During the last decade, an increased awareness of the importance of breastfeeding patterns in determining maternal fecundity and infant health has led to a dramatic rise in surveys that query mothers on suckling duration and frequency. However, with few exceptions, observational studies have not been undertaken, and the accuracy of the recalled information is not known. This analysis ascertained the concordance of maternal recall and observational data collected for a single sample of Andean women practicing on-demand nursing. Interviews of 30 women were conducted in their native language; for 10 mother-infant pairs, breastfeeding behavior was precisely timed during the course of normal daily activities for a total of 86 hours. There is virtually no agreement of recall and observational data, either for each individual or for sample statistics. Rounding is the rule; suckling duration is reported in 5-minute units when, in fact, women often nurse for only 2–3 minutes. Overestimation is also very common; suckling durations of 30 minutes are frequently reported but almost never observed. These reporting patterns obscure the considerable actual variation in suckling duration—exposed by observation—and bring into serious question the conclusions of studies based on mothers' recall. In particular, given that analyses of these observational data indicate otherwise, the claim that suckling duration is unimportant in the regulation of maternal fecundity should be re-examined. Further, the design, implementation, and evaluation of breastfeeding promotional programs intended to increase child spacing requires data of greater accuracy than that obtainable from maternal recall. © 1994 Wiley-Liss, Inc.