a whole. They have tended to occur first among those women at the forefront of changes in dominant social values and among those with the resources (whether it is time, energy, or money) to permit adoption of new feeding practices.

Examples of alternative feeding practices, such as the use of wet nurses or human milk substitutes, occur throughout recorded history (Fildes, 1986). The early twentieth century was marked, however, by an unprecedented increase in formula feeding, in part because the development of nutrition science coincided with a pervasive increase in the value placed on scientific products and processes (Rosenberg, 1976; Starr, 1982). As women asserted their rights for self-determination in public life, those who were wealthy enough and sufficiently in tune with contemporary values were adopting formula feeding for their infants.

A corresponding and consistent decline in breastfeeding is evident in data from U.S. fertility surveys (Hendershot, 1980, 1981; Hirschman and Hendershot, 1979), specific studies of infant feeding (Meyer, 1958, 1968), and market research surveys conducted by Ross Laboratories, 1 a manufacturer of infant formula (Martinez and Krieger, 1985; Martinez and Nalezienski, 1979, 1981; Martinez et al., 1981). Seventy-seven percent of the infants born between 1936 and 1940 were breastfed; the incidence declined during the subsequent decades to about 25% by 1970 (Hendershot, 1980, 1981; Hirschman and Hendershot, 1979; Meyer, 1958, 1968). Duration of breastfeeding declined as well, dropping from a mean of 4.2 months in the early 1930s to 2.2 months in the late 1950s.

Because this decline in breastfeeding was not uniform in all segments of the U.S. population, the demographic characteristics of the group of mothers who breastfed changed substantially (Table 3-1). Ethnic differences in rates of decline are especially striking. Rates fell sharply among blacks; there was a less pronounced decline among whites and Hispanics (Hirschman and Hendershot,


Although the subcommittee considers the data compiled annually by Ross Laboratories to be the best data on breastfeeding rates in the United States, these data have two limitations common to many survey data: sampling bias and response bias. The list from which the Ross sample is derived represents 85% of all new mothers in the United States. This list probably underrepresents any unregistered births and those not occurring in hospitals. Such births are more likely to be to economically disadvantaged mothers, including illegal aliens.

The response rate for the 1987 survey was 54% (Ryan and Martinez, 1989), a rate fairly typical of past surveys (Fomon, 1987). Responses are weighted before analysis to attempt to compensate for nonrespondents. Compared with all mothers in the U.S. population, the unweighted sample has a lower percentage of mothers who are black or Hispanic, unmarried, and less than age 24; have high school or lower education levels; and have family incomes of less than $25,000; but it is comparable in terms of maternal employment status and geographic region of residence.

The weighting procedures cannot correct for, and indeed may amplify, any response bias associated with breastfeeding status. If, for example, those black or Hispanic mothers who breastfeed are more likely to respond than those who do not breastfeed, the weighted data will give the false impression that a higher percentage of black and Hispanic women breastfeed than is actually the case.

The Ross Laboratories data probably are least representative of that segment of the U.S. population with the lowest income. Therefore, it is likely that the breastfeeding rates among participants in the Special Supplemental Food Program for Women, Infants, and Children (WIC), blacks, and mothers of lower socioeconomic class are even lower than the Ross data indicate.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement