When you hold a phone number in mind while walking across the room to write it down, you’re using working memory. When you follow a complex argument and track how each point connects to the conclusion, you’re using working memory. When you calculate a tip without paper, hold intermediate results, and apply them to the next step — working memory.
It is the cognitive workspace where active thinking happens. And it is severely limited.
The Limits of Working Memory
George Miller’s 1956 paper — “The Magical Number Seven, Plus or Minus Two” — established the classic claim: humans can hold approximately 7 items (plus or minus 2) in short-term memory. More recent work by Nelson Cowan suggests the true capacity is closer to 4 items, with Miller’s 7 being partially explained by chunking (grouping items into meaningful units).
Chunking is significant. A chess master can hold far more information about a chess position than a novice because they perceive it as patterns and chunks (“kingside castle with fianchettoed bishop”) rather than individual piece locations. Their working memory capacity hasn’t changed — their chunking ability has.
This is important because it means working memory effectiveness depends on both raw capacity and the quality of chunks stored in long-term memory. Experts think more efficiently not because they have more working memory, but because they use it more economically.
Individual Variation and What It Predicts
Working memory capacity varies significantly across individuals, and this variation is a strong predictor of:
- Reading comprehension (tracking the referents of pronouns, holding earlier context)
- Mathematical ability (holding intermediate results during computation)
- Fluid intelligence (the ability to reason in novel situations)
- Language acquisition (tracking syntactic structures)
Importantly, working memory capacity also predicts susceptibility to distraction. People with lower working memory capacity have more difficulty maintaining attention in the face of irrelevant stimuli — not because they lack discipline, but because suppressing irrelevant information is itself a working memory demand.
Can Working Memory Be Trained?
This is one of the most contested questions in cognitive psychology.
The most famous working memory training program is dual n-back (simultaneously tracking two streams of stimuli n steps back). Studies by Susanne Jaeggi in the late 2000s claimed dual n-back training transferred to improvements in fluid intelligence — an extraordinary claim that generated enormous excitement and significant controversy.
Subsequent attempts to replicate these findings have produced inconsistent results. A 2014 meta-analysis by Melby-Lervåg and colleagues found “no convincing evidence” that working memory training transfers to fluid intelligence or everyday cognitive tasks. The 2020 Consensus Statement from cognitive neuroscientists echoes this.
What is consistently found: people get better at the specific task they train on. Whether that improvement transfers is the question, and the evidence for far transfer remains weak.
What Does Help
Sleep. The hippocampus — critical for short-term memory encoding — is particularly sensitive to sleep deprivation. One night of poor sleep demonstrably reduces working memory performance. Chronic sleep restriction produces cumulative deficits.
Reduced cognitive load from external systems. Externalizing information — writing it down, using a trusted task list, organizing your workspace — reduces the working memory demands of your environment. GTD (Getting Things Done) is essentially a system for keeping your working memory clear by trusting the external system to hold context.
Chunking and expertise. The most durable way to expand effective working memory capacity is to develop expertise in a domain. Expert chunks compress information and free working memory for higher-level operations. A programmer who knows patterns deeply can hold more of a complex algorithm in mind than someone who is still thinking at the syntax level.
Reduced distraction. Working memory is easily disrupted by irrelevant stimuli. An interruption doesn’t just pause a task — it can displace the working memory contents that were active at the moment of interruption. Protecting working memory means protecting the environment from distractions.
Mindfulness meditation. Several studies have found that mindfulness training improves working memory performance, likely through its effects on attentional control and the ability to recover attention after mind-wandering. A 2010 study by Jha, Stanley, Kiyonaga, Wong, and Gelfand found working memory improvements in marines who completed an 8-week mindfulness course before deployment.
The Memory Grid as a Benchmark
The Memory Grid on this site is a simplified sequential memory task — not a clinical working memory assessment, but a reasonable benchmark for tracking your own performance over time. The sequence length at which you reliably succeed approximates your effective working memory span for this type of task.
Notice how your performance varies with sleep, stress, and time of day. These fluctuations in working memory performance are more informative than any single test score.
Test your working memory span with the Memory Grid — see how far you can push the sequence.