212 women with twin pregnancies were randomly allocated either to receive advice to rest in hospital from 32 weeks' gestation until delivery, or to be part of a control group in which hospital admission was offered selectively (and, on average, 5 weeks later). Preterm delivery was more common among women admitted routinely for bed rest than among controls, and this difference was unlikely to have occurred by chance. There is at present no scientifically acceptable evidence that this common, disruptive, and expensive obstetric policy does more good than harm.