The aim of this study was to determine the predictive value of cervical length as a risk factor for spontaneous pre-term birth in twin gestations. A retrospective chart review was carried out on patients with twin pregnancies referred to our multiples' clinic. Cervical length was measured by transvaginal ultrasonography. Patients with an indicated pre-term delivery or intervention were excluded from the analysis. Outcomes included preterm delivery < 28 and < 35 weeks gestation. After extracting the data, 2 × 4 tables were constructed. Likelihood ratios were then generated for cervical lengths ≤2.0 cm, ≤2.5 cm, ≤3.0 cm, and > 3.0 cm. Because of the limited number of measurements taken < 25 weeks gestation, we elected to collapse the tables, thereby achieving more meaningful results. For measurements taken before 30 weeks gestation, a shorter cervix did predict delivery < 28 weeks gestation (likelihood ratios for cervical lengths ≤2.0 cm, ≤2.5 cm, ≤3.0 cm, and > 3.0 cm were 4.43, 1.94, 0.97, and 1.02, respectively). The probability of preterm delivery < 35 weeks gestation increased with decreasing cervical length (likelihood ratios for cervical length ≤2.0 cm, ≤2.5 cm, ≤3.0 cm, and > 3.0 cm were 2.58, 1.66, 1.38, and 0.81, respectively). A shorter cervix measured before 30 weeks gestation was a stronger predictor of preterm delivery < 28 weeks compared to < 35 weeks gestation. Cervical length was not predictive of preterm delivery if measured after 30 weeks. Cervical length is predictive of preterm delivery < 28 weeks and < 35 weeks gestation when measured before 30 weeks gestation. No trend was seen when measured after 30 weeks gestation. A prospective study is currently underway to confirm these results. Twin Research (2000) 3, 213–216.