How to Improve Working Memory: 7 Science-Backed Strategies

Working memory is the cognitive system that holds and manipulates a small amount of information in an active, immediately accessible state. Baddeley and Hitch's 1974 model — still the dominant framework — divides it into a central executive (an attentional controller), a phonological loop (for verbal and auditory information), and a visuospatial sketchpad (for spatial and visual information). A fourth component, the episodic buffer, was added in 2000 to explain how information from multiple sources is integrated.

Working memory is fundamentally different from long-term memory. Long-term memory stores vast amounts of information for years or decades; working memory holds roughly 4±1 chunks for seconds before the information is either transferred, rehearsed, or lost. This small workspace is the bottleneck of cognition: reading comprehension requires holding earlier parts of a sentence while processing later parts; multi-step mental arithmetic requires retaining intermediate results; planning requires maintaining sub-goals while executing steps. When working memory fails, these processes all degrade together.

The first strategy is consistent daily practice at moderate duration. Research consistently shows that 10–15 minutes of focused working memory training per day outperforms longer sporadic sessions. The reason is consolidation: the brain requires time between sessions to stabilise the synaptic changes that training induces. Cramming a week's training into one session overloads the same circuits repeatedly without allowing recovery and integration.

The second strategy is progressive overload — always operating near capacity. Easy working memory tasks create no meaningful adaptation, just as lifting light weights creates no muscular growth. The key signal for the brain to invest in circuit-level changes is operating at or slightly beyond current capacity. If a task feels comfortable, increase the difficulty. The third strategy is dual-task training — managing two streams simultaneously. Tasks like tracking a sequence of numbers while solving a problem place greater demands on the central executive than either task alone, and studies show larger central executive gains from dual-task practice than from single-task training matched for total working memory load.

Sleep is the fourth strategy and the most undervalued. During slow-wave sleep, the hippocampus replays the day's memory traces and transfers them to cortical storage, a process called memory consolidation. Cutting sleep from 8 to 6 hours per night for two weeks produces deficits in working memory capacity — measurable on standard span tasks — that are equivalent to two full nights of total sleep deprivation. The subjective feeling of adaptation to mild sleep restriction is illusory; objective performance continues to decline even as people report feeling fine.

Aerobic exercise is the fifth strategy with the strongest biological mechanism. Twenty minutes of moderate-intensity cardio elevates levels of BDNF (brain-derived neurotrophic factor) — a protein that promotes the growth and maintenance of synaptic connections — specifically in the hippocampus and prefrontal cortex, the key working memory structures. Randomised controlled trials in children, young adults, and older adults all find that regular aerobic exercise improves working memory performance, with effect sizes that rival cognitive training directly.

Chunking is the sixth strategy and the highest return on investment. It involves reorganising incoming information into meaningful units that the brain encodes as single items. A phone number is not ten random digits but an area code, an exchange, and a subscriber number — three chunks instead of ten. Expert chess players perceive board positions as recognisable configurations of pieces rather than 32 independent locations; studies show their working memory span for chess positions is no larger than a novice's, but their chunk library is far richer, so fewer memory slots are needed to represent a full game position.

The seventh strategy is aggressive reduction of cognitive load through environmental design. Multitasking does not exist as a parallel processing mode — it is rapid task-switching, and each switch incurs a cognitive cost as working memory is partly cleared and reloaded with a new task context. Studies by David Meyer and colleagues estimate that task-switching degrades performance on the switched task by up to 40% in complex cognitive work. Removing distractions before a focus session — phone out of sight, notifications off, single browser tab — is not a preference but a performance choice with a measurable impact.

Near transfer — improvement on tasks closely similar to the trained task — is robust and well-replicated. Train digit span and your digit span improves; train spatial sequences and your spatial sequence performance improves. The more practically important question is far transfer: does working memory training make you more intelligent or capable in unrelated domains? Here the evidence is more modest.

A 2016 meta-analysis by Melby-Lervåg and colleagues concluded that far transfer to general cognitive ability is small and short-lived for most training programmes. The contexts where meaningful transfer is consistently found are: reading comprehension in children with language difficulties (where working memory is the bottleneck), symptom reduction in ADHD (where the trained inhibitory control is directly relevant), and age-related memory complaints in older adults (where the trained skill addresses a real functional deficit). The goal of working memory training should be improving the specific capacity you need, not expecting general intelligence enhancement.