International health organizations recommend six months of exclusive breastfeeding (EBF) as optimal for infant health. Complementary feeding (CF) with any liquids or solids before this age risks increasing infant pathogen exposure and offsetting breast milk intake, which may increase risk of nutritional morbidity and early weaning. Globally, however, most infants begin CF well before six months---including in non-industrialized populations in which breastfeeding initiation and prolonged, on-demand nursing are universal. Why is early CF so common, even among populations in which prolonged EBF would be most protective for infants and feasible for mothers? Epidemiological research generally frames this discordance in terms of barriers to optimal practice and evolutionary research in terms of maternal-infant conflict---i.e. the benefits of reducing EBF costs for mothers may outweigh any risks for infants. An alternate and less-explored hypothesis is that early CF, if introduced without reducing breastfeeding intensity, may ultimately benefits infants.

I examine evidence of maternal-infant conflict and congruence in shaping patterns of early CF among the Tsimane. The Tsimane are an indigenous, high-fertility, high-mortality, forager-horticulturalist population residing in the Bolivian Amazon. Interviews, anthropometric, behavioral, and biomarker data were collected from a mixed-longitudinal sample of Tsimane mother-infant pairs from August 2012 -- April 2013. Tsimane mothers exhibited universal and prolonged breastfeeding, with EBF durations of about 4 months on average. Analysis of predictive factors and outcomes associated with age of CF introduction generally supported a Feeding Augmentation rather than a Feeding Substitution model of early CF. Mothers' reasons for introducing CF more often emphasized perceptions of infant needs than their own time or energy constraints. Earlier CF introduction (0-3 vs. 4-6 months) was associated with increased CF frequency at later ages, but not with subsequently lower frequency of breastfeeding bouts or earlier weaning.

Among mothers, shorter EBF durations were not associated with biological indicators of reduced lactational costs. While primiparity and high parity were both associated with age of CF introduction, these factors, and not age of CF introduction, were associated with timing of resumption of menstruation. Finally, while before six months of age infants who were introduced CF earlier were relatively smaller as compared to EBF counterparts, after six months of age mean height-for-age was higher for infants introduced CF at 0-3 vs. 4-6 months. Earlier CF was not associated with greater likelihood of reported illness before or after six months of age. Among the Tsimane early CF does not appear to benefit mothers at a cost to infants, and likely supplements, but does not supplant intensive breastfeeding. While the Feeding Augmentation Model was supported in this study, the Feeding Substitution Model may have more predictive power in populations in which bottle and formula feeding are more common. However, both models ultimately emphasize variability in optimal EBF duration, dually shaped by the needs of infants and mothers. This emphasis need not contradict current recommendations or related public health campaigns, but may be helpful in prompting a parallel dialog about optimal infant feeding practices for individual families.