This article examines retest reliability and digit preference in retrospective survey data on breastfeeding duration and type of supplementary food, covering three decades and reported by more than 1200 Malaysian women. Women with little or no education, rural residents, and those of Malay ethnicity are found to give less reliable data. In a logistic regression analysis, these respondent characteristics are more important determinants of data quality than the length of the recall period.
Data from the Malaysian Family Life Survey show an increase in the percentage of infants breastfed, at least initially, from 75 per cent in 1970-74 to 79 per cent in 1975-77. Contrary to what would be expected if Malaysia were following the trends observed in the United States and Western Europe, the increase has occurred among poor and uneducated women as well as among the more fortunate. The increase was especially marked for infants born in hospitals and private clinics, which had very low rates of breastfeeding in the early 1970s. The change may be due partly to a shift in the practices and recommendations of health professionals. Trends in infant feeding practices in Malaysia during the whole period 1950-77 are reviewed. Reasons for thinking the increase in the mid-1970s an artifact of the survey are presented and provisionally rejected. The implications of these findings for child health policy in Malaysia and for theories of infant feeding trends in developing countries are discussed.