De Wolff and van IJzendoorn's meta-analysis supported the relation of parental behavior to attachment security; however, they concluded that sensitivity was not the best predictor of attachment security. This secondary analysis of their data re-examined their conclusion and tested moderators of the relation between sensitivity and attachment, such as socioeconomic status and study methodology. A different grouping of studies combining behavior types of sensitivity, mutuality, and synchrony showed significantly higher correlations with attachment than did other types of maternal behavior. An analysis of a similar dataset of individual correlations defined sensitivity as synchronous, responsive, and appropriate behavior, producing similar results. Tests of moderators validated differences between studies using measures of sensitivity as compared to other measures of maternal behavior. Sensitivity was a weaker predictor of attachment in studies of low-income families.