Mothers of infants (n = 328) born between 1996 and 1999 in four Guatemalan communities were interviewed every 2 wk until 6 mo postpartum (pp) to collect prospective breast-feeding data and assess the association between delayed (>3 d pp) onset of lactation (OL) and the risk of ending full breast-feeding. Cox proportional hazards regression was used to examine the association between delayed OL and the hazard of ending full breast-feeding in the first 6 mo, adjusted for potential confounders and effect modifiers. A significant interaction with community was found (community-specific hazard ratios: HR(A) = 2.87, 95% CI = 1.25, 6.60; HR(B) = 3.43, 95% CI = 1.55, 7.59; HR(C) = 0.26, 95% CI = 0.06, 1.14; HR(D) = 1.11, 95% CI = 0.44, 2.77). Supplementation before OL (preonset supplementation) was associated with delayed OL [odds ratio (OR) = 4.87, 95% CI = 2.29, 10.36] and an increased risk of ending full breast-feeding (HR = 1.49, 95% CI = 1.05, 2.11). In the two communities in which mothers who experienced delayed OL had a significantly greater risk of ending full breast-feeding than did mothers experiencing normal OL, the association between delayed OL and full breast-feeding was mediated in part by preonset supplementation. Further analyses suggest that this was due to delayed OL leading to nonbreast milk feeds, rather than to introduction of supplements delaying OL. We conclude that some mothers who experience delayed OL are less likely to continue full breast-feeding in the first 6 mo and that further research should examine the contextual factors that modify this relationship.