Working memory is the mental workspace where you hold whatever you’re actively thinking about — a phone number while dialing, the first half of this sentence while reading the second. George Miller’s famous 1956 estimate of its capacity: about seven items, plus or minus two. (Modern estimates for genuinely independent items run closer to four. Either way: brutally small.)
The escape hatch is what Miller called chunking: recoding many raw items into one familiar unit. Try holding these twelve letters:
F B I C I A N S A I R S
Twelve items — over budget, painful. Now regroup: FBI · CIA · NSA · IRS. Four chunks, easy. Nothing about your memory capacity changed; the units changed. A chunk is a pointer into long-term memory — the pattern is stored out there, and working memory holds only the handle. Capacity is fixed in chunks but unbounded in content per chunk, which is why the same buffer that struggles with 12 letters comfortably holds four three-letter agencies (12 letters), or a chess master’s view of a whole midgame.
That last claim was tested in one of the most famous experiment series in cognitive psychology (Adriaan de Groot in the 1940s, sharpened by Chase & Herbert Simon in 1973 — yes, the AI pioneer). Show a real midgame position (~25 pieces) for 5 seconds: masters reconstruct nearly the whole board; novices manage 4–5 pieces. The question above asks what happened in the control condition — boards with the same pieces scattered at random. Your answer is a theory of what chess mastery is.