Working Memory: What It Is and How to Improve It

Working memory is the small, temporary workspace your mind uses to hold and manipulate information while you're actively using it — the mental scratchpad where you keep a phone number long enough to dial it, follow the thread of a sentence, or do a calculation in your head. It's not the same as long-term memory, which is your vast archive of facts and experiences. Working memory is fast, fragile, and strikingly limited in capacity. And the honest answer to how you improve it is less about expanding the workspace and more about using a small one cleverly.

That distinction matters, because the way working memory gets marketed — "train it and watch your whole brain get faster" — overpromises in ways the research doesn't support. The good news is that the genuine, evidence-based strategies are more useful than the hype anyway.

What working memory actually is

The dominant model in cognitive psychology comes from Alan Baddeley and Graham Hitch, who proposed a multi-component picture of working memory in 1974. Rather than a single "buffer," they described a system with parts that each do a specific job:

• The central executive — the control system. It directs attention, decides what to focus on, and coordinates the other components. Think of it as the manager, not the storage.
• The phonological loop — a short-term store for verbal and acoustic information. It's why you can repeat a phone number to yourself; it holds sound-based information and refreshes it through silent rehearsal.
• The visuospatial sketchpad — the equivalent store for visual and spatial information: picturing a route, rotating a shape in your mind, remembering where something is on a page.

Baddeley later added an episodic buffer (around 2000) to explain how these streams get bound together into coherent, multi-modal chunks. But the original three-part structure is still the workhorse model taught and tested today. The key insight is that working memory isn't one thing — it's an attention-controlled system with limited stores, and attention is the bottleneck.

The real limit: not seven, but closer to four

You may have heard that human memory holds "seven, plus or minus two" items. That figure comes from George Miller's famous 1956 paper, "The Magical Number Seven, Plus or Minus Two." It's one of the most cited numbers in all of psychology — and it's also widely misunderstood.

Miller himself was somewhat playful about the number, and later research suggests the true capacity of focused working memory is smaller. In an influential 2001 review, Nelson Cowan argued that once you strip away rehearsal and grouping tricks, the genuine limit of what we can hold in the focus of attention is closer to four chunks. Much of what looked like a seven-item span was people quietly rehearsing and re-grouping, not holding seven separate things at once.

The practical takeaway isn't that your memory is worse than you thought. It's that capacity is measured in chunks, not items — and chunks are something you can build.

A chunk is a meaningful unit. The sequence F-B-I-C-I-A-N-A-S-A is ten letters — well past anyone's raw capacity. Grouped as FBI, CIA, NASA, it's three chunks, trivially easy. The letters didn't change; your encoding did. Chunking is the single most reliable way to do more with a fixed-size workspace, and it's why experts in any field seem to have superhuman memory for their domain: a chess master sees one meaningful position where a novice sees twenty scattered pieces.

Can you train working memory itself?

This is where honesty matters, because it's the most oversold claim in the brain-training world.

Here's what the research shows. In 2008, Susanne Jaeggi and colleagues published a study in PNAS reporting that training on a demanding "dual n-back" task — a working-memory exercise — produced gains on tests of fluid intelligence, with more training yielding bigger effects. The paper was exciting and widely cited, and it helped launch a wave of cognitive-training products.

But the follow-up evidence has been sobering. Later meta-analyses found that working-memory training reliably makes you better at the trained task and closely related tasks (near transfer) — but the evidence for far transfer, meaning gains that spill over into general intelligence, reading, or real-world performance, is weak. A 2013 meta-analysis by Monica Melby-Lervåg and Charles Hulme found short-term improvements on working-memory measures but "no convincing evidence" of reliable far transfer to broader abilities. A 2016 consensus review by Daniel Simons and colleagues reached a similarly cautious conclusion about brain-training programs in general.

So the n-back makes you better at the n-back. That's real, but it's not the IQ boost the marketing implies. We unpack this whole near-transfer-versus-far-transfer question in more detail in does brain training actually work?.

What actually helps in practice

If you can't easily expand the workspace, the smart move is to reduce the load on it and protect the attention that drives it. These strategies are mundane and well-supported — the opposite of magic, which is exactly why they work.

1. Chunk deliberately. Group information into meaningful units. Break a long number into pairs; turn a to-do list into a story; find the pattern instead of memorizing the raw sequence. You're trading raw capacity for structure.
2. Offload what you don't need to hold. Writing something down isn't cheating — it frees the central executive for the part that actually needs thinking. The goal is to spend your four chunks on the hard part, not on remembering step two.
3. Protect attention. Working memory is attention-limited, so anything that fragments attention shrinks it. Single-task the hard things. Every notification you respond to is a chunk evicted from the workspace.
4. Sleep. Sleep loss measurably degrades working-memory performance and the attentional control behind it. A tired central executive is a leaky one. (More on the daily basics in daily habits for a sharper mind.)
5. Practice the real task, not a proxy. If you want to follow complex arguments or hold a multi-step plan in your head, practice doing exactly that. Specificity is the whole lesson of the transfer research: you get good at what you actually do. The flip side is encouraging — your brain genuinely adapts to repeated, effortful practice, which is the core idea behind neuroplasticity.

This is part of why we built Rusty around short, varied daily reps rather than a single endlessly-repeated puzzle. Its memory and focus circuits give you targeted practice with the kinds of holding-and-manipulating tasks working memory uses — and because the format changes, you're training the underlying skill in context, not grinding one narrow game in hopes the benefit leaks somewhere useful.

Key takeaways

• Working memory is your mind's small, temporary workspace, best described by Baddeley and Hitch's model: a central executive directing a phonological loop and a visuospatial sketchpad.
• Its real capacity is smaller than the famous "7±2" — closer to four chunks of meaningful information at the focus of attention (Cowan, 2001).
• You can't easily enlarge the workspace, but chunking lets you pack far more into it.
• Training tasks like the n-back improve the trained task; far transfer to general intelligence is weak and contested, so be skeptical of "raise your IQ" claims.
• The reliable wins are practical: chunk, offload, protect attention, sleep, and practice the real skill you care about.

Want a few focused minutes of memory and focus practice a day? Download Rusty free on the App Store and run your first circuit in five minutes.