A college runner with an average memory learned to hold 80 digits
Psychologists tested him first to prove a limit. He spent 230 hours proving the limit was the wrong thing to measure.
In the late 1970s, psychologists William Chase and Anders Ericsson recruited a college student, known in the literature only as S.F., for a memory experiment. His intelligence and memory tested as unremarkable. The task was simple and merciless: listen to a string of random digits read at one per second, then repeat it back. Most people top out around seven digits, the famous limit of short-term memory.
S.F. trained on the task several times a week for more than 230 hours across twenty months. By the end, he could hold and recall strings of nearly 80 random digits in a row.
His short-term memory hadn't actually expanded — when researchers tested him on random letters instead of numbers, he was back to ordinary. What he'd built was a translation system. S.F. happened to be a competitive runner, so he started converting digit strings into running times: 3492 became "3 minutes 49.2 seconds, near world-record mile pace." He chunked those times into groups, then nested the groups into larger structures, building an entire private architecture out of a domain he already knew cold.
The number 80 isn't the point. The point is that he didn't get a better brain. He got better filing, built entirely from what he already understood, applied to something that had nothing to do with running at all.