Why a rounded figure travels further than a precise one

The study reported 38.7 per cent with a confidence interval. Within a week it is forty per cent everywhere, and within a month it is nearly half. Nobody decided to round it. Each teller passed on the version they could say without looking anything up.

What everyone sees

A figure is assumed to travel as published, with distortion attributed to carelessness or bad faith. Most of the loss is structural. Each retelling is a memory task, and what survives is what can be recalled and produced fluently — which systematically favours round values over exact ones.

What is actually happening

Heath, Bell and Sternberg studied why some accounts spread and found that transmission acts as a selection process: versions with properties that make them easy to retell are passed on more often, so the surviving form is shaped by the requirements of retelling rather than by the original. Dehaene and Mehler documented the underlying regularity in language itself, showing that number words follow a steep frequency distribution across languages with strong peaks at round values — the vocabulary of quantity is already biased towards the numbers that are easy to say and remember. Put together, a decimal enters a channel that has no slot for it. The round neighbour is available, sufficient and free to produce.

Why it stays hidden

The drift hides because every intermediate step is honest and approximately right. Rounding 38.7 to forty is a defensible simplification; rounding forty to nearly half is another. No single hop is wrong by much, and the chain is not recorded, so the final figure arrives with the authority of the study and none of its number.