PIP: )reliable statistical reporting systems for a few countries extending back to World War II show that perinatal mortality rates have continuously fallen. To understand the reasons for this trend, 2 kinds of studies must be distinguished: studies which relate observed perinatal mortality rates or changes in rates to social, demographic, and other factors; and studies which set out "experimentally" to manipulate related factors and to study the effect on perinatal mortality rates. This discussion focuses exclusively on observational studies, covering factors associated with birth weight; components of the birth weight distribution; population studies; joint effects of several factors; birth weight, gestation, and mortality; and perinatal mortality. There are 2 distinct approaches to the study of birth weight: one that uses a threshold value, usally 2500 gm, and studies the proportion of births below that value; and one that studies differences in mean birth weight. Since, for many purposes, there is a particular interest in the very light babies, a direct measurement of their number is often the most appropriate approach. Which component procedures may be able to provide useful summary graduations of a birth weight distribution, their claims to represent substantive reality need ot be viewed critically. There are numerous studies of birth weight and perinatal mortality, but there are relatively few which have used adequate random sampling techniques allowing valid statements to be made about well defined populations. Attention is directed to results from a few large scale population studies, particularly the resultsof the 1958 British Perinatal Mortality survey. For both age and parity there was a change in the spread of birth weights as the mean birth weight rises, and this is reflected in the relationship between these factors and perinatal mortalty. The common practice of presenting percentiles of birth wieght for given gestation length is questionble since it suggests that babies at the same percentile value are in some sense equivalent. The use of birth weight alone gives a better prediction of perinatal mortality as can be seen by the manner in which the high mortality contours tend to become parallel to the gestation axis. For purposes of comparability, the 1973 World Health Organization (WHO) studies shows clearly that the use of separate late fetal death rates and neonatal death rates can be misleading. The use of a perinatal rate overcomes this problem. A table shows how perinatal mortality rates vary across the categories of several factors for Cuba, New Zealan, and Sweden in the 1973 WHO study. From the point of view of predicting high perinatal mortality risk, the data suggest that twins, high age and high parity mothers, and short interpregnancy intervals are common to all countries.