In our 12-month observational study of three medical–surgical units, nurses responded to a median of 187 device alarms per shift, of which fewer than 12 percent required any clinical action. Response latency lengthened significantly over the course of a shift, particularly after the seventh hour. We do not claim that alarm volume causes patient harm; our design cannot establish that. What our data do suggest is that current default alarm thresholds impose a substantial and largely uncompensated attentional cost on bedside staff, and that this cost is invisible in most staffing models.Ninety-two words holding a number, a proportion, a scope limit, a refusal to claim cause, and one phrase worth borrowing. Each of those points to a different move.