Working memory is the small, loud workspace where thinking actually happens: the few things you're holding right now while you reason about them. It is not storage. It's a desk, and the desk is much smaller than anyone expects.
Most people know the number seven, from Miller's 1956 paper. The better estimate is worse. Nelson Cowan's 2001 review, gathering evidence from experiments designed to stop people from quietly grouping items, put the real limit at around four — three to five, depending on the task and the person. Seven was measured under conditions where participants were already compressing. Four is closer to the raw capacity.
Four. That's the workspace you run your entire intellectual life from.
It isn't one thing
Baddeley and Hitch's model, still the standard framework, splits working memory into parts rather than treating it as a single pool:
- The phonological loop — the inner voice, holding words and sounds. It's why you can repeat a phone number to yourself, and why someone talking near you wrecks your reading.
- The visuospatial sketchpad — the inner eye, holding shapes, layouts, and where things are.
- The central executive — the controller that allocates attention between them and reaches into long-term memory.
- The episodic buffer, added later, which binds the streams into a single scene.
The practical consequence is that the two stores interfere within themselves more than across themselves. Listening to a lecture while reading text competes hard; listening while glancing at a diagram competes much less. That's the mechanism underneath dual coding: a word and a picture use different channels, so you get two routes to the memory for less than double the load.
Every open loop costs a slot
Here's the part that matters outside a psychology lab. Working memory doesn't only hold what you're thinking about. It holds what you're trying not to forget.
"Reply to that email." "Book the thing before Friday." "Ask them about the invoice." Each of those is a small background process, and each one is spending capacity from a budget of about four. That's why a day with six unresolved commitments feels foggy even when you're doing nothing hard: you're thinking with two slots because four are busy holding your to-do list.
This is the honest, unglamorous case for writing things down. Not tidiness, not productivity aesthetics — offloading storage to something outside your head gives the slots back. People describe a trusted capture habit as calming, and this is why. The relief is measurable capacity, not a mood.
The condition is that the external store has to be believed. If some part of you suspects the note won't be findable, the loop stays open and keeps its slot — you've done the work and kept the cost. That's the failure mode behind digital amnesia, and it's why retrieval, not capture, is the real test of a notes system.
What a four-slot desk implies
- Capture must be near-instant. A thought competing for a slot is lost to the next interruption. Fast capture isn't a convenience feature; it's the whole mechanism.
- Chunk, don't cram. You can't add slots, but you can make each one carry more — that's chunking, and it's the reason experts appear to have larger memories than they do.
- Design instruction for the ceiling. Material that arrives faster than it can be chunked overwhelms the workspace and nothing lands. That ceiling is the entire subject of cognitive load theory.
- Don't split the loop. Switching tasks doesn't hand you a clean desk; part of the last task stays on it as attention residue.
Working memory is the constraint the rest of memory and focus works around. You can't expand it. You can stop spending it on things a note would hold for free.