To understand the factors contributing to changes in breastfeeding duration, we analyzed data from seven countries in Latin America and from Haiti to document changes in breastfeeding duration between 1986 and 2005. We used a novel method that permits the overall change to be separated into the portion attributable to changing population characteristics (e.g., greater urban population or increased maternal employment) and the portion resulting from changing breastfeeding behaviors within population subgroups (e.g., more breastfeeding among urban women). Our results indicate that in the low-to-middle-income countries studied, which are experiencing socioeconomic and demographic changes, improvements in breastfeeding duration occurred. These improvements are explained almost entirely by changing breastfeeding behaviors, which were particularly evident in certain subgroups of women, such as those with higher levels of education, and very little by changing population characteristics. The socioeconomic and demographic changes we studied that were previously associated with less breastfeeding no longer appear to have a large negative effect. Our findings show that individual behaviors are amenable to change and that changes in individual behaviors collectively contribute to positive national trends in breastfeeding.