What this Library got wrong.
Every correction made after publication is listed here, in full, with the date and what changed. Not a changelog of typos — a record of the times a claim in a finished, paid product turned out to be wrong.
A publisher who tells you to check their sources has to be checkable themselves. That means this page exists whether or not it flatters anyone.
2026
Short-sleeper prevalence: a trait figure attached to the wrong thing
What it said. That people with the DEC2 gene mutation function on six hours or fewer without impairment, and that they represent roughly one to three per cent of the population.
Why that was wrong. One to three per cent is an estimate for the trait — familial natural short sleep. It is not the prevalence of any single mutation. Attaching it to DEC2 implies one to three people in a hundred carry that specific variant, which the evidence does not support. The individual variants are far rarer: the NPSR1 short-sleep variant has been reported in fewer than one person in four million.
What it says now. The trait is named as the trait, estimated at roughly one per cent with some sources going to three. Four associated genes are named rather than one — DEC2/BHLHE41, ADRB1, NPSR1, GRM1. And it now records the finding neither version mentioned: biobank data turned up carriers of these same high-impact mutations sleeping perfectly normal hours. The gene does not guarantee the phenotype.
Confidence grade changed. This claim carried Settled. It should never have. Estimates vary and rest on small familial studies. It is now graded Emerging.
The argument survives. Take the most generous figure on the table — three per cent — and ninety-seven people in a hundred still are not short sleepers. The trainee who has trained himself to need less sleep is not making a genetic claim. He is making a guess, and the odds are ninety-seven to three against him.
Found something wrong? It belongs on this page. Send it and it gets an entry, with credit if you want it.