It is hypothesized that ampicillin may treat subclinical deciduitis and prolong the "effective" latent period in patients with preterm premature rupture of the membranes. We studied 82 patients with preterm rupture of membranes who were managed expectantly and were randomly assigned either to receive ampicillin prophylaxis (n = 43) or not to receive ampicillin prophylaxis (n = 39). Patients were excluded from study entry on admission if they had suspected or frank chorioamnionitis, active preterm labor, a history of penicillin allergy, multiple gestation, or cervical cerclage. There were no significant differences between the groups in duration of membrane rupture prior to admission, gestational age at membrane rupture, use of steroids and tocolysis, and demographic factors. Life-table analysis showed that the risk of delivery was significantly lower for the group of patients receiving prophylactic ampicillin. The incidence of neonatal infection was significantly lower in the ampicillin group, 1 (2%) versus 6 (17%), p less than 0.04.