We used decision analysis to examine the strategies of amniocentesis, chorionic villus sampling, and no prenatal testing for a pregnant woman who would be 35 years of age at the expected date of delivery. Probabilities were obtained from the obstetric and genetic literature, and utilities from previously published standard reference gambles and from responses of obstetric residents and students recorded on a linear rating scale. The expected utility of amniocentesis exceeded that of chorionic villus sampling by 0.1 utility units, and of no prenatal testing by 0.12 utility units. The decision was insensitive to clinically plausible values for the probabilities of spontaneous abortion after amniocentesis and chorionic villus sampling, the probabilities of abnormal and indeterminate chorionic villus sampling results, the probability of an abnormal amniocentesis result after an indeterminate chorionic villus sampling, the sensitivities and specificities of amniocentesis and chorionic villus sampling, and the probabilities of significant maternal morbidity after first- and second-trimester therapeutic abortion. Chorionic villus sampling was preferred to amniocentesis when the utility of a first-trimester therapeutic abortion exceeded that of a second-trimester abortion by 23.2 utility units, or when the anxiety "cost" of awaiting second-trimester amniocentesis results exceeded 0.1 utility unit. We conclude that over a range of assumptions concerning the probabilities involved in the prenatal testing decision, amniocentesis was preferred to chorionic villus sampling. However, for a decision maker for whom a second-trimester therapeutic abortion would be significantly less acceptable than a first-trimester procedure, or for whom the anxiety of awaiting second-trimester chromosomal diagnosis might be an important consideration, chorionic villus sampling could become the procedure of choice.