National statistics on the risk of infant mortality by birth weight were collected most recently in 1980 and 1960. (Infant mortality risk is the number of deaths of infants under 1 year of age per 1,000 live births.) In this 20-year period, the infant mortality risk (IMR) for single-delivery infants declined 53 percent, from 23.3 deaths per 1,000 live births to 11.0; 91 percent of this decline was due to lower IMRs within birth weight categories, and 9 percent was due to reduced frequency of low birth weight. The greatest reduction in neonatal mortality (under 28 days)--73 percent--occurred among infants of 1,500-1,999 grams (g) birth weight, whereas the greatest reductions in postneonatal mortality (28 days to under 1 year)--51 percent to 54 percent--occurred among infants of 3,500 g or more birth weight. Trends in IMR for black and white infants were similar, and the twofold gap between the races in IMR persisted from 1960 to 1980. For whites, reductions in the frequency of low birth weights contributed to the decline in the IMR. For blacks, the percentage of infants with birth weights of less than 1,500 g increased, and the total reduction in the IMR was attributable to lower birth weight-specific mortality risks. In some regions of the United States, failure to observe an increase in birth weight for blacks may be a reporting artifact, reflecting improved reporting of births of very small black infants in 1980. Examination of changes in perinatal mortality risks (from 20 weeks gestation to less than 28 days of life) did not suggest that infant mortality trends were substantially affected by changes in the distinction between fetal and neonatal deaths over the 20-year period. Reducing the number of low birth weight infants remains the greatest potential for future reductions in infant mortality.