Crystal Ball

What if the purpose of memory is not to remember things?

We generally judge our memory on accuracy and completeness – and we are generally disappointed. Memory is jumbled, retouched and unreliable as a definitive record of the past. But a recent New Scientist suggests perhaps that’s because remembering is not what it’s for.

Thinking in evolutionary terms, what use is a perfect record of your entire past on the Serengeti plains? Not much. There would have been precious little time for introspection with four-legged food to chase and four-legged death to avoid. Not to mention increasingly cunning two-legged competition alongside.

Memory must have conferred a survival advantage – so it seems reasonable to think it developed from what other mammals probably have: recall of close shaves, sources of food and – if elephants are anything to go by – key life events: births, deaths and marriages.

And this is why dates get jumbled, memories get intertwined and autobiographical narratives develop in our heads – to guide us on what to do next, not produce a perfect historical record. Memory exists to better predict and guide our future.

Memory tells us who to trust, how to act and what might happen. Yes it’s flawed by inductive logic. Past performance is stricto sensu no guide to the future. But we remember what we need to remember – what’s useful for our future.

This difference in purpose is the big difference between computer memory and ours. Ours is constantly shuffled, refined and selected for its Bayesian predictive power, not its precision.

No wonder it’s sometimes cloudy; human memory isn’t a time capsule, it’s our crystal ball.