Objective: To examine the frequency and timing of fetal death and its association with maternal serum alpha-fetoprotein (MSAFP) levels. Methods: Pregnancy outcomes were evaluated in 6927 predominantly middle-class women (83% white, 17% black) who had second-trimester MSAFP determinations performed in our laboratory. All cases of multiple gestation, preexisting fetal death, and fetal malformation were excluded. Results: The overall fetal death rate was 13 per 1000 (n = 90). Black women had a higher fetal death rate than white women (35.6 per 1000 versus 8.4 per 1000; P <.001). One hundred forty-eight women (2.1%) had an adjusted MSAFP multiples of the median (MoM) value of at least 2.5, which was not explained by multiple gestation, congenital anomaly, or preexisting fetal death. As the MSAFP increased, the fetal death rate increased (MoM less than 2.0,11 per 1000; MoM 2–2.49,29 per 1000; MoM 2.5 or greater, 95 per 1000; P <.001). Despite the increased risk of fetal death in the elevated MSAFP group, most fetal deaths (84%) occurred in women with levels below 2.5 MoM. Furthermore, the timing of fetal loss was significantly different between the group less than 2.5 MoM and the group at or above 2.5 MoM. Fetal death occurred at or after 26 weeks in 45% of the women with normal MSAFP, compared with only 14% of women with high MSAFP (at least 2.5 MoM) (P =.023). Conclusions: Women with unexplained elevations in MSAFP are at increased risk for fetal loss, with most of the losses occurring in the second trimester. Because many of these fetal deaths occur at gestational ages when the neonatal survival is very low, it is unlikely that antepartum fetal surveillance aimed at early delivery would substantially increase fetal salvage. © 1994 by The American College of Obstetricians and Gynecologists.