The "fetuses at risk" concept of gestational-age-specific mortality proposed by Yudkin et al. (Lancet 1987;1:1192-4) and extended by various researchers is becoming popular in perinatal and pediatric epidemiology. However, the definitions using this concept have led to a puzzling phenomenon in which mortality rates appear to increase monotonically with advancing gestational age. While proponents of this concept have highlighted the rationale for using as the denominator the number of fetuses at risk rather than that of fetuses delivered at a particular gestational week, many have not discussed how the numerator is derived. This article reviews various definitions of gestational-age-specific mortality rates. It points out that the definitions based on the concept of Yudkin et al. are indicators of short-term risk only. Furthermore, the seemingly monotonic positive association between mortality rates and gestational age is a misinterpretation. All of the definitions are valid when used and interpreted correctly. The choice of which definition to adopt depends on the specific issues one attempts to address. Calls for abandoning the conventional definitions are not justified.