ability, academic performance, and job performance across a wide range of roles. Its correlation with general intelligence (r ≈ 0.50) is among the highest of any single cognitive factor.

When people say they are "not good at multitasking" or "can't hold all the pieces in mind at once," they are often describing working memory limitations. When children struggle with multi-step instructions in school, the mechanism is usually working memory. When you lose your train of thought mid-sentence, it is working memory.

Given this importance, an entire industry has grown up around "working memory training." The claims are significant. The evidence is more complicated.

The Training Controversy

The most prominent working memory training programme — Cogmed — claimed in the mid-2000s that targeted working memory training produced improvements that transferred broadly to general cognitive ability. The original studies were promising enough to drive widespread adoption in schools and clinical settings.

Then the replication attempts came in. A 2013 meta-analysis and subsequent large-scale studies found that working memory training produced reliable improvements on the trained tasks themselves ("near transfer") but minimal reliable transfer to untrained tasks and virtually no transfer to general cognitive ability like IQ ("far transfer").

The n-back task became the focal point of this research partly because of a 2008 paper by Jaeggi and colleagues claiming that adaptive dual n-back training improved fluid intelligence — a claim that would have been genuinely significant if it replicated, since fluid intelligence was widely considered immutable. In a dual n-back task, participants monitor two simultaneous streams (visual position and auditory letters) and respond whenever the current stimulus matches what appeared n trials ago. The task is genuinely demanding and loads heavily on the central executive component of working memory. The Jaeggi claims prompted a wave of replication attempts; most failed to find the far transfer effect. The near transfer finding — practice makes you better at n-back — remained robust. That turned out not to be what anyone needed.

What Transfers to Real-World Performance

The picture is not entirely negative. Some interventions do show meaningful effects on working memory performance and appear to have real practical value:

Aerobic exercise. This is the clearest winner in the working memory literature. Multiple meta-analyses confirm that regular aerobic exercise improves working memory performance, with effect sizes in the moderate range (d ≈ 0.30–0.50). The effect appears to be mediated by prefrontal cortex blood flow, BDNF-driven neuroplasticity, and reduced stress hormones. Exercise also shows far transfer to other executive functions in a way that computerised training does not.

Sleep. Working memory is acutely sensitive to sleep deprivation — one of the most reliably degraded functions under sleep restriction. Protecting sleep quality and duration effectively raises functional working memory performance, even if it does not increase the underlying capacity ceiling.

Mindfulness meditation. A growing body of evidence suggests that sustained mindfulness practice improves the attentional control component of working memory — particularly the ability to prevent intrusive thoughts and irrelevant information from occupying working memory capacity. The effect size is modest but appears to transfer beyond the meditation task itself.

Strategies That Reduce Working Memory Demands

Rather than trying to increase the capacity of the container, another approach is to reduce how much you put in it. External working memory strategies are well-supported and widely underutilised:

• Externalise information — write things down, use visual diagrams, and create external representations of multi-step problems rather than trying to hold all elements in mind simultaneously
• Chunk information — group individual items into meaningful higher-order units before processing them (see: cognitive load theory)
• Reduce environmental load — limit notifications, background noise, and switching demands during cognitively intensive work
• Build domain expertise — expert knowledge compresses information into chunks, effectively expanding usable working memory within the domain

The Honest Prescription

If you want to perform better on tasks requiring working memory — tests, complex reasoning, multi-step problem-solving — the evidence-based interventions are:

1. Get adequate, consistent sleep (the highest leverage intervention)
2. Exercise regularly (aerobic, 150+ minutes per week)
3. Reduce acute cognitive load through external tools and environmental design
4. Practice mindfulness if attention regulation is a bottleneck

What the evidence does not support is buying a working memory training app and expecting broad cognitive gains. Train the system directly — through physiology and environment — rather than trying to expand the container through practice alone.

One population where the picture is somewhat different: children and adults with ADHD. Working memory deficits are a core feature of ADHD, and studies of Cogmed in ADHD populations have found improvements in parent and teacher ratings of attention and daily functioning — not IQ gains, but functional improvements in the contexts where working memory limitations are most disruptive. This does not validate the broad claims made for neurotypical populations. High-baseline deficits may respond to targeted remediation differently than normal working memory responds to enhancement, and the same intervention that does little for someone with typical working memory may provide meaningful benefit to someone whose working memory is substantially below typical.