To evaluate the feasibility of the use of serial ultrasound measurements of cervical length, membrane protrusion, and dilatation to discriminate between the competent and the incompetent cervix, 107 at-risk patients and 30 control subjects were examined prospectively. Patients were divided into five groups based on treatment and method of diagnosis. Epidemiologic, ultrasound, and outcome data were analyzed. Means and standard deviations for ultrasound measurements were established. Highly significant differences between all prediagnostic and postdiagnostic-pretreatment measurements were found (p less than 0.001). Highly significant differences were also found between all postdiagnostic-pretreatment and postdiagnostic-posttreatment measurements (p less than 0.001). No significant differences between prediagnostic and postdiagnostic-posttreatment measurements were noted. The incidence of preterm delivery was significantly higher among untreated diagnosed patients (p less than 0.01). By combined clinical and ultrasound criteria 51 patients (47.7%) were identified as not having cervical incompetency. Fifty-six patients (52.3%) were diagnosed.