We used linear regression analysis to determine if a previously reported negative association between BF and length gain was maintained throughout the second year of life. BF frequency, time between length measurements, diet, and diarrheal morbidity were used to predict 3-mo length gain for each 3-mo period. For 12-21 mo, an increase in BF was negatively associated with linear growth for toddlers who had low complementary diet intake and high incidence of diarrhea. This association disappeared only at 21-24 mo Then, BF had a positive (p<0.05) main effect on linear growth: an increase in BF from the 25th to 75th percentile values was associated with a 0.3-cm increase in 3-mo length gain. Mothers purposefully postponed weaning young toddlers in poor health, thus producing the observed negative association between BF and growth. By 21 mo, however, toddlers' diet and morbidity no longer demonstrated a similar effect on maternal feeding decisions. The results show that reverse causality (where poor child health determines BF status rather than BF causing poor child outcomes) is less common in the analysis of data from older children.