Objective: The aim of this study was to determine the value in the prediction of spontaneous preterm delivery of ultrasonographically measured cervical length measured between 14 and 24 weeks' gestation. Study Design: A retrospective cohort study examined cervical length by means of a two-stage procedure, transabdominal ultrasonography followed by transvaginal ultrasonography if cervical length was <30 mm. Results: A total of 6877 patients met inclusion criteria. Mean cervical length was 37.5 mm. Odds ratios for early preterm delivery (≤32 weeks' gestation) for patients with cervical lengths ≤10, ≤15, ≤20, ≤25, and ≤30 mm were, respectively, 29.3 (95% confidence interval, 11.3-75.8), 24.3 (95% confidence interval, 12.9-45.9), 18.3 (95% confidence interval, 10.8-31.0), 13.4 (95% confidence interval, 8.8-20.6), and 3.2 (95% confidence interval, 2.4-4.4). For early preterm delivery a cervical length of ≤15 mm had a positive predictive value of 47.6%, a negative predictive value of 96.7%, a sensitivity of 8.2%, and a specificity of 99.7%. Conclusions: A short cervix seen on a second-trimester sonogram was a powerful predictor of early spontaneous preterm delivery (≤32 weeks' gestation). Nearly 50% of patients with a cervical length ≤15 mm had an early spontaneous preterm delivery, which suggests that clinical trials of interventions (eg, cerclage) in this population are urgently needed. (Am J Obstet Gynecol 2000;182:1458-67.)