Abstract Large surveys tend to be strong on reliability and generalizability but weak on validity. The recently collected Early Childhood Longitudinal Study—Birth Cohort, a large and representative sample of over 10,000 infants assessed at 9 and 24 months, took a provocative step toward trying to improve the validity of measures of infants’ environmental conditions. Instead of relying solely on self-reported surveys, the study design employed two additional data collection methods—interviewer observations and a videotaped mother/infant interaction. We assess how much additional and unique information about mothering is gained by having this extra data. The percentage of variation in infants’ cognitive development we can explain with mothering behaviors doubles when using a multi-method approach versus relying solely on mothers’ self-reports. However, our multi-method model of mothering still only explains about two percent of the variation in infants’ cognitive development.